JP6044977B2 - Optimal index generation device, optimal index generation method, optimal index generation program, and optimal index generation server - Google Patents

Description

  The present invention relates to an optimum index generating apparatus for supplying index information related to a plurality of apparatuses to a production management system that controls a plurality of apparatuses to produce a wide variety of products.

Conventionally, in a manufacturing process in semiconductor / liquid crystal manufacturing (hereinafter referred to as semiconductor manufacturing, etc.), if only processing conditions such as processing time are designed and managed, product quality as designed can generally be ensured.
However, as processing dimensions become smaller (in the order of several tens of nanometers) and unit products become larger, product quality as designed unless the residence time between one process and the start / end of another process is strictly controlled. Can no longer be achieved. This upper limit value or lower limit value of the staying time is called a Q-time constraint (or time constraint), which is an indispensable manufacturing condition for manufacturing all products.

Semi-finished products that fail to comply with Q-time constraints during production are reworked (reprocessed) or discarded, so Q-time constraint cracking is a waste of quality, value, cost, time, resources, energy, etc. Has been a big loss.
As such a loss, for example, the unit price of a large-diameter / highly integrated product wafer is high from 1 million yen to several tens of million yen. In addition, the unit price of production equipment is as high as several tens of millions of yen to several billions of yen, and the depreciation costs account for more than 50% of the annual production cost, so it is extremely difficult to secure the production volume of good products per unit time This is an important issue.
In addition, the occurrence of rework increases the processing load and leads to an increase in secondary rework.

  Therefore, in semiconductor manufacturing, compliance with Q-time constraints is regarded as the first priority production target and is more important than other major production targets (observation of delivery date, improvement of throughput, reduction of setup, etc.) Often (from interviews with domestic factories and prior literature).

  Problems related to the Q-time constraint are becoming more and more important and becoming a common issue in the industry as processing dimensions are reduced according to the technology roadmap and processing units are enlarged (expensive). Since the mid-2000s, it has been actively researched mainly by companies as a production scheduling problem.

In this way, while the importance is recognized, the optimal production management method (especially the production scheduling method) for achieving other production goals while observing the Q-time constraint has not yet been established. is there.
The reason for staying in this situation is mainly due to the following three problems.

  As a first problem, it is necessary to limit the production flow rate in order to comply with the Q-time constraint, and this limitation is a requirement contrary to achievement of other production targets (for example, throughput improvement). Therefore, multipurpose optimization of multi-product production is difficult. In addition, analytical processing is difficult because the processing unit of the manufacturing process is a batch, the processing unit is changed, or a multitask type in which a plurality of different processes are performed internally.

As a second problem, the preventive maintenance of the manufacturing apparatus, which is the main cause of the Q-time constraint cracking, has not been optimized.
During the preventive maintenance period, the number of operating units decreases, and the production capacity of the process (device group) significantly decreases. In addition, after the preventive maintenance is completed, the apparatus is operated at a high operation rate, which causes a large change in the logistics of subsequent processes. In such a situation in which fluctuations in production capacity and physical distribution are large, Q-time constraint cracking is likely to occur, and it has a secondary effect on other production targets. However, in the conventional production management system, the primary management mainly handling production logistics and the secondary management dealing with preventive maintenance of the apparatus are independent in business, and comprehensive management of both is not performed. For example, in the preventive maintenance plan, the purpose is to minimize the cost related to preventive maintenance, and the Q-time constraint is not taken into consideration [Non-Patent Document 2 etc.].

As a third problem, since several tens of Q-time constraints are scattered in the manufacturing flow and mutually interfere with each other, the entire simultaneous management is indispensable. In addition, a high-speed calculation method is required to manage the entire manufacturing flow of several hundred processes in real time. There is no general-purpose methodology itself for the structure analysis and simultaneous management methods of multiple Q-time constraints.
As for the Q-time constraint, for example, there are 60 or more Q-time constraints on the memory manufacturing flow of about 700 steps. In the case of mutual interference, there arises a problem that one Q-time constraint can be observed but the other Q-time constraint cannot be observed.
As a fourth problem, a problem common to the prior art is that a calculation time adaptable to real-time management cannot be realized.
For example, the calculation order of the mixed integer programming applied in Non-Patent Document 1 and Non-Patent Document 2 is exponential time, and therefore limited to a production schedule for a production period of about one week for a part of the production process. Otherwise, it cannot be calculated in a realistic time. Like semiconductor manufacturing, the production period is as long as 1.5 to 2.5 months, and the number of processes included in the production process is 300 to 1000 processes. A simple calculation method is required. In addition, high-speed calculation is required to realize real-time recalculation to cope with dynamic changes that occur daily.

  In the prior art, there are many prior arts related to the first problem. Broadly speaking, there are a method for managing the stay time of each lot (Non-Patent Document 1) and a method for managing the stay time by the number of lots (Patent Document 1, Patent Document 2).

In Patent Document 1, a production management system limited to Q-time constraints from the completion of the previous process to before the start of construction for the purpose of observing the Q-time constraint values and achieving the target production volume for semiconductor pretreatment factories. Has proposed.
After calculating the average processing time of the device group i from the average processing time of all the processes sharing the same device group i, the in-process lot number upper limit WL (i) for keeping the Q-time constraint time is calculated, The magnitude of risk is determined by comparing the expected value Wρ of the number of waiting lots of the device group i calculated using the queue M / M / 1 model with the upper limit value WL (i). Quantitative evaluation is performed using the slack and operation rate of the process, and the equipment is classified according to the level of possibility (risk) of not complying with the Q-time constraint as an improvement guideline. With respect to the process processing time, the number of processes, and the number of devices, a plurality of improvement plan candidates are calculated and their feasibility is determined.

  In Patent Document 2, work information from a processing device is linked to a host computer of a conventional transfer system, an appropriate reference stock is calculated, and the computer issues a transfer instruction by itself to automatically increase the work quality assurance time. maintain. As a result, an automatic management system capable of solving the work quality assurance time, the so-called Q-time overtime, and carrying the work capable of maintaining the quality assurance time in the conveyance of the work used for manufacturing the liquid crystal display device. It has been reported.

  Non-Patent Document 1 reports the results of joint research at National Taiwan University, Connecticut University, and Inota Terra Memories. In the three-step model in series, two Q-time constraints from the end point of the first step to the start point of the second step and from the end point of the second step to the start point of the third step are continuous. A single Q-time constraint section is targeted. We propose a method that solves the assignment of jobs to devices as a mixed integer programming problem, and shows an improvement effect compared with heuristics. The heuristic that is the object of comparison is an empirical method that starts the process of the second process when the remaining process time of the third process reaches 60 minutes. Results of improved evaluation index compared to heuristics in the proposed method (3rd operation rate increased by 16%, production volume increased by 15%, second process waiting time is 1/4, average waiting time is approximately 1 / 2). Data from actual factories are used for comparison experiments.

  Non-Patent Document 2 reports the results of joint research with US universities, companies, and ISMI members. An architecture and an example of use of a preventive maintenance optimization software tool (PMOST) based on an algorithm for optimal scheduling of preventive maintenance (PM) tasks in semiconductor manufacturing are shown. The PM schedule is calculated as a result of adjusting the base schedule given from the outside using a mixed integer programming problem for the purpose of minimizing the cost of the PM task. The simulation results of four cases using actual industrial data and factory models were shown. From these results, it was shown that the PM work was strengthened and the equipment productivity was improved. It also provides templates and guidelines for implementing other PM optimization algorithms used in PMOST.

JP 2008-77370 A JP 2003-108213 A

Optimal Wet-Furnace Machine Allocation for Daily Fab Production Proceedings of the 18th IEEE International Symposium on Semiconductor Manufacturing (ISSM2010), pp.47-50, 2010 October 18-20. Optimal Preventive Maintenance Scheduling in Semiconductor Manufacturing Systems: Software Tool and Simulation Case Studies IEEE Transaction on Semiconductor Manufacturing, vol.23 No.3, pp.477-489, 2010 August.

  However, even if these methods are used, it is difficult to say that a method for achieving good product throughput improvement and cost reduction while reliably observing the Q-time constraint in multi-product production has been established, maintaining and improving productivity. It is a factor that hinders.

For example, in patent document 1, since the modeling precision is low, there existed a problem that a Q-time restrictions crack could not be prevented.
In Patent Document 2, an emphasis is placed on a management system (information system) such as conveyance, and the technique lacks practicality from the viewpoint of improving production efficiency.

  In Non-Patent Document 1, the situation is that Q-time constraints are easily broken, and optimization is aimed at without dealing with the lower limit of Q-time constraints. However, when a large number of Q-time constraints are handled, the number of combinations becomes enormous. There was a problem that high-speed calculations that could be used in real time could not be expected.

On the other hand, there is no similar technique related to the second and third problems.
For example, in Non-Patent Document 1, only one typical type of Q-time constraint section is an issue. In Patent Document 1, four types of structures are taken up as problems, but only one type of structure provides the management method. As a result, neither of the prior arts provides a method for comprehensively defining and managing all Q-time constraint structures.

  The present invention has been made in view of the above circumstances, and even in a situation where Q-time constraint cracking, which has not been achieved by the conventional method, is likely to occur, the Q-time constraint is reliably observed, and good product throughput is improved. The objective is to provide optimum operation management conditions that can achieve cost reduction and reduction of environmental load and resource / energy consumption.

  In addition, in relation to the above purpose, the analysis method and appropriateness are also considered for the main causes of Q-time constraint cracking, such as handling of multi-product production, preventive maintenance of equipment, and mutual interference of multiple Q-time constraints. It is an object of the present invention to provide an easy management method to improve the throughput of non-defective products while simultaneously reducing the cost and reducing the environmental load while complying with the Q-time constraint.

  In order to solve the above-mentioned problem, the invention according to claim 1 is an optimal method for supplying index information relating to the plurality of devices and products to a production management system that controls a plurality of devices to produce a wide variety of products. An index generation device, an information input unit for inputting information about the plurality of devices and products via the production management system, and a limit value of a staying time between processing steps by the plurality of devices from the input information Q-time structure analysis unit for analyzing the Q-time structure indicating the optimal number of kanbans and buffer size for each product type based on the input information and the Q-time structure An index calculation unit that outputs the calculated information as index information, a batch assembly waiting time related to product arrival, and an upper limit value of a Q-time constraint An optimal load rule determination unit that calculates a boundary value of an arrival rate at which the load is switched and determines an appropriate load rule, and includes the optimal index information and the load rule calculated by the index calculation unit in the production management system It is an optimal index generation device characterized by being supplied.

  The invention according to claim 3 is an optimum index generation method for supplying index information about the plurality of devices and products to a production management system that controls a plurality of devices to produce a wide variety of products, An information input step for inputting information on a plurality of devices and products via the production management system, and analyzing a Q-time structure indicating a limit value of a staying time between processing steps by the plurality of devices from the input information Q-time structure analysis step, an optimal theoretical value calculation step for calculating the optimum number of kanbans and buffer sizes for each product type based on the input information and the Q-time structure, and the calculated information Magnitude relationship between index calculation step output as index information, batch assembly waiting time related to product arrival and upper limit value of Q-time constraint An optimal load rule determination step for calculating a boundary value of the arrival rate to be switched and determining an appropriate load rule, and executing the optimal index information and the load rule calculated by the index calculation step in the production management system It is an optimal index generation method characterized by supplying.

  The invention according to claim 11 is an optimum index generation server for transmitting index information about the plurality of devices and products via a communication line to a production management system that controls a plurality of devices to produce a wide variety of products. An information receiving unit that receives information about the plurality of devices and products via the production management system, and Q indicating a limit value of a stay time between processing steps by the plurality of devices from the received information A Q-time structure analysis unit that analyzes a time structure, an optimal theoretical value calculation unit that calculates an optimal number of kanbans and a buffer size for each product type based on the received information and the Q-time structure, and The magnitude relationship between the index calculation unit that outputs the calculated information as index information, the batch assembly waiting time related to product arrival, and the upper limit value of the Q-time constraint is broken. An optimal load rule determination unit that calculates boundary values of alternative arrival rates and determines appropriate load rules, and transmits the optimal index information and load rules calculated by the index calculation unit to the production management system This is an optimal index generation server characterized by

  The invention according to claim 12 is executed by a server that transmits index information about the plurality of devices and products via a communication line to a production management system that controls a plurality of devices to produce a wide variety of products. An optimal index generation method, wherein an information receiving step of receiving information about the plurality of devices and products via the production management system, and a restriction on a stay time between processing steps by the plurality of devices from the received information A Q-time structure analysis step for analyzing a Q-time structure indicating a value, and an optimal theoretical value calculation for calculating an optimal number of kanbans and a buffer size for each product type based on the received information and the Q-time structure An index calculating step for outputting the calculated information as index information, a batch assembly waiting time related to product arrival, and Q an optimal load rule determination step of calculating a boundary value of an arrival rate at which a magnitude relationship with the upper limit value of the time constraint is switched, and determining an appropriate load rule, and performing optimal index information calculated by the index calculation step And the load rule are transmitted to the production management system.

According to a thirteenth aspect of the present invention, optimal index generation for supplying index information about a plurality of devices to a production management system for controlling a plurality of devices provided in a production execution system for producing a wide variety of products. An information input unit for inputting information on a plurality of devices included in the production execution system via the production management system, and a bottleneck in a processing step by the plurality of devices from the input information A bottleneck analysis unit that analyzes a location and a maximum flow rate, and a Q-time constraint structure analysis that analyzes a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the input information A device classification calculation unit that divides the plurality of devices into a regular device having a regular capability and a substitution device having a alternate capability, and the calculated device category To calculate the operation number and operation timing of the plurality of conventional devices and alternative device from the input information, the bottleneck position, maximum flow rate, the start and end structure and said device classification sections according to the Q-time constraints based on the timing information, the optimum theoretical value calculator for calculating the kind-based optimal Kanban number and buffer size, and time index calculation unit for calculating the preventive maintenance start time related with the conventional device and the alternate device, before Symbol calculated An index calculation unit that outputs the optimal theoretical value as index information, and a procedure execution control unit that simultaneously manages all Q-time constraints scattered in the entire manufacturing flow included in the production execution system, The start / end time information calculated by the time index calculation unit and the optimum index information calculated by the index calculation unit are used as the production management system and the energy. It is optimal indicator generation unit and supplying the-engineering business system.

According to a fourteenth aspect of the present invention, optimal index generation for supplying index information relating to a plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. An information input step of inputting information on a plurality of devices included in the production execution system via the production management system, and a bottleneck in a processing step by the plurality of devices from the input information A bottleneck analysis step for analyzing a location and a maximum flow rate, and a Q-time constraint structure analysis for analyzing a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the input information A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and a alternate device having a alternate capability; To calculate the estimated by a device category as the input information to the plurality of common devices and operating times of the substitute system operating time as operating conditions, the bottleneck position, maximum flow rate, interval according to the Q-time constraints The optimum theoretical value calculating step for calculating the optimum number of kanbans and buffer size for each product type based on the structure of the apparatus, the device classification and the time index, and the time index for calculating the preventive maintenance start timing for the regular device and the alternative device a calculation step, and the index calculation step of outputting the optimum theoretical value before Symbol is calculated as the index information, and the execution, the time index calculation start end timing information calculated by the step, which is calculated by the index calculation step optimum Optimal index characterized by supplying accurate index information to the production management system and engineering business system Is an adult way.

The invention according to claim 18 is directed to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products, and provides index information relating to the plurality of devices via a communication line. An optimal index generation server for transmitting, an information receiving unit for receiving information on a plurality of devices included in the production execution system via the production management system, and a processing step by the plurality of devices from the received information A bottleneck analysis unit for analyzing a bottleneck location and a maximum flow rate in the network, and a Q for analyzing a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the received information A time constraint structure analysis unit, a device classification calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability, and Calculates the running count and the operation timing and operation conditions of the from the information received and the constant by the apparatus divided plurality of conventional devices and alternative device, the bottleneck position, maximum flow rate, the section related to the Q-time constraints Based on the structure, the device classification, and the start / end time information, the optimum theoretical value calculation unit for calculating the optimum number of kanbans and the buffer size for each product type, and the start / end time information of preventive maintenance related to the regular device and the alternative device and time index calculation unit that calculates a before and Symbol index calculation unit for outputting the calculated optimum theoretical value as the index information includes a starting end timing information calculated by the time index calculation unit, by the index calculation unit An optimal index generation server that transmits the calculated optimal index information to the production management system or the engineering business system.

  According to a nineteenth aspect of the present invention, index information relating to a plurality of devices is provided via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. An optimal index generation method executed by a server that transmits, an information reception step of receiving information on a plurality of devices included in the production execution system via the production management system, and a plurality of the plurality of information from the received information The bottleneck analysis step for analyzing the bottleneck location and the maximum flow rate in the machining process by the apparatus, and the structure of the section related to the Q-time constraint in which one or more Q-time constraints exist from the received information Q-time constraint structure analysis step of analyzing the bottleneck location, the maximum flow rate and the section related to the Q-time constraint An index calculation step for calculating the optimum number of kanbans and buffer sizes for each product type and outputting them as index information based on the structure, and dividing the plurality of devices into a regular device having a normal capability and a substitute device having a replacement capability A device category calculation step, an optimal theoretical value calculation step for calculating the number of operations, operation timing, and operation conditions of the plurality of regular devices and alternative devices from the calculated device category and received information, and the regular device A time index calculating step for calculating a preventive maintenance start time related to the alternative device, and executing the start / end time information calculated by the time index calculating step and the optimum index information calculated by the index calculating step. It is an optimal index generation method characterized by being transmitted to a management system or an engineering business system.

  The invention according to claim 20 is an optimum index generation device that supplies index information related to the plurality of devices to a production management system that controls a plurality of devices for producing a wide variety of products. An information input unit for inputting information on the devices of the system through the production management system, a device category calculating unit for classifying the plurality of devices into a regular device having a regular capability and a substitute device having a alternate capability, and the calculated The optimum theoretical value calculation unit for calculating the number of operations, operation timing, and operation conditions of the plurality of regular devices and alternative devices from the input device classification and the input information, and the calculation from the calculated operation frequency, operation timing, and operation conditions A start / end time calculation unit for calculating start / end time information on the regular device and the alternative device, and the optimum start / end time information calculated by the start / end time calculation unit It is optimal indicator generation unit and supplying the production management system and engineering business system.

  The invention according to claim 22 is an optimum index generation method for supplying index information relating to the plurality of devices to a production management system that controls a plurality of devices for producing a variety of products. An information input step for inputting information on the device through the production management system, a device category calculation step for classifying the plurality of devices into a regular device having a regular capability and a substitute device having a alternate capability, and the calculated The optimum theoretical value calculating step for calculating the number of operations, the operation timing, and the operation conditions of the plurality of regular devices and alternative devices from the input device classification and the input information, and the calculation from the calculated operation count, operation timing, and operation conditions A start / end time calculation step of calculating start / end time information regarding the regular device and the alternative device, and the calculation is performed by the start / end time calculation step. It is optimal index generation method and supplying the optimal start and end time information to the production management system and engineering business system.

  The invention described in claim 25 is an optimum index generation server that transmits index information about a plurality of devices via a communication line to a production management system that controls a plurality of devices for producing a wide variety of products. An information receiving unit that receives information related to the plurality of devices via the production management system, and a device classification calculating unit that divides the plurality of devices into a regular device having a regular capability and a substitute device having a alternate capability. An optimal theoretical value calculation unit for calculating the number of operating times, operating time and operating conditions of the plurality of regular devices and alternative devices from the calculated device classification and the received information, and the calculated operating frequency and operating time And a start / end time calculation unit for calculating start / end time information on the regular device and the alternative device from the operating conditions, and an optimum opening calculated by the start / end time calculation unit. It is optimal indicator generation server and transmits a completion timing information to the production management system and engineering business system.

  The invention described in claim 26 is the optimum that is executed by a server that transmits index information about the plurality of devices via a communication line to a production management system that controls a plurality of devices for producing a wide variety of products. An index generation method, wherein an information receiving step of receiving information on the plurality of devices via the production management system, and dividing the plurality of devices into a regular device having a regular capability and a substitute device having a alternate capability A device category calculation step, an optimum theoretical value calculation step for calculating the number of operations, operation timing, and operation conditions of the plurality of regular devices and alternative devices from the calculated device category and the received information; and the calculated operation A start / end time calculating step of calculating start / end time information regarding the regular device and the alternative device from the number of times, the operation time, and the operation condition, It is optimal index generation method and transmitting the optimal start and end time information calculated by the end time calculating step in the production management system and engineering business system.

According to a twenty-seventh aspect of the present invention, optimal index generation for supplying index information about a plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. An information input unit for inputting information on a plurality of devices included in the production execution system via a production management system, and a restriction on a stay time between processing steps by the plurality of devices from the input information A Q-time structure analysis unit that analyzes a structure of a section related to a Q-time constraint indicating a value, and a device category calculation unit that classifies the plurality of devices into a regular device having a regular capability and a substitution device having a alternate capability , calculates the running count and the operation timing and operation conditions of said plurality of conventional devices and alternative device from the filled in calculated by the apparatus sorting information, which is the input Distribution, based on the structure and the device category as the start and end time information of a section according to the Q-time constraints, the optimal theoretical value calculator for calculating the kind-based optimal Kanban number and buffer size, and the conventional apparatus the start end timing calculating section for calculating the start and end time information on a replacement device, and the index calculation unit for outputting the optimal theoretical value before Symbol is calculated as the index information comprises, calculated by the start and end timing calculating section Optimal start / end time information is supplied to the production management system and engineering business system, and optimal index information is produced by the index calculation unit for each period with different process capabilities determined by the calculated start / end time information. An optimal index generation apparatus characterized by being supplied to a management system.

According to a twenty-ninth aspect of the present invention, optimal index generation for supplying index information about a plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. An information input step for inputting information on a plurality of devices included in the production execution system via a production management system, and a limitation on a stay time between processing steps by the plurality of devices from the input information. A Q-time structure analysis step for analyzing a structure of a section related to a Q-time constraint indicating a value, and a device classification calculation step for dividing the plurality of devices into a normal device having a normal capability and an alternative device having a replacement capability , together calculating the running count and the operation timing and operation conditions of said plurality of conventional devices and alternative device from the filled in calculated by the apparatus type information Information said inputted, based on the Q-time structure of the section according to the constraints and the device category as the start and end time information, the optimum theoretical value calculation step of calculating the kind-based optimal Kanban number and buffer size executes a start end timing calculating a start and end time information on said common device and an alternative apparatus, the index calculation step of outputting the optimum theoretical value before Symbol is calculated as the index information, and the start and end time Optimal start / end time information calculated in the calculation step is supplied to the production management system and engineering business system, and optimal index information is provided for each period having different process capabilities determined by the calculated start / end time information. Optimal index generation method characterized in that the index calculation step supplies the production management system with the index calculation step It is.

According to a thirty-second aspect of the present invention, index information relating to a plurality of devices is provided via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. An optimal index generation server for transmitting, an information receiving unit for receiving information on a plurality of devices included in the production execution system via a production management system, and between processing steps by the plurality of devices from the received information A Q-time structure analysis unit for analyzing a structure of a section related to a Q-time constraint indicating a limit value of a staying time of the user, and classifying the plurality of devices into a normal device having a normal capability and an alternative device having a replacement capability together device and division calculation unit, said calculation by a device divided from the received information and the plurality of common devices and operating times of the substitute system operating timing and calculating the operating conditions , The received information, based on the structure and the device category as the start and end time information of a section according to the Q-time constraints, the optimal theoretical value calculator for calculating the kind-based optimal Kanban number and buffer size the includes conventional device and the start and end timing calculation unit for calculating the start and end time information on a replacement device, and the index calculation unit for outputting the index information to the optimal theoretical value before Symbol is calculated, the calculated the start and end time The optimum start / end time information calculated by the section is transmitted to the production management system and the engineering business system, and the optimum index information is obtained for each period having different process capabilities determined by the calculated start / end time information. An optimum index generation server characterized by being transmitted to a production management system by an index calculator.

  According to a thirty-third aspect of the present invention, index information relating to a plurality of devices is provided via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a variety of products. An optimal index generation method executed by a transmitting server, the information receiving step for receiving information on a plurality of devices included in the production execution system via a production management system, and the plurality of devices from the received information Based on the Q-time structure analysis step for analyzing the structure of the section related to the Q-time constraint indicating the limit value of the stay time between the machining processes, and the received information and the structure of the section related to the Q-time constraint An index calculating step for calculating an optimum number of kanbans and buffer sizes for each product type and outputting them as index information; and the plurality of devices A device category calculation step for classifying into a normal device having a utility capacity and an alternative device having a substitution capability; and from the calculated device category and the received information, the number of operations and the operation timing of the plurality of regular devices and alternative devices; An optimal theoretical value calculating step for calculating operating conditions, and a start / end time calculating step for calculating start / end time information regarding the regular device and the alternative device, and the optimal end value calculated by the start / end time calculating step is executed. The start / end time information is transmitted to the production management system and the engineering business system, and the optimum index information for each period having different process capabilities determined by the calculated start / end time information is obtained by the index calculation step. It is the optimal index generation method characterized by transmitting to.

  According to the present invention, even in a situation where Q-time constraint cracking that has not been achieved by the conventional method is likely to occur, the Q-time constraint is reliably observed to improve the throughput of non-defective products, reduce costs, and reduce environmental impact. It is possible to provide optimum operation management conditions that can be performed.

  In addition, in relation to the above effects, the analysis method and appropriateness are also considered for the main causes of Q-time constraint cracking, such as handling of multi-product production, preventive maintenance of equipment, and mutual interference of multiple Q-time constraints. By providing a simple management method, it is possible to simultaneously improve the throughput of non-defective products, reduce costs, and reduce the environmental load while observing the Q-time constraint.

It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 1st Embodiment of this invention. It is a figure which shows the Q-time restriction | limiting area of a semiconductor pre-process as a specific example of the process shown in FIG. It is a figure for demonstrating a performance evaluation parameter | index and its calculation method. It is a figure which shows the apparatus (cleaning process) of batch continuous multitask processing. It is a figure which shows a batch apparatus (heat treatment process). It is a figure which shows an in-process control area. It is a figure which shows a load rule. It is a figure which shows batch group waiting time (factor). It is a figure which shows batch group waiting time (calculation). It is a figure which shows a single kind basic model. It is a figure which shows a single kind basic model (process). It is a figure which shows a single kind basic model (kind). It is a figure which shows evaluation comparison (single kind, periodical input, non-bottleneck, ρ = 0.8). It is a figure which shows evaluation comparison (single kind, periodical input, bottleneck, ρ = 1.2). It is a figure which shows evaluation comparison (single kind, index input, non-bottleneck, ρ = 0.8). It is a figure which shows evaluation comparison (single kind, index input, bottleneck, ρ = 1.2). It is a figure which shows a multiple kind basic model. It is a figure which shows a multiple product basic model (process). It is a figure which shows a multiple types basic model (type). It is a figure which shows a processable apparatus (cleaning). It is a figure which shows a processable apparatus (heat processing). It is a figure for demonstrating the determination of an arrival rate. It is a figure which shows evaluation comparison (plurality kind, index input, non-bottleneck, (rho) = 0.8). It is a figure which shows evaluation comparison (plurality kind, index input, bottleneck, (rho) = 1.2). It is a figure which shows the rework condition in (rho) = 0.3. It is a figure which shows the rework factor in (rho) = 0.3. It is a figure which shows rule comparison (plurality kind, index input, non-bottleneck, ρ = 0.3). It is a figure which shows rule comparison (plurality kind, index input, non-bottleneck, ρ = 0.8). It is a figure which shows rule comparison (plurality kind, index input, bottleneck, ρ = 1.2). It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 2nd Embodiment of this invention. It is a figure which shows a lot single task type | mold apparatus. It is a figure which shows the example of fixed quantity maintenance. It is a figure which shows the apparatus group (m1-m5) at the normal time. It is a figure which shows when the apparatus m6 substitutes m2. It is a figure for demonstrating the case where two or more units | sets of PM occur at the same time. It is a figure for demonstrating the case where PM occurs only at the same time. It is a figure which shows simple parallel processing. FIG. 6 is a job allocation diagram that takes into account quantitative maintenance. It is a figure for demonstrating [2] of the conditions (3) of a JAMP method. It is a figure for demonstrating the apparatus operation guideline 1 when perfect dispersion | distribution cannot be carried out in 1 period. It is a figure for demonstrating the apparatus operation guideline 2 when perfect dispersion | distribution cannot be carried out in 1 period. It is a figure for demonstrating complete dispersion | distribution (after 2nd period) of the pointer | guide 3. FIG. It is a figure which shows the setting value of a model. It is a figure for demonstrating the setting of the frequency | count of substitution in a JAMP method. It is a figure for demonstrating the comparison of a simple parallel rule and a JAMP rule. It is a figure which shows the arrival time of a lot, and leaving (comparison of a JAMP method and a simple parallel rule). It is a figure which shows the fluctuation | variation (when simple parallel MTTR = 1380 [minute]) of the operating condition of the number of apparatuses. It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 3rd Embodiment of this invention. It is a figure for demonstrating CKB + JAMP method. It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 4th Embodiment of this invention. It is a figure for demonstrating the kind of Q-time restriction | limiting as MNA1 method. It is a figure for demonstrating the structure (4 types of serial types) which a some Q-time constraint comprises. It is a figure for demonstrating the structure (5 types of parallel types) which several Q-time restrictions comprise. It is a flowchart (the 1) of a procedural CKB method (standard). It is a flowchart (the 2) of a procedural CKB method (standard). It is a figure which shows a parallel network flow and a minimum cut. It is a figure which shows a parallel network flow. It is explanatory drawing which shows the production process model of a nonpatent literature 1. It is explanatory drawing which shows that the result of having conducted the simulation experiment according to the procedural CKB method is superior to the result by the conventional Q-time constraint section management rule described in Non-Patent Document 1. It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 5th Embodiment of this invention. FIG. 61 is a diagram showing a processing flow in the system of FIG. 60. 6 is a diagram illustrating a processing flow of a setup change cycle calculation unit 53. FIG. It is a figure which shows the processing flow of the apparatus classification calculation part. It is a figure which shows the shift | offset | difference of the maintenance time by setup time. It is a figure which shows the in-process upper limit which satisfy | fills process stay time restrictions. (2) It is a figure which shows the calculation method of the maximum period which minimizes setup time in a multi-product production period calculation part. 60 is a diagram showing a (2) minimum setup cycle calculation method of the setup change cycle calculation unit 53. FIG. 63 is a diagram exemplifying a method for determining the apparatus classification and maintenance order (3) apparatus classification and maintenance order. 60 is a diagram showing a setup utilization method (specified by the user with the auxiliary input unit 52) in the JAMP-wS optimum logical value calculation unit 55. It is a figure which shows the performance comparison with a typical rule: A setup rate and a non-defective product production rate (conventional method, this invention). It is a figure which shows the performance comparison with a typical rule: Total equipment efficiency (conventional method, this invention). It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 6th Embodiment of this invention. It is a figure which shows the processing flow of the system which can apply the optimal index generation apparatus which concerns on 6th Embodiment of this invention. It is a figure (production spike figure) explaining the production spike in an apparatus group. It is a figure which shows a production risk. It is a figure which shows the operating condition of an apparatus, and the relationship of a production spike. It is a figure explaining the method of decomposing | disassembling appropriately the spike of the production capacity of an apparatus group in multi-product and repeated production. It is a figure which shows the formation method of the composite spike which compounded the spike of production capacity across a some apparatus group and a production process. It is a figure which shows the system configuration | structure which can apply the optimal index generation apparatus which concerns on 7th Embodiment of this invention. It is a figure which shows the processing flow of the system which can apply the optimal index generation apparatus which concerns on 7th Embodiment of this invention. It is a figure which shows the information processing for Q-time restrictions structure analysis. It is a figure which shows the production capacity distribution and production spike in the parallel structure of a production process. It is a figure which shows the example 1 (arrival linked centralized inventory type system) of the spike interlock of a process area. It is a figure which shows the example 2 (staged fluctuation absorption dispersion | distribution stock type system) of the spike interlock of a process area. It is a flowchart (the 1) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 2) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 3) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 4) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 5) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 6) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 7) for demonstrating the optimal index generation method which concerns on 8th Embodiment of this invention. It is a flowchart (the 1) for demonstrating the optimal index generation method which concerns on 9th Embodiment of this invention. It is a flowchart (the 2) for demonstrating the optimal index generation method which concerns on 9th Embodiment of this invention. It is a flowchart (the 3) for demonstrating the optimal index generation method which concerns on 9th Embodiment of this invention. It is a flowchart (the 4) for demonstrating the optimal index generation method which concerns on 9th Embodiment of this invention. It is a flowchart (the 1) for demonstrating the optimal index generation method which concerns on 10th Embodiment of this invention. It is a flowchart (the 2) for demonstrating the optimal index generation method which concerns on 10th Embodiment of this invention. It is a flowchart (the 3) for demonstrating the optimal index generation method which concerns on 10th Embodiment of this invention. It is a flowchart (the 1) for demonstrating the optimal index generation method which concerns on 11th Embodiment of this invention. It is a flowchart (the 2) for demonstrating the optimal index generation method which concerns on 11th Embodiment of this invention. It is a flowchart (the 3) for demonstrating the optimal index generation method which concerns on 11th Embodiment of this invention.

Embodiments of the present invention will be described with reference to the drawings.
<First Embodiment>
With reference to FIG. 1, an optimal index generation apparatus according to a first embodiment of the present invention will be described.
The optimum index generation device 11 according to the first embodiment of the present invention, for example, produces a processing step (device group) 2 and a processing step (device group) 3 provided in a production step # 1 for producing a semiconductor. The production management system 10 that manages the schedule is the target of control.

The production management system 10 is a data input / output destination of the production logistics control method (hereinafter referred to as CKB method), extracts various input data, passes them to the input units 12 and 13 (CKB method), and calculates the CKB method. To create a schedule.
The production management system 10 inputs various production management data and CKB method theoretical value calculation results from the optimum index generation device 11, extracts data necessary for CKB method calculation, passes them to the input units 12, 13, and CKB method theory. The production flow schedule including the Q-time constraint section is calculated using the value. The production management system 10 calculates the optimal schedule based on the CKB method theoretical value calculation result and the scheduling algorithm (dispatching rule, setup rule, etc.). The production management system 10 outputs information related to the production schedule to the input units 12 and 13.

As shown in FIG. 1, the optimal index generation device 11 includes an input unit 12, an auxiliary input unit 13, a Q-time structure analysis unit 14, a CKB optimal logical value calculation unit 15, a sensitivity analysis unit (CKB characteristic) 16, and an index calculation. Unit 17 and a sensitivity analysis unit (performance characteristic) 18.
The input unit 12 is information related to the production plan, product information, process plan, etc. via the production management system 10 and the production planning device, the production time / quantity of each product type, the manufacturing flow (by product type), and the device of each process. Type and number, type-device correspondence table, processing time for each type, load size, load interval, minimum required time other than main processing (transport time, setup time, etc.), Q-time constraint time (upper limit, lower limit) ).

  The auxiliary input unit 13 receives the setup time (manufacturing conditions), delivery date (production conditions), equipment failure / maintenance plan / actual data, equipment stop time data (facility management information), equipment standby location from the production management system 10. Enter the presence / absence and the capacity of the standby location (equipment conditions). The auxiliary input unit 13 inputs non-essential input data. However, in practice, the CKB optimum logical value calculation unit 15 at the next stage collectively performs preprocessing and performs an average setup time (product type / device). -Outputs by process), delivery date coefficient (by product type / process), MTBF (average failure interval), MTTR (average repair time), presence / absence of equipment and capacity value. When data is input from the auxiliary input unit 13, the CKB optimum logical value calculation unit 15 performs more detailed calculation.

The Q-time structure analysis unit 14 analyzes the structure of the Q-time constraint, detects the structure of the Q-time constraint section, performs conversion as necessary, and outputs Q-time constraint section process information. The analysis result is recorded in the internal storage unit and can be used as master data for the Q-time constraint section. Here, when a new Q-time constraint is added or when an existing Q-time constraint is changed, the internal storage unit is updated.
The CKB optimum logical value calculation unit 15 inputs all the data from the input unit 12, the data from the auxiliary input unit 13, and the data from the Q-time structure analysis unit 14. The optimum kanban number is calculated from the input data, the optimum buffer size is calculated, and the optimum kanban number and the optimum buffer size are output. At this time, by using the information stored in the internal storage unit of the sensitivity analysis unit (CKB characteristic) 16, higher-speed calculation is possible.

  The sensitivity analysis unit (CKB characteristic) 16 inputs Q-time section process information, traffic intensity (variable), lot arrival rate (by product type) (variable), process processing capacity (variable), load size (variable), and the like. The change characteristic of the result of the CKB method theoretical value calculation when the arrival rate of the lot or the processing capacity of the process fluctuates is calculated. This sensitivity analysis result is recorded in the internal storage unit, and can be immediately used as update data when the arrival rate of the lot or the processing capacity of the process dynamically changes in the Q-time constraint section. . The sensitivity analysis unit (CKB characteristic) 16 includes, as performance characteristics, the number of kanbans (by product type) and buffer size for traffic intensity, in-process arrival rate, and process processing capability, and the number of kanbans and buffer size for load size (variable). For the above characteristics, the traffic intensity, arrival rate, process capability, load size, and Q-time time length corresponding to the respective kanban number and buffer size threshold are recorded in the internal storage unit. The input / output result of the CKB optimum logical value calculation unit 15 is also recorded in the internal storage unit. The information in the internal storage unit is also used for determining the necessity of execution of the optimum index generation device 11. This determination of necessity corresponds to various changes in the production process # 1 to be managed (changes in processing capacity of the device group, changes in the target production rate, changes in traffic, changes / additions of Q-time constraints, etc.) To be judged. When the necessity is confirmed, a calculation instruction is issued to the optimum index generation device 11.

The index calculation unit 17 outputs the calculation result of the CKB optimum logical value calculation unit 15 to the production management system 10 in an appropriate data format, the information in the internal storage unit of the sensitivity analysis unit 16, and the current CKB optimum logical value calculation. It also has a function of comparing the calculation conditions of the unit 15 and sending update data to the production management system 10 more quickly.
The sensitivity analysis unit (performance characteristics) 18 inputs the input data used for the CKB method and the condition of the CKB method optimum theoretical value, and evaluates the performance of the process under the condition of the optimum CKB method optimum theoretical value. The calculation is performed, and various performance evaluation indexes for the traffic intensity, the arrival rate, and the process throughput are output to the production management system 10 as performance characteristics.

  The optimum load rule determination unit 19 determines the boundary of the arrival rate λ ≧ L where TB ≦ T_ (Q +) where the magnitude relationship between the batch set waiting time TB related to the arrival of the product and the Q-time constraint (upper limit) T_ (Q +) is switched. A value L is calculated and an appropriate load rule is determined.

With reference to FIG. 1, the operation of the optimum index generation device 11 according to the first embodiment of the present invention will be described.
Information regarding a plurality of devices included in the production execution system is input by the information input unit 12 via the production management system 10, and Q-time indicating a limit value of stay time between machining steps by the plurality of devices from the input information The constraint structure is analyzed by the Q-time structure analysis unit 14, the optimum number of kanbans and the buffer size for each product type are calculated based on the input information and the Q-time structure, and the index calculation unit 17 uses the information as index information. The optimum index information calculated by the index calculation unit 17 is supplied to the production management system 10.

As a result, even in situations where Q-time constraint cracking, which has not been achieved by the conventional method, is likely to occur, the Q-time constraint can be reliably observed, and good product throughput improvement, cost reduction, and environmental load reduction can be achieved. Possible optimal operation management conditions can be provided.
In addition, Q-time constraints are provided by providing an analysis method and an appropriate management method for dealing with multi-product production and the mutual interference of multiple Q-time constraints, which are the main causes of Q-time constraint cracks. Can be achieved while improving the throughput of non-defective products and reducing costs and environmental impact.

Hereinafter, the operation of each unit constituting the optimum index generation device 11 according to the first embodiment of the present invention will be described.
For example, the optimum index generation device 11 controls the Q-time constraint section of the semiconductor preprocess shown in FIG. This process section is a process section in which the cleaning process and the heat treatment process are continuous, and the following can be mentioned as the feature. In the present embodiment, the control target of the optimal index generation device 11 is treated as the Q-time constraint section of the semiconductor pre-process. However, the present invention is not limited to such a target. For example, the liquid crystal manufacturing process, It can be applied to information communication processes.
(1) Since this process section may become a bottleneck of the entire production execution system due to the presence of the Q-time constraint, it affects the throughput.
(2) Since it is a batch processing line, the load size (batch) determined by the load rule affects the throughput, the usage rate of the apparatus, and the cost for processing per unit product.

Here, the semiconductor pre-process is a manufacturing process for forming an electronic circuit having a predetermined specification by performing chemical physical processing on a disk (called a wafer) obtained by thinly slicing a semiconductor single crystal. Point to. It is a very important management object because it is a central process of semiconductor manufacturing and requires a large initial investment in a single factory.
A product has various processing units such as “part (wafer)”, “lot”, and “batch”.
In the present embodiment, when 25 parts, which are general factory processing and transport units, are set as one lot and a plurality of lots can be processed simultaneously, the plurality of lots that can be processed are collectively referred to as a batch. The maximum number of batches that can be processed simultaneously (batch size) varies depending on the process. For example, the batch size of the cleaning process targeted by this embodiment is 2 lots, and the batch size of the heat treatment process is 6 lots. A processing line in which processes that can be processed in batch units in this way are continuous is referred to as a “batch processing line”. The cleaning process and the heat treatment process will be described later.
The load rule is a rule for determining a batch group policy such as how many lots can be started in a batch processing apparatus, and whether or not other types can be processed at the same specification (processing conditions).

The production target and performance evaluation index in the process of the representative Q-time constraint section shown in FIG. 2 will be described.
Production targets that are important in actual factories are:
(1) Maximize the compliance rate for Q-time constraints (2) Maximize the compliance rate for delivery dates (3) Maximize throughput (4) Maximize setup costs (5) Optimizing device utilization (6) Minimize total costs It is. In the present embodiment, production targets (1), (2), (3), (5), and (6) (other than setup costs) are taken up. FIG. 3 shows a performance evaluation index that is a key for achieving these goals and its calculation method.

An apparatus provided in the production execution system will be described.
[Washing process]
The cleaning process is always used before or after each process, and is a process that is repeatedly used several tens of times. Contamination of the surface of a semiconductor product has a great influence directly on the yield (non-defective product rate) and the like as the semiconductor manufacturing technology becomes finer. The purpose of cleaning is to remove contaminants from the surface.
FIG. 4 is a diagram showing a batch continuous multitask processing apparatus (cleaning process).
In the cleaning process, an apparatus called a batch multitasking (batch multichamber) type is used. As shown in FIG. 4, the batch multitask (batch multichamber) type has a plurality of cleaning tanks (called units) in one apparatus, and the batches are arranged in a sequence determined by processing conditions. When processing is completed in the (tank), the batch is ejected from the apparatus. Accordingly, the processing time from the start of processing in the device to the end of processing differs from the interval between batch input to the device and batch output from the device.
In the batch continuous multitask processing apparatus, the processing proceeds through a plurality of predetermined units while the WIP (batch) is stored in the apparatus in units. Batches are submitted at the load interval (Load Interval), not at the processing time interval. Generally, the processing capacity of the cleaning process is higher than that of the next heat treatment process.

[Heat treatment process]
The purpose of the heat treatment process is the formation of an oxide film on the surface, the densification of a sparse film, and the like.
FIG. 5 is a diagram showing a batch apparatus (heat treatment step).
In the heat treatment process, an apparatus called a batch is used. As shown in FIG. 5, this batch apparatus refers to an apparatus capable of processing in batch units. Further, in this apparatus, lots can be input at processing time intervals.
In such a batch processing line in which apparatuses (processes) loaded with different batch sizes are continuous, a queue is generated even by a batch group. Moreover, since the batch group waiting time changes depending on the load rule, the staying time changes.

Next, the basics of Kanban and queuing theory will be described.
In the present embodiment, a production scheduling method for suppressing deterioration of other process performances while observing the Q-time constraint in the batch processing line by applying the kanban method and queuing theory is considered. Here, as a preliminary knowledge, I will explain Little's formula, which is particularly applied in Kanban and queuing theory.

[Kanban method]
Here, the Kanban system is an inventory system as part of the Toyota production system, particularly the just-in-time system. This is a dependent demand version of a kind of quantitative ordering system based on the principle of returning to inventory. In addition, the kanban used in the present embodiment refers to “in-process kanban” that can be used repeatedly for production instructions. The basis of the Kanban system is “take-back of the post-process”, that is, “take a part that requires the downstream process to the upstream process as much as necessary”. In other words, it is not permissible for the pre-process to start production and ship without permission from the post-process. In the sense that it is pulled from the subsequent process, it is also called “pull system” (pull system).

[Little's formula]
Little's formula is that L is the average number of people, U is the average stay time, and λ is the arrival rate.
L = λU (1)
Is an equation showing that the average number of people in the system can be expressed by a proportional expression of the average staying time.

Hereinafter, the CKB method, which is a physical distribution control method for observing Q-time constraints, will be described.
[Compliance with Q-time restrictions and upper limit of in-process number]
In this embodiment, as an overall function of the production system, a production scheduling method is considered in which the Q-time constraint is maintained and other performance evaluation indexes are not deteriorated. Q-time constraint cracking occurs as a result of WIP being processed one after another in the cleaning process and arriving at the heat treatment process without considering the processing capability of the heat treatment process, and waiting for many WIPs before the heat treatment process. Therefore, it is considered that the Q-time constraint cracking can be suppressed by taking into consideration the processing capability of the heat treatment process, limiting the number of waiting WIPs, and waiting in advance before the cleaning process. To achieve this, the maximum number of in-process devices that can comply with the Q-time constraint is obtained from Little's formula in the queuing theory, and control of the number of in-process devices using the Kanban method is considered.

  In addition, there are cases where only the same type of processing time flows in the process, there are different types, or even if the type is the same due to repeated processing, there are mixed specifications (layers) and different processing times. May flow. In such a case, it is necessary not only to limit the number of in-process items in the process, but also to limit the number of in-process items individually for different processing specifications. In the present embodiment, products with different processing specifications due to such repetition are considered as different varieties and are expressed in a pseudo manner, considering the limitation on the number of devices in progress for both single varieties and multiple varieties. Experiment.

The maximum number of work in progress is limited in three steps. A detailed method will be described later.
(1) The maximum number of waiting lots that can comply with the Q-time constraint is calculated using Little's formula.
(2) Based on the maximum number of waiting lots, the number of kanbans is determined so as to control the in-process from the cleaning process to the heat treatment process. Here, in the CKB method, the number of kanbans is set for each type (spec) of kanban, and in the CKB- (minus) method, the number of kanbans is set collectively in the process without distinguishing the types. Therefore, we decided to use it to confirm the effect of management by product type.
(3) Due to the high processing capacity of the cleaning process, the maximum number of in-process items is not sufficient with only the number of kanbans. The maximum number of waiting lots to be set) is ensured to ensure the maximum number of work in progress.

を単位時間当たりの待ちロット数［ｌｏｔ］、Ｗを１ロットあたりの平均待ち時間［分］、λを単位時間当たりの平均到着ロット数（到着率）［ｌｏｔ／分］とすると、

で表される。 [Calculation of maximum number of waiting lots and determination of number of kanbans]
(1) Calculation of the maximum number of waiting lots (Little's formula)
The maximum number of waiting lots is calculated using Little's formula. In the present embodiment, the definition of Little's official staying time is replaced with the waiting time definition.
The number of waiting lots is

Is the number of waiting lots per unit time [lot], W is the average waiting time per minute [minutes], and λ is the average number of arrival lots per unit time (arrival rate) [lot / min].

It is represented by

を求める。これが熱処理工程の前で待つことができる最大ロット数を表す。
なお、
トラフィック強度（ρ）＝（到着率（λ））／｛（サービス率（μ））×（装置台数（ｍ））｝
バッチ装置のサービス率（μ）＝（ロードサイズ）／（平均処理時間）
バッチ連続マルチタスク装置のサービス率（μ）＝（ロードサイズ）／（ロードインターバルＬＩ） （３） In the above calculation formula, W is an allowance time of the Q-time constraint time, λ is the arrival rate of the heat treatment step (however, when the traffic intensity (ρ) is larger than 1, λ is the arrival when the traffic intensity is 1 The maximum number of waiting lots

Ask for. This represents the maximum number of lots that can wait before the heat treatment step.
In addition,
Traffic intensity (ρ) = (arrival rate (λ)) / {(service rate (μ)) × (number of devices (m))}
Batch device service rate (μ) = (load size) / (average processing time)
Service rate of batch continuous multitasking device (μ) = (load size) / (load interval LI) (3)

(2) Determination of the number of kanbans The number of kanbans is determined based on the maximum number of waiting lots obtained above. As can be seen from FIG. 6, the number of lots in the in-process controllable section by kanban includes not only the maximum waiting lot number but also the number of lots processed in the cleaning process at that time. Therefore, the number of kanbans must be determined in consideration of the number of lots being processed in the cleaning process on average. The average number of lots during the cleaning process is determined by the following formula.

The average number of lots in the cleaning process is as follows: LW is the average number of lots in the cleaning process [lot], Tp is the effective processing time [minute / unit] of the cleaning process, and LI is the load interval of the cleaning device [Load Interval] [ Min / unit], Tp / LI is the number of units used (漕), ρ1 is the traffic intensity (utilization rate) of the cleaning process, L1 is the maximum load size of the cleaning device [lot / unit], and ρ1 × L1 is the average of the cleaning device When the load size [lot / unit] and m1 is the number of devices in the cleaning process [unit]
LW = Tp / LI × ρ1 × L1 × m1 (4)
It is expressed.
Here, the average load size of the cleaning apparatus indicates an average value when a lot is introduced so that the cleaning apparatus operates 100%.

FIG. 6 is a diagram illustrating an in-process control section. FIG. 7 is a diagram showing a typical load rule. FIG. 8 is a diagram showing the batch group waiting time (factor).
Since Little's formula presupposes processing of one lot at a time, there is no time to wait for a batch set generated by batch processing. In batch processing, processing is performed for each load size (batch), not for each lot. Therefore, as shown in FIG. 8, even if the lot arrives at the heat treatment process from the cleaning process, it waits more than the waiting time defined by Little's formula until the specified load size is reached. The waiting time required before starting the batch processing will be referred to as a batch group waiting time.
In the following formulas (5) to (7), for each load rule of the heat treatment process, the maximum batch group waiting time for the lot that arrives first in the batch group is obtained.
The specified load size varies depending on the load rule. FIG. 7 shows a load rule used in this embodiment.

FIG. 9 is a diagram showing the batch group waiting time (calculation).
(A) In the case of “cleaning process: full load heat treatment process: full load” TB is the batch assembly waiting time [minutes], L1 is the load size of the cleaning process apparatus [lot / unit], and L2 is the load of the heat treatment process apparatus. Size [lot / unit], L2 / L1 the number of arrivals from the cleaning process until the load size that can be processed by the apparatus of the heat treatment process is satisfied, λ is the arrival rate of the cleaning process [lot / min], and L1 / λ is cleaned Assuming that the arrival interval of the lots from the process [min / lot], as shown in FIG. 9, the lots that arrive at the cleaning process at the arrival interval (1 / λ) are processed with the load size L1, and the arrival interval (L1 / λ) is obtained. ) To the heat treatment process.

In batch assembly, lots that arrived first in the heat treatment process arrived from the cleaning process (L2 / L1-1) times, except for the arrival of the first lot until the batch assembly was completed. I will wait to do it.
From the above, the maximum batch set waiting time for the first lot that arrives in the batch set is
TB = (L2 / L1-1) × (L1 / λ) (5)
It can be expressed as.

Similarly,
(A) In the case of “cleaning process: Full load heat treatment process: Partial load allowed” TB is the batch assembly waiting time [minutes], L1 is the load size of the cleaning process apparatus [lot / unit], and L2 is the heat treatment process apparatus. Load size [lot / unit], ρ2 is the traffic intensity (utilization rate) of the heat treatment process, ρ2L2 / L1 is the number of arrivals from the cleaning process until the heat treatment process can be processed, and λ is the arrival rate of the cleaning process [Lot / min], where L1 / λ is the arrival interval of lots from the cleaning process [min / lot],
TB = (ρ2L2 / L1-1) × (L1 / λ) (6)
It can be expressed as.

(C) In the case of “cleaning process: Full load, heat treatment process: Attempt full load” TB is the batch assembly waiting time [minutes], L1 is the load size of the cleaning process [lot / unit], and Avg · L2 is the average of the heat treatment process Load size [lot / unit], Avg · L2 / L1 the number of arrivals from the cleaning step until the load size that can be processed in the heat treatment step is satisfied, λ is the arrival rate of the cleaning step [lot / min], and L1 / λ is When the arrival interval of lots from the cleaning process [min / lot],
TB = (Avg · L2 / L1-1) × (L1 / λ) (7)
It can be expressed as.

As described above, the batch group waiting time varies depending on the load rule.
Taking the above two points into consideration, the number of kanbans is obtained by the following calculation.
N ** is the number of kanbans [lot], N * is the maximum number of waiting lots considering the batch assembly waiting time [lot], TQ + is Q-time (maximum stay time) [minutes], and TQ- is Q-time (minimum) Stay time) [minutes], TB is waiting time for batch assembly [minutes], E is the minimum required transport time, λ is the arrival rate [lot / minute], LW is the average number of lots being processed [lot] in the cleaning process Then,
N * = (TQ−max (TB, TQ −) − E) × λ (8)
And
N ** = N * + LW (9)
Represents the number of kanbans.
However, when the traffic intensity (ρ) is greater than 1, λ is the arrival rate when the traffic intensity (ρ) is 1.
Further, in the section determined by the subsequent Q-time constraint that interferes, the number of kanbans is calculated by Expression (10).

(3) Buffer size The buffer size is used in order to ensure the logistics control for observing the Q-time constraint. As shown in FIG. 6, the number of in-process devices that can be controlled by the Kanban method is only the sum of the number of waiting lots in the heat treatment process and the number of lots being processed in the cleaning process. Therefore, if the cleaning process has a higher processing capacity than the heat treatment process, all the lots processed in the cleaning process are output early, and all the lots given kanban are the waiting lots for the heat treatment process. It may become.
In such a case, the maximum number of lots that can be waited for before the heat treatment process determined by Little's formula is exceeded, resulting in rework. In actual factories, the cleaning process generally has a higher processing capacity than the heat treatment process, and it is considered that this situation often occurs. However, it is added that the buffer size is not always necessary because the time for which the heat treatment apparatus is in an idle state does not occur when the relationship between the processing capabilities is reversed. Note that this is not the case in the case of dealing with fluctuations in production logistics or in the third embodiment.

  From the above, a buffer size is provided before the heat treatment step, and the number of waiting lots is limited so as not to exceed the maximum number of lots that can be waited before the heat treatment step obtained by Little's formula. The value of the buffer size (the number of lots to be limited) here uses the value of the maximum number of waiting lots obtained without considering the batch assembly waiting time using the equation (L = λU) obtained in (1). The reason for defining the buffer size in this way is that the Q-time constraint breakage (occurrence of rework, etc.) is suppressed even if the cleaning process is not stopped if it is defined as a result of taking into account the batch assembly waiting time. This is to prevent a situation where the cleaning process is stopped despite the situation.

  For example, the average number of lots being processed in the cleaning process is 11 lots, the maximum number of waiting lots in the heat treatment process taking into account the batch assembly waiting time is 86 lots, and the maximum waiting lot in the heat treatment process not taking into account the batch assembly waiting time When the number is 90, the number of Kanbans is 97, and the buffer size is 90 lots. At this time, if the buffer size is set to 86 lots, and the number of waiting lots in the heat treatment process is 86 lots, the cleaning process is stopped even if the cleaning process is in operation even though the Q-time constraint crack does not occur. End up. In order to prevent such a situation, the buffer size is set to 90 lots, which is the maximum number of waiting lots in the heat treatment process that does not take into account the batch assembly waiting time.

The buffer size is defined in units of processes. When a plurality of types are mixed, the buffer size is obtained for each type and the sum is set as the buffer size.
That is, if B * is the buffer size [lot], TQ is Queue Time (maximum stay time) [minutes], and λ _i is the arrival rate [lot / minute] of the product _i ,

でバッファサイズが表される。ここで、ｎは品種数とする。

Represents the buffer size. Here, n is the number of varieties.

Since the control of the upper limit of the number of work in progress by the buffer size is forced to stop (wait for processing) the previous process (in this case, the cleaning process), it is not very desirable for smooth production. In the embodiment, it is positioned as a function used for more reliably controlling Q-time compliance by the number of in-process devices in the Kanban method.
The in-process upper limit value is calculated by the above procedure.
In the Q-time constraint section where a plurality of Q-time constraints interfere, the preceding Q-time constraint that is rate-limited to the subsequent production process with the lowest production rate (hereinafter referred to as the bottleneck process) The calculation method of the number of sheets is different. The kanban number Ns ** (j) in the Q-time constraint section j before the bottleneck process is expressed as the sum of the buffer size B * (j) in the Q-time constraint section j and the load size in the bottleneck process (10 ).
(Number of Kanban Ns ** (j)) = (Buffer size B * (j)) + (Load size L of bottleneck process) (10)

Next, the performance of the case where the upper limit value is set and the case where the upper limit value is not set are compared.
Hereinafter, a single product type simulation experiment will be described.
[Purpose and method]
The purpose of conducting a single product type simulation experiment is to first confirm that, in a single product type, the setting of the calculated number of kanbans and the buffer size does not deteriorate other performance indexes while keeping the Q-time constraint.
FIG. 10 is a diagram showing a single product basic model. FIG. 11 is a diagram showing a single product basic model (process). FIG. 12 is a diagram showing a single product basic model (product).
As an experimental method, as shown in FIGS. 10, 11 and 12, a batch processing line for cleaning process and heat treatment process, one type, a group of four apparatus for cleaning process, and a group of five apparatuses for heat treatment process are assumed. Conducted the simulation experiment. The cleaning process is composed of a batch / multitask type apparatus group, and an effective processing time is 90 minutes, and the other heat treatment process is composed of a batch apparatus and has an effective processing time of 240 minutes. Further, as shown in FIGS. 11 and 12, the batch size and the number of apparatuses are also different.

  Using the model as described above, the experiment is performed by changing the arrival rate. The arrival rate of the varieties is 0.3, 0.6, 0.8, 1.0, 1.2, and 1.5 for traffic intensity (utilization rate) of the smaller processing capacity of the two processes of cleaning and heat treatment. Decided to be. Based on the determined arrival rate, a simulation experiment is performed for two cases of periodic input and index input. In all experiments in the present embodiment, the heat treatment process has a smaller processing capability, and therefore the arrival rate is determined based on the traffic intensity of the heat treatment process.

Based on these considerations, a model (QT) that is set to rework when the Q-time constraint is broken (QT), a model that sets the calculated number of kanbans and buffer size to the QT model ( CKB method), and 3 of the conventional model (CT) that determines whether or not the product can be introduced by judging whether or not the Q-time constraint can be observed from the apparatus state and the number of preceding WIPs (number of devices in progress) at the start of processing. Perform simulation experiments on two models and compare their performance.
In addition, the lot processing order is first-in first-out order (First In First Out), and the load rule uses a full load that starts processing only when the number of lots meets the batch size in both the cleaning process and the heat treatment process. .

[result]
From the results obtained by the simulation experiment, the service capability decreases due to a non-bottleneck state with a traffic intensity of 0.8 assuming a steady state of the factory where Q-time constraint cracking has not occurred, and failure, The results of a bottleneck state with a traffic intensity of 1.2 assuming that the Q-time constraint is likely to break because the arrival rate temporarily exceeds the service capability are shown below.
FIG. 13 is a diagram showing an evaluation comparison (single product type, periodic input, non-bottleneck, ρ = 0.8). FIG. 14 is a diagram showing an evaluation comparison (single product type, periodic input, bottleneck, ρ = 1.2). FIG. 15 is a diagram showing an evaluation comparison (single product type, index input, non-bottleneck, ρ = 0.8). FIG. 16 is a diagram showing evaluation comparison (single product type, index input, bottleneck, ρ = 1.2).
FIGS. 13 and 14 show the results in the case of regular input, and FIGS. 15 and 16 show the results in the case of index input. Here, those with good performance are shown in underline, and those with poor performance are shown in bold.

[Discussion]
From the above experimental results, in either case of periodic input or index input, in a non-bottleneck state with a traffic intensity of 0.8, the performance deteriorates even if the number of kanbans and the buffer size are set. I can see that there is nothing. In this state, since no rework has occurred in any function, it can be seen that no Q-time constraint cracking has occurred. In the case of a bottleneck with a traffic intensity of 1.2, many Q-time constraint cracks occur in the QT model regardless of periodic inputs and index inputs, but rework is suppressed to zero in the CKB method and CT models. ing.

The cycle time has a large value for the CKB method and the CT model, which is considered to be a result of intentionally waiting the output lot before the cleaning process. However, as a result, the throughput is not lowered, so the performance is not necessarily deteriorated. Regarding the availability, the CKB method and the CT model both achieve the same amount of throughput with a lower availability than the QT model, so the cost performance is rather better.
In an experiment with a single product, it was possible to prevent the Q-time constraint cracking without significantly degrading performance with kanban and buffer size, but this is not necessarily the best method at this stage. Experiments are performed even when multiple varieties are mixed, and the performance is compared.

Hereinafter, the multi-product simulation experiment will be described.
[Purpose and method]
The purpose of conducting a multi-type simulation experiment is to confirm that an experiment conducted with a single type can achieve the same purpose even when a plurality of types are mixed and devices are shared. This experimental model is a continuous process of batch processing of cleaning process and heat treatment process used in actual factories.
FIG. 17 is a diagram showing a multiple product basic model. FIG. 18 is a diagram showing a multi-model basic model (process). FIG. 19 is a diagram showing a plurality of types of basic models (types). FIG. 20 is a diagram showing a processable device (cleaning). FIG. 21 is a diagram showing a processable apparatus (heat treatment).
As shown in FIGS. 17, 18, and 19, as in the case of a single product, the experimental method is a continuous process of batch processing of the cleaning process / heat treatment process, three types, five cleaning processes / number of devices, heat treatment process / A simulation experiment assuming a plurality of device groups with six devices is performed.

  The cleaning process is composed of a batch sequential type apparatus group, and the other heat treatment process is composed of a batch apparatus, and the batch size and the number of apparatuses are different as shown in FIGS. The effective processing time and the number of used devices differ depending on the type, and the type 1 (p1) has an effective processing time of 90 minutes for the cleaning process, 5 used devices, an effective processing time of 242 minutes for the heat treatment process, and 4 used devices. . Type 2 (p2) has an effective processing time of 90 minutes for the cleaning process, 4 units used, 240 minutes for the effective processing time of the heat treatment step, and 5 units used. The product type 3 (p3) has an effective processing time of 90 minutes for the cleaning process, 5 units of equipment used, an effective processing time of 195 minutes for the heat treatment process, and 5 units of equipment used. In addition, a device capable of processing each type is determined. In the cleaning process, as shown in FIG. 20, the devices used for p1, p2, and p3 are different. Even in the heat treatment process, as shown in FIG. 21, there are apparatuses that use different types of p1, p2, and p3.

Using the model as described above, the experiment is performed by changing the arrival rate. As with the experiment in the previous section, the arrival rate of the type is 0.3, 0.6, 0.8, 1.0, 1 .2 and 1.5 are determined. In this embodiment, since all the heat treatment processes have a smaller processing capacity, the traffic intensity (ρ: utilization rate) is the same as that of the heat treatment process.
However, when a plurality of types are mixed, the effective processing time and the number of devices that can be used are different for each type, so it is necessary to obtain the arrival rate separately. As shown in FIG. 15, the arrival rate of the product is determined by calculating the traffic intensity in advance, and calculating backward from the processing capacity (μ: (μ = LTp / L: load size Tp: processing time)) and the number of devices used (m). To decide. As a premise for determining the arrival rate, it is assumed that when the apparatus processes a plurality of varieties, they are processed at an equal rate.

The procedure for determining the input rate is shown below. FIG. 22 is a diagram for explaining the determination of the input rate.
(1) When ρ> 1, calculate with ρ = 1 (2) Calculate the input rate of each product from the production ratio of each product in each shared device (3) Input rate in all devices for each product (4) Calculate the input rate of the product from the added input rate. Based on these factors, set the number of kanban for each QT and CT model and product type, and set the buffer size. The CKB method model provided with the CKB method, the number of kanbans was set throughout the process without providing discrimination for each product type, and experiments were performed on the four models of the CKB-method model provided with the buffer size, and their performances were compared.

[result]
FIG. 23 is a diagram showing evaluation comparison (multiple types, index input, non-bottleneck, ρ = 0.8). FIG. 24 is a diagram showing evaluation comparison (multiple types, index input, bottleneck, ρ = 1.2).
First, in the case where the traffic intensity is both 0.8 and 1.2, the CKB method suppresses Q-time constraint cracking.
In an experiment in a model in which a plurality of types are mixed, Q-time constraint cracking occurred in all models when the traffic intensity was 0.3 and 0.6 when the index was input.

  Therefore, the non-bottleneck state of traffic strength 0.8 where Q-time constraint cracking does not occur and the bottleneck state of traffic strength 1.2 where Q-time constraint cracking occurs as in the previous section. In addition to the results, an experimental result with a traffic intensity of 0.3 in the case of index input is also shown. However, in the case of multiple varieties, since the number of inputs differs depending on the varieties, an index for verifying whether or not all varieties are processed equally is added. If the number of processes is biased, it can be considered that the delivery date is adversely affected. Therefore, when the throughput differs for each product type, the ratio of the throughput (output rate) to the input rate of each product type is newly provided as an index. For the model in which rework has occurred, the ratio of throughput to the input of the entire first process in addition to the input of rework is used as a performance evaluation index. FIG. 23 and FIG. 24 show the results when the index is input.

[Discussion]
In a non-bottleneck state with a traffic intensity of 0.8 after regular input, the throughput per 1% operation is high in the CT, CKB method, and CKB- method, and the average work in progress is also kept low. None have reworked, and no performance degradation has been observed. Here, if one of cycle time and in-process is smaller, the other will be smaller, but the behavior is not seen from the result because the difference is less than the second decimal place. It is. At traffic intensity 1.2 (bottleneck state), Q-time constraint cracking is greatly suppressed by CT, CKB method, and CKB-method as compared with QT as shown in FIG. In particular, in the CKB method, the Q-time crack is suppressed to zero, and rework is completely suppressed as compared with the reference QT, while the throughput is increased, the average cycle time is shortened, and the number of devices in progress is reduced.

  On the other hand, the CT rule of the conventional method causes Q-time constraint cracking although it is smaller than the QT rule, and has side effects such as the inability to process each type (p1, p2, p3) in a balanced manner. In addition, in the CKB-method in which settings are not managed for each product type, each product type is processed in a well-balanced manner, but the performance such as the number of rework is not good, and it can be seen that the management of the number of kanbans by product type in the CKB method is effective. These simulation results are performance evaluation results based on net data for 40 days (simulation period 50 days), and the difference due to the rule further increases in proportion to the period length.

  The basic QT model has a high throughput per 1% operation in the cleaning process, but it has decreased in the heat treatment process. This is thought to be due to the fact that the heat treatment process, which should have been a bottleneck, has not been fully operated, and it can be said that the performance has deteriorated from the whole factory. Above all, a lot of rework (32%) has occurred as a performance. The CT, CKB method, and CKB- method do not show any deterioration in performance in terms of average work in progress compared to QT. The difference in final work is expressed by converting the difference in throughput into lots. This difference is thought to be the result of uneven treatment of varieties.

  For index input, as a feature of the non-bottleneck state with traffic intensity of 0.3, when using a load rule that requires simultaneous processing of the maximum quantity (Full or p1Full4 in FIG. 27), rework occurs in any model. It is mentioned. This cause will be analyzed in detail later. In a non-bottleneck state with a traffic intensity of 0.8, the CT, CKB method, and CKB- method do not show any deterioration in performance compared to QT, and the cycle time and average work in progress for each type vary. However, the overall average has not deteriorated. In the bottleneck state with a traffic intensity of 1.2, it is clear that rework is completely suppressed only by the CKB method, and if the arrival of CT varies, there is a limit to compliance with Q-time constraints. .

  The ratio of processing including rework is higher in the CKB method and CKB-method than in CT and QT, and the overall throughput is higher in all of the CT, CKB method, and CKB-method than in QT. None of the throughput per 1% operation is worse than QT. It can be seen that the overall average is not worse than QT even though the cycle time and the average work in progress vary from one product to another.

Set the maximum number of waiting lots and in-process lots calculated using Little's method for each type of Kanban, and the CKB model with a buffer size is completely Q- except when the arrival of WIP is extremely small. It was confirmed that the time constraint cracking can be suppressed, the varieties are processed uniformly, the performance does not deteriorate, and the performance index is improved. In the case where Q-time cracking is likely to occur (ρ = 1.2), when the CKB method is not applied, the rework rate (Q-time constraint cracking rate), which is about 32%, is reliably suppressed to zero and the product quality deteriorates. In addition, the throughput, average cycle time, in-process inventory, etc. can be improved, and a result of achieving a production rate (= (production amount) / (input amount)) superior to the conventional method was obtained.
From the above experimental results, it has become clear that the performance of the CKB method for setting the upper limit value of the number of devices in progress for each product type is the best, except when the arrival of WIP is extremely small.

Hereinafter, we will consider based on the model of the CKB method.
From now on, the rework that occurs when the traffic intensity is 0.3 with the above-mentioned index input will be considered. The occurrence of rework in this case is not directly related to the control of the in-process upper limit value according to this embodiment. So we will solve the problem by another approach.
First, the reason why rework occurs when the traffic intensity is 0.3 is because the arrival of the lot is slow, so the batch cannot be assembled within the time of the Q-time constraint (the lot for the batch size has not arrived), and the process starts. It is thought that it is because it is not possible.

  In a QT model of traffic intensity 0.3 of multiple types, an experiment was performed without the condition of “rework when the time of Q-time constraint is exceeded”, and a histogram showing the inter-process residence time of the lot at that time is shown in FIG. It is. From this figure, it can be seen that there are lots whose stay time exceeds 720 minutes (12 hours), which is certainly a Q-time constraint crack. The only lots that had Q-time constraint cracks were Variety 1 and Variety 2. In order to investigate the cause of the Q-time constraint cracking of these lots, the lots causing the Q-time constraint cracking and the lots that were batch-assembled with the lots were extracted, and two of them were shown in FIG. From this table, if it is a regular input, a batch can be assembled before the Q-time constraint crack occurs, and even though the lot should be able to start processing, the arrival of the lot varies due to the index input. It can be seen that the Q-time constraint cracking has occurred as a result of the late arrival of the batch and the batch assembly time becoming longer.

FIG. 25 is a diagram showing a rework situation when ρ = 0.3. FIG. 26 is a diagram illustrating rework factors when ρ = 0.3. FIG. 27 is a diagram showing rule comparison (multiple types, index input, non-bottleneck, ρ = 0.3). FIG. 28 is a diagram showing rule comparison (multiple types, index input, non-bottleneck, ρ = 0.8). FIG. 29 is a diagram showing rule comparison (multiple types, index input, bottleneck, ρ = 1.2).
In order to prevent Q-time constraint cracking caused by variations in arrival in such a situation where the arrival of lots is small, experiments are performed by changing the load rule. The description of each load rule is shown in FIG. In previous experiments, all load rules have been performed at full load.

  However, this causes a Q-time constraint crack by waiting for a batch set as before. Therefore, the experiment is performed by changing only the load rule of the heat treatment process. The models to be compared are three models in which the heat treatment load rules are Full, PLA, and AFL. Further, when the traffic intensity was 0.3, the number of kanbans of p1 in Full was 4 sheets smaller than the batch size of 6 lots in the heat treatment process. Therefore, a similar experiment was also performed on a model (p1Full4) in which only p1 was set to process in four lots. However, when the traffic intensity became 0.6 or more, the p1Full4 model had more than 6 kanbans, and all the settings were the same as those of Full, and the same result was obtained. In the experiment, the arrival rate was determined based on the traffic intensity of the heat treatment process as before. The results obtained for traffic intensity of 0.3, 0.8, and 1.2 are shown in FIGS. 27, 28, and 29. FIG.

  From the above experimental results, when the traffic intensity is 0.3, rework is suppressed when the load rule of the second process (heat treatment) is PLA or AFL, and other performance evaluation indexes such as throughput and cycle time are also good. I can read. In p1Full4, rework is not completely suppressed. This indicates that if processing can be performed with a theoretical setting of the number of kanbans, it is not always possible to suppress Q-time constraint cracking. However, it became clear that a neutral evaluation can be obtained by reducing the load size.

Therefore, it was found that when the arrival of WIP is late, rework cannot be controlled only by the number of kanbans, and it is better to control rework by changing the load rule. When the traffic intensity is 0.8, Full has better performance evaluation indexes such as throughput and cycle time. When the traffic intensity was 1.2, the results were all the same.
From the above, it became clear that even when the traffic intensity is 0.3, if the PLA and AFL load rules are used, the Q-time constraint is not broken and the performance evaluation index is not deteriorated. However, when the traffic intensity is 0.8, the Full load rule has better performance. Therefore, in order to perform better production scheduling, the process state is monitored and the load rule is changed according to the change of the traffic intensity. It is considered necessary.

[Conclusion]
In the present embodiment, the Q-time constraint section of the continuous process of the batch process of the cleaning / heat treatment process, which is particularly important, is taken up, focusing on the Q-time constraint that is highly important in semiconductor manufacturing. The prior art related to the Q-time constraint up to now has insufficient strictness and accuracy, and the influence on other performance indexes other than the compliance with the Q-time constraint has not been considered. In this embodiment, as a production scheduling method that keeps the Q-time constraint and does not deteriorate other performance indexes as much as possible, a CKB method for controlling the upper limit of the number of devices in progress in the Q-time constraint section is constructed, and the performance is improved. It was evaluated by a probabilistic simulation experiment. The CKB method is highly reliable and versatile because it theoretically calculates the kanban number and the upper limit of the number of devices in progress based on the kanban method and queuing theory. In addition, it was confirmed that the method is particularly effective particularly in a situation where the process utilization rate is 60% or more which is often used in a normal factory.

  In the present embodiment, even a situation where the arrival rate is low, which is the second cause of Q-time constraint cracking (in the case of traffic intensity of 0.3, etc.), it is ensured by simple change of the load rule or the like. Considering a countermeasure to comply with Q-time, we confirmed from the simulation results that Q-time cracking was prevented. In the present embodiment, even if the arrival rate is low, the Q-time constraint may break due to the setting of the load rule or the like in the batch type or batch multitask type device. By judging this case and appropriately changing the load rule as described below, it is possible to improve productivity while avoiding Q-time constraint cracking (used in combination with the CKB method). The optimal load rule determination unit 19 determines the boundary of TB ≦ T_ (Q +) where the arrival rate λ ≧ L where the magnitude relation between the batch set waiting time TB considering the arrival variation and the Q-time constraint (upper limit) T_ (Q +) is switched. The value L is calculated and an appropriate load rule is used. If λ ≧ L, full load (forces full-size batch set). If λ <L, use a method such as PLA or AFL that does not force a full-size batch set. This embodiment takes into account the fact that the arrival rate differs depending on the product type and process, in addition to the case where the arrival rate is low and varies.

In addition, failure or maintenance of the device may occur, and the processing capacity of the device may fluctuate. Even in consideration of these situations, the determination of the number of kanbans and the buffer size used in this embodiment can be calculated with high speed and high accuracy with respect to fluctuations in the arrival rate and processing capacity, and Q- Time constraint cracking can be prevented.
As described above, as a production scheduling method that observes the Q-time constraint and does not deteriorate other performance indicators, the goal is achieved through the construction and performance evaluation of the CKB method with high qualitative evaluations such as calculation speed, accuracy, and practicality. We were able to.

In the first embodiment, the first problem is solved by the CKB method.
In multi-product production, a practical production logistics management method that achieves high throughput while observing the Q-time constraint is named CKB (Control of Kanban and Buffer sizes) method. In the first embodiment, a CKB method program and a performance evaluation result using the program are based on a stochastic simulation using actual factory data. From this performance evaluation result, when the CKB method is not applied, the rework rate (Q-time constraint cracking rate) which is about 32% is surely suppressed to zero to prevent deterioration of the product quality, and the throughput, average cycle time, In-process inventory, etc. can be improved, and results have been achieved that achieve a production rate (= production amount / input amount) superior to that of the conventional method.

Second Embodiment
With reference to FIG. 30, the optimal index production | generation apparatus which concerns on 2nd Embodiment of this invention is demonstrated. The optimum index generation device 21 according to the second embodiment of the present invention manages the maintenance schedule of the device group 5 provided in the production process # 1 for the operation method for decentralizing the preventive maintenance time of the manufacturing device, for example. The production management system 10 and the engineering business system 29 are controlled. In the present embodiment, a production load distribution method (hereinafter referred to as JAMP method) for distributing maintenance time will be described.

As shown in FIG. 30, the optimal index generation device 21 includes an input unit 22, an auxiliary input unit 23, a device category calculation unit 28, a JAMP optimal logical value calculation unit 24, a sensitivity analysis unit (JAMP characteristic) 25, and a start / end time calculation. Section 26 and a sensitivity analysis section (performance characteristic) 27.
The input unit 22 receives the maintenance time from the production management system 10 and the engineering business system 29, maintenance guidelines (the number of processes until the start of maintenance in quantitative maintenance, the maintenance cycle in periodic maintenance, etc.), the number of devices in the device group (process), the devices The production quantity or arrival rate for each product type of the group (process), the processing time for each type / process / device, the product type information that can be processed by the device, and the device type are input.

  The auxiliary input unit 23 receives setup time from the production management system 10 and the engineering work system 29, information on the batch type device (batch size, load rule, load size, load interval), information on the multitask type device (standby point in the device) All of the information obtained from the input section (device type, number of devices, processing time, production quantity or arrival rate by device group (process), product type that can be processed by the device) input. The auxiliary input unit 23 inputs non-essential input data. However, in practice, the JAMP optimal logic value calculation unit 24 at the next stage collectively collects important data and performs preprocessing, and the setup time (product type, device, By process), information on batch devices (average load size, load interval, batch assembly time, device execution service rate), MTBF (average failure interval), MTTR (average repair time), information on multitasking devices (device Outputs information such as whether there is a standby and the capacity value). When data is input from the auxiliary input unit 23, the JAMP optimum logical value calculation unit 24 performs more detailed calculation.

The JAMP optimum logical value calculation unit 24 inputs maintenance time, maintenance guideline, number of devices, production quantity or arrival rate by product type, processing time by type / process / device, type information that can be processed by the device, and device type. The JAMP method theoretical value calculation result, the processing load (number of times) of each device and its processing time are calculated and output based on the setup time and other information about the device.
The sensitivity analysis unit (JAMP characteristic) 25 calculates the JAMP method theoretical value calculation result when the JAMP method theoretical value calculation result input from the JAMP optimum logical value calculation unit 24, the processing load (number of times) of each device, and the processing time thereof vary. Calculate the change characteristics of the results. The sensitivity analysis unit (JAMP characteristic) 25 records the sensitivity analysis result in the internal storage unit, and the JAMP method theoretical value calculation result, the processing load (number of times) of each device, and the processing time thereof dynamically change. In some cases, it can be used immediately as update data.

The start / end time calculation unit 26 inputs the lot information and engineering information to be processed, and calculates the start / end times of processing and preventive maintenance of each device from the JAMP method theoretical value calculation result. Then, as the preventive maintenance schedule, the processing of the regular device and the alternative device and the start / end time of the preventive maintenance are output.
The sensitivity analysis unit (performance characteristics) 27 inputs the input data used for the JAMP method and the conditions of the JAMP method optimum theoretical value, and for the performance evaluation of the process under the condition of the determined JAMP method optimum theoretical value. And the traffic strength v. s. Various performance evaluation indexes are output to the production management system 10.

台となる。この時、代替能力は（ｍ−Λ／μ）で、代替装置は、

台となる。 Based on information input from the input unit 22 and the auxiliary input unit 23, the device category calculation unit 28 classifies the device into a regular device and an alternative device (normal capability and alternative capability), and outputs device category information. The regular device (ordinary capability) refers to a device capability that is regularly required so that the waiting of a lot does not become long, and the alternate device (alternative capability) refers to the other average surplus capability. If the number of devices included in the device group is m, and the weighted average of the service rates of each device is μ, the service capability of the device group (process) is な る = mμ. Here, when the input rate of the device group product is Λ, the ordinary capacity is Λ / μ, and the ordinary device is

It becomes a stand. At this time, the alternative capability is (m−Λ / μ), and the alternative device is

It becomes a stand.

  The engineering work system 29 is a system that mainly manages work such as preventive maintenance and failure handling of production facilities, and manages general conditions, schedules, and results of engineering work that are not held in the production management system 10. For example, master information such as the range of responsibility of the equipment engineer, execution conditions such as shift, actual data of preventive maintenance and failure response, and the like. Depending on the production management system, the engineering business system 29 may be incorporated as a function of the production management system 10.

With reference to FIG. 30, the operation of the optimum index generation device 31 according to the second embodiment of the present invention will be described.
Information relating to a plurality of devices included in the production execution system is input by the input unit 22 via the production management system 10, and a plurality of devices are replaced by the device classification calculation unit 28 with a normal device having a normal capability and an alternative having a replacement capability The number of operations and operation times of multiple regular devices and alternative devices (and regular capabilities and alternate capabilities) are calculated from the device capability classification and input information, and preventive maintenance related to regular devices and alternate devices. The start / end time is calculated by the start / end time calculation unit 26, and the optimum start / end time information calculated by the start / end time calculation unit 26 is supplied to the production management system 10 and the engineering business system 29.

  As a result, compliance with Q-time constraints is provided by providing an analysis method and appropriate management method for preventive maintenance of equipment, which is the main cause of Q-time constraint cracking that has not been achieved by conventional methods. At the same time, it is possible to improve the throughput of non-defective products, reduce costs, and reduce environmental impact. This reduces the situation where Q-time constraints are likely to break, and provides optimal operation management conditions that can improve non-defective product throughput, reduce costs, and reduce environmental impact while ensuring compliance with Q-time constraints. can do.

Hereinafter, the operation of each unit constituting the optimum index generation device 21 according to the second embodiment of the present invention will be described.
First, an operation method for decentralizing the preventive maintenance time of the manufacturing apparatus by the optimum index generation apparatus 21 will be described.
[Importance of production equipment management]
In semiconductor manufacturing, 40% to 6.5% of manufacturing cost is depreciation of machinery and equipment, which is an important management target. Generally, in a production facility that performs chemical-physical processing, it is necessary to scrap or recycle the semi-finished product (called rework) in the event of a failure or abnormal condition. become.
In addition, compared to the time spent for subsequent maintenance to restore normal operation after a failure or abnormality occurs in the equipment state, the equipment operation time is more effective when the equipment is maintained before failure or abnormality occurs. Therefore, preventive maintenance (referred to as PM) is performed.
In preventive maintenance, regular maintenance is often performed to perform predetermined maintenance at regular intervals. However, quantitative maintenance that performs maintenance according to the number of treatments is performed for devices that perform chemical physical processing. Often. In this embodiment, the quantitative maintenance of the semiconductor factory is mainly taken up (note that it can also be applied to regular maintenance).

[Relationship between factory traffic and preventive maintenance]
During the preventive maintenance period (called Mean Time to Repair: MTTR), the production facility is stopped and work-in-process (WIP) cannot be processed. Therefore, during MTTR, it is necessary to process WIP using other production facilities. Usually, the degree of congestion (traffic) of physical distribution in a factory is planned with the process utilization rate (traffic intensity) being less than 1 (100%) as an average in a steady operation state. Here, when the number of devices that become PMs at the same time is large when the process is actually used, the remaining utilization devices alone cannot process the WIP and the process utilization rate exceeds 100%. In this situation, it is known that the waiting time of WIP increases abruptly. Increasing waiting time has a negative effect on cost increase, delivery rate compliance rate and Q-time constraint compliance.
For this reason, it is desirable to disperse the PMs of a plurality of apparatuses so that they do not occur at the same time.

[Effect of preventive maintenance on quality and cost]
Semiconductor processing dimensions are becoming finer in accordance with the technology roadmap. At present, the processing dimension has been reduced to 1 / 1000th of the thickness of the hair. This means that if the semiconductor processing size is halved, the speed will be doubled and the power consumption will be reduced to a quarter. For example, high performance and low cost can be realized just by miniaturizing the semiconductor. Because it can. However, such miniaturization of processing dimensions has made it difficult to ensure yield (non-defective product rate) in terms of manufacturing. In order to produce a product (non-defective product) having the size and function as designed, the residence time between processes as well as the processing time had to be strictly controlled. For this reason, production scheduling considering the Q-time constraint (upper limit value and lower limit value of the stay time between processes) is important. As described in the first embodiment, in the situation where the waiting time becomes longer due to the preventive maintenance (PM) of the manufacturing apparatus, the Q-time constraint crack is likely to occur. This is a very important issue.

In this embodiment, we take a semiconductor manufacturing process consisting of multiple devices that perform maintenance quantitatively, and develop a new load balancing method that assigns jobs to devices so that the number of devices that become PMs at the same time is reduced The purpose is to confirm the effect.
The feature of quantitative maintenance is that maintenance is started when the number of times processed by the apparatus reaches a specified value. In other words, there is a possibility that the number of processes of each device is controlled by controlling the assignment of WIP to the devices so that the preventive maintenance periods of a plurality of devices do not overlap. In this embodiment, paying attention to this point, the device is divided into a regular device and an alternative device, the alternative device is used to decentralize the timing of PM, and an extra waiting time is also given to the job at times other than PM. Consider a load balancing method that does not occur.

  The performance of the load balancing method is confirmed by a simulation experiment with reference to an actual factory process model (process for processing one product with a given processing time using six devices, regular input). Here, a comparison is made with a simple parallel rule that uses devices sequentially without intentional load balancing. In the experiment, there are two cases where PM can be completely dispersed by reducing the number of devices that become PM at the same time by the load balancing method, and when PM cannot be completely dispersed in one cycle. Set and verify the effect of load balancing on each.

  Here, especially when the preventive maintenance period is long, the waiting time of the lot due to PM becomes long, and a situation where rework occurs due to the Q-time constraint is likely to occur. The occurrence of rework is only a loss in production and is undesirable. By distributing the load in advance so as to reduce the number of units that become PM at the same time, the traffic intensity of the process is always kept as low as possible, and the number of Q-time split lots is suppressed by not generating extra waiting time. Also make sure you can.

The following describes the target process model and quantitative maintenance, and then describes the load distribution method for parallel devices that takes quantitative maintenance into account. The conclusion of embodiment is summarized.
[Problem settings]
In the present embodiment, a single process composed of a plurality of lot / single task type apparatuses requiring quantitative maintenance will be taken up as an object.
A product has various processing units such as “part (wafer)”, “lot”, and “batch”. In the present embodiment, “lot” is used as a processing unit with 25 parts as one lot. In a semiconductor factory, a lot is usually a unit of conveyance.
A lot / single-task type apparatus refers to an apparatus whose processing unit is one WIP (for example, a lot) as shown in FIG. After a predetermined processing time has elapsed, when the WIP is swept out and becomes free, processing of the next lot can be started.

[Performance evaluation index]
There are various performance evaluation indexes for improving the factory's ability in a semiconductor production line. The performance evaluation index and its definition taken up in this research are shown.
In the performance index, the total waiting time [minutes] is calculated as the sum of waiting times generated in each lot, and is preferably smaller. The number of waiting lots [lot] is calculated as the total number of lots in which a waiting time has occurred, and it is preferable that the waiting lot number be smaller. The Q-time constraint crack lot number [lot] is calculated as the number of lots that have Q-time constraint cracks, and the smaller one is better.

[Quantitative maintenance]
FIG. 32 is a diagram illustrating an example of quantitative maintenance.
In the example shown in the figure, a maintenance method is called quantitative maintenance every time 100 lots are processed each time the apparatus reaches a preset processing count (number of lots). As shown in FIG. 32, when the maintenance is started, the period of PM becomes a certain period (on average, MTTR), and the apparatus is in a resting state during that period. The processing is started again from the time when PM ends.
A feature of quantitative maintenance is that maintenance is required when the number of times processed by the apparatus reaches a specified value. In other words, there is a possibility that the number of processes of each device is controlled by controlling the assignment of WIP to the devices so that the preventive maintenance periods of a plurality of devices do not overlap.

Next, a method for assigning jobs to parallel devices in quantitative maintenance will be described. [Structure of overall function and sub function]
FIG. 33 is a diagram illustrating a normal device group (m1 to m6). FIG. 34 is a diagram illustrating when the device m6 substitutes for m2 during preventive maintenance (m1, m3, m4, m5, m6 other than WIP2).
In the present embodiment, a load distribution method (Job Allocation for Maintenance in Parallel (hereinafter abbreviated as JAMP)) that takes into account newly considered quantitative maintenance (hereinafter referred to as PM) is illustrated in the order of overall functions and sub-functions. A description will be given by taking an example of a model of six devices shown in FIG. 33 (an actual factory model).

First, in this embodiment, the overall configuration of a newly developed rule will be described using an actual factory process model.
As a premise, in the situation where no device is performing PM, the traffic intensity Λ / Μ (indicating the degree of congestion of the process) that is the ratio of the arrival rate Λ of the job to the process and the service capability の of the process is 1 ( Equipment plans and load plans are made to be smaller than 100%. In other words, there is surplus service capacity. For example, in the process shown in FIG. 33, there are six devices with the same capability, and normally, if five of them (m1, m2,..., M5) are processed, all lots can be processed without delay, and the rest One (m6) device can be used as an alternative device (alternative capability). This is because when the traffic intensity in this process is 0.8, the number of apparatuses necessary for processing a job is theoretically 4.8, that is, five. The remaining device can be used as an alternative device.

FIG. 35 is a diagram for explaining a case where two or more PMs occur at the same time. FIG. 36 is a diagram for explaining a case where PM occurs only at the same time.
Here, in response to the arrival of a job, when five normally used devices (referred to as regular devices) are used sequentially in parallel, multiple devices are used at the same time. A period during which the apparatus becomes PM (MTTR: Mean time to repair) occurs. In the process model shown in FIG. 33, since the processing capacity ratio with respect to job arrival is 4.8, at least five devices are required, and two or more devices are stopped at the same time (that is, four devices can be used). The traffic intensity exceeds 1 (FIG. 35). It is known that when the traffic intensity is greater than 1, the waiting time of the job increases abruptly. Increasing waiting time has a negative impact on cost increase, compliance with deadlines, and compliance with Q-time constraints.

Therefore, when the number of devices in the process is m, it is desirable to keep the number of devices to be maintained at the same time to (m- [Λ / μ]) or less (1 in the case of FIG. 33) (FIG. 36). ). A feature of quantitative maintenance is that maintenance is started when the number of times of processing by the apparatus reaches a specified value. In other words, there is a possibility that the allocation of jobs to devices is controlled to control the number of processing times in the devices so that the preventive maintenance times of a plurality of devices do not overlap.
In this embodiment, paying attention to this point, a method is considered in which alternative devices are used to distribute PM timings, and an extra waiting time is not generated in a job at times other than PM. Here, the case of six apparatuses shown in FIG. 33 will be described as an example.

In the following, conditions necessary for PM dispersion in the JAMP method are listed.
[Load distribution method 1]
(1) Before the first device starts PM, during the MTTR period after the first device starts PM, before the first device starts PM, The PM start time is shifted backward by allowing the alternative device to substitute the process in advance for the time overlapped with the MTTR. At this time, if the total number of alternative processes in the alternative apparatus is equal to or less than the predetermined number of remaining processes until the start of PM of the first regular apparatus, the PM timing of the regular apparatus is shifted.
(2) In addition to the fact that the substitute device does not start PM within the period when the regular device performs PM,
(3) The PM timing of the alternative device can be adjusted so that it does not overlap with the PM timing of other devices.

When the conditions (1) and (2) are combined and the number of substitution processes of the substitution device is equal to or less than the specified number at which PM starts, the PM of the regular device can be completely dispersed (the number of devices that become PM at the same time) It can be kept below the number of alternative devices). If this condition is not satisfied, the number of devices that require maintenance at the same time must be greater than (m− [Λ / μ]), and an additional waiting time is generated in the job.
As described above, when the conditions (1), (2), and (3) are satisfied, the regular device and the alternative device PM can be completely dispersed in one cycle. Even when it cannot be completely distributed, the period in which the PM periods overlap is shortened by performing load distribution as much as possible.

[Load distribution method 2 (when conditions (1) and (2) are not satisfied)]
The case where the conditions (1) and (2) are not completely satisfied corresponds to the case where the PM overlap time of all the regular devices cannot be made zero in one cycle. Here, one cycle refers to a period from when the first device (m1) starts to operate until all regular devices finish PM once. For example, this corresponds to a case where the time length of MTTR is relatively long.
In the first period, load distribution by the JAMP method is applied as much as possible to reduce and distribute the PM overlap time.
In the second and subsequent cycles, substitution processing is performed within the range of conditions to achieve complete dispersion.

[Guideline 1: Reduction of overlapping units and time length]
Based on the concept of the JAMP method, the load is distributed using an alternative device, and the number of PMs and the time length that become PMs at the same time are suppressed as much as possible (first cycle).
[Guideline 2: Overlapping time distribution]
The overlapping times are dispersed so that the overlapping periods of the PM times are not continuous as much as possible (first period).
[Guideline 3: PM decentralization]
In the second and subsequent cycles, substitution processing is performed for complete dispersion, and the PM timing of each device is shifted.

Each of the above conditions and guidelines will be described in detail. FIG. 37 is a diagram illustrating simple parallel processing. FIG. 38 is a job allocation diagram in consideration of quantitative maintenance.
[Requirements for decentralization of PM]
Here, the above conditions (1), (2), and (3) will be described in detail.
With respect to the condition (1), in the case of the model shown in FIG. 38, λ _j is the arrival rate per device j [lot / min], Λ is the lot arrival rate to the process [lot / min], and μ is a single unit. Average service rate of the device 1 / Tp [lot / min], M is the process service rate == mμ = m / Tp [lot / min] (m: number of devices, Tp: average processing time), where λ _j = Λ / [Λ / μ]), N is the reference value [lot] for mass maintenance, MTTR _j is the required time [minutes] for preventive maintenance (PM) of the device j (the device stops during this period, and the production lot Cannot be processed). The number of regular devices is [Λ / μ] units, and the number of alternative devices is (m− [Λ / μ]) units.

(1) First, the number of substitutions (basic number) n _j necessary for the regular device j (j = 2, 3,... [Λ / μ]) is calculated.

（２）実際には、整数値への切り上げの差分による影響があるため下式の調整が必要になる。
切り上げの差分

(2) Actually, since there is an influence due to the difference of rounding up to an integer value, adjustment of the following equation is necessary.
Rounding difference

The replacement number n _j of the j-th device is

このとき、ｊ＋１台目以降の常用装置に対しては、

At this time, for the j + 1 and subsequent devices,

This adjustment is repeated (j−1) times for each j from j = 2 to j = [Λ / μ]. From the above calculation, the number of alternative processes necessary for the device j is calculated. (This is the number of lots to be replaced in advance within the MTBF period until the first device starts PM).
As described above, if the number of substitutions required for the device j obtained as a result of the calculations of (1) and (2) is n _j again, the total sum needs to be (N−1) or less (condition ( 1))

で条件（１）が表される。

The condition (1) is expressed as follows.

The condition (1) is that in the model shown in FIG. 38, m6 is performing m2 to m5 substitution processing so that other devices do not become PM during the PM period of m1, and the total time is (1), which indicates that the alternative process shown on the left side of the condition (1) is completed before the start of the PM of m1.
Condition (2) means that the alternative device does not start PM during the period from when the first device starts PM until the devices up to the [Λ / μ] -th device finish PM. This alternative is shown on the right side of FIG.
Condition (2) is expressed as the following equation.

ここで、実際には、代替装置は、条件（１）に必要な処理回数を消化しているため、これら（１）と（２）の条件を合わせて、代替装置の処理回数がＰＭを始める規定数以下にならなければ常用装置のＰＭの完全な分散化はできない。この条件式は、下式のようになる。

Here, since the alternative device actually digests the number of processes necessary for the condition (1), the number of processes of the alternative device starts PM by combining the conditions (1) and (2). Unless the number is less than the specified number, the PM of the regular device cannot be completely dispersed. This conditional expression is as follows.

条件（３）は、代替装置のＰＭと他の装置のＰＭが重ならないように、代替装置のジョブの割り付けを調整できることである。本実施形態では、以下２つの場合に分けて考える。
［１］常用装置で最後にＰＭを開始する装置（図３３に示すｍ５）でＰＭを終了した時点に、代替装置（図３３に示すｍ６）がＰＭを開始する場合、下式として表せる。

The condition (3) is that the job assignment of the alternative device can be adjusted so that the PM of the alternative device and the PM of the other device do not overlap. In the present embodiment, the following two cases are considered separately.
[1] When the alternative device (m6 shown in FIG. 33) starts PM at the point of time when PM is ended by the last device (m5 shown in FIG. 33) that is a regular device, it can be expressed as the following equation.

［２］代替装置（図３３に示すｍ６）が、常用装置で最後にＰＭを開始する装置（図３３ではｍ５）でＰＭを終了した後に、後続周期で１回以上処理を行った後にＰＭを行う場合、代替装置が第１周期で代替する回数は、条件（１）と（２）を合わせて、

[2] After the alternative device (m6 shown in FIG. 33) finishes PM at the last device (m5 in FIG. 33) that is a regular device and then performs PM at least once in subsequent cycles When performing, the number of times the substitute device substitutes in the first period is the sum of the conditions (1) and (2)

であり、第２周期以降は、このうち第２項目のみが必要となる。

In the second cycle and thereafter, only the second item is required.

  Regarding condition (3), in the alternative device, the remaining number of possible alternatives at the end of the (k−1) period is represented by the left side of the following equation. Since the number of possible substitutions must be a positive value, the following condition is satisfied.

この左辺で示した残り代替可能回数が、次の１周期で必要な代替回数より小さくなる周期で代替装置はＰＭになる。

The substitution device becomes PM at a period in which the remaining number of possible substitutions indicated on the left side is smaller than the number of substitutions required in the next one period.

となる最小のｋ’周期目に、代替装置のＰＭをする必要がある。
さて、ここで、図３８の右側で示したように常用装置のＰＭを順次実施し、その期間に処理すべきロットを代替装置で処理していくと、［Λ／μ］台目の装置のＰＭを終了する前に、代替装置でＰＭが必要になる。これは、常用装置と代替装置が同時期にＰＭになることを意味するため、望ましくない。そこで、代替装置のＰＭが必要になるｋ’周期目では、常用装置のＰＭが開始する前に、代替装置のＰＭを終える必要がある。

It is necessary to perform PM of the alternative device in the minimum k ′ period.
Now, as shown in the right side of FIG. 38, when the PM of the regular device is sequentially performed and the lot to be processed in that period is processed by the alternative device, the [Λ / μ] th device Before the PM is finished, the alternative device needs the PM. This is undesirable because it means that the regular device and the alternate device become PM at the same time. Therefore, in the k ′ period when the PM of the alternative device is required, it is necessary to finish the PM of the alternative device before the PM of the regular device starts.

  The number of times N ′ that can be processed by the alternative device before the PM starts in the k ′ period is expressed by the following equation.

この回数分を、常用装置がＰＭを開始する前に、意図的に処理し終えて、代替装置がＰＭを終了することができれば、代替装置のＰＭと常用装置のＰＭを同時期に発生せずに済む。
そこで、代替装置の残り可能処理数を常用装置の代替処理にあて、代替装置を意図的に事前にＰＭさせることを考える。常用装置は、第１周期ですでにそれぞれのＰＭが分散化するように調整がなされているため、最後にＰＭを終了する装置から順次、代替装置に処理を代替させる。
ここで、常用装置の台数は、［Λ／μ］であるので、通常最初にＰＭを開始する常用装置（図４０に示す例ではｍ１）に対する代替回数Ｎ’_１は、

If the number of times is intentionally finished before the normal device starts PM, and the alternative device can finish PM, the PM of the alternative device and the PM of the normal device are not generated at the same time. It will end.
Therefore, it is considered that the number of remaining possible processes of the alternative device is assigned to the alternative processing of the regular device, and the alternative device is intentionally PM in advance. Since the regular device has already been adjusted so that the respective PMs are dispersed in the first period, the alternate device sequentially substitutes the processing from the device that finally ends the PM.
Here, since the number of regular devices is [Λ / μ], the number of alternatives N ′ ₁ for the regular device (m1 in the example shown in FIG. 40) that normally starts PM _first is

その他の常用装置（図４０に示す例ではｍ２、ｍ３、ｍ４、およびｍ５）に対して代替する合計回数は、Ｎ’−Ｎ’_１となる。

(In the example shown in FIG. 40 m @ 2, m3, m4, and m5) other conventional devices total number of times that alternate relative becomes N'-N _'1.

  Further, when the number of processing completions of the alternative device reaches N ′, PM of the alternative device is performed. The number of lots that can be processed in the PM period (MTTR) of the alternative device k is

である。
ここで、ある周期において、常用装置の中でＰＭを最初に終了する装置（図４０に示す例ではｍ１）が、ＰＭを最後に終了する常用装置（図４０に示す例ではｍ５）のＰＭ終了時刻までに処理するロット数は、下式で表される。

It is.
Here, in a certain cycle, the device that ends PM first in the regular device (m1 in the example shown in FIG. 40) ends the PM of the regular device that ends PM last (m5 in the example shown in FIG. 40). The number of lots processed by the time is expressed by the following formula.

ここで、次の周期（前の周期でＰＭを最後に終了する常用装置のＰＭ終了時刻以降）に、常用装置の中で最初にＰＭを開始する装置が処理できるロット数は、

Here, in the next cycle (after the PM end time of the regular device that ends PM last in the previous cycle), the number of lots that can be processed by the device that starts PM first among the regular devices is:

上式のロット数から代替装置による代替回数を引いた回数が、代替装置がＰＭ終了時までの処理回数以上であれば、代替装置のＰＭが常用装置のＰＭと重ならずに済むので、下式が条件（３）の［２］の条件となる。

If the number of times of substituting the substitute device by the number of lots in the above formula is equal to or greater than the number of times the substitute device has processed until the PM finishes, the PM of the substitute device will not overlap with the PM of the regular device. The expression becomes the condition [2] of the condition (3).

なお、条件（３）の［１］は、条件（３）の［２］でＮ’＝Ｎ’_１＝０となる場合の特殊例で、常用装置による処理の代替を必要としない場合である。

The condition (3) [1] is a special case in which N ′ = N ′ ₁ = 0 in the condition (3) [2], and does not require a substitute for the processing by the regular device. .

  As described above, when the conditions (1), (2), and (3) ([1] or [2]) are satisfied, the PM can be distributed so that the traffic intensity does not exceed 1. First, it is examined whether or not these conditions are satisfied. If it is satisfied, the timing of PM can be distributed by allowing the alternative device to process a predetermined number of alternatives in advance.

When there are a plurality of alternative devices, the conditions (1), (2), and (3) are respectively expressed by the following equations.
Condition (1) is

条件（２）は、

Condition (2) is

条件（１）および（２）は、

Conditions (1) and (2) are:

Regarding conditions [3] [1] (1) and conditions (3) [2] (2), it is necessary that the conditions are satisfied for the portion where each alternative device is responsible for the load.
[1] of condition (3) is

条件（３）の［２］は、

Condition (3) [2] is

Next, a description will be given of a method for reducing the negative influence by PM by replacing the load in advance with an alternative device when the condition cannot be satisfied and PM cannot be completely dispersed.
In the present embodiment, the basic concept of the load distribution method 2 (when the conditions (1) and (2) are not satisfied) is summarized in the following two points.
In the first period, load distribution by the JAMP method is applied as much as possible to reduce and distribute the PM overlap time.
In the second and subsequent cycles, substitution processing is performed within the range of conditions to achieve complete dispersion.
[Guideline 1: Reduction of overlapping units and time length]
FIG. 40 is a diagram for explaining the apparatus operation guideline 1 when complete dispersion cannot be performed in one cycle.
Based on the concept of the JAMP method, the load is distributed by using an alternative device, and the number of PMs and the length of time that become PM at the same time are suppressed as much as possible.
However,

ロットについて同一周期（例えば第１周期）では代替できない。

The lot cannot be replaced with the same period (for example, the first period).

[Guideline 2: Overlapping time distribution]
FIG. 41 is a diagram for explaining the apparatus operation guideline 2 when complete dispersion cannot be performed in one cycle.
The overlapping time lengths are distributed so that the overlapping periods of PM times are not as continuous as possible.

ロットを各装置に分散化する。
［指針３：ＰＭの分散化］
２周期目以降で、完全分散化のために代替処理を施し、各装置のＰＭ時期をずらす。

Distribute lots to each device.
[Guideline 3: PM decentralization]
In the second and subsequent cycles, substitution processing is performed for complete dispersion, and the PM timing of each device is shifted.

Here, the dispersion of the overlapping time (first period) in the pointer 2 will be described with reference to FIG.
It cannot be completely replaced in the first cycle

回の処理がある。
これを、１台目以外の残りの装置（［Λ／μ］−１）台と代替装置（ｍ−［Λ／μ］）台でのＰＭ時期の重なりとして分散し、工程性能が低下した状態が連続することを防ぐ。
常用装置において、ＰＭ期間が重なる処理回数

There are times of processing.
This is dispersed as PM time overlap between the remaining devices ([Λ / μ] -1) and the alternative device (m- [Λ / μ]) except for the first device, and the process performance is degraded. To prevent the continuous.
In regular equipment, the number of times PM periods overlap

装置ｊの代替回数（最大）を、ｎ_ｊ’＝ｎ_ｊ’−ｒ（ｊ＝２、・・・［Λ／μ］）とする。全てｒ回分ずれることになるので、実際に待ちロットを発生させるのはｍ１とｍ２の重なり部分になる。
また、残りの重なり期間の処理回数

The number of substitutions (maximum) of the device j is n _j ′ = n _j ′ −r (j = 2,... [Λ / μ]). Since all are shifted r times, it is the overlapping portion of m1 and m2 that actually generates the waiting lot.
In addition, the number of processing for the remaining overlap period

の重なり分を以下のように配分する。なお、より細かな設定条件もあるが、ここでは省略する。
（１）ｅ≦（ｍ−［Λ／μ］）ならば、代替装置に残り不足回数によるＰＭ重なり期間を分配する。
（２）ｅ＞（ｍ−［Λ／μ］）ならば、常用装置に残り不足回数によるＰＭ重なり期間を分配する。ｒ’＝ｒ＋１とし、ｎ_ｊ’’＝ｎ_ｊ’−ｒ’とする。

The overlap is distributed as follows. Although there are more detailed setting conditions, they are omitted here.
(1) If e ≦ (m− [Λ / μ]), the PM overlap period due to the remaining shortage is distributed to the alternative device.
(2) If e> (m− [Λ / μ]), the PM overlap period due to the remaining shortage is distributed to the regular device. Let r ′ = r + 1 and n _j ″ = n _j ′ −r ′.

FIG. 42 is a diagram for explaining complete dispersion of the hands 3 (after the second cycle).
Overlapping time of PM time is eliminated including R times that could not be completely replaced in the first cycle.
The number of replacements (maximum) of the device j is n _j ″ = n _j ′ −r or n _j ′ −r ′. This means that each device is shifted by at least r (or r ′) processing times before the number of times required for complete PM dispersion.
Here, after the second cycle, the shift is shifted backward to achieve complete dispersion.
However, replacement is performed within the range satisfying the conditions of the JAMP method.

Next, the effect of the JAMP method is confirmed by a simulation experiment.
[Purpose and method]
The purpose of this simulation experiment is to distribute the load to each device in advance in consideration of quantitative maintenance when simply processing in parallel in a process with multiple devices (hereinafter referred to as simple parallel rule) The purpose is to confirm the effect of the JAMP rule by comparing the performance of the JAMP rule.
As an experiment method, a simulation experiment is performed assuming six lot / single task type apparatuses and one product type as shown in FIG. Here, the lot / single-task type device means a device for processing one lot at a time. Lots shall arrive at regular intervals according to traffic intensity by periodic insertion. As for the operation of the 6 devices based on the simple parallel rule, when the traffic intensity is 0.8, normally 5 devices are used as shown in FIG. 33, and when any device cannot be used due to PM or the like, an alternative device Use the remaining one.

FIG. 43 is a diagram illustrating model setting values.
FIG. 43 shows the processing time, the traffic intensity of the process, the number of lots (PM guide lot number) serving as a guide for starting quantitative maintenance, and the time required for preventive maintenance (MTTR).
In the setting of MTTR, there are two cases where the conditions (1) and (2) of the JAMP rule can be satisfied (S1, S2), and the case where the condition combining the conditions (1) and (2) cannot be satisfied (S3) A total of three scenarios are used. The difference between the scenarios S1 and S2 is that, in S1, the PM of the alternative device occurs at the end of the PM of the regular device, and corresponds to the condition [3] [1].
S2 is a case where it is necessary to suppress the occurrence of waiting time by applying the condition (3) [2] because the PM of the alternative device overlaps with the timing of PM of other devices if no adjustment is made. ing. S3 is a case where two or more PMs must be generated at the same time. S3 corresponds to the actual PM condition setting of the factory (pauses for 24 hours for PM every time 100 lot processing is finished).
Using the above model, experiments are performed on two rules, simple parallel and JAMP, and the results are compared.

The setting of two rules in this simulation experiment will be described below. FIG. 44 is a diagram for explaining the alternative number setting in JAMP.
As shown in FIG. 37, the simple parallel rule has five units m1, m2,..., M5 at first, and there is a shift of 1 / Λ due to the arrival interval. In the period in which the parallel processing is simply performed on the stand, and any one of the first to fifth units performs PM, the device that can start the processing earlier among m6 or other devices substitutes the processing.

  As shown in FIG. 38, the JAMP rule is a rule for distributing the PM by performing alternative processing of the job in advance while taking quantitative maintenance into consideration. In order to prevent two or more PMs from overlapping, before the m1 starts the PM as shown in FIG. 38, the alternative device m6 substitutes the processing quantity of the m2 to m5 so that the PM of the five devices m1 to m5. The load is distributed so that the number of devices that are suspended by PM is always kept to one or less. FIG. 39 is an explanatory diagram of [2] of the condition (3) of the JAMP method. FIG. 43 shows the number of alternatives when the JAMP rule is used under the setting conditions of this experiment.

Here, one cycle refers to a period from when m1 starts processing the first lot until m5 finishes the MTTR (see FIG. 40). Scenario S3 corresponds to the case where complete decentralization is not possible in one cycle because it exceeds 100 of the PM guideline.
[Results and Discussion]
FIG. 44 is a diagram for explaining the setting of the number of substitutions in the JAMP method. 45 and 46 are diagrams for explaining comparison between the simple parallel rule and the JAMP rule.
43 and 44, simulation experiments were conducted in three scenarios for two rules. FIG. 45 shows the results up to the end of the cycle in which the alternative device performs PM. Here, the performance evaluation index includes three points: the number of lots waiting due to PM, the cycle time, and the number of lots exceeding the Q-time constraint.

As shown in FIG. 45, as a result up to the cycle of completing load distribution and PM in the alternative device, the results up to 9 cycles are shown in S1, and the results up to the 3rd and 4th cycles are shown in S2 and S3, respectively.
As the production period (observation period) becomes longer, the difference between simple parallel and JAMP increases in proportion to the values in the above table.
From the above experimental results, it can be seen that in the JAMP rule, in the scenarios S1 and S2, the waiting for the lot by PM does not occur, and there is no lot exceeding the Q-time constraint. In particular, it can be seen from the total waiting time that the simple parallel rule does not generate a wait in the JAMP rule while the long wait occurs in the simple parallel rule.
Further, from the total number of lots in which waiting has occurred, in the simple parallel rule, a waiting lot has occurred every cycle, and this difference increases with time. From the results of these performance evaluations, in the JAMP rule, by distributing PM, the traffic intensity is prevented from becoming larger than 1, the waiting time of the lot due to PM is suppressed, and the lot exceeding the Q-time constraint We were able to confirm that the occurrence of

  If the conditions (1) and (2) are combined in the JAMP rules as in scenarios S1 and S2, the alternative process for the condition (1) is performed by the alternative device in the first period and the PM is started in advance. In order to adjust the time point, the processing relating to the condition (1) is not necessary after the second round. In the second and subsequent rounds, the processing for the condition (2) is performed by the alternative device, and the processing is performed between the devices so that the PM timing of the alternative device does not overlap with the PM timing of the regular device as shown in the condition (3). Perform an operation to substitute for. By these operations, it was confirmed that the waiting time in the JAMP rule does not occur.

  Regarding the JAMP rule, the scenario S1 corresponds to [1] of the condition (3). That is, since the alternative device m6 completes the processing of the 100th lot and enters the PM, the alternative processing frequency in the alternative device m6 is 30 times in the first cycle, and 10 times in each cycle after the second week. The time when the number of processes in the alternative device m6 reaches the 100th lot is the time 30 + 10 + 10 + 10 + 10 + 10 + 10 + 10 = 100 when the PM of all five regular devices ends in the eighth cycle, and the alternate device m6 enters the PM, so there is no effect on the regular device. At the time when the alternative device m6 starts PM, the device m1 has already been processed 2 × 4 = 8 lots for 4 MTTR periods, so the number of remaining processings possible until the next PM starts is 100−8 = 92 times It is.

  On the other hand, the total number of times of substitution of 10 times after the second cycle and the number of times the substitution device is processed twice per device during the PM period is 12 times the remaining number of times of processing of the regular device m1 that starts the next PM Since it is small, PM can be finished with a sufficient margin. In the second and subsequent rounds, it is not necessary to replace the processing with the normal devices m2 to m5 in advance, so that the PM of all the devices is distributed without any problem, and one device that performs PM at the same time It became the following.

  On the other hand, using the case of scenario S2, the result when the alternative device m6 reaches 100 lots of the alternative device in the middle of the second and subsequent cycles and enters the PM is shown. In this case, if no action is taken, the PM of the alternative device is started during the PM of the regular device, so that a waiting lot is generated. In order to deal with this case, the condition (3) [2] is required. In [2] of this condition (3), the MTTR of a plurality of devices is prevented from overlapping by intentionally substituting another device in order to adjust the PM timing of the alternative device after the second lap. Can be adjusted. In the case of S2, the number of prior process substitutions for shifting the PMs of the devices m1 to m5 so as not to overlap at the same time in the first round is 60 times, and the devices m1 to m5 are substituted for processing during the MTTR period. 35 alternatives are required. The substitute device needs to start PM at the time when the remaining 100-95 = 5 times are substituted in the second round.

  In the second lap, there is no need to replace the process in advance, so if nothing is done, the m6 device must start the PM when the process is replaced five times during the PM of m1. Two devices become PM during the PM period MTTR. In order to prevent this, it is necessary to complete PM of the alternative device m6 before m1 starts PM in the second period. Therefore, from the time when the PM of m5 is completed (m1 has been processed 28 lots by this time), the device m6 substitutes the devices of m1 to m5 once (by this time, m1 is 28 + 4 = 32 lots). When this is finished, m6 starts PM. After MTTR elapses, m6 becomes usable (by this time, m1 is processed 32 + 7 = 39 lots). At this time, even with the device m1 where PM is most likely to occur, only 39 lots have been processed, and the number of guide lots for quantitative maintenance is 100 or less, so the PM of all devices can be distributed. From the experimental results, it was confirmed that the PM can be dispersed even when the condition (3) [2] is satisfied, and the generation of the waiting time of the lot due to the PM can be completely suppressed.

  Finally, in scenario S3, the condition that combines conditions (1) and (2) is not satisfied. That is, a situation occurs in which a plurality of devices must enter the PM in the first cycle. However, even in this case, it was confirmed that the dispersion of the PM as much as possible by using the alternative device can suppress the waiting time of the lot due to the PM and suppress the Q-time constraint crack to zero. Scenario S3 is a setting for quantitative maintenance in an actual factory, and an effect can be expected.

The JAMP method is a method aimed at leveling production by minimizing a decrease in production capacity of a process due to PM to suppress fluctuations. Therefore, in the subsequent process, the arrival of lots is leveled (FIG. 46: end time).
FIG. 46 is a diagram showing the time of arrival and withdrawal of a lot (comparison between JAMP method and simple parallel rule). On the other hand, as a simple parallel problem, the increase or decrease (fluctuation) in production capacity is large (lower figure), resulting in a rapid increase in the number of work in progress and fluctuations in flow rate to the subsequent process (upper figure).
FIG. 47 is a diagram showing fluctuations in the operating status of the number of devices (in the case of simple parallel MTTR = 1380 [minutes]). From the above, the effect of the job allocation method JAMP in consideration of quantitative maintenance was shown.

[Conclusion]
In the present embodiment, a processing process with quantitative maintenance is taken up, and a device load distribution method for reducing the number of devices that become preventive maintenance at the same time as much as possible has been developed, and research has been conducted for the purpose of confirming the effect.
In the present embodiment, a load balancing method JAMP (Job Allocation for Maintenance in Parallel) method is newly proposed. In order to confirm the effect of the JAMP method, a comparison with a simple parallel rule was performed by a simulation experiment with reference to an actual factory process model. As the simulation model, a continuous processing model was used in which a process was processed with a single device using six devices, and periodic input was performed so that the traffic intensity of the process became 0.8.
In the experiment, the number of devices that become PM at the same time by JAMP can be reduced to 1 or less and completely distributed (traffic intensity can always be less than 1), and PM cannot be completely distributed (always always) Two cases were set in which the traffic intensity could not be suppressed to less than 1, and the effect of JAMP in each case was confirmed.

As a result, when PM can be completely decentralized, the waiting time due to PM can be kept to zero, and even when PM cannot be completely decentralized, the number of devices that JAMP becomes PM at the same time is possible As a result, it was confirmed that it is superior to the simple parallel rule because it works to reduce as much as possible.
Here, especially when the MTTR is large, the waiting time of the lot due to PM becomes long, and a situation in which rework occurs due to the Q-time constraint is likely to occur. The occurrence of rework is only a loss in production and is undesirable because it obstructs the production flow of regular lots. In order to prevent this Q-time constraint cracking, the number of PMs that become PM at the same time is reduced to keep the traffic intensity of the process as low as possible, and no extra waiting time is generated. It was confirmed that the number could be reduced.
The proposed method can be used not only for quantitative maintenance but also for decentralized regular maintenance.

In practical use, it is necessary to consider setup time in multi-product production and batch assembly time in a batch process.
The conditions in that case are shown below.
λ _j is the arrival rate per device j [lot / min] λ _j = Λ / [Λ / μ], Λ is the entire device group arrival rate [lot / min], and M is the process service capability = m / Tp [Lot / min], (m: number of devices, Tp: average processing time), N is the reference value [lot] for quantitative maintenance, MTTR is the time required for preventive maintenance (PM) [min], and Ts1 and Ts2 are setup times The conditions when considering the setup in multi-product production can be expressed as follows.
Condition (1) is

条件（２）は、

Condition (2) is

条件（１）および（２）は、

Conditions (1) and (2) are:

となる。

It becomes.

Moreover, the conditions which considered the batch assembly time in a batch process are as follows.
In a batch apparatus, the processing start interval (load interval) and the processing unit (load size) differ depending on the load rule. The execution service rate per unit can be expressed by the following formula.
Execution service rate per unit = (load size) / (load interval)
The load interval when viewed from the apparatus can be calculated as follows when the load size is La.
Since the time interval at which one lot arrives is 1 / Λ, the time interval until one batch of size La is assembled is La / Λ. The load interval TL of the device is
When La / Λ> Tp / m (ie, ρ <1), TL = Tp + (La / Λ−Tp / m)
When La / Λ ≦ Tp / m (ie, ρ ≧ 1), TL = Tp

Here, M is the number of devices, Tp is the average processing time, La is the average load size [lot], Λ is the arrival rate of the entire device group [lot / min], and λ _j = Λ / [Λ / μ].
By using this TL in place of Tp to calculate the process capability Μ, it is possible to consider the batch assembly time.
La depends on the load rule and traffic intensity. If the load rule is such that the processing can be started even if the maximum batch size is not satisfied, the processing in the device starts as soon as the device is available (the device always operates).
The average time interval when one device is free is Tp / m, and the number of WIPs arriving in that period is Tpm × Λ.
Planned service rate μ = LTp

このＬａを用いる。
なお、各変数の定義は、次の通りである。
ρ：工程利用率、Ｌ：最大バッチサイズ［ｌｏｔ］、Ｌａ：平均ロードサイズ［ｌｏｔ］、Λ：装置群全体の到着率［ｌｏｔ／分］、λ_ｊ＝Λ／［Λ／μ］、ρ＝Λ／ｍμ

This La is used.
The definition of each variable is as follows.
ρ: process utilization rate, L: maximum batch size [lot], La: average load size [lot], Λ: arrival rate of the entire device group [lot / min], λ _j = Λ / [Λ / μ], ρ = Λ / mμ

In the second embodiment, the second problem is solved by the JAMP method.
The load distribution method to the device for decentralizing the preventive maintenance time of the device, which is the main cause of the Q-time constraint cracking, is named JAMP (Job Allocation for Maintenance in Parallel) method. The JAMP method minimizes the number of equipment that stops at the same time, maintains high production capacity and leveles production, and drastically prevents unnecessary waiting of product lots and cracking of Q-time constraints. The second embodiment is a JAMP method program and a performance evaluation result using the program (by deterministic simulation using actual factory data). From the performance evaluation results, the JAMP method reduces 188 lots of rework per 2.5 months compared to the simple parallel rule that uses parallel devices evenly in order, and has the effect of reducing rework costs of 20 million yen or more. In addition, since the number of waiting lots is expected to be reduced to 1/172 and the total waiting time of the lots is reduced to 1/13805, it is effective in meeting the delivery date. By using a combination of the CKB method and the JAMP method, optimal management of a single Q-time constraint section is realized.

<Third Embodiment>
With reference to FIG. 48, the optimal index generation apparatus which concerns on 3rd Embodiment of this invention is demonstrated. The optimum index generation device 31 according to the third embodiment of the present invention, for example, produces the processing step (device group) 2 and the processing step (device group) 3 provided in the production step # 1 for producing a semiconductor, respectively. The production management system 10 that manages the schedule is the target of control. At the same time, the production management system 10 and the engineering work system 29 that manage the maintenance schedule of the apparatus 5 provided in the production process are controlled by the operation method for distributing the preventive maintenance time of the manufacturing apparatus. In the present embodiment, a production distribution control method and a production load distribution method (hereinafter referred to as CKB-JAMP method) will be described.

As shown in FIG. 48, the optimal index generation device 31 includes an input unit 12, an auxiliary input unit 13, and a Q-time structure analysis unit 14 in the same manner as the optimal index generation device 11 (first embodiment) shown in FIG. I have.
Moreover, the optimal index generation device 31 is similar to the optimal index generation device 21 (second embodiment) shown in FIG. 30, and includes an input unit 22, an auxiliary input unit (JAMP method) 23, a start / end time calculation unit 26, and device classification. A calculation unit 28 is provided.
Compared with the first and second embodiments, as shown in FIG. 48, the characteristic configuration of the third embodiment is a fluctuation monitoring detection unit 32, an input change detection unit 33, a calculation necessity determination unit 34, a sensitivity. An analysis unit (CKB + JAMP characteristic) 35, a CKB + JAMP optimum theoretical value calculation unit 36, and a sensitivity analysis unit (CKB + JAMP performance characteristic) 37 are provided.

Here, the above-described configuration characteristic of the third embodiment will be described.
The change monitoring detector 32 monitors and detects changes due to event / state data updates such as start / end of lot processing, changes in device status (according to failures, etc.) and changes in the processing capability of the device group, and changes information is obtained. It outputs to the calculation necessity judgment part 34.

  The input change detection unit 33 inputs input data related to the CKB method and input data related to the JAMP method, detects a changed portion from the previous input data related to the CKB method and input data related to the JAMP method, and there is a change. Output the input data.

  The calculation necessity determination unit 34 inputs an integer threshold value in the sensitivity analysis result calculated in advance stored in the storage unit, and recalculation is necessary for the fluctuation detected by the input change detection unit 33. Judge whether there is. The calculation necessity determination unit 34 determines that recalculation is necessary when the upper and lower thresholds of the previous range are exceeded.

The CKB + JAMP optimal theoretical value calculation unit 36 receives input data from the Q-time constraint structure analysis unit 14 and the input unit (JAMP method) 22, and performs a JAMP method theoretical value calculation on each device group in the Q-time constraint section. The processing load (number of times) of the device and its processing time are calculated, and the time transition of the processing capacity in the device group is calculated for one cycle. Also, the CKB method optimum theoretical value calculation is performed every time the processing capacity is switched in one cycle. Here, one cycle is defined as a period (or a separately designated production period) until all regular devices complete preventive maintenance once or more and all alternative devices complete once.
Furthermore, the CKB + JAMP optimal theoretical value calculation has a function of changing the input amount in the Q-time constraint section in advance as a value that can comply with the Q-time constraint in accordance with a change in the processing capability of the device group. Realize section management.

  With reference to FIG. 49, the operation of the CKB + JAMP optimum theoretical value calculator 36 will be described. In a certain process (for example, a bottleneck apparatus group in the Q-time restriction section), the traffic intensity of the process is kept below 1 (100%) on average during a period in which all apparatuses can operate. (A production plan is made so as to do so.) At this time, all lots arriving at the processing step can finish the processing within a certain stay time. Here, during a period when any of the devices is in preventive maintenance, the processing capability of the process (device group) is reduced. In some cases, the traffic intensity of the process exceeds 1 (100%) during this period. And if this period becomes long, the Q-time constraint will not be able to be observed. In order to avoid such a situation, decentralization of preventive maintenance is attempted by the JAMP method. However, when the preventive maintenance time is relatively long with respect to the operation time, the traffic intensity of the process is set to 1 in all cases. It is not necessarily limited to the following. For this reason, a function is provided for reviewing the production flow rate in the Q-time restricted section in accordance with changes in the processing capacity of the process.

Further, in order to increase the throughput of the Q-time constraint section as much as possible while maintaining the Q-time constraint as in the CKB method, the maximum lot that can satisfy the constraint is allowed to stay in the Q-time constraint section. In such a case, a lot entered in the Q-time constraint section before the processing capacity of the process is reduced cannot comply with the Q-time constraint.
For this reason, before the processing capacity decreases, the rate of lot injection into the Q-time constraint section is reduced in advance at an appropriate time.
On the other hand, after the preventive maintenance is completed, the processing capacity is improved and the apparatus is operated as much as possible to eliminate the accumulated intermediate stock. Also at this time, if the lot input to the process is not increased in advance, the processing capacity and time after the apparatus restoration is wasted.
Taking these into account, the calculation of the appropriate theoretical values for each period with different process capabilities, and the revision of the theoretical values including the transition period to adjust the input according to the change in the process capability Appropriate timing is calculated.

Here, with reference to FIG. 49, a method of dividing each period and transition period having different process capabilities will be described.
First, the time of preventive maintenance and the process capability of each period are determined by the JAMP method. Based on this calculation result, the periods with different process capabilities are divided into three major categories. The first period is a stationary period. When the arrival rate of the lot is Λ and the service rate of the processing process in the stationary period is μ1, the average condition is μ1 = Λ.
Next, the second period is a period of decline, in which the service rate of the processing process is reduced due to preventive maintenance or the like. Assuming that the service rate in the decline period is μ2, it is in an average state of μ2 <Λ (when such a state cannot be prevented by the JAMP method or the like).
Next, the third period is the recovery period, and the processing process is operated at the maximum service rate in order to quickly eliminate the intermediate inventory that has accumulated in the second period. When the service rate in the recovery period is μ3, the average condition is μ3> Λ.

Here, in the section with Q-time constraint, the first period (stationary period) to the second period (decreasing period), the second period (decreasing period) to the third period (recovery period), the third period ( At the time of Td, Tr, Tk before the transition from the recovery period) to the first period (stationary period), it is necessary to appropriately change the arrival rate to the process in advance.
Here, the time point when the stationary phase is switched to the lowering phase is T1, the time point when the lowering phase is switched to the stationary phase is T2, and the time point when the recovery phase is switched to the stationary phase is T3. In the stationary period, the stay time from the arrival of the lot to the start of processing is set as Tw, and the time limit due to the Q-time constraint in this section is Tq.

i) Regarding switching from the stationary phase to the decreasing phase (A) In the case of Tw = Tq, Td = T1-Tw
(B) When Tw <Tq, Td = T1 + (μ2Tq−μ1Tw) / (μ1-μ2)
In order to keep the stay time below Tq, the arrival rate is controlled to be changed from μ1 to μ2 at the time of Td at the latest.
In addition,
(C) When Tw> Tq, the Q-time constraint cannot be always complied with. In this case, it is first necessary to control the logistics so that the condition of (A) or (B) is satisfied by the CKB method. is there.

ii) Regarding switching from the recovery period to the recovery period,
(A) When Tw = Tq, Tr = T2-Tw
(B) If Tw <Tq, Tr = T2-Tq
Control is made so that the arrival rate is changed from μ2 to μ3 at the time after Tr at the earliest.

iii) Regarding switching from the recovery phase to the stationary phase,
(A) When Tw = Tq, Tk = T3-Tw
(B) When Tw <Tq, Tk = T3− (μ3Tq−ΛTw) / (μ3−Λ)
Control is performed so that the arrival rate is changed from μ3 to Λ (= μ1) as early as after Tk.

Sensitivity analysis unit (CKB + JAMP characteristics) 35, CKB method is based on the sensitivity analysis of JAMP method, features of the sensitivity analysis (CKB + JAMP characteristics), as well as the features in the CKB + JAMP optimal theoretical value calculating section 36 Sensitivity analysis is performed for all target periods for each period with different process capability and its transition period. The information in the internal storage unit is also used for determining the necessity of executing the optimum index generation 31. This necessity determination is based on various changes in the production processes # 1 and # 2 to be managed (changes in the processing capacity of the device group, changes in the target production rate, changes in traffic, changes / additions of Q-time constraints, etc. ). When the necessity is confirmed, a calculation instruction is issued to the optimum index generation 31.
Sensitivity analysis unit (CKB + JAMP performance characteristics) 37, CKB method is based on the sensitivity analysis of JAMP method, features of the sensitivity analysis (CKB + JAMP performance characteristics) is characterized in the sensitivity analysis section (CKB + JAMP characteristics) 35 Similarly, the performance evaluation in the sensitivity analysis is performed for all the target periods for each period with different process capability and its transition period.

With reference to FIG. 48, the operation of the optimum index generation device 31 according to the third embodiment of the present invention will be described.
Information about a plurality of devices included in the production execution system is input by the input unit 12 via the production management system 10, and Q indicating the upper limit value and the lower limit value of the stay time between the machining steps by the plurality of devices from the input information. -A time structure is analyzed by the Q-time structure analysis unit 14, and for each Q-time constraint, the arrival rate at which the magnitude relationship between the batch set waiting time related to product arrival and the upper limit value of the Q-time constraint is switched. A plurality of devices are determined from information on a plurality of regular devices and alternative devices included in the production execution system input from the input unit 22 via the production management system 10 The category calculation unit 28 classifies the device into a regular device and an alternative device, calculates the number of operations and the operation timing of a plurality of devices, and predicts the regular device and the alternate device. The maintenance start time is calculated, and based on the calculated preventive maintenance time, the input information and the Q-time structure, the optimum number of kanbans and the buffer size for each product type are calculated and output as index information by the index calculation unit 17 Then, the number of operations, the operation timing, and the preventive maintenance start time relating to the regular device and the alternative device are calculated by the start / end timing calculation unit 26, and the optimum index information calculated by the index calculation unit 17 is supplied to the production management system 10. At the same time, the optimum start / end time information calculated by the start / end time calculation unit 26 is supplied to the production management system 10.

As a result, even in situations where Q-time constraint cracking, which has not been achieved by the conventional method, is likely to occur, the Q-time constraint can be reliably observed, and good product throughput improvement, cost reduction, and environmental load reduction can be achieved. Possible optimal operation management conditions can be provided.
Also, by providing analysis methods and appropriate management methods for the main causes of Q-time constraint cracking, handling of multi-product production, preventive maintenance of equipment, and mutual interference of multiple Q-time constraints. , While complying with the Q-time constraint, it is possible to simultaneously improve the throughput of non-defective products and reduce costs and environmental loads.

<Fourth embodiment>
With reference to FIG. 50, the optimal index generation apparatus which concerns on 4th Embodiment of this invention is demonstrated. The optimum index generation device 41 according to the fourth embodiment of the present invention is, for example, the production of the processing step (device group) 2 and the processing step (device group) 3 provided in the production step # 1 for producing a semiconductor. The production management system 10 that manages the schedule is the target of control. At the same time, the production management system 10 that manages the maintenance schedule of the apparatus 5 provided in the production process # 1 is controlled by the operation method for decentralizing the preventive maintenance time of the manufacturing apparatus. In the present embodiment, an interlocked production logistics control method (hereinafter referred to as a procedural CKB method) will be described.

As illustrated in FIG. 50, the optimal index generation device 41 includes an input unit 12, an auxiliary input unit 13, and an index calculation unit 17, similar to the optimal index generation device 11 (first embodiment) illustrated in FIG. .
Further, the optimal index generation device 41 includes an input unit 22, an auxiliary input unit 45 , and a start / end time calculation unit 26, as in the optimal index generation device 21 (second embodiment) shown in FIG.
Further, the optimal index generation device 41 is similar to the optimal index generation device 31 (third embodiment) shown in FIG. 48. The fluctuation monitoring detection unit 32, the input change detection unit 33, the calculation necessity determination unit 34, the sensitivity analysis unit (CKB + JAMP characteristic) 35, CKB + JAMP optimum theoretical value calculation unit 36, and sensitivity analysis unit (CKB + JAMP performance characteristic) 37.
Compared with the first, second, and third embodiments, as shown in FIG. 50, the characteristic configuration of the fourth embodiment is a production flow bottleneck / flow rate analysis unit 42, a Q-time structure analysis unit 43. A device classification calculation unit 44, an auxiliary input unit (JAMP method) 45, a drive control unit 46, a margin time distribution analysis unit 47, and a procedure execution control unit (CKB + PS method) 48. The drive control unit 46 includes a fluctuation monitoring detection unit 32, an input change detection unit 33, a calculation necessity determination unit 34, and a sensitivity analysis unit (CKB + JAMP characteristic) 35.

Here, the above-described configuration characteristic of the fourth embodiment will be described.
The production flow bottleneck / flow rate analysis unit 42 manages the process connection matrix (production flow in order to simultaneously manage the Q-time constraints corresponding to the fluctuations in the production flow rate and process capability in the entire production flow included in the production execution system. ), Processing capacity matrix of process, allowable flow rate of Q-time constraint interval, input TOC theory or maximum flow problem / minimum cut theorem, etc., to perform integrated production flow analysis, bottleneck location detection, maximum flow rate calculation . The production flow bottleneck / flow rate analysis unit 42 outputs the bottleneck location (process group corresponding to the minimum cut) and the flow rate (processing load) in each processing step.

The Q-time structure analysis unit 43 includes CKB method input data, JAMP method input data, process connection data (production flow), process throughput data, Q-time constraint data, and processing load for each production type in the process. Enter. Then, the Q-time structure analysis unit 43 detects which Q-time constraint section corresponds to the structure (4 types in series and 5 types in parallel) and analyzes the structure of the Q-time constraint. At the same time, the Q-time structure analysis unit 43 checks whether or not a Q-time constraint section that requires special processing (exclusion over three processes or more and intermediate process overlap) is required.
The Q-time structure analysis unit 43 outputs the range and structure of each Q-time constraint section (start / end process of Q-time constraint section, structure No., etc.) and sets a special structure flag (ON = 1). .

  The device category calculation unit 44 is the same as the device category calculation unit 28 used in the second embodiment.

The auxiliary input unit 45 inputs data used in the CKB + JAMP optimum theoretical value calculation unit 36, the margin time distribution analysis unit 47, and the drive control unit 46 from the production management system or the engineering work system.
The drive control unit 46 inputs the latest production plan / actual information from the production management system by the fluctuation monitoring detection unit 32 and manages each change by the input change detection unit 33 in order to perform management in accordance with the dynamic change of the production process. A change with respect to the Q-time constraint section is detected, and the calculation necessity determination unit 34 determines whether or not recalculation of the CKB method is necessary. If it is determined that it is necessary, a recalculation instruction is input to the procedure execution control unit 48, etc., and the processes from S22 to S25 of the procedural CKB method shown in FIG. Recalculate the value.

The surplus time distribution analysis unit 47, in response to the input from the auxiliary input unit 45, determines the location (between processes) and the time length of the surplus time distribution in the Q-time constraint section, the number of devices in each process, the service capability, and the product logistics. Calculate from the state of A structure that requires an additional procedure in the procedural CKB method (standard) is managed by a procedure with conditional branching as shown in FIG. The structure requiring an additional procedure is a case where a plurality of Q-time constraints corresponding to the serial type (C) overlap in FIG. 52 or a single Q-time constraint section spans three or more steps (in a parallel structure, When it is converted to the serial type, it corresponds to the above structure).
In the case of a production process including a structure requiring these additional procedures, the process flow of FIG. 55 is used. The characteristic difference from FIG. 54 lies in the determination of the time allocation location in S35 and the determination of the allowance time allocation in S36. The allowance time distribution analysis unit 47 assumes these processing functions.

The procedure execution control unit 48 controls and organizes the processing flow of the procedural CKB method (CKB-PS method). For this purpose, the execution of the main processing flow shown in FIGS. 54 and 55 is controlled.
The basic processing flow of the procedural CKB method will be described using the processing flow of FIG.
In FIG. 54, first, a process is performed in S21. This is because the Q-time structure analysis unit 43 detects all the Q-time constraint sections, and sets the detected information and the set of processing steps included in the Q-time constraint section as a sub-set of the production process. Then, the target data of the main processing flow of the procedural CKB method is initialized. While the structure, time length, etc. of all the Q-time constraint sections are not changed, a set sub-set including information on all the Q-time constraint sections detected in S21 is recorded in the internal storage unit S21a. . In next S22, the bottleneck portion in the production process is detected and the maximum flow rate in each processing process is calculated. In addition to the production flow bottleneck / flow rate analysis unit 42 shown in FIG. 50, S22 can be calculated more easily from the production plan and the service capability of the processing process in the case of a serial production process flow. The calculation result of S22 is stored in the internal storage unit S22a.

Next, S23 and S24 are processes for each Q-time constraint section. In S23, the bottleneck location in each Q-time constraint section is detected and the maximum flow rate is calculated in the same manner as S22. At this time, the production flow rate input into the Q-time restriction section is read from the internal storage unit S22a. Thereby, it becomes possible to manage each Q-time restriction section according to the bottleneck location and the maximum flow rate of the entire production process. In S24, the CKB method is calculated for the Q-time constraint section based on the calculation result of S23. For S24, the CKB + JAMP optimal theoretical value calculator 36 shown in FIG. 50 is used. The calculation result of S24 is recorded in the internal storage unit.
In S25, all the Q-time constraint sections recorded in the internal storage unit S21a are read, and it is determined whether or not the processes of S23 and S24 have been performed for all the Q-time constraint sections. In S25, while there are one or more Q-time constraint sections in which the processes in S23 and S24 are not completed, the processes in S23 and S24 are executed.
Thus, after the optimal index calculation is completed for all the Q-time constraint sections, the optimal index calculation results of all the Q-time constraint sections are output to the production management system in S26, and management is executed.

After that, in order to manage according to the dynamic change of the production process, it is necessary to acquire the production history information from the production management system in S27, monitor the change of each Q-time constraint section, and recalculate. If it is determined, the processes from S22 to S25 are re-executed, and the optimum index is recalculated. The change monitoring detector 32 and the input change detector 33 shown in FIG. 50 are used for monitoring changes, and the calculation necessity determining unit 34 is used for determining the necessity of recalculation.
50 uses the sensitivity analysis result based on the previous production process model (particularly, a threshold value as a guideline for recalculation) and the production process model to be evaluated this time.
In FIG. 50, the previous sensitivity analysis result (particularly, a threshold value that serves as a guideline for recalculation) is input to the calculation necessity determination unit 34 from the internal storage unit of the sensitivity analysis unit 35, and the current production process is performed from the fluctuation monitoring detection unit 32. Information about the model is entered.

A structure that requires an additional procedure in the procedural CKB method (standard) is managed by a procedure with conditional branching as shown in FIG. The structure requiring an additional procedure is a case where a plurality of Q-time constraints corresponding to the serial type (C) overlap in FIG. 52 or a single Q-time constraint section spans three or more steps (in a parallel structure, When it is converted to the serial type, it corresponds to the above structure).
In the case of a production process including a structure requiring these additional procedures, the process flow of FIG. 55 is used. The characteristic difference from FIG. 54 lies in the determination of the time allocation location in S35 and the determination of the allowance time allocation in S36.
In S <b> 35, the surplus time distribution analysis unit 47 determines between which process steps the surplus time in the Q-time constraint section is allocated. In general, it is desirable to assign more preferentially before and after the bottleneck portion, but in the Q-time constraint section, two points are taken into consideration: compliance with the Q-time constraint and improvement of the production rate at the bottleneck portion.
In S36, the allowance time distribution analysis unit 47 calculates the allowance time allowance in the Q-time constraint section from the determination result in S35, the number of devices in each process, the service capability, the state of the product distribution, and the like.

With reference to FIG. 50, operation | movement of the optimal index generation apparatus 41 which concerns on 4th Embodiment of this invention is demonstrated.
Information regarding a plurality of devices included in the production execution system is input by the input unit 12 and the auxiliary input unit 13 via the production management system, and the bottleneck location and the maximum flow rate in the production process by the plurality of devices are input from the input information. Is analyzed by the production flow bottleneck / flow rate analysis unit 42, and the Q-time structure analysis unit 43 converts the Q-time structure indicating the upper limit value or the lower limit value of the stay time between processing steps by a plurality of apparatuses from the input information. Analyze and store in internal memory, and for each Q-time constraint, calculate the boundary value of the arrival rate at which the magnitude relationship between the batch set waiting time related to product arrival and the upper limit value of the Q-time constraint switches From the information regarding a plurality of devices included in the production execution system that is determined by the input unit 22 and that is determined by an appropriate load rule, the device classification calculation unit 4 Using the CKB + JAMP optimal theoretical value calculation unit 36, the number of operations and the operation times of the plurality of regular devices and alternative devices are calculated from the calculated device classification and the input information. In addition, the preventive maintenance start time for the regular device and the alternative device is stored in the internal storage 26a of the start / end time calculation unit 26, and the optimum load rule, bottleneck location, and maximum flow rate determined as the calculated preventive maintenance time are stored. Based on the Q-time structure, the optimum number of kanbans and buffer size for each product type are calculated. As in the third embodiment, the optimum index information calculated by the index calculation unit 17 is supplied to the production management system 10 for each period in which the process capability calculated by the start / end time calculation unit 26 is different, and the start / end is started. Optimal index information calculated by the time calculation unit 26 is supplied to the production management system 10.

As a result, even in situations where Q-time constraint cracking, which has not been achieved by the conventional method, is likely to occur, the Q-time constraint can be reliably observed, and good product throughput improvement, cost reduction, and environmental load reduction can be achieved. Possible optimal operation management conditions can be provided.
Also, by providing analysis methods and appropriate management methods for the main causes of Q-time constraint cracking, handling of multi-product production, preventive maintenance of equipment, and mutual interference of multiple Q-time constraints. , While complying with the Q-time constraint, it is possible to simultaneously improve the throughput of non-defective products and reduce costs and environmental loads.
The optimum operation management conditions can be updated in response to changes in the production logistics of the dynamic production process, and the optimum operation management conditions consistent with the state of the production process can be always provided.

Hereinafter, the operation of each unit constituting the optimum index generation device 41 according to the fourth embodiment of the present invention will be described.
First, a structure analysis and simultaneous management method for a plurality of Q-time constraints and a policy for dealing with changes in production logistics and production capacity in the manufacturing flow according to the fourth embodiment will be described.
A method for analyzing and simultaneously managing a structure formed by a plurality of Q-time constraints is the MNA1 method, and a method for dynamically managing a plurality of Q-time constraints simultaneously based on the detection result of the production flow of the entire production line is MNA2. The MNA1 method and the MNA2 method are collectively referred to as the MNA method.
And the method of managing the Q-time restriction section of the whole production flow using the MNA method (MNA1 method and MNA2 method) is defined as a procedural CKB (CKB with Procedure and Structure: CKB-PS) method. As a result, it is possible to reduce the amount of calculation by distributing and managing independent Q-time constraint sections, and it is possible to achieve overall optimization in consideration of the interference relationship.

FIG. 50 is a diagram showing an overview of a management system that integrates the MNA method and the procedural CKB method.
Next, the MNA1 method, the MNA2 method, and the procedural CKB method will be described below.
(1) The MNA1 method is a structure analysis of a plurality of Q-time constraints and a simultaneous management method according to the structure.
When a plurality of Q-time constraints have a structure in which they interfere with each other, if they are not managed at the same time, there is a possibility that one can be protected but the other cannot be protected.
Therefore, as the MNA1 method, after defining the type of Q-time constraint and the conversion method (FIG. 51), a structure formed by a plurality of Q-time constraints (4 types in series) (FIG. 52) and 5 types in parallel (FIG. 53) defined exhaustively, and determined a method for simultaneously managing a plurality of Q-time constraints in each structure by utilizing the CKB-PS method.
This makes it possible to comprehensively manage all the Q-time constraints scattered in the production flow. A newly added Q-time constraint can be handled in the same manner. Thus, since there is no prior art that comprehensively defines the structure of the Q-time constraint and organizes the management method, the novelty is high. By combining the MNA1 method and the CKB-PS method, Q-time constraints with interference on the production flow can be managed simultaneously, and compliance with the Q-time constraints of all structures and an improvement in productivity can be expected.

In FIG. 51, the type of Q-time constraint: since only the process start time is technically controllable, it indicates that the start point of the Q-time constraint is converted into a process start event (structure definition preprocessing). FIG. FIG. 52 is a diagram showing the structure of Q-time constraints (four types in series: (A), (B) -1 (1), (C) -1 (2), (C) -2)). FIG. 53 is a diagram showing a structure of Q-time constraints (parallel type 5 types: (D), (E) -1, (E) -3, (F) -1, (F) -3).
Here, the management method of each structure will be briefly described.
(ST1) In the serial type (A) exclusion, each Q-time constraint section can be managed by the CKB method.
(ST2) In the case of series type (B) common point, it is introduced according to the bottleneck location, and each Q-time constraint section is managed by the CKB method. Through the CKB method, the preceding Q-time constraint section is linked with the subsequent Q-time constraint section. (Linkage control by buffer size works immediately when the input rate in the following section decreases, and linked control by Kanban works immediately when it increases)
(ST3) In the serial type (C) -1 process overlap (envelopment), this “process overlap (enclose)” structure is converted into a “point common” structure, and the encircling Q- Allocate the remaining time of the time constraint time. Input according to the bottleneck location, and manage each Q-time constraint section by CKB method (same as point common).

(ST4) In the serial type (C) -2 intermediate process duplication, it is added in accordance with the bottleneck location, and each Q-time constraint section is managed by the CKB method. However, kanban release and buffer size filling rate at the end of all Q-time constraint sections are passed to the start point of all Q-time constraints, and the kanban and buffer size conditions for all Q-time constraint sections are aligned. The process can be started only in the case. (Additional items: Kanban copy and distribution, etc.)
(ST5) In parallel type (D) complete parallel, each independent line section is handled in the same manner as in series type.
(ST6) Parallel type (E) -1 (1) Point merging = (2) Process merging is handled in the same manner as the multi-product CKB method. The ability distribution ratio in the merge start process is determined and each is managed independently (as a serial type).
(ST7) Parallel type (E) -2 section merging is handled in the same way as the multi-product CKB method. The ability distribution ratio is determined in the overlapping process after the start of merging, and each is managed independently (in series).
(ST8) The parallel type (F) -1 (1) point branch = (2) process branch is handled in the same manner as the multi-product CKB method. The capacity distribution ratio in the branch start process is determined and managed independently (as a serial type).
(ST9) In the parallel type (F) -2 section branch, it is handled in the same manner as the multi-product CKB method. The capacity allocation ratio is determined in the overlapping process before the start of branching, and each is managed independently (in series).

(2) The MNA2 method is an analysis method for detecting, monitoring, and predicting a dynamic bottleneck location.
In an actual production system, the dynamic situation changes and the bottleneck location may also change. Therefore, it is necessary in practice to monitor the bottleneck location and take appropriate measures in accordance with the bottleneck location.
Therefore, the number of arrivals of lots per unit time and the process capability were calculated, and a method for detecting the bottleneck portion was determined. Depending on whether the production flow is serial or parallel, the method of detecting the bottleneck can be divided to reduce the calculation load in the case of the serial type.
(ST21) In the case of the serial type, the part having the minimum value of the arrival rate and the process service rate is calculated.
(ST22) In the case of the parallel type, the bottleneck portion is extracted by a high-speed algorithm applying the maximum flow problem-minimum cut theorem.

  There are no examples of applying the maximum flow problem-minimum cut theorem to semiconductor manufacturing, etc., but for the management of the entire manufacturing flow, management guidelines for all bottleneck locations and each Q-time constraint section The visualization based on network calculation is useful for the overall optimization in practice, so the method based on network theory is suitable. Here, when calculating the number of arrivals, in addition to the actual value, a predicted value based on a prediction method such as exponential smoothing is used, and the actual value as well as the predicted value based on the JAMP method or failure analysis is used for the process capability. It can also be used for prediction.

A handling guideline according to a change in the positional relationship between the bottleneck location and the Q-time constraint will be described.
When the bottleneck portion of the production line is in front of the Q-time section, the arrival rate Λ increases or decreases, but there is no particular problem. The arrival rate Λ is always monitored (or predicted) and can be adjusted according to the arrival rate Λ.
When the bottleneck part of the production line is in the Q-time section and the service capacity Μ increases or decreases. This is no problem. Always monitor (or predict) cocoons and make adjustments to match them.
When the bottleneck part is behind the Q-time section, a request to increase or decrease the throughput occurs. The increase will be adjusted according to the review of the input plan. The reduction needs to be taken into account when it must be reduced. Although it is desirable to readjust after reviewing the input plan, it is necessary to adjust the real time in the section so as not to reduce productivity. In this case, it is necessary to adjust the number of kanbans and the buffer size of the CKB method by factors other than the arrival of the lot in the section and the fluctuation of the process processing capacity. Therefore, it is necessary to have a function of calculating the bottleneck portion of the entire production process and the production flow rate corresponding to the bottleneck portion by the MNA2 method.

(3) Procedural CKB (CKB-PS) method MNA1 method and MNA2 method are merged to simultaneously manage Q-time constraints based on dynamic state change monitoring / prediction for the entire manufacturing flow. The procedure (standard, with conditional branching (serial type, parallel type)) was uniformly defined as the “procedural CKB method”.
FIG. 54 is a flowchart (part 1) of the procedural CKB method (standard). The optimum index generation device 41 includes an HDD, a RAM, and a CPU that store a control program represented by a flowchart. The CPU reads the control program from a storage medium such as an HDD, a ROM, or a CDROM, and controls the device 41. I will do it.

First, in step S21, the control unit detects the structure of the Q-time constraint section based on the MNA1 method based on the information input from the input unit 12, and detects a sub-set of the Q-time constraint section. At this time, the counter value of the repetition loop is initialized (in terms of the number of sections).
Next, in step S22, the control unit detects the flow rate to each Q-time constraint section (detects the bottleneck portion (BN) of the entire production line and the flow rate of each section). Here, statically determined from the production plan, process capability, and the like. Dynamically, it is calculated using the MNA2 method.

Next, in step S23, the control unit detects the bottleneck portion (BN) of the Q-time constraint section and calculates the total flow rate.
Next, in step S24, the control unit calculates the CKB method for the Q-time constraint section.
Next, in step S25, the control unit determines whether or not the above processing has been executed for all sections. If the above processing has not been executed for all sections, the counter value is decremented (−1) with respect to the counter of the iteration loop, and the process returns to step S23 to perform steps for all Q-time constraint sections (by structure). S23 to S25 are executed.
When the above process is executed for all the sections, in step S26, management is executed based on the calculation result.
In dynamic management, changes are monitored, and readjustments are made by repeating S22 to S27.

A structure that requires an additional procedure in the procedural CKB method (standard) is managed by a procedure with conditional branch as follows. A structure that requires an additional procedure is a case where a plurality of Q-time constraints corresponding to serial (C) overlap mentioned in the MNA1 method, or a case where a single Q-time constraint section spans three or more steps (in a parallel structure) Is the case where the above structure is met when converted to the serial type.
The procedural CKB method (with conditional branching) will be described with reference to the flowchart (No. 2) shown in FIG.
First, in step S31, the control unit detects the structure of the Q-time constraint section based on the MNA1 method based on the information input from the input unit 12, and detects the Q-time constraint sub-set. At this time, the counter value of the iteration loop is initialized.
Next, in step S32, the control unit detects the flow rate to each Q-time constraint section (detects the bottleneck portion (BN) of the entire production line and the flow rate of each section). Statically, it is determined from the production plan, process capability and the like. On the other hand, dynamically, it is calculated using the MNA2 method.

Next, in step S33, the control unit sets the total throughput matched to the bottleneck location (BN) and the input rate for that purpose.
Next, in step S34, the control unit calculates the traffic intensity, the number of waiting WIPs, and the waiting time (expected value) of all the processes in the section from the set input rate.
Next, in step S35, the control unit determines where to allocate the Q-time constraint margin time.
Next, in step S36, the control unit determines the allocation of the Q-time constraint margin time based on the results of S34 to S35 and the allocation constraint / rule.
Next, in step S <b> 37, the control unit calculates the CKB method for each s-s section in the section (from the start of the first process to the start of the second process in two consecutive processing processes). .

Next, in step S38, the control unit determines whether or not the above processing has been executed for all sections. If the above process is not executed for all sections, the counter value is not zero (0), so the process returns to step S34, and steps S34 to S38 are executed for all Q-time constraint sections (by structure).
When the above process is executed for all the sections, in step S26, management is executed based on the calculation result. In dynamic management, the actual result is monitored, and S32 to S38 are repeated to readjust.
As described above, the entire production flow can be managed uniformly by the procedural CKB method that combines the MNA1 method and the MNA2 method, and further practical effects can be expected.

Here, the maximum flow problem-minimum cut theorem will be described. 56 and 57 are diagrams showing a parallel network flow.
In order to evaluate the bottleneck location of the production flow rate, it is necessary to find a location that determines the maximum flow rate on the production flow. However, as shown in FIGS. 56 and 57, when a parallel structure is included, It is not easy to specify the bottleneck portion (because the bottleneck portion is formed by a plurality of branches). Therefore, it is formulated as a maximum flow problem, and the maximum flow-minimum cut theorem is applied to a high-speed calculation algorithm to identify the bottleneck location and the flow rate of each branch. In the maximum flow-minimum cut theorem, the maximum value of the flow between the start point s and end point t of the network is the minimum value of the cut capacity (minimum cut) of the cut set separating the start point s and end point t as shown by the dotted line in FIG. ). In other words, it means that the maximum flow rate of the manufacturing network is limited to the minimum cut (= bottleneck).

  There is capacity for all cuts (combinations of branches) between the start point s and end point t of the network,

容量に基づいて、可能流が決まる。

The possible flow is determined based on the capacity.

Maximum−flow Minimum−cut Theorem（最大流-最小カットの定理）では、可能流の最も小さいカットを最小カットとする。

In Maximum-flow Minimum-cut Theorem, the cut with the smallest possible flow is the minimum cut.

Ford Fulkerson(1956)が発表したMaximum-flow minimum-cut Theoremとして、

As Maximum-flow minimum-cut Theorem announced by Ford Fulkerson (1956),

を用いても良い。
この最小カットの検出のためのアルゴリズムはラべリング法をはじめとして、いくつか存在し、検出の高速化に寄与している。

May be used.
There are several algorithms for detecting this minimum cut, including the labeling method, which contributes to speeding up the detection.

The fourth embodiment solves the third problem by the MNA method and the procedural CKB method.
A method (procedural CKB method) that simultaneously manages the Q-time constraint section of the entire manufacturing flow based on the structural analysis of multiple Q-time constraints (MNA1 method) and the analysis of the product flow of the entire manufacturing flow (MNA2 method) ) Was proposed. Here, the MNA method is an abbreviation for Manufacturing Network Analysis, and the generic name of the MNA1 method and the MNA2 method is the MNA method. In the fourth embodiment, these specifications, program flow, and the like are targeted.

Here, the MNA1 method is a method of structural analysis and simultaneous management of a plurality of Q-time constraints. The MNA2 method is an analysis method for detecting, monitoring, and predicting a dynamic bottleneck. The procedural CKB method is a procedural method that optimally and simultaneously manages all Q-time constraints in the production system using the MNA1 method, the MNA2 method, and the CKB method.
With these three methods, distributed management of each Q-time constraint section is realized according to the total production flow of the manufacturing flow, and both high-speed calculation and total optimization that can respond to production fluctuations in real time can be achieved. . The MNA1 method and the MNA2 method can be used alone.

Next, the effect of the linked logistics control method (procedural CKB method) is confirmed by a simulation experiment.
[Purpose and method]
The purpose of this simulation experiment is described in Non-Patent Document 1 in which intentional production logistics control is performed without considering an interference structure formed by a plurality of Q-time constraints in a production process having a plurality of Q-time constraints. By comparing the performance of the procedural CKB method that simultaneously manages in consideration of the interference structure formed by a plurality of Q-time constraints, in the case of the above method (hereinafter referred to as a conventional Q-time constraint section management rule), The purpose is to confirm the effectiveness of the procedural CKB method.
The experimental method includes the model and setting conditions of the production process of the actual factory described in Non-Patent Document 1 in which five device groups, seven processing steps, and five Q-time constraints shown in FIG. 58 are set. Perform the simulation experiment used. In this model, the same device group is shared for repeated processing. The traffic in the production process was set in the same manner as in Non-Patent Document 1, and a simulation experiment was conducted on the assumption that lots arrived regularly and arrived Poisson.

As a result of the simulation experiment, as shown in FIG. 59, the procedural CKB method has an interprocess waiting time of about 60 while improving the production rate as compared with the conventional Q-time constraint section management rule described in Non-Patent Document 1. It was confirmed to shorten by ~ 80%.
Further, from the performance evaluation result regarding the calculation time required for generating the optimum index, it is possible to reduce the calculation time which took 3 seconds in Non-Patent Document 1 to about 0.005 seconds and realize a calculation time of 1/600. It was confirmed. This performance evaluation is a result of an experiment using a computer (personal computer) having the same or lower performance.

<Fifth Embodiment>
[Problems and problems in the second embodiment]
The second embodiment has a problem in that the shift in maintenance time due to the setup time as shown in FIG. 64 and the influence thereof are not taken into consideration. Moreover, the setup time in semiconductor manufacturing is an important management target that accounts for a quarter of the production period, and there remains a problem of providing a method for reducing the setup time, which is the maximum non-production period. The fifth embodiment is a modification of the second embodiment, and solves problems and problems by an apparatus business management method that integrates a setup business minimization method with a maintenance management method. It has the feature of reducing the setup time of a manufacturing apparatus that occupies a quarter of the production period in the production process of semiconductor manufacturing and utilizing it in the opposite direction. In the present embodiment, a device business management method (hereinafter referred to as a JAMP-wS method) will be described.

With reference to FIG. 60, the optimal index generation apparatus which concerns on 5th Embodiment of this invention is demonstrated.
The optimum index generation device 51 according to the fifth embodiment of the present invention is provided in the production process # 1 for the operation method of the manufacturing engineering work such as the maintenance of the device and the setup for the process switching in the multi-product production. A production management system 10 that manages a schedule of engineering work related to the device group 5 is a control target. At the same time, the production management system 10 that manages the plan, instructions, and results of the equipment-related work of the equipment group 5 provided in the production process # 2 for the operation method for optimizing the timing of preventive maintenance and setup of the production equipment and engineering The business system 29 is a control target.

  As illustrated in FIG. 60, the optimal index generation device 51 includes an input unit 12 and a Q-time structure analysis unit 14, as with the optimal index generation device 11 (first embodiment) illustrated in FIG. Moreover, the input part 22 and the auxiliary | assistant input part 23 are provided similarly to the optimal parameter | index production apparatus 21 (2nd Embodiment) shown in FIG. Compared with the first embodiment and the second embodiment, as shown in FIG. 60, the characteristic configuration of the fifth embodiment includes an auxiliary input unit (JAMP-wS method) 52, a setup change cycle calculation unit 53, A device classification calculation unit 54, a JAMP-wS optimum logical value calculation unit 55, a sensitivity analysis unit (JAMP-wS characteristic) 58, a start / end time calculation unit 56, and a sensitivity analysis unit (performance characteristic) 57. .

Here, the above-described configuration characteristic of the fifth embodiment will be described.
Auxiliary input unit (JAMP-wS method) 52 is compatible with device types / specs, set-up time in multi-product / repetitive processing, set-up utilization method (select from 3 types, etc.), device-type spec allocation rules (3 types) Information such as device-type assignment (unique condition), multi-product work in progress, etc. is input.
The calculation result in the JAMP-wS method optimum logical value calculation unit 55 varies depending on the setting of the setup utilization method. The setup utilization method stipulates the inevitable setup time utilization method. For example, the purpose of increasing the maintenance time dispersion ratio in the JAMP method, the purpose of arranging the spare time corresponding to the lengthening of the maintenance time, and both For the purpose of fusing. The setup utilization rule selected by the user of the optimum index generation device 51 is input from the auxiliary input unit (JAMP-wS method) 52.

  The setup change cycle calculation unit 53 calculates a multi-product production order (hereinafter referred to as production cycle) and a production cycle in which the setup rate of the device group 5 is minimized while observing the Q-time restriction and the process delivery date of the production process # 1. To do. The setup change cycle calculation unit 53 performs the procedure processing shown in FIG. 62 ((0) device-variety assignment determination, (1) multi-product production cycle detection, (2) multi-product production cycle calculation, (3) minimum setup cycle calculation). I do. In (0), when specific device-type assignment information is input from the auxiliary input unit 52, the input information is used.

  The device classification calculation unit 54 calculates information for associating the calculation result of the JAMP-wS method optimum logical value calculation unit 55 with the device entity by using each input unit and the calculation result of the setup change cycle calculation unit 53 as inputs. . The apparatus category calculation unit 54 performs the procedure processing ((1) calculation of necessary number by apparatus category, (2) sorting in order of setup time, (3) determination of apparatus category and maintenance order) shown in FIG.

  The JAMP-wS method optimum logical value calculation unit 55 receives the calculation results of the input unit 22, the auxiliary input unit 23, the setup change cycle calculation unit 53, and the device classification calculation unit 54 as input of the maintenance time of the devices in the device group 5. The decentralization is performed, and the maintenance time of each device and the production capacity of the device group in each period are calculated so that the device group 5 always maintains a high processing capacity. For example, information such as maintenance start / end timing, multi-product production cycle, setup cycle, etc. of each device is calculated and supplied to the production management system 10 and the engineering business system 29.

The start / end time calculation unit 56 stores the calculation result of the JAMP-wS method optimum logical value calculation unit 55 in the internal storage unit. In addition, the start / end time calculation unit 56 collates the production information of the internal storage unit, the production management system 10 and the engineering business system 29, and the information on the equipment-related business results to produce the production instructions and the engineering business system 29 in the production management system 10. Outputs information for business instructions. A unique identification number or the like is given to each of these production instructions and business instructions, and can be used for collation between the production results and business results and the internal storage unit.
Unlike the start / end time calculation unit 26 of the second embodiment, the start / end time calculation unit 56 adds setup work information in addition to maintenance.

  The sensitivity analysis unit (performance characteristics) 57 inputs the input data used for the JAMP-wS method and the conditions of the JAMP-wS method optimum theoretical value, and produces the data under the conditions of the determined JAMP-wS method optimum theoretical value. Calculations for performance evaluation of the process are performed, and various performance evaluation indexes with respect to the processing capacity and production load (in-process arrival rate) of the apparatus group are output to the production management system 10 and the apparatus engineering business system 29 as performance characteristics.

  The sensitivity analysis unit (JAMP-wS characteristic) 58 is configured to calculate the JAMP-wS method theoretical value calculation result input from the JAMP-wS method optimum logical value calculation unit 55, the processing load (number of times), the setup cycle, and the timing of each device. The change characteristic of the result of JAMP method theoretical value calculation when it fluctuates is calculated. Further, the sensitivity analysis unit (JAMP-wS characteristic) 58 records the sensitivity analysis result in the internal storage unit 58a, and the JAMP method theoretical value calculation result, the processing load (number of times) of each device, and its processing time are determined. When it changes dynamically, it can be used as update data immediately. The information in the internal storage unit 58a is also used for determining the necessity of execution of the optimum index generation device 51. This determination of necessity is made in response to various changes in the production process # 1 to be managed (changes in the processing capacity of the device group, changes in the target production rate, changes in traffic, etc.). When the necessity is confirmed, a calculation instruction is issued to the optimum index generation device 51.

With reference to FIG. 60 and FIG. 61, the operation of the optimum index generation device 51 according to the fifth embodiment of the present invention will be described.
In step S <b> 51, information related to the Q-time constraint is input by the input unit 12 via the production management system 10. In step S <b> 52, information calculated through the Q-time structure analysis unit 14 and information regarding a plurality of devices included in the production execution system are input by the input unit 22 and the auxiliary input unit 23 via the production management system 10.
Furthermore, in step S53, the information regarding the setup change and setup utilization method of the apparatus is input from the auxiliary input unit 52, and the setup change cycle calculation unit 53 uses the information about the setup change cycle and the process delivery date while complying with the Q-time constraint and the process delivery date. The multi-product production cycle and the setup cycle that minimize the setup rate are calculated and input to the device category calculation unit 54.
Next, in step S54, the device classification calculation unit 54 classifies the plurality of devices into a regular device having a regular capability and a substitute device having a alternate capability. In step S55, the production load / operation for a plurality of regular devices and alternative devices (and regular capabilities and alternate capabilities) are performed by the JAMP-wS method optimum logical value calculation unit 55 from the device capability classification and the input information and the setup information of each device. Calculate the number of times and operation time.
In step S56, the preventive maintenance start / end timing, the setup cycle and the product production cycle of each device regarding the regular device and the substitute device are calculated by the start / end timing calculation unit 56 and stored in the internal storage unit.
In step S57, the optimum start / end time information of the product processing / setup / maintenance calculated by the start / end time calculation unit 56 is supplied to the production management system 10 and the engineering work system 29.
Each of these optimum start / end timing information of production and work is given a unique identification number, etc., and used for collation between the production results and work results managed by the production management system 10 and the engineering work system 29 and the internal storage unit. be able to.

The setup change cycle calculation unit 53 in FIG. 60 performs calculation according to the processing flow shown in FIG. The calculation procedure of the setup change cycle calculation unit 53 has four steps: (S61) device-product assignment determination, (S62) multi-product production cycle detection, (S63) multi-product production cycle calculation, and (S64) minimum setup cycle calculation. Composed.
First, in step S61, when specific assignment information is input from the auxiliary input unit 52 in the pre-processing for the setup change cycle calculation, the input information is used. When the specific assignment is not designated, the product-device assignment is calculated according to the device-product assignment rule input from the auxiliary input unit (JAMP-wS method) 52 and the device usable product / spec correspondence. In the device-product type allocation rule, a production risk that a production rate of a product type is significantly reduced or zero over a certain period and a reduction in setup time are taken into consideration.
Next, in step S62, in multi-product production cycle detection, (n-1)! Each production cycle is extracted, and the total setup time in one cycle in which each setup change is performed once is calculated for each cycle.
Next, in step S63, in the multi-product production cycle calculation, for each production cycle extracted in step S62, a production cycle that minimizes the setup is calculated while complying with the Q-time constraint of the multi-product lot. Finally, the minimum setup rate cycle is calculated from the multi-product production cycle calculation and the setup time in one cycle.
In step S64, the setup change cycle calculation unit 53 enables the setup to be minimized while complying with the stay time constraint such as the Q-time constraint in the manufacturing apparatus shared by three or more kinds of production lots. .

60 performs calculation according to the process flow shown in FIG.
The procedure of the device category calculation unit 54 is composed of three steps: (S71) Calculation of required number by device category → (S72) Sort order of setup → (S73) Device category and maintenance order determination. First, in (S71) necessary number calculation for each apparatus category, as in the second embodiment, the quantity for each category of the regular device (ordinary capability) and the alternate device (alternative capability) is calculated. (S72) In the setup order sort, the devices belonging to the device group 5 of the production process # 1 are rearranged in order of the setup rate calculated by the setup change cycle calculation unit 53. (S73) In apparatus classification and maintenance order determination, (S72) the apparatus classification and maintenance order of each apparatus are determined based on the result of the setup order sort.

  In the calculation procedure of the setup change cycle calculation unit 53 (S61), in the device-product type assignment determination, if no specific assignment is designated from the auxiliary input unit (JAMP-wS method) 52, the auxiliary input unit (JAMP-wS method) The type-device allocation is calculated according to the device-type allocation rule input from 52 and the device-usable type / spec correspondence. In the device-product type allocation rule, a production risk and a reduction in setup time are taken into consideration so that the production rate of a product type is significantly reduced or zero over a certain period. For example, there are three types: (A) a setup minimization rule, (C) a production risk minimization rule, and (B) a setup-production risk adjustment type rule. (A) The setup minimization rule calculates the required number of devices for each product type from the production load and service rate in order to minimize setup, assigns the integer units to dedicated devices as much as possible, and sets the remaining fractions Assign to the shared device of the product type. (C) The production risk minimization rule is a method in which the production load is evenly distributed to the usable devices of each product type in order to minimize the production risk. Will be assigned to process. (B) The setup-production risk adjustment type rule is a rule located in the middle of the (A) setup minimization rule and (C) production risk minimization rule. The allocation is performed by combining devices. The required number of devices is calculated from the production load of each product type and the average load per device, and a predetermined margin (for example, plus one) is added to the number of devices. Note that the function of (S61) device-type assignment determination can be used alone.

In (2) multi-product production cycle calculation in the calculation flow of the setup change cycle calculation unit 53 shown in FIG. 62, an in-process upper limit value that satisfies the process stay time constraint is calculated as shown in FIG. The production cycle as shown in FIG. 67 is calculated so as to satisfy the in-process upper limit value. First, an in-process upper limit value shown in FIG.
When the continuous production period Tp, the waiting time Tw, the product lot arrival rate λ ₀ , the product service rate μ, the production cycle Tc, and the basic allocation device number n, the average stay time WB in the buffer of the work in progress of a certain product is 38.

で表せる。ここで、バッファに蓄積されることになる仕掛品の最大量とこの時の待ち時間上限時間（平均）は、それぞれ、

It can be expressed as Here, the maximum amount of work in progress that will be stored in the buffer and the waiting time upper limit (average) at this time,

と表せる。

It can be expressed.

（ｉ≠ｊ）、は制約条件で、

(I ≠ j) is a constraint condition,

は、k種類の品種交換に必要とされる時間の総和である。

Is the total time required for the exchange of k types.

The maximization problem formulated by the following equations 43 to 47 is solved to obtain the maximum cycle Tc.
Formulation of optimization problem (for example, when the production order is A, B, and C with 3 types as shown in FIG. 66)

ここで、Ａ品種の処理後のアイドル時間、Ｂ品種の処理後のアイドル時間、Ｃ品種の処理後のアイドル時間（スラック）をそれぞれをｔ_Ａ、ｔ_Ｂ、ｔ_Ｃとすると、

Here, assuming that the idle time after processing of the A type, the idle time after processing of the B type, and the idle time (slack) after processing of the C type are t _A , t _B , and t _C , respectively.

改めて、

again,

であるから

Because

ここで、

here,

制約条件下で、Ｔｃの最大生産周期を算定する。
３品種以上の場合には、全２通りの品種サイクルに対する最大サイクルを求める。

Under the constraint conditions, the maximum production cycle of Tc is calculated.
In the case of three or more varieties, the maximum cycle for all two varieties cycles is obtained.

  Here, the difference in processing specifications that require setup is not limited to the product type. If setup occurs even in the case of repeated processing, a setup minimization cycle is determined for each spec (pair of product type, repeated processing stage, processing conditions, etc.). Here, in order to realize a smooth product distribution in the repeated processing lot and maximize the throughput, the service rate for the repeated processing lot may be distributed to be equal.

In (3) the minimum setup cycle calculator on the processing flow of the setup change cycle calculation unit 53, as shown in FIG. 67, the sum t ₀ of time required to k types of breeds replacement of each cycle, each kind Divide by the cycle Tc to calculate the setup rate. The production cycle and production cycle with the minimum setup rate are adopted.

The JAMP-wS method optimum logical value calculation unit 55 calculates the JAMP-wS method theoretical value based on information input from the input unit 22, the auxiliary input unit 23, the setup change cycle calculation unit 53, and the device category calculation unit 54. As a result, the processing load (number of times) of each device and its processing time, production cycle and setup time, maintenance time, etc. are calculated and output.
The sensitivity analysis unit (JAMP-wS characteristic) 58 is used when the processing load (number of times) input from the JAMP-wS method optimal logical value calculation unit 55 and its processing time, production cycle, setup time, and maintenance time are changed. The change characteristic of the result of JAMP-wS method theoretical value calculation is calculated. The sensitivity analysis unit (JAMP-wS characteristic) 58 records the sensitivity analysis result in the internal storage unit 58a, and the JAMP-wS method theoretical value calculation result, the processing load (number of times) of each device and its processing time, When the production cycle, setup time, or maintenance time changes dynamically, it can be used immediately as updated data. The information in the internal storage unit 58a is also used for determining the necessity of execution of the optimum index generation device 51. This determination of necessity is made in response to various changes in the production process # 1 to be managed (changes in the processing capacity of the device group, changes in the target production rate, changes in traffic, etc.). When the necessity is confirmed, a calculation instruction is issued to the optimum index generation device 51.

となる。ここで、

の場合、負になるため代替回数の軽減が可能になる。これは、即ち、段取り時間を逆手に活用することによって、保全時期の分散化の可能性を高められることを意味する。常用装置ｍ’台全体としては代替処理個数は、段取りを活用することにより、

回減らすことが可能となる。これを利用して、図６９に示す段取り活用法（１）を定めた。また、一方で、実際の工場では、保全時間が予定より長大化することが少なくない。ＪＡＭＰ法が提供する保全時期分散化方法は、ある保全が終了した後に次の保全が開始するまでの時間間隔をゼロから可能な範囲まで代替回数や段取り時間を用いて柔軟に調整できる。そこで、保全時間の長大化へ対応するために段取り時間を活用する図６９に示す段取り活用（２）を定めた。段取り活用法（３）は、段取り活用方法（１）と段取り活用法（２）を合体させたものである。 In the JAMP-wS method optimal logical value calculation unit 55, unlike the calculation in the JAMP optimal logical value calculation unit 24 of the second embodiment, the shift of the operation time length due to the setup time in multi-product production or from the regular device to the alternative device Consider changes in the number of alternative processes to be transferred. For example, when the regular device j (j = 1,..., M ′) is used, if the total setup time in the operation period of the device j is Ts (j), the continuous operation period of the device j is the setup time Ts ( j) shift backwards. Further, when the device number when the regular devices are arranged in the maintenance order is j, the necessary number of substitutions in the JAMP method calculation of the regular device j is

It becomes. here,

In the case of, since it becomes negative, the number of substitutions can be reduced. This means that the possibility of decentralizing maintenance time can be increased by utilizing the setup time in reverse. The number of alternative processing units as a whole for the normal equipment m 'can be adjusted by using the setup.

It is possible to reduce the number of times. Using this, the setup utilization method (1) shown in FIG. 69 was determined. On the other hand, in an actual factory, the maintenance time is often longer than planned. The maintenance time distribution method provided by the JAMP method can flexibly adjust the time interval from the completion of a certain maintenance to the start of the next maintenance from zero to a possible range using the number of substitutions and the setup time. Therefore, the setup utilization (2) shown in FIG. 69 that uses the setup time in order to cope with the lengthening of the maintenance time is determined. The setup utilization method (3) is a combination of the setup utilization method (1) and the setup utilization method (2).

  Unlike the second embodiment, the fifth embodiment takes into account the shift in maintenance time due to the setup time as shown in FIG. 64 and the influence thereof, and uses the inevitable setup time as shown in FIG. I will provide a. It has the feature of reducing the setup time, which accounts for a quarter of the production period in the production process, and using it in the opposite direction to suppress the operation time and production capacity of the device.

Next, the effect of the JAMP-wS method will be confirmed by numerical experiments.
[Purpose and method]
Using the actual production process model shown in FIG. 17 (Q-time constraint section), typical rules (first-come-first-served basis (FCFS)) and first-come-first-served basis with Q-time constraint section input control (hereinafter FCFS-CT) The performance of the conventional method such as the shortest processing time order with input control (hereinafter referred to as SPT-CT) and the setup reduction / utilization method of the present invention will be compared. The performance index includes the setup rate defined as the ratio of setup time per unit production period, the production rate that represents the ratio between the production lot input rate and the non-defective product output rate, and the total production rate and operating rate. The overall equipment efficiency (hereinafter referred to as OEE), which is a general index, was used. Total equipment efficiency is an important performance index for practical use, and is regarded as important in industries where the depreciation of equipment in the manufacturing cost is very high, such as semiconductor manufacturing. Here, a non-rework lot that complies with the Q-time constraint is regarded as a non-defective product. Also, the calculation time is compared with the prior art (Non-Patent Document 1, Non-Patent Document 2, etc.).

The results of the performance evaluation are shown in FIGS. FIG. 70 shows a comparison result of (a) setup rate and (ii) non-defective product production rate when the average traffic intensity is set to 0.8. From these results, it was confirmed that the setup reduction method of the present invention significantly reduced setup compared to the other methods and achieved the target production rate of all the varieties. In SPT-CT, the difference in production rate between varieties is large, and in FCFS-CT, the setup rate is high. In addition, the non-defective product rate is remarkably lowered in the FCFS in which the input control for the Q-time constraint section is not added.
As shown in FIG. 71, it was confirmed that the proposed method of the present invention is a method capable of improving the overall equipment efficiency (OEE).
Regarding the calculation time, the calculation that required 3 seconds in Non-Patent Document 1 was suppressed to within 0.01 seconds, that is, a calculation time of 1/300 was realized. This enables real-time optimal management of production operations and engineering operations for a complete production period including maintenance, etc. (In a real factory, each production instruction can be made within about 3 seconds for real-time management. Expected to be fast). The calculation order of the mixed integer programming applied to Non-Patent Document 1 and Non-Patent Document 2 follows exponential time, whereas the calculation order of the fifth embodiment of the present invention is a method that realizes linear time. Yes, high speed calculation is realized. For example, in the calculation of Non-Patent Document 1, it takes 3 seconds to calculate a production schedule for a production period of 10 hours for a part of the target production process (5 device groups and 7 processes), and the calculation time follows an exponential time. In principle, it is difficult to finish calculations over 6 days within 6 days. However, according to the method of the fifth embodiment of the present invention, it is possible to calculate a production work and an engineering work plan for a production period of several months within a few seconds. In this way, the ability to immediately perform recalculation to cope with dynamic changes that occur in reality is also realized.

The fifth embodiment is an application example of the second embodiment, and is an apparatus operation management method that integrates the setup operation minimization method with the maintenance management method. It solves the problem of optimizing the maintenance time by minimizing the setup time and considering the shift of the maintenance time due to the setup time. It has the feature of reducing the setup time of a manufacturing apparatus that occupies a quarter of the production period in the production process of semiconductor manufacturing and utilizing it in the opposite direction. Performance evaluation experiments have shown that compared to typical methods and conventional technologies, the setup time can be greatly reduced to increase the operating rate of production facilities and increase the yield of non-defective products for various types of products.
In particular, the photolithography apparatus group is expensive and has a limited number of units, which is a bottleneck that determines the production rate of the entire production process. Also, in the photolithography process, an integrated circuit manufacturing master called a mask (called a reticle in a factory) is shared by a plurality of apparatuses, and the setup time for the replacement affects the production rate. Costs such as reducing the number of appropriate reticles based on increasing the production rate of such bottlenecks, drastically improving the production rate of manufacturing equipment in the production process, and minimizing setup work in appropriate multi-product production It helps to solve the reduction problem.

  As a modification of the optimum index generation device according to the fifth embodiment of the present invention, the program shown in the flowcharts of FIGS. 61, 62, and 63 may be executed by a processor provided in the server. In this modification, a communication control unit is provided in the server, and the communication control unit communicates information with a production management system or an engineering business system via a communication line. As a result, the server can transmit the index information regarding the plurality of devices and products via the communication line to the production management system and the engineering business system that control the plurality of devices in order to produce a wide variety of products.

<Sixth Embodiment>
[Problems and problems in the third and fifth embodiments]
In the third embodiment, a method of integratedly managing production logistics / inventory management and a manufacturing facility maintenance work optimization method in a production process in which a stay time is restricted is shown. Further, in the fifth embodiment, the method for integratively optimizing the setup work and the maintenance work related to the manufacturing facility has been shown. However, the problem is that the integrated optimization method for the operations related to the manufacturing facilities and the optimal management method for production logistics and inventory are not integrated as a comprehensive management technology. On the other hand, responding to unscheduled shutdowns of manufacturing facilities due to process sophistication and automation is also an important issue. In addition to planned maintenance and setup, dynamic failures due to sudden failures and predictive maintenance are also included. There is a need for technology improvements that can take into account the decline in production capacity due to a significant equipment shutdown. Therefore, in the sixth embodiment, as a modified example of the third embodiment and the fifth embodiment, a manufacturing operation integrated management method in which a setup operation and a general maintenance management method are integrated with production logistics / inventory management is shown. How to solve the problem.
Based on these, the optimum management method of the dynamic device business / production logistics of the sixth embodiment is referred to as a CKB = JAMP-with Setup and Stop (hereinafter, CKB = JAMP-wS ² ) method.
In the sixth embodiment, the implementation timing of setup and maintenance work is optimized while complying with the Q-time restrictions of multi-product / repetition processing lots, and the improvement of the equipment operation rate, the improvement of the non-defective product and the improvement of the target production achievement rate are achieved at the same time. The purpose is to do.

  With reference to FIG. 72, the optimal index generation apparatus 61 which concerns on 6th Embodiment of this invention is demonstrated. The optimum index generation device 61 according to the sixth embodiment of the present invention is, for example, the production of the processing step (device group) 2 and the processing step (device group) 3 provided in the production step # 1 for producing a semiconductor. The production management system 10 that manages the schedule is the target of control. At the same time, the device group 5 provided in the production process includes an operation method for optimizing the timing of preventive maintenance and setup of manufacturing equipment, and a method for appropriately managing post-maintenance and predictive maintenance based on failure monitoring and prediction. The production management system 10 and the engineering business system 29 that manage plans, instructions, and results of the apparatus-related business, and the engineer / operator who performs the business of the production process # 1 are controlled.

As shown in FIG. 72, the optimum index generation device 61 is similar to the optimum index generation device 31 (third embodiment) shown in FIG. 48, the input unit 12, the auxiliary input unit 13, the fluctuation monitoring detection unit 32, the input change. A detection unit 33 and a calculation necessity determination unit 34 are provided. Furthermore, the optimal index generation device 61 includes a Q-time structure analysis unit 43.
Moreover, the optimal index generation device 61 is the same as the optimal index generation device 51 (5th Embodiment) shown in FIG. 60, the input part 22, the auxiliary input part (JAMP method) 23, and the auxiliary input part (JAMP-wS method). 52, the setup change cycle calculation unit 53, the device classification calculation unit 54, the JAMP-wS method optimal logical value calculation unit 55, the sensitivity analysis unit (JAMP-wS characteristic) 58, and the like. 72, the input unit 12 and the auxiliary input unit 13 are combined into an input module (CKB method) 12A, and the input unit 22, the auxiliary input unit (JAMP method) 23, and the auxiliary input unit (JAMP-wS method) 52 are combined. The input module 22A is collectively shown.
Compared with the third and fifth embodiments, the characteristic configuration of the sixth embodiment is that, as shown in FIG. 72, an auxiliary input unit (CKB = JAMP-wS ² method) 62, device group production cycle phase setting Unit 63, production spike calculation unit 64, CKB = JAMP-wS ² optimum theoretical value calculation unit 65, index calculation unit 66, sensitivity analysis unit (CKB = JAMP-wS ² performance characteristics) 67, sensitivity analysis unit (CKB = JAMP- (wS ² characteristic) 68.

The main features of the sixth embodiment will be described in comparison with the third embodiment and the fifth embodiment.
In the fifth embodiment, when determining the preventive maintenance time, the setup time is minimized while observing the Q-time restriction of the multi-product production lot processed by each device, and the inevitable setup time is used in reverse. In this way, we showed how to promote decentralization of preventive maintenance periods and make production risk recommendations. With the optimum index generation device 51 of the fifth embodiment, production spikes as shown in FIG. 75 can be reduced as much as possible, and the production capacity can always be kept high. Here, by using the production load distribution method of each device in consideration of the multi-product production setup and preventive maintenance calculated in the fifth embodiment, the first embodiment, the third embodiment, and the fourth embodiment are used. A method for improving the functions and performances of the optimal index generation device 11, the optimal index generation device 31, and the optimal index generation device 41 shown in the sixth embodiment will be described. In the sixth embodiment, a method for managing a production process to which a stay time between processes such as a Q-time restriction section is added in consideration of maintenance of manufacturing equipment and setup of multi-product / repetitive production will be described.
On the other hand, in the optimal index generation device 31 (third embodiment) and the like, the optimal index and the replacement condition for production logistics / inventory management in the Q-time constraint section according to the bottleneck of the production process to be managed are generated. did. In the sixth embodiment, based on the information on the production capacity and production load of the device group at each time point associated with maintenance and setup generated by the optimum index generation device 51 of the fifth embodiment and the production load distribution to each device, (1) Anticipate / analyze changes in bottlenecks and production capacity (production spike calculation unit 64), specifically indicate the cause and analysis method of production spikes shown in the third embodiment and FIG. An application example of production logistics / inventory management in a -time constraint section is materialized (CKB = JAMP-wS ² optimal theoretical value calculation unit 65). (2) At this time, by introducing a phase into the multi-product production cycle of each device calculated by the optimum index generation device 51 (fifth embodiment), to reduce the production risk of the production process and increase the production rate, The production spike is deformed (device group production cycle phase setting unit 63).
(3) In addition to the preventive maintenance, setup, and production load distribution taken up in the fifth embodiment, predictive maintenance based on monitoring and prediction of manufacturing equipment abnormalities, and dynamic changes in production spikes due to sudden failures Add a description of how to deal with it.

With reference to FIG. 72 and FIG. 73, the operation of the optimum index generation device 61 according to the sixth embodiment of the present invention will be described.
First, main processing of the optimum index generation device 51 of the fifth embodiment is executed. Specifically, in step S81, the Q-time structure analysis unit 43 creates a Q-time structure based on information input from the production management system 10 or the engineering business system 29 via the input unit 12 or the auxiliary input unit 13. analyse.
In step S82, the result of the Q-time structure analysis and input from the production management system 10 and the engineering work system 29 via the input unit 22, the auxiliary input unit (JAMP method) 23, and the auxiliary input unit (JAMP-wS method) 52 are input. Based on the information thus obtained, calculations of the setup change cycle calculation unit 53, the device classification calculation unit 54, and the JAMP-wS method optimum logical value calculation unit 55 are executed.
In step S83, the calculation result of the JAMP-wS optimal logic value calculation unit 55 and the calculation result of the sensitivity analysis unit (JAMP-wS characteristic) 58 are stored in the internal storage 58a.
In step S84, the number of operation times and operation timings of a plurality of regular devices and alternative devices (and normal capability and alternative capability) calculated from the JAMP-wS method optimum logical value calculation unit 55, and preventive maintenance on the regular devices and alternative devices are started. The end time, the setup cycle of each device and the product production cycle are input via the start / end time calculation unit 56, and the production spike calculation unit 64 calculates the production spike and stores it in the internal storage 64a.
In step S85, the production spike calculation unit 64 uses the information input from the auxiliary input unit (CKB = JAMP-wS ² method) 62 and the internal storage 64a to store the internal memory of the sensitivity analysis unit (JAMP-wS characteristic) 58. In comparison with 58a, a dynamic production spike or the like is calculated and stored in the internal storage 64a.
In step S86, the production spike can be transformed using the calculation result of the device group production cycle phase setting unit 63 for the calculation by the production spike calculation unit 64.
In steps S87 and S88, CKB = JAMP-wS ² optimal theoretical value calculation unit based on information input from each input unit, Q-time structure analysis unit 43, optimal load rule determination unit 19 and production spike calculation unit 64 In 65, an optimum index for production logistics / inventory management in the Q-time constraint section of the target production process is generated, and the index is given to the production management system 10, the engineering business system 29, or the engineer or operator of the production process # 1. While outputting via the calculation part 66, it memorize | stores in the internal memory 66a.
Further, in step S89, the index calculation unit 66 collates the internal storage unit 66a with the production record of the production management system 10 and the engineering work system 29 and the information related to the equipment related work record, Information for business instructions in the engineering business system 29 is output. A unique identification number or the like is assigned to each of these production instructions and business instructions, and can be used for collation between the production results and business results and the internal storage unit 66a.
The index calculation unit 66 is the same as the calculation unit of the third embodiment and 56 of the fifth embodiment. However, the index calculation unit 66 includes comprehensive information including information on maintenance and setup work and information on production logistics and inventory management. Different production / operation instruction information is output.

Here, the above-described configuration characteristic of the sixth embodiment will be described.
In the auxiliary input unit 62, the location and timing of occurrence of dynamic failure or predictive maintenance, the scheduled maintenance end timing, the target production load (input) plan change, and production are performed via the production management system 10 and the engineering business system 29. Enter information such as the number of in-process quantities for the multi-product lot in the process. Also, information related to the production cycle phase setting rule used by the production cycle phase setting unit 63 is input.

In the production spike calculation unit 64, for the device group of the production process # 1, a change in the production capacity of each device group using the optimum index generation device 31 of the third embodiment in FIG. The information regarding the fluctuation of the production capacity or the production load is inputted, and the size of the production spike as shown in FIG. 74 is analyzed. The production spike calculation unit 64 stores the calculated production spike and information related to the production risk caused by the production spike in the internal storage 64a.
The production spike indicates a difference in production capacity (equipment group and process throughput) with respect to the production target rate and the actual production input rate. As production risks, two types of risks are analyzed: production risk relating to production rate (and inventory) and production risk relating to time (and inventory) as shown in FIG.

There are two types of production spikes calculated by the production spike calculation unit 64: static production spikes and dynamic production spikes. For static production spikes, production capacity spikes (time transition of maximum production rate) of equipment groups before, during and after the maintenance period of each equipment group based on the JAMP-wS method, and production capacity of equipment groups are produced in a variety of products. The multi-product production load distribution spike assigned to the load (time transition of the production rate of each product based on the lot production cycle) is used.
A production spike when a production spike in the storage unit 64a is updated upon receiving information on dynamic failure or predictive maintenance from the auxiliary input unit 62 is called a dynamic production spike.
Production spike calculation units include devices, device groups, multiple device groups, processing steps, multiple processing steps, and the like.

The production spike is expressed as a difference in production capacity with respect to the target production rate, as shown in FIG. The magnitude of the production spike is expressed by the area of the spike determined by the depth of the spike in the decline period and the length of the recovery period. The degree of influence of production spikes is determined by the shape and size of the spikes.
A single spike may have several different decline and recovery periods. Such a production spike occurs, for example, when the number of maintenance target devices increases in stages, and then the equipment that has completed maintenance restarts the manufacturing process in stages, as shown in FIG. It is expressed in a composite spike in which multiple steps of spikes are superimposed.

Using the production spike diagram as shown in FIG. 74, the following procedure is used for analysis, and the production spike and production risk are quantitatively analyzed as shown in FIG.
(1) Input the calculation result of each device group based on the JAMP-wS method. (2) Depth and length of spike k (decrease period and recovery period) generated by preventive maintenance, failure, predictive maintenance, etc. in each device group j. Calculate the size.
(A) Decreasing period end point and target production rate line determined by the depth (μ ₁ (j, k) −μ ₂ (j, k)) and length T _r (j, k) of the decreasing period Time difference of production process capacity rate etc.):

（Ｂ）復旧期の長さ：

(B) Length of recovery period:

（Ｃ）生産スパイクの大きさ（＝スパイクの面積）：

(C) Production spike size (= spike area):

この面積分が在庫ロスや生産期間ロスを引き起こす。これは、理想の生産能力や生産目標に対する生産実績の差を表し、各種生産目標割れの可能性（生産リスク）を定量化したものである。
（Ｄ）生産スパイクを用いた生産リスクの分析：
図７４や図７５に示すような生産スパイク図を用いて、生産リスクを以下（あ）と（い）の観点で定量的に分析する。
（あ）生産率（と在庫）に関する生産リスク：生産スパイクの面積
生産スパイクに与える目標生産率を、工程区間や生産工程全体のボトルネックにおける生産率とした場合、工程区間や生産工程全体の生産率を更に低下させる生産リスクを定量的に分析することができる。当該リスクの大きさＳ_Ａ（ｊ，ｋ）は、数式５０で示した式で算定する。

（い）時間（と在庫）に関する生産リスク：限界線と実績の差分
対象とする装置群ｊ、工程ｋにおける滞在時間の制約ｔ_ｄｄ（ｊ，ｋ）とｔ_ｄ（ｊ，ｋ）を比較してｔ_ｄｄ（ｊ，ｋ）＜ｔ_ｄ（ｊ，ｋ）である場合、滞在時間の制約を超過する生産リスクが発生する。この生産リスクの大きさを面積の差分Ｓ_Ｂ（ｊ，ｋ）として計算する。

（３）（２）で算定された各装置群ｊで最大スパイクや生産リスクの大きな順に並べる。
（４）（２）で算定された各装置群ｊのスパイク（時刻ｔに依存）や生産リスクを重ね合わせ、図７８に示すような生産工程を通した複合生産スパイクや複合生産リスクを算定する（Ｑ−ｔｉｍｅ制約区間、ボトルネック−ボトルネック区間、生産工程全体、など）。
なお、最適指標生成装置２１（第２実施形態）の性能評価事例で図４６に示したように、ＪＡＭＰ法を用いて保全時期を分散化（すなわち生産能力の平準化）を図ることによって生産スパイクが発生しない場合もある。このように、生産スパイク図は、最適指標生成装置の性能評価指標としても用いることができる。

This area causes inventory loss and production period loss. This represents the difference in production performance with respect to the ideal production capacity and production target, and quantifies the possibility of production target breakage (production risk).
(D) Analysis of production risk using production spikes:
Production risk is quantitatively analyzed from the viewpoints of (a) and (ii) below using production spike diagrams as shown in FIGS.
(A) Production risk related to production rate (and inventory): If the target production rate given to the production spike area production spike is the production rate at the bottleneck of the process section or the entire production process, the production of the process section or the entire production process The production risk that further reduces the rate can be analyzed quantitatively. The magnitude of risk S _A (j, k) is calculated by the equation shown in Equation 50.

(Ii) Production risk related to time (and inventory): device group j that is the target of difference between limit line and actual result, comparison of stay time constraints t _dd (j, k) and t _d (j, k) in process k If t _dd (j, k) <t _d (j, k), a production risk that exceeds the stay time constraint occurs. The magnitude of this production risk is calculated as an area difference S _B (j, k).

(3) Arrange in the descending order of maximum spike and production risk in each device group j calculated in (2).
(4) The spikes (depending on time t) and production risk of each device group j calculated in (2) are overlapped to calculate the complex production spike and complex production risk through the production process as shown in FIG. (Q-time constraint section, bottleneck-bottleneck section, entire production process, etc.).
Incidentally, as shown in FIG. 46 in the performance evaluation example of the optimum index generation device 21 (second embodiment), the production spike is obtained by distributing the maintenance time (that is, leveling the production capacity) using the JAMP method. May not occur. Thus, the production spike diagram can also be used as a performance evaluation index of the optimal index generation device.

The device group production cycle phase setting unit 63 performs a production spike in multi-product production with respect to the multi-product lot production cycle of each manufacturing device calculated by the setup change cycle calculating unit 53 in the optimum index generating device 61 of the sixth embodiment.・ Calculate the phase to reduce production risk. By using the device group production cycle phase setting unit 63, production spikes and production risks can be reduced.
For example, there are three devices (624, 625, 627) that are shared by the three types of products p1, p2, and p3 in the device group (processing step) to which device-type assignment is made as shown in FIG. In these apparatuses, the optimum index generation apparatus 51 (fifth embodiment) is used to calculate the production cycle that minimizes the setup while complying with the stay time constraint such as the Q-time constraint. For example, assume that the production order of p1, p2, and p3 is calculated as optimal. Here, since all three units have the same setting conditions, the optimal production order is also the same order of p1, p2, and p3, and the sustainable production time of each product type is Tp (p1), Tp (p2), Tp (p3 ), When the same production cycle is used in all apparatuses under the same production cycle start point setting, the first Tp (p1) time is the time when all three machines process the product p1 and the other product p2. The in-process lot of p3 waits for a time of Tp (p1). Then, in the processing steps of the varieties p2 and p3 during this Tp (p1) period, the production capacity is reduced, and the possibility of production spikes and production risks is increased. In order to avoid such a situation and level the production capacity to be allocated to all product lots, the device group production cycle phase setting unit 63 is used to calculate an appropriate phase. For the setting of this phase, the production cycle phase setting rule input by the auxiliary input unit 62 is used.

CKB = JAMP-wS ² Optimal theoretical value calculator 65 calculates the processing order and setup cycle of multi-product production, the processing load (number of times) and the processing in each device calculated by the optimal index generation device 51 (fifth embodiment). The timing, the phase set by the device group production cycle phase setting unit 63, and the production spike calculated by the production spike calculation unit 64 based on the time transition of the processing capacity in the device group for a predetermined cycle are input. CKB method optimal theoretical value calculation is performed each time the processing capacity or production load (or production target) of the target production process is switched within a predetermined cycle. Here, the predetermined cycle includes, for example, a period (one cycle) until all the regular devices complete at least once and all the alternative devices complete the preventive maintenance once, a limit cycle of the maintenance time distribution of the device group ( The period necessary for optimizing the maintenance schedule) or a production period specified separately.
Furthermore, in the CKB = JAMP-wS ² method optimal theoretical value calculation, as in the optimal index generation device 31 (third embodiment), the input amount in the Q-time constraint section in advance according to the change in the processing capability of the device group. Is provided as a value that can comply with the Q-time constraint, and smooth production distribution / inventory management in the Q-time section is realized.
Here, the difference from the CKB = JAMP-wS method optimal theoretical value calculation of the fifth embodiment is that the phase of the production cycle is appropriately adjusted based on the analysis of the production spike and the production risk, and the multi-product / repetitive production leveling is performed. In addition to preventive maintenance and setup, in addition to preventive maintenance and set-up, it is necessary to dynamically determine the timing of the introduction and production of various types of products, as well as dynamic changes associated with changes in production targets, etc. This is in consideration of changes in processing capacity and production load. This dynamic change is input to the device group production cycle phase setting unit 63, the production spike calculation unit 64, and the like via the auxiliary input unit (CKB = JAMP-wS ² method) 62.

Index calculating unit 66, the calculation result of the production spike calculation unit 63, and stores the calculation results of CKB = JAMP-wS ² optimum theoretical value calculating section 65 in the internal storage unit 66a. The index calculation unit 66 collates the internal storage unit 66a with the production management system 10, the engineering work system 29, and the information on the production results and device-related work results of the engineer / operator to produce production instructions and engineering in the production management system 10. Information for business instructions in the business system 29 is output. A unique identification number or the like is assigned to each of these production instructions and business instructions, and can be used for collation between the production results and business results and the internal storage unit 66a.

Sensitivity analyzer (CKB = JAMP-wS ² performance characteristic) 67 inputs the input data used for CKB = JAMP-wS ² method, production spike calculation result, CKB = JAMP-wS ² method optimal theoretical value calculation result, Under the conditions of the determined CKB = JAMP-wS ² method optimal theoretical value, calculations for performance evaluation of the production process are performed, and various performance characteristics and production loads (in-process arrival) of the equipment group are calculated as performance characteristics. The performance evaluation index is output to the production management system 10 and the equipment engineering business system, and stored in the internal storage unit 67a.

The sensitivity analysis unit (CKB = JAMP-wS ² characteristic) 68 is input from the CKB = JAMP-wS ² optimum theoretical value calculation unit 65. The theoretical value calculation result of the CKB = JAMP-wS ² method, the production capacity of each device group And change characteristics when the processing load and processing time fluctuate. In addition, the sensitivity analysis unit (CKB = JAMP-wS ² characteristic) 68 records the sensitivity analysis result in the internal storage unit 68a, and calculates the theoretical value calculation result of the CKB = JAMP-wS ² method. When the production capacity, processing load and processing time fluctuate, it can be used immediately as update data.

The sixth embodiment is a modification of the third embodiment and the fifth embodiment, and integrates operations related to manufacturing facilities by a manufacturing operation integrated management method in which a setup operation and a maintenance management method are integrated with production logistics / inventory management. It solves the problem of comprehensively managing the optimization method and the optimal management method for production logistics and inventory.
In addition, in addition to scheduled maintenance and setup, the sixth embodiment takes into account the dynamic production capacity decline due to sudden failures and predictive maintenance, and the production of equipment groups and production processes based on production spikes. Equipped with an evaluation index for quantitative analysis of risk. Provide a method for optimal management of production logistics and inventory based on analysis of production spikes and production risks caused by equipment shutdown.

  Note that as a modification of the optimum index generation device according to the sixth embodiment of the present invention, the program shown in the flowchart in FIG. 73 may be executed by a processor provided in the server. In this modification, a communication control unit is provided in the server, and the communication control unit communicates information with a production management system or an engineering business system via a communication line. As a result, the server can transmit the index information regarding the plurality of devices and products via the communication line to the production management system and the engineering business system that control the plurality of devices in order to produce a wide variety of products.

<Seventh embodiment>
[Problems and problems in the fourth and sixth embodiments]
In the fourth embodiment, the information processing procedure related to the structural analysis of the Q-time constraint section and the production capacity allocation method for converting the parallel structure into the serial structure are not clearly described. In particular, when multiple processes share the same device group and Q-time constraint times in each process are different, management that simultaneously reduces setup time, adheres to Q-time constraints, and maximizes the production rate Defining the method was an issue. Further, in the sixth embodiment, problems such as construction of a method for shortening the production period of the multi-variety lot in the entire target production process remain.
In the seventh embodiment, as a modification of the fourth embodiment and the sixth embodiment, static production that occurs due to maintenance and setup time calculated by the JAMP-wS method for all device groups in the production process to be managed. Analyze composite spikes that combine dynamic production spikes combined with dynamic failures, etc., for multiple production processes, absorb production spikes in the target production process with the minimum safety stock, and overall production process The purpose is to reduce the inventory and the production period while preventing the decline in the production rate at the resource bottleneck that determines the production rate.

  With reference to FIG. 79, the optimal index generation apparatus 71 which concerns on 7th Embodiment of this invention is demonstrated. The optimum index generation device 71 according to the seventh embodiment of the present invention controls, for example, the production management system 10 that manages the production schedule of the processing step (device group) of the production step # 1 for producing a semiconductor. Yes. At the same time, the operation method of optimizing the timing of preventive maintenance and setup of manufacturing equipment, and the method of appropriately managing post-maintenance and predictive maintenance based on failure monitoring and prediction, The production management system 10 and the engineering business system 29 that manage the plan / instruction / result of the apparatus-related business, and the engineer / operator who performs the business of the production process # 1 are controlled.

As shown in FIG. 79, the optimal index generation device 71 is similar to the optimal index generation device 61 (sixth embodiment) shown in FIG. 72, the input unit 12, the auxiliary input unit 13, the input unit 22, and the auxiliary input unit 23. , Fluctuation monitoring detection unit 32, input change detection unit 33, auxiliary input unit (JAMP-wS method) 52, auxiliary input unit (CKB = JAMP-wS ² method) 62, Q-time structure analysis unit 43, setup change cycle calculation Unit 53, device classification calculation unit 54, JAMP-wS method optimum logical value calculation unit 55, sensitivity analysis unit (JAMP-wS characteristic) 58, device group production cycle phase setting unit 63, production spike calculation unit 64, etc. Yes. All of these are referred to as an optimal index generation device 61 calculation module. 79, as in FIG. 72, the input unit 12 and the auxiliary input unit 13 are combined into the input module (CKB method) 12A, the input unit 22, the auxiliary input unit (JAMP method) 23, and the auxiliary input unit (JAMP). -WS method) 52 is collectively shown as an input module 22A.
Further, the optimum index generation device 71 is the same as the optimum index generation device 41 (fourth embodiment) shown in FIG. 50, the optimum load rule determination unit 19, the production flow bottleneck / flow rate analysis unit 42, the auxiliary input unit (procedure Type CKB method) 45, drive control unit 46, margin time distribution analysis unit 47, and procedure execution control unit (CKB + PS method) 48. All of these are referred to as an optimal index generation device 41 calculation module. The drive control unit 46 includes a fluctuation monitoring detection unit 32, an input change detection unit 33, and a calculation necessity determination unit 34.
Compared with the fourth and sixth embodiments, as shown in FIG. 79, the characteristic configuration of the seventh embodiment is an optimal index generation device 61 calculation module, an optimal index generation device 41 calculation module, and an optimal index calculation unit. 75, an index calculation unit 76, a sensitivity analysis unit (procedural CKB = JAMP-wS ² method) 77, and a sensitivity analysis unit (procedural CKB = JAMP-wS ² performance) 78.

With reference to FIG. 80, the operation of the optimum index generation device 71 according to the seventh embodiment of the present invention will be described.
First, in the optimal index generation device 61 calculation module, characteristic processing of the optimal index generation device 61 (sixth embodiment) is executed.
Specifically, in step S91, the Q-time structure is analyzed by the Q-time structure analysis unit 43 based on information input from the production management system 10 or the engineering business system 29 via the input unit 12 or the auxiliary input unit 13. analyse.
Next, in step S92, the results of the Q-time structure analysis and the production management system 10 and the engineering business system 29 via the input unit 22, the auxiliary input unit (JAMP method) 23, and the auxiliary input unit (JAMP-wS method) 52 are used. Based on the input information, the setup change cycle calculation unit 53, the device classification calculation unit 54, and the JAMP-wS method optimum logical value calculation unit 55 are calculated. Here, the calculation result of the JAMP-wS method optimum logical value calculation unit 55 and the calculation result of the sensitivity analysis unit (JAMP-wS characteristic) 58 are stored in the internal storage 58a.
Next, in step S93, the number of operation times and operation times of the plurality of normal devices and alternative devices (and normal capability and alternative capability) calculated from the JAMP-wS method optimum logical value calculation unit 55, prevention related to the normal device and the alternative device. The maintenance start / end time, the setup cycle of each device and the product production cycle are input via the start / end time calculation unit 56, and the production spike calculation unit 64 calculates the production spike and various production risks, and the internal storage 64a. To remember.
Next, in step S94, the production spike calculation unit 64 uses a sensitivity analysis unit (JAMP-wS characteristic) 58 based on information input from the auxiliary input unit (CKB = JAMP-wS ² method) 62 and the internal storage 64a. A dynamic production spike or the like is calculated in comparison with the internal storage 58a and stored in the internal storage 64a.
In step S95, the production spike can be deformed by using the calculation result of the device group production cycle phase setting unit 63 for the calculation by the production spike calculation unit 64.

Next, a characteristic process of the optimum index generation device 41 (fourth embodiment) is executed in the optimum index generation device 41 calculation module.
Specifically, in step S96, information on a plurality of devices included in the production execution system is input via the production management system by the input unit 12, the auxiliary input unit 13, and the auxiliary input unit (procedural CKB method) 45, Further, the bottleneck location and the maximum flow rate in the production process by the plurality of devices analyzed by the production flow bottleneck / flow rate analysis unit 42 from the information input from the production spike calculation unit 64 and the analysis by the Q-time structure analysis unit 43 The result is input, and an instruction to calculate the optimum management index for all the Q-time intervals is issued using the procedure execution control unit (CKB + PS method) 48.
Next, in step S97, in the calculation of the optimal management index, information from the optimal load rule determination unit 19, the Q-time structure analysis unit 43, and the production spike calculation unit 64 is input, and the optimal index calculation unit 75 An optimal index for production logistics / inventory management in all Q-time constraint sections of the production process is generated, and the index calculation unit 76 is sent to the production management system 10, the engineering business system 29, and the engineer and operator of the production process # 1. And stored in the internal storage 76a.
In addition, the index calculation unit 76 collates the internal storage unit 76a with information on production results and device-related business results such as the production management system 10 and the engineering business system 29, and the production instruction in the production management system 10 and the engineering business system 29. Outputs information for business instructions. A unique identification number or the like is assigned to each of these production instructions and business instructions, and can be used for collation between the production results and business results and the internal storage unit 76a.

The index calculation unit 76 is the same as the calculation unit 66 of the sixth embodiment and the calculation unit and start / end time calculation unit 26 of the fourth embodiment, but engineering work such as maintenance and setup for the entire production process. And general production / business instruction related information including information on production logistics and inventory management is output.
The sensitivity analysis unit (procedural CKB = JAMP-wS ² method) 77 and the sensitivity analysis unit (procedural CKB = JAMP-wS ^{2 method} ) 78 are the sensitivity analysis unit (CKB = JAMP-wS ² method) 67 and the sensitivity analysis unit. (CKB = JAMP-wS ² performance) It has the same function as 68, but the sensitivity analysis result for the entire production process is stored in the internal storage unit 77a.

Here, the above-described configuration characteristic of the seventh embodiment will be described.
The optimal index generation device 61 calculation module is connected to the input unit 22, the auxiliary input unit (JAMP method) 23, the auxiliary input unit (JAMP method) 45, and the auxiliary input unit (JAMP-wS) via the production management system 10 and the engineering business system 29. Method) The information input from 52 and the result of the Q-time structure analysis unit 43 based on the information input from the input unit 12 are input, the setup change cycle calculation unit 53, the device category calculation unit 54, JAMP The calculation of the wS method optimum logical value calculation unit 55 is executed. The calculation result of the JAMP-wS optimal logic value calculation unit 55 and the calculation result of the sensitivity analysis unit (JAMP-wS characteristic) 58 are stored in the internal storage 58a. Number of operation times and operation times of a plurality of regular devices and alternative devices (and normal capability and alternative capability) calculated from the JAMP-wS method optimum logical value calculation unit 55, preventive maintenance start / end timings for regular devices and alternative devices, and The setup cycle and product production cycle of each device are input via the start / end time calculation unit 56, and the production spike calculation unit 64 calculates production spikes and various production risks, and stores them in the internal storage 64a. The production spike calculation unit 64 collates with the internal storage 58a of the sensitivity analysis unit (JAMP-wS characteristic) 58 based on the information input from the auxiliary input unit (CKB = JAMP-wS ² method) 62 and the internal storage 64a. Then, a dynamic production spike or the like is calculated and stored in the internal storage 64a. Further, the production spike can be deformed by using the calculation result of the device group production cycle phase setting unit 63 for the calculation in the production spike calculation unit 64.

  The optimal index generation device 41 calculation module executes characteristic processing of the optimal index generation device 41 (fourth embodiment). Specifically, information regarding a plurality of devices included in the production execution system is input via the production management system by the input unit 12, the auxiliary input unit 13, and the auxiliary input unit (procedural CKB method) 45, and further, production spike calculation is performed. The bottleneck location and the maximum flow rate in the production process by a plurality of devices analyzed by the production flow bottleneck / flow rate analysis unit 42 from the information input from the unit 64 and the analysis result by the Q-time structure analysis unit 43 are input. The procedure execution control unit (CKB + PS method) 48 is used to issue an instruction to calculate the optimum management index for all the Q-time intervals.

  The optimal index calculation unit 75 inputs information from the optimal load rule determination unit 19 and the margin time distribution analysis unit 47, and the optimal index calculation unit 75 performs production in all Q-time constraint sections of the target production process. An optimum index is generated for logistics / inventory management and related operations of the device group 5. The optimal index for each Q-time constraint section is the same as the index calculated by the optimal index generation device 61 (sixth embodiment).

  The calculation result of the optimum index calculation unit 75 is output to the production management system 10 and the engineering work system 29 (or the engineer or operator of the production process # 1) via the index calculation unit 76 and is also stored in the internal storage 76a. .

  In the seventh embodiment, information processing of the Q-time structure analysis unit 43 will be described with reference to FIG. First, the manufacturing process flow for each product, the start point and end point of the Q-time constraint, the time length of the Q-time constraint, and the like are input from the input unit 12 and the auxiliary input unit 13. Here, the start point and the end point are designated at the start or end of processing of a certain process step. In general, the start point is often set to the process end point of a certain process step 2, and the end point is often set to the process start point of a subsequent process step 3. After appropriate pre-processing as shown in FIG. 51 is performed on the information on these Q-time constraints, the device group (or processing step) and end point of the starting point are determined from the information on the Q-time constraints as shown in FIG. It is expressed by a connection matrix of the device groups (or processing steps). From the conversion using the processing step sequential device group Q-time matrix, a Q-time constraint over three or more steps is detected. In addition, the device group Q-time matrix in which the information of the Q-time constraint is written in the cell with the pair of the device group of the start point and the end point, and the number of positive value cells in the row or column having the same device group is counted By checking whether the count value is 2 or more, it is possible to detect a branch structure having the same Q-time constraint start point and a merge structure having the same Q-time constraint end point.

In the seventh embodiment, a production capacity distribution method for converting a production process having a parallel structure to a serial structure to speed up the calculation will be described.
A parallel structure of production processes is generated by a multi-product lot or a repeated processing lot sharing the same device group. The parallel structure includes merging and branching as shown in the fourth embodiment. This is because in-process lots with different attributes merge for processing using the same device (same device group), or conversely, in-process lots processed by the same device (same device group) This corresponds to branching to a different device (device group) for different processing. In general, differences in the manufacturing process flow of multi-product production lots appear as merging and branching in the production process. Furthermore, in the manufacture of electronic devices (semiconductors, flood panel displays, etc.), processing lots are merged and branched to repeat the process of forming each layer to form a three-dimensional structure.

Converting the parallel structure to the serial structure means that the production capacity in the shared manufacturing apparatus is allocated to the multi-product / repetitive production lot. The distribution of the optimum production capacity for the multi-product production lot is indicated by the optimum index generation device 61 (sixth embodiment). The seventh embodiment further shows a method for optimizing the allocation of production capacity to repeated processing lots having different Q-time constraints for the purpose of maximizing the production rate. In the case of the purpose of maximizing the production rate in the allocation of the production capacity of the apparatus group to the repeated processing lot of the same product type, the production capacity is evenly distributed to the repeated processing lot by the proof as shown in FIG. In other words, it has been shown that it is optimal to make the production rate (also called throughput) of repeated processing lots equal. However, this equality is an average value for a certain period. Here, this production capacity distribution rule is referred to as REFS (Re-entrant flow Smoothing). While maximizing the production rate, it is also necessary to achieve multipurpose purposes that reduce setup time while complying with restrictions on stay time between processes (Q-time restrictions, delivery date, etc.). Here, (1) the distribution of the production capacity is equalized, and (2) the switching period of the shared device that reduces the setup time while complying with the restriction of the stay time between the processes as shown in FIG. By defining the above, optimal production logistics and inventory management that can achieve multiple purposes will be realized.
An optimal method for allocating production capacity to a plurality of Q-time constraint sections is realized using the cases (i) and (ii) based on the following assumptions (A) and (B). As described above, the case where a plurality of processes share the same device group and the -time constraint time in each process is different is defined.

  In the seventh embodiment, based on the production spike and risk calculation in the optimum index generation device 61 (sixth embodiment) shown in FIG. 72, an appropriate inventory location and quantity are determined to shorten the production period. The method is shown. First, the importance of shortening the production period will be described. In a production process, it is important to maintain a high production capacity of a resource bottleneck (for example, a photolithography apparatus in semiconductor manufacturing) that determines the production rate of the entire production process when the number is expensive and limited. However, as a matter of fact, the production capacity of the equipment group is reduced due to equipment stoppage due to maintenance or failure of equipment in resource bottlenecks and non-resource bottlenecks, and fluctuations in production logistics (particularly equipment Production rate of production bottleneck is caused by causing production rate to drop to the bottleneck (target or maximum) production rate and production spikes formed by the subsequent recovery period. May decrease the rate. In addition, a Q-time constraint section exists before and after the resource bottleneck, and the production distribution control in the Q-time constraint section may affect the production rate of the resource bottleneck. In order to absorb such changes in the production capacity of devices and the effects of product distribution control in the Q-time constrained section, semiconductor manufacturing processes generally involve a large amount of intermediate inventory (actually, (Stored in the stocker in the middle and the buffer on the transport route). Due to the influence of the intermediate inventory, the processing waiting time between processes accounts for about 40 to 65% of the production period in semiconductor manufacturing, which causes a long production period (generally 1.5 to 2.5 months). As a result, the time competitiveness is not improved, and it is also a cause of product death.

  In the seventh embodiment, in order to solve the problem of the production period, two types of methods are shown: an arrival-linked centralized inventory type method shown in FIG. 83 and a stepwise fluctuation absorption distributed inventory type method shown in FIG. By using these methods as the management unit (called process management module) between resource bottlenecks, between dynamic bottlenecks and between Q-time constraint sections, the arrangement and quantity of the minimum safety stock amount are determined.

I. Centralized inventory method linked to arrival The production logistics rate spike caused by equipment stoppage (during preventive maintenance) that occurred in the production process is put together in a process section that is not subject to time constraints such as Q-time restrictions. This refers to a method in which inventory is removed and the distribution fluctuation due to the stop of the apparatus is not propagated to the subsequent resource bottleneck process (lithography process) or the like. Here, it is essentially a problem that a factory has “a large amount of inventory” based on the same concept. Therefore, it is possible to optimize (minimize) the inventory for removing spikes. It becomes a challenge. In the arrival-linked centralized inventory method, safety stock is placed in a process section with low production risk where there is little fluctuation and no Q-time restrictions etc. are set, and lot removal (production logistics rate) of the process in which capacity decline has occurred Regardless of changes, control is performed so that the production rate is constant (leveled) after the process of placing safety stock. In this method, in the Q-time constraint section, in principle, the procedural CKB = JAMP-wS ² method is used, and each Q is linked to the production capacity (production rate) in the decline period and the restoration period caused by equipment shutdown. -The CKB method optimal management value of the time constraint is recalculated and reset. If no extra inventory is placed between each process, the system operates in conjunction with the arrival of the lot unless the process utilization rate exceeds 100%. As shown in FIG. 83, WIP _{L r} to remove spikes generated by the apparatus concentrated down time _{t d} is the _{L r} = _{λ 0} t _r. Here, t _r = t _d is set in order to remove the process sections without restrictions.
The advantage of the arrival-linked centralized inventory method is that the same service level can be achieved with a smaller inventory by inventory aggregation. It is in the point which absorbs the influence of the apparatus stop in a plurality of front processes at hand. The shortcomings assumed for the arrival-linked centralized inventory type method cannot be handled when the intermediate inventory of a process that requires a certain amount of inventory, such as a batch processing process, is not used, or when all sections have Q-time constraints. A point is mentioned.

(Case) Production process model of Non-Patent Document 1 shown in FIG.
In photolithography ~ "DRY1-WET (1) -Fur ( 1) -DRY2-WET (2) -N 2 -Fur (2) " - photolithography interval, example where preventive maintenance of Fur (1) takes place Add a description to.
Fur (1) has a utilization rate of about 80% in a steady state.
Here, in Fur (1), it is assumed that an apparatus stop occurs that exceeds the utilization rate of Fur (1) 100%. As described above, this is a case where a certain number or more of them stop at the same time due to maintenance, failure, or the like (maintenance times are dispersed so that such a decrease does not occur, but in the event that it occurs).
If it is known in advance that such a fall period will occur, as shown in the third embodiment, the Q-time constraint section relating to Fur (1): DRY1-WET ( 1) By recalculating and replacing the optimum in-process management value of -Fur (1), the Q-time constraint is prevented from being exceeded and the target production rate is achieved (third embodiment).
Here, the lot exit rate of the final process: Fur (1) in the Q-time constraint section is the same as that at the start of the process, and changes greatly between the stationary period, the decreasing period, and the recovery period. The subsequent non-Q-time constraint section: Fur (1) -Dry (2) is quickly removed so that the production spike of the product distribution rate is not propagated to the next resource bottleneck device group (photolithography process). . Therefore, the in-process lot (intermediate stock) of Dry (2) is stored to the extent that the largest production spike is absorbed.
Such intermediate stock need not always be saved. If the implementation time, such as preventive maintenance, is determined in advance, it is only necessary to collect it according to that time (for example, in Dry (2), an accidental failure occurs, so the amount that could not be processed at that time There is also an opportunity to accumulate.)

  As the semiconductor processing dimensions become finer (sophisticated), the Q-time constraint section in the production process is steadily increasing. If it is limited not to correspond to this, there is a possibility that this spike reduction will be limited. In addition, the process of providing intermediate inventory is suitable for processes that require a certain amount of inventory from the beginning, such as a batch type, or a device group in which devices are unlikely to accidentally fail. This is not necessarily a Q-time restricted section. Not necessarily. Speaking of the example of the previous process model, WET and FUR are typical examples of devices that are batch-type and have few accidental failures. Therefore, consider the use of the following stepwise variable absorption inventory method.

となる。
段階的変動吸収分散在庫型方式の利点は、プロセスの高度化の影響で増加の一途にあるＱ−ｔｉｍｅ制約が点在する生産工程でも、各工程の中間在庫を活用して可能な限り生産スパイクを段階的に低減し、資源的ボトルネック工程へのスパイクの伝搬を早い段階で防ぐ可能性を高める点にある。また、バッチ工程でのローディングを生産工程全体に合致させるよう適正化する。段階的変動吸収分散在庫型方式の想定される欠点は、バッチ工程でローディングサイズを小さくすることを許容すると、同一期間長におけるローディング回数を増やすことになり、熱処理炉（Ｆｕｒ）などの装置群では予防保全周期が短くなることがあり、注意を払う必要がある。 II. Stepwise fluctuation absorption distributed inventory type method The difference from the arrival-linked centralized inventory type method is that as much as possible by utilizing the inventory of process sections that require a certain amount of inventory, such as batch processing, regardless of whether there is a Q-time constraint. The point is to reduce production spikes. In order to reduce the deviation from the target production rate of the resource bottleneck process, the load size in the batch process is appropriately changed.
The intermediate stock L _ri for removing spikes generated by the apparatus concentration stop time t _d as shown in FIG. 84 is L _ri = λ ₀ t _ri .
Here, since the constraint is added, _{tr 1} = T _q1 when the constraint time T _qi <t _r1 .
And the time length _{tr (i + 1)} of the successor constraint section is

It becomes.
The advantage of the gradual absorption and absorption distributed inventory method is that even in the production process dotted with the increasing Q-time constraint due to the process sophistication, the production spike is possible as much as possible by utilizing the intermediate inventory of each process. Is to increase the possibility of preventing the propagation of spikes to the resource bottleneck process at an early stage. Also, optimize the batch process loading to match the entire production process. The assumed disadvantage of the step-variation absorption distributed inventory method is that if the loading size is allowed to be reduced in the batch process, the number of loadings in the same period length is increased. The preventive maintenance cycle may be shortened and care must be taken.

The operation of the stepwise fluctuation absorption distributed inventory type method will be described using examples.
Since the production process models DRY1 and DRY2 in FIG. 58 are devices using an unstable plasma system process, an accidental failure occurs. It should be noted that since both DRY1 and DRY2 are the top steps of the Q-time constraint section, there is no problem in complying with the Q-time constraint. A lot (wafer group) forced to stop processing due to a failure is lost.
As a general operating condition, if the Fur operation rate is 80% (normal operation, no stop) and the average production rate of the lithography process, the drop period occurs in DRY1 when 10 or more units are down at the same time It is. (In total, DRY1 has 16 devices and DRY2 has 16 other devices)
In the previous step j (for example, DRY1), due to the influence of the stop of the device due to the failure, the period of decline (production rate μ_ (d, j) <initial target production rate (average) λ) is reached. Even after the decline period begins, the same production is maintained until the intermediate inventory (in-process lot) is below a certain level (zero unless otherwise specified), while maintaining the Q-time constraint after the next process (WET (1)). Continue processing at the rate (initial target production rate (average) λ).
After the intermediate stock falls below a certain level (zero unless otherwise specified), processing is started in accordance with the arrival lot (until the end of the decline period = the start of the recovery period).
If the production rate μ_ (r, j) of DRY1 in the recovery period (= lot arrival rate to WET (1)) exceeds the average processing capacity of WET (1), the lot in process at WET (1) The number will increase. When the number of devices in progress reaches the maximum number of lots to comply with the Q-time constraint, the upper limit management of the CKB method starts to function again (change point in FIG. 84).
In this case, the average arrival rate used for the CKB method optimum value calculation is the production rate μ_ (r, j) (> λ) in the recovery period of the previous step (Dry1). Since the period td of the decline period and the initial target production rate (average) λ are known, the length Y of the restoration period can be estimated. Y = td (λ−μ_ (d, j)) / μ_ (r, j). Even in case of accidental failure, it can be estimated at the end of recovery.
In this way, in the subsequent process of stopping the apparatus, the Q-time restriction is maintained even in the Q-time restriction section, and the spikes in production logistics are gradually reduced.

  In the seventh embodiment, as a modification of the fourth embodiment and the sixth embodiment, static production that occurs due to maintenance and setup time calculated by the JAMP-wS method for all device groups in the production process to be managed. Analyze composite spikes that combine dynamic production spikes combined with dynamic failures, etc., for multiple production processes, absorb production spikes in the target production process with the minimum safety stock, and overall production process The method of reducing inventory and shortening the production period was shown while preventing the decline in the production rate in the resource bottleneck that determines the production rate.

<Eighth Embodiment>
With reference to FIG. 85, the optimal index production | generation method which concerns on 8th Embodiment of this invention is demonstrated. The present embodiment is characterized in that the processor executes a program shown in the following flowchart for the operation of each unit constituting the optimum index generation device according to the first embodiment. Each step indicates the operation content of the software module, and the detailed operation content is described in the first embodiment, and thus the description thereof is omitted.
When the optimum index generation device is powered on, the CPU boots and executes the OS read from the auxiliary storage unit on the RAM. Then, the CPU executes the following optimum index generation application program as an example of application software read from the auxiliary storage unit to the RAM on the OS.

The present embodiment is characterized in that index information relating to the plurality of devices and products is supplied to a production management system that controls a plurality of devices in order to produce a wide variety of products.
First, in step S301, information regarding a plurality of devices and products is input via the production management system.
Next, in step S302, a Q-time structure indicating a limit value of stay time between processing steps by a plurality of apparatuses is analyzed from the input information and stored in the internal storage unit .
Next, in step S303, the boundary value of the arrival rate at which the magnitude relationship between the batch set waiting time related to the arrival of the product and the upper limit value of the Q-time constraint is switched is calculated, and an appropriate load rule is determined.
Next, in step S304, based on the input information and the Q-time structure, the optimum number of kanbans and buffer size for each product type are calculated.
Next, in step S305, the calculated information is output as index information. Next, in step S306, the calculated optimum index information and load rule are supplied to the production management system.

Referring to the flowchart shown in FIG. 86, in step S311, sensitivity analysis is performed in advance on the optimal index information calculated in step S304 and stored in the internal storage unit. In step S312, the optimum index information corresponding to the variation of various input information is immediately supplied to the production management system.
Referring to the flowchart shown in FIG. 87, in step S315, the number of kanbans and the buffer size are changed in conjunction with the change in throughput in the subsequent process section or the preceding process section.
Referring to the flowchart shown in FIG. 88, in step S317, the size of the buffer provided between the machining steps by a plurality of devices is limited so as not to exceed the maximum number of lots that can be waited between the machining steps. Thus, the output of the product in the preceding process is controlled.

Referring to the flowchart shown in FIG. 89, in step S318, control is performed so that the traffic intensity at the bottleneck in the Q-time constraint section does not exceed 1. In step S319, if the traffic intensity is greater than 1, λ is the input rate when the traffic intensity (ρ) is 1.
Referring to the flowchart shown in FIG. 90, in step S320, the number of lots staying in the Q-time constraint section is limited by the number of kanbans according to product types and processing conditions, and the processing start time of each lot is set to an empty kanban. Control by.

Next, with reference to FIG. 91, a modification example of the optimal index generation method according to the eighth embodiment of the present invention will be described on the assumption that a program provided in a server is executed by a processor provided in a server.
In this modification, a server that transmits index information about a plurality of devices and products via a communication line is executed for a production management system that controls a plurality of devices to produce a wide variety of products. It is characterized by doing. The server is provided with a communication control unit, and the communication control unit communicates information with the production management system via a communication line.

First, in step S321, information about a plurality of devices and products is received via the production management system. Next, in step S322, the Q-time structure indicating the limit value of the stay time between the machining steps by the plurality of apparatuses is analyzed from the received information and stored in the internal storage unit.
Next, in step S323, the calculated information is output as index information.
Next, in step S324, the boundary value of the arrival rate at which the magnitude relationship between the batch set waiting time related to the arrival of the product and the upper limit value of the Q-time constraint is switched is calculated, and an appropriate load rule is determined.
Next, in step S325, based on the received information and the Q-time structure, the optimum number of kanbans and buffer size for each product type are calculated. Next, in step S326, the calculated optimal index information and load rule are transmitted to the production management system.

  As described above, the objective is achieved through the construction and performance evaluation of the CKB method, which has high qualitative evaluation such as calculation speed, accuracy, and practicality as a production scheduling method that keeps Q-time constraints and does not deteriorate other performance indicators. We were able to.

<Ninth Embodiment>
With reference to FIG. 92, the optimal index production | generation method which concerns on 9th Embodiment of this invention is demonstrated. Note that the present embodiment is characterized in that the processor executes the program shown in the following flowchart for the operation of each unit constituting the optimum index generation device according to the fourth embodiment. Each step indicates the operation content of the software module, and the detailed operation content is described in the first embodiment, and thus the description thereof is omitted.
When the optimum index generation device is powered on, the CPU boots and executes the OS read from the auxiliary storage unit on the RAM. Then, the CPU executes the following optimum index generation application program as an example of application software read from the auxiliary storage unit to the RAM on the OS.

In this embodiment, index information relating to a plurality of devices is supplied to a production management system that controls a plurality of devices provided in a production execution system for producing a variety of products. Features.
Next, in step S331, information regarding a plurality of devices and products included in the production execution system is input via the production management system.
Next, in step S332, the bottleneck location and the maximum flow rate in the machining process by a plurality of apparatuses are analyzed from the input information.
Next, in step S333, the structure of the Q-time constraint section where one or more Q-time constraints exist is analyzed from the input information and stored in the internal storage unit.
Next, in step S334, using the information stored in the internal storage unit in S333, it is determined whether or not the optimal index calculation has been completed for all the Q-time constraint sections. When the optimal index calculation is completed, the process proceeds to step S339. If the optimal index calculation has not been completed, the process proceeds to step S335.

Next, in step S335, the boundary value of the arrival rate at which the magnitude relationship between the batch set waiting time related to the arrival of the product and the upper limit value of the Q-time constraint is switched is calculated, and an appropriate load rule is determined.
Next, in step S336, the plurality of devices are classified into a regular device having a regular capability and an alternate device having a alternate capability.
Next, in step S337, the bottleneck location and the maximum flow rate in the Q-time constraint section are analyzed.
Next, in step S338, based on the bottleneck location, the maximum flow rate, and the Q-time structure, the optimum number of kanbans and buffer sizes for each product type, the production load and the preventive maintenance start timing for the regular device and the alternative device are calculated. And it memorize | stores in an internal memory part as parameter | index information, and it progresses to step S334.
Next, in step S339, the optimal index information stored in the internal storage unit in S338 is supplied to the production management system and the engineering business system, and the optimal index information calculated in the time index calculation step is supplied to the production management system. Then, the process ends.

Referring to the flowchart shown in FIG. 93, in step S341, all the Q-time constraints scattered in the entire manufacturing flow included in the production execution system are managed simultaneously.
Further, referring to the flowchart shown in FIG. 94, in step S343, in order to simultaneously manage the Q-time constraints corresponding to the fluctuations in the production flow rate and process capability in the entire production flow included in the production execution system, Analyze the bottleneck and the maximum flow rate of each process.

Next, with reference to FIG. 95, a modified example of the optimum index generation method according to the ninth embodiment of the present invention will be described as having a processor provided in a server execute a program shown in the following flowchart.
Note that, in this modification, for a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products, an index related to the plurality of devices via a communication line. A server that transmits information is executed. The server is provided with a communication control unit, and the communication control unit communicates information with the production management system via a communication line.

Next, in step S351, information regarding a plurality of devices and products included in the production execution system is received via the production management system.
Next, in step S352, the bottleneck location and the maximum flow rate in the machining process by the plurality of apparatuses are analyzed from the received information.
Next, in step S353, the structure of the Q-time constraint section where one or more Q-time constraints exist is analyzed from the received information and stored in the internal storage unit.
Next, in step S354, using the information stored in the internal storage unit in S353, it is determined whether or not the optimal index calculation has been completed for all the Q-time constraint sections. When the optimal index calculation is completed, the process proceeds to step S359. If the optimal index calculation has not been completed, the process proceeds to step S355.

Next, in step S355, the boundary value of the arrival rate at which the magnitude relationship between the batch set waiting time related to the arrival of the product and the upper limit value of the Q-time constraint switches is calculated, and an appropriate load rule is determined.
Next, in step S356, the plurality of devices are classified into a regular device having a regular capability and an alternate device having a alternate capability.
Next, in step S357, the bottleneck location and the maximum flow rate in the Q-time constraint section are analyzed.
Next, in step S358, based on the bottleneck location, the maximum flow rate, and the Q-time structure, the optimum number of kanbans and buffer size for each product type are calculated and stored in the internal storage unit as index information, and the process proceeds to step S354. .
Next, in step S359, the optimal index information stored in the internal storage unit in S358 is transmitted to the production management system and the engineering business system, and the optimal index information calculated in the time index calculation step is transmitted to the production management system. Then, the process ends.

As a result, even in situations where Q-time constraint cracking, which has not been achieved by the conventional method, is likely to occur, the Q-time constraint can be reliably observed, and good product throughput improvement, cost reduction, and environmental load reduction can be achieved. Possible optimal operation management conditions can be provided.
Also, by providing analysis methods and appropriate management methods for the main causes of Q-time constraint cracking, handling of multi-product production, preventive maintenance of equipment, and mutual interference of multiple Q-time constraints. , While complying with the Q-time constraint, it is possible to simultaneously improve the throughput of non-defective products and reduce costs and environmental loads.

<Tenth Embodiment>
With reference to FIG. 96, the optimal index production | generation method which concerns on 10th Embodiment of this invention is demonstrated. The present embodiment is characterized in that the processor executes a program shown in the following flowchart for the operation of each unit constituting the optimum index generation device according to the second embodiment. Each step indicates the operation content of the software module, and the detailed operation content is described in the first embodiment, and thus the description thereof is omitted.
When the optimum index generation device is powered on, the CPU boots and executes the OS read from the auxiliary storage unit on the RAM. The CPU executes the following optimum index generation application program as an example of application software read from the auxiliary storage unit to the RAM on the OS.

In the present embodiment, index information relating to the plurality of devices is supplied to a production management system that controls a plurality of devices for producing a wide variety of products.
First, in step S361, information regarding a plurality of devices is input via the production management system. Next, in step S362, the plurality of devices are classified into a regular device having a regular capability and a substitute device having a alternate capability. Next, in step S363, the number of operations (production load) and operation timing of a plurality of regular devices and alternative devices are calculated from the calculated device classification and the input information.
Next, in step S364, the preventive maintenance start / end timing for the regular device and the alternative device is calculated from the calculation result. Next, in step S365, the calculated optimum start / end time information is supplied to the production management system and the engineering business system.

  Referring to the flowchart shown in FIG. 97, in step S371, sensitivity analysis is performed in advance on the optimal start / end timing information calculated in the optimal theoretical value calculation step, and the result is stored in the internal storage unit. Next, in step S372, optimal index information corresponding to fluctuations in various input information is immediately supplied to the production management system.

Next, with reference to FIG. 98, a modified example of the optimal index generation method according to the tenth embodiment of the present invention will be described as causing the processor provided in the server to execute the program shown in the following flowchart.
In this modification, a server that transmits index information about a plurality of devices via a communication line is executed for a production management system that controls a plurality of devices for producing a wide variety of products. It is characterized by that. The server is provided with a communication control unit, and the communication control unit communicates information with the production management system via a communication line.
First, in step S381, information about a plurality of devices is received via the production management system. Next, in step S382, the plurality of devices are classified into a regular device having a regular capability and an alternate device having a alternate capability. Next, in step S383, the number of operations (production load) and operation timing of a plurality of regular devices and alternative devices are calculated from the calculated device classification and the received information.
Next, in step S384, the preventive maintenance start / end timing for the regular device and the alternative device is calculated from the calculation result. Next, in step S385, the calculated optimum start / end time information is transmitted to the production management system and the engineering work system.

<Eleventh embodiment>
With reference to FIG. 99, the optimal index production | generation method which concerns on 11th Embodiment of this invention is demonstrated. The present embodiment is characterized in that the processor executes the program shown in the following flowchart for the operation of each unit constituting the optimum index generation apparatus according to the third embodiment. Each step indicates the operation content of the software module, and the detailed operation content is described in the first embodiment, and thus the description thereof is omitted.
When the optimum index generation device is powered on, the CPU boots and executes the OS read from the auxiliary storage unit on the RAM. Then, the CPU executes the following optimum index generation application program as an example of application software read from the auxiliary storage unit to the RAM on the OS.

In the present embodiment, index information relating to the plurality of devices is supplied to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. It is characterized by.
In step S391, information about a plurality of devices and products included in the production execution system is input via the production management system.
Next, in step S392, a Q-time constraint indicating a limit value of stay time between machining steps by a plurality of apparatuses is analyzed from the input information and stored in the internal storage unit.
Next, in step S393, information related to a plurality of devices included in the production execution system is input from the production management system. Next, in step S394, the plurality of devices are classified into a regular device having a regular capability and an alternate device having a alternate capability.

Next, in step S395, the operation count and operation timing of a plurality of regular devices and alternative devices are calculated from the calculated device classification and the input information. Next, in step S396, the preventive maintenance start timing for the regular device and the alternative device is calculated.
Next, in step S397, based on the input information, the Q-time structure, and the preventive maintenance start time, the optimum number of kanbans and buffer size for each product type are calculated and output as index information.
Next, in step S398, the calculated optimum start / end time information is supplied to the production management system and the engineering business system, and the optimum index information is obtained for each period in which the process capability determined by the calculated start / end time is different. Supply to production management system.

Referring to the flowchart shown in FIG. 100, in step S401, sensitivity analysis is performed in advance for the calculated optimum index information and start / end time, and the result is stored in the internal storage unit.
Next, in step S402, optimum index information and start / end timing information corresponding to fluctuations in various input information are immediately supplied to the production management system.

Next, with reference to FIG. 101, a modified example of the optimal index generation method according to the eleventh embodiment of the present invention will be described as causing a processor provided in a server to execute a program shown in the following flowchart.
Note that, in this modification, for a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products, an index related to the plurality of devices via a communication line. A server that transmits information is executed. The server is provided with a communication control unit, and the communication control unit communicates information with the production management system via a communication line.
First, in step S411, information regarding a plurality of devices included in the production execution system is received via the production management system. Next, in step S412, the Q-time constraint indicating the limit value of the stay time between the machining steps by the plurality of apparatuses is analyzed from the received information and stored in the internal storage unit.

Next, in step S413, information regarding a plurality of devices included in the production execution system is received from the production management system. Next, in step S414, the plurality of devices are classified into a regular device having a regular capability and a substitute device having a alternate capability. Next, in step S415, the number of operations and the operation timing of a plurality of regular devices and alternative devices are calculated from the calculated device classification and the received information.
Next, in step S416 , the preventive maintenance start timing for the regular device and the alternative device is calculated.
Next, in step S417, based on the received information, the Q-time structure and the preventive maintenance start timing, the optimum number of kanbans and buffer size for each product type are calculated and output as index information.
Next, in step S418, the calculated optimum start / end time information is transmitted to the production management system and the engineering business system, and the optimum index information is obtained for each period having different process capabilities determined by the calculated start / end time. Send to production management system.

  The present invention can be used for production simulation, schedule planning (production scheduling), production planning, facility personnel planning, load calculation, etc., and in addition to the semiconductor and liquid crystal industries, it can also be applied to information communication, wholesale, medical, etc. it can.

2 processing steps (device group), 3 processing steps (device group), 4 buffer size, 10 production management system, 11 optimal index generation device, 12 input unit, 12A input module, 13 auxiliary input unit, 14 Q-time structure analysis , 15 CKB optimal logical value calculation unit, 16 sensitivity analysis unit (CKB characteristic), 17 index calculation unit, 18 sensitivity analysis unit (performance characteristic), 19 optimal load rule determination unit, 21 optimal index generation device, 22 input unit, 22A input module, 23 auxiliary input unit, 24 JAMP optimum logical value calculation unit, 25 sensitivity analysis unit (JAMP characteristic), 26 start / end time calculation unit, 27 sensitivity analysis unit (performance characteristic), 28 device classification calculation unit, 29 engineering Business system, 31 Optimal index generation device, 32 Fluctuation monitoring detection unit, 33 Input change detection unit, 34 Calculation necessity determination unit, 35 Sensitivity Analyzing unit (CKB + JAMP characteristics), 36 CKB + JAMP optimal theoretical value calculation unit, 37 sensitivity analysis unit (CKB + JAMP performance characteristics), 38 calculation instruction unit, 41 optimal index generation apparatus, 42 production flow bottleneck flow analyzer, 43 Q-time Structure analysis unit, 44 device classification calculation unit, 45 auxiliary input unit, 46 drive control unit, 47 margin time distribution analysis unit, 51 optimum index generation device, 52 auxiliary input unit (JAMP-wS method), 53 stage change cycle calculation unit , 54 device classification calculation unit, 55 JAMP-wS method optimum logical value calculation unit, 56 start / end time calculation unit, 57 sensitivity analysis unit (performance characteristics), 58 sensitivity analysis unit (JAMP- wS characteristic), 61 optimum index generation device 62 Auxiliary input unit (CKB = JAMP-wS ² method), 63 Device group production cycle phase setting unit, 64 Production spike calculation unit 65 CKB = JAMP-wS ² optimal theoretical value calculation unit, 66 index calculation unit, 67 sensitivity analysis unit (CKB = JAMP-wS ² performance characteristic), 68 sensitivity analysis unit (CKB = JAMP-wS ² characteristic), 71 optimal Index generating device , 76 Index calculating unit

Claims (33)

An optimum index generation device that supplies index information about a plurality of devices and products to a production management system that controls a plurality of devices to produce a wide variety of products,
An information input unit for inputting information on the plurality of devices and products via the production management system;
A Q-time structure analysis unit for analyzing a Q-time structure indicating a limit value of a stay time between processing steps by the plurality of devices from the input information;
An optimal theoretical value calculator for calculating the optimum number of kanbans and buffer size for each product type based on the input information and the Q-time structure;
An index calculation unit that outputs the calculated information as index information;
An optimal load rule determination unit that calculates a boundary value of an arrival rate at which the magnitude relationship between the batch assembly waiting time related to product arrival and the upper limit value of the Q-time constraint is switched, and determines an appropriate load rule;
An optimal index generation apparatus, characterized in that the optimal index information and load rules calculated by the index calculation unit are supplied to the production management system. A sensitivity analysis unit that performs a sensitivity analysis in advance for the optimal index information calculated by the optimal theoretical value calculation unit,
The index calculation unit immediately supplies optimal index information corresponding to fluctuations in various input information to the production management system, and changes the number of kanbans and the buffer size in conjunction with changes in the throughput of the subsequent process section or the preceding process section. The optimal index generation device according to claim 1, wherein: An optimum index generation method for supplying index information about a plurality of devices and products to a production management system that controls a plurality of devices to produce a wide variety of products,
An information input step of inputting information on the plurality of devices and products via the production management system;
A Q-time structure analysis step of analyzing a Q-time structure indicating a limit value of a stay time between processing steps by the plurality of devices from the input information;
An optimal theoretical value calculation step of calculating an optimal number of kanbans and a buffer size for each product type based on the input information and the Q-time structure;
An index calculating step of outputting the calculated information as index information;
An optimal load rule determination step that calculates a boundary value of an arrival rate at which the magnitude relationship between the batch assembly waiting time related to product arrival and the upper limit value of the Q-time constraint switches, and determines an appropriate load rule; ,
An optimal index generation method characterized in that the optimal index information and load rules calculated in the index calculation step are supplied to the production management system. For the optimal index information calculated by the optimal theoretical value calculation step, further performs a sensitivity analysis step of performing a sensitivity analysis in advance,
4. The optimal index generation method according to claim 3, wherein the index calculation step supplies optimal index information corresponding to fluctuations in various input information to the production management system immediately.   4. The optimal index generation method according to claim 3, wherein the index calculation step changes the number of kanbans and the buffer size in conjunction with a change in throughput in a subsequent process section or a preceding process section.   The index calculation step restricts the size of the buffer provided between the processing steps by the plurality of devices so as not to exceed the maximum number of lots that can be waited between the processing steps, so that the product of the preceding step The optimal index generation method according to claim 3, wherein the output is controlled.   The optimal index generation method according to claim 3, wherein control is performed so that traffic intensity at a bottleneck in the section related to the Q-time constraint does not exceed 1.   The optimal index generation method according to claim 7, wherein when the traffic intensity is greater than 1, the rate of lot injection into the section related to the Q-time constraint is controlled so that the traffic intensity becomes 1. .   4. The number of lots staying in a section related to the Q-time constraint is limited by the number of kanbans according to product types and processing conditions, and the processing start time of each lot is controlled by empty kanbans. The optimal index generation method.   10. An optimal index generation program that causes a processor to execute the steps in the optimal index generation method according to claim 3. An optimum index generation server that transmits index information about a plurality of devices and products via a communication line to a production management system that controls a plurality of devices to produce a wide variety of products,
An information receiving unit for receiving information on the plurality of devices and products via the production management system;
A Q-time structure analysis unit for analyzing a Q-time structure indicating a limit value of a stay time between processing steps by the plurality of apparatuses from the received information;
An optimal theoretical value calculation unit for calculating an optimal number of kanbans and a buffer size for each product type based on the received information and the Q-time structure;
An index calculation unit that outputs the calculated information as index information;
An optimal load rule determination unit that calculates a boundary value of an arrival rate at which the magnitude relationship between the batch assembly waiting time related to product arrival and the upper limit value of the Q-time constraint is switched, and determines an appropriate load rule;
An optimal index generation server, wherein optimal index information and load rules calculated by the index calculation unit are transmitted to the production management system. An optimum index generation method executed by a server that transmits index information about a plurality of devices and products via a communication line to a production management system that controls a plurality of devices to produce a wide variety of products,
An information receiving step of receiving information about the plurality of devices and products via the production management system;
A Q-time structure analysis step of analyzing a Q-time structure indicating a limit value of a stay time between processing steps by the plurality of devices from the received information;
An optimal theoretical value calculating step of calculating an optimal number of kanbans and a buffer size for each product type based on the received information and the Q-time structure;
An index calculating step of outputting the calculated information as index information;
An optimal load rule determination step that calculates a boundary value of an arrival rate at which the magnitude relationship between the batch assembly waiting time related to product arrival and the upper limit value of the Q-time constraint switches, and determines an appropriate load rule; ,
An optimal index generation method, comprising: transmitting optimal index information and load rules calculated in the index calculation step to the production management system. An optimum index generation device that supplies index information about the plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products,
An information input unit for inputting information on a plurality of devices included in the production execution system via the production management system;
A bottleneck analysis unit for analyzing a bottleneck location and a maximum flow rate in a machining process by the plurality of devices from the input information;
A Q-time constraint structure analysis unit that analyzes a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the input information;
A device category calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
While calculating the operation frequency and operation timing of the plurality of regular devices and alternative devices from the calculated device classification and input information , the bottleneck location, the maximum flow rate, the section structure relating to the Q-time constraint, and Based on the device classification and the start / end time information, an optimal theoretical value calculation unit that calculates the optimal number of kanbans and buffer size for each product type ,
A time index calculating unit for calculating start / end time information of preventive maintenance related to the regular device and the alternative device ;
And index calculation unit for outputting the optimal theoretical values pre SL calculated as index information,
A procedure execution control unit that simultaneously manages all Q-time constraints scattered in the entire manufacturing flow included in the production execution system,
An optimal index generation apparatus, characterized in that the start / end time information calculated by the time index calculation unit and the optimal index information calculated by the index calculation unit are supplied to the production management system and engineering business system. An optimum index generation method for supplying index information related to a plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a variety of products,
An information input step for inputting information on a plurality of devices included in the production execution system via the production management system;
A bottleneck analysis step for analyzing a bottleneck location and a maximum flow rate in a machining process by the plurality of devices from the input information;
A Q-time constraint structure analysis step of analyzing a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the input information;
A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
The operation number, operation time, and operation condition of the plurality of regular devices and alternative devices are calculated from the calculated device classification and input information, and the bottleneck location, the maximum flow rate, and the interval relating to the Q-time constraint An optimal theoretical value calculation step for calculating the optimum number of kanbans and buffer size for each product type based on the structure of the device and the device classification and time index ;
A time index calculating step for calculating start / end timing information of preventive maintenance related to the regular device and the alternative device ;
An index calculation step of outputting the calculated optimum theoretical value as index information; and
An optimal index generation method, characterized in that the start / end time information calculated in the time index calculation step and the optimal index information calculated in the index calculation step are supplied to the production management system and the engineering business system.   The optimal index generation method according to claim 14, further comprising a procedure execution control step for simultaneously managing all Q-time constraints scattered in the entire manufacturing flow included in the production execution system.   In the bottleneck analysis step, in order to simultaneously manage Q-time constraints in response to fluctuations in production flow rate and process capability in the entire manufacturing flow included in the production execution system, the bottleneck location in each manufacturing flow and each process 15. The optimum index generation method according to claim 14, wherein the maximum flow rate is analyzed.   The optimal index generation program which makes a processor perform the step in the optimal index generation method of any one of Claims 14 thru | or 16. An optimum index generation server that transmits index information about a plurality of devices via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. And
An information receiving unit for receiving information on a plurality of devices included in the production execution system via the production management system;
A bottleneck analysis unit for analyzing a bottleneck location and a maximum flow rate in a processing step by the plurality of devices from the received information;
A Q-time constraint structure analysis unit that analyzes a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the received information;
A device category calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
The number of operation times, operation timings and operation conditions of the plurality of regular devices and alternative devices are calculated from the calculated device classification and the received information, and the bottleneck location, the maximum flow rate, and the section relating to the Q-time constraint An optimal theoretical value calculation unit for calculating the optimum number of kanbans and buffer size for each product type based on the structure of the device and the start / end time information ;
A time index calculating unit for calculating start / end time information of preventive maintenance related to the regular device and the alternative device ;
An index calculation unit that outputs the calculated optimum theoretical value as index information,
An optimal index generation server that transmits start / end time information calculated by the time index calculation unit and optimal index information calculated by the index calculation unit to the production management system or engineering business system. Optimal index executed by a server that transmits index information about the plurality of devices via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. A generation method,
An information receiving step for receiving information on a plurality of devices included in the production execution system via the production management system;
A bottleneck analysis step for analyzing a bottleneck location and a maximum flow rate in a processing step by the plurality of devices from the received information;
A Q-time constraint structure analysis step of analyzing a structure of a section related to a Q-time constraint in which one or more Q-time constraints exist from the received information;
Based on the bottleneck location, the maximum flow rate, and the structure of the section related to the Q-time constraint, an index calculation step for calculating the optimum number of kanbans and buffer size for each product type and outputting as index information;
A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
An optimal theoretical value calculating step for calculating the number of times of operation of the plurality of regular devices and alternative devices, operating time, and operating conditions from the calculated device classification and the received information;
Performing a time index calculating step of calculating start / end time information of preventive maintenance related to the regular device and the alternative device, and
A method of generating an optimum index, comprising: transmitting start / end time information calculated in the time index calculation step and optimal index information calculated in the index calculation step to the production management system and the engineering business system. An optimum index generation device that supplies index information about the plurality of devices to a production management system that controls a plurality of devices for producing a wide variety of products,
An information input unit for inputting information on the plurality of devices via the production management system;
A device category calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
An optimal theoretical value calculator that calculates the number of operations, operation timing, and operation conditions of the plurality of regular devices and alternative devices from the calculated device classification and input information;
A start / end time calculating unit for calculating start / end time information on the regular device and the alternative device from the calculated operation frequency, operation time, and operation condition;
An optimum index generation apparatus, characterized in that the optimum start / end time information calculated by the start / end time calculation unit is supplied to the production management system and engineering business system. A sensitivity analysis unit that performs a sensitivity analysis in advance for the optimal start / end time information calculated by the optimal theoretical value calculation unit,
21. The optimum index generation device according to claim 20, wherein the start / end time calculation unit immediately supplies optimum index information corresponding to fluctuations of various input information to the production management system. An optimum index generation method for supplying index information about a plurality of devices to a production management system that controls a plurality of devices for producing a variety of products,
An information input step of inputting information on the plurality of devices via the production management system;
A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
An optimal theoretical value calculating step for calculating the number of operations, operation timing and operation conditions of the plurality of regular devices and alternative devices from the calculated device classification and input information;
A start / end time calculating step for calculating start / end time information on the regular device and the alternative device from the calculated operation number, operation time, and operation condition;
An optimum index generation method, characterized in that the optimum start / end time information calculated in the start / end time calculation step is supplied to the production management system and the engineering business system. A sensitivity analysis step for performing a sensitivity analysis in advance for the optimal start / end time information calculated by the optimal theoretical value calculation step;
23. The optimal index generation method according to claim 22, further comprising a start / end time calculation step of immediately supplying optimal index information corresponding to fluctuations of various input information to the production management system.   24. An optimal index generation program causing a processor to execute steps in the optimal index generation method according to claim 22 or 23. An optimum index generation server that transmits index information about the plurality of devices via a communication line to a production management system that controls a plurality of devices for producing a wide variety of products,
An information receiving unit for receiving information on the plurality of devices via the production management system;
A device category calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
An optimal theoretical value calculation unit for calculating the number of operations, operation timing, and operation conditions of the plurality of regular devices and alternative devices from the calculated device classification and received information;
A start / end time calculating unit for calculating start / end time information on the regular device and the alternative device from the calculated operation frequency, operation time, and operation condition;
An optimum index generation server, wherein the optimum start / end time information calculated by the start / end time calculation unit is transmitted to the production management system and the engineering business system. An optimum index generation method executed by a server that transmits index information related to a plurality of devices via a communication line for a production management system that controls a plurality of devices for producing a wide variety of products,
An information receiving step for receiving information on the plurality of devices via the production management system;
A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
An optimal theoretical value calculating step for calculating the number of times of operation of the plurality of regular devices and alternative devices, operating time, and operating conditions from the calculated device classification and the received information;
A start / end time calculating step for calculating start / end time information on the regular device and the alternative device from the calculated operation number, operation time, and operation condition;
An optimal index generation method characterized by transmitting the optimal start / end time information calculated in the start / end time calculation step to the production management system or the engineering business system. An optimum index generation device that supplies index information about the plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products,
An information input unit for inputting information on a plurality of devices included in the production execution system via a production management system;
A Q-time structure analysis unit that analyzes a structure of a section related to a Q-time constraint indicating a limit value of a stay time between processing steps by the plurality of apparatuses from the input information;
A device category calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
The operation count, operation timing, and operation condition of the plurality of regular devices and alternative devices are calculated from the calculated device classification and input information, and the input information and the structure of the section relating to the Q-time constraint And an optimum theoretical value calculation unit for calculating the optimum number of kanbans and buffer sizes for each product type based on the device classification and the start / end time information ,
A start / end time calculating unit for calculating start / end time information on the regular device and the alternative device ;
An index calculation unit that outputs the calculated optimum theoretical value as index information,
The optimum start / end time information calculated by the start / end time calculation unit is supplied to the production management system and the engineering business system, and the process capability determined by the calculated start / end time information is optimal for each period. An optimum index generation device, characterized in that the index calculation unit supplies the index information to a production management system. A sensitivity analysis unit that performs a sensitivity analysis in advance for the optimal index information and start / end time information calculated by the index calculation unit and the start / end time calculation unit,
28. The optimum index generation apparatus according to claim 27, wherein the start / end time calculation unit immediately supplies optimum index information and start / end time information corresponding to changes in various input information to the production management system. An optimum index generation method for supplying index information related to a plurality of devices to a production management system that controls a plurality of devices provided in a production execution system for producing a variety of products,
An information input step of inputting information on a plurality of devices included in the production execution system via a production management system;
A Q-time structure analysis step of analyzing a structure of a section related to a Q-time constraint indicating a limit value of a stay time between processing steps by the plurality of apparatuses from the input information;
A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
The operation count, operation timing, and operation condition of the plurality of regular devices and alternative devices are calculated from the calculated device classification and input information, and the input information and the structure of the section relating to the Q-time constraint And an optimal theoretical value calculating step for calculating an optimal number of kanbans and buffer sizes for each product type based on the device classification and the start / end time information ,
A start / end time calculating step for calculating start / end time information on the regular device and the alternative device ;
An index calculation step of outputting the calculated optimum theoretical value as index information; and
Optimal start / end time information calculated by the start / end time calculation step is supplied to the production management system and the engineering business system, and the process capability determined by the calculated start / end time information is optimal for each period. A method for generating an optimal index, characterized in that correct index information is supplied to a production management system through the index calculation step. For the optimal index information and start / end time information calculated by the index calculation step and the start / end time calculation step, a sensitivity analysis step for performing sensitivity analysis in advance is further performed,
30. The optimum index generation method according to claim 29, wherein the start / end time calculation step immediately supplies optimum index information and start / end time information corresponding to fluctuations of various input information to the production management system.   31. An optimal index generation program that causes a processor to execute steps in the optimal index generation method according to claim 29 or 30. An optimum index generation server that transmits index information about a plurality of devices via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. And
An information receiving unit for receiving information on a plurality of devices included in the production execution system via a production management system;
A Q-time structure analysis unit that analyzes a structure of a section related to a Q-time constraint indicating a limit value of a stay time between processing steps by the plurality of apparatuses from the received information;
A device category calculation unit that divides the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
The operation count, operation timing, and operation conditions of the plurality of regular devices and alternative devices are calculated from the calculated device classification and the received information, and the received information and the structure of the section related to the Q-time constraint And an optimum theoretical value calculation unit for calculating the optimum number of kanbans and buffer sizes for each product type based on the device classification and the start / end time information ,
A start / end time calculating unit for calculating start / end time information on the regular device and the alternative device ;
An index calculation unit that outputs the calculated optimum theoretical value as index information,
Optimal start / end time information calculated by the start / end time calculation unit is transmitted to the production management system and engineering business system, and the process capability determined by the calculated start / end time information is optimal for each period. The optimal index generation server, wherein the index calculation unit transmits the index information to the production management system. Optimal index executed by a server that transmits index information about the plurality of devices via a communication line to a production management system that controls a plurality of devices provided in a production execution system for producing a wide variety of products. A generation method,
An information receiving step of receiving information on a plurality of devices included in the production execution system via a production management system;
A Q-time structure analysis step for analyzing a structure of a section related to a Q-time constraint indicating a limit value of a stay time between processing steps by the plurality of apparatuses from the received information;
Based on the received information and the structure of the section relating to the Q-time constraint, an index calculation step of calculating the optimum number of kanbans and buffer sizes for each product and outputting as index information;
A device category calculation step for classifying the plurality of devices into a regular device having a regular capability and an alternate device having a alternate capability;
An optimal theoretical value calculating step for calculating the number of times of operation of the plurality of regular devices and alternative devices, operating time, and operating conditions from the calculated device classification and the received information;
Performing a start / end time calculation step of calculating start / end time information regarding the regular device and the alternative device, and
Optimal start / end time information calculated by the start / end time calculation step is transmitted to the production management system and engineering business system, and the process capability determined by the calculated start / end time information is optimal for each period. The optimal index generation method is characterized in that accurate index information is transmitted to the production management system through the index calculation step.

Images