TWI826087B - Dispatching system and dispatching method - Google Patents

Abstract

A dispatching system and a dispatching method are provided, including: obtaining a first feature set, a second feature set, and a third feature set; inputting the first feature set to a product line cycle time (CT) model to obtain a processed amount of work in process (WIP) of a first machine; inputting the second feature set to an activation model to obtain an activation of the first machine; inputting the third feature set, the processed amount of WIP, and the activation to a move model to obtain a move of the first machine; calculating a move distribution rate corresponding to the first machine and the first product according to the processed amount of WIP, the activation, the move, and a target move of the first product; and performing a dispatching process according to generate a optimal scheduling according to the move distribution rate and outputting the optimal scheduling.

Description

Work dispatch system and work dispatch method

The invention relates to a work dispatching system and a work dispatching method.

In order to improve product production efficiency, factories often use rule-based dispatch systems to allocate work to production machines. The work dispatch system will first collect information such as product type or quantity of work in process (WIP) of each machine, and then sort WIP work orders according to specific dispatch criteria. However, the quantity of work-in-progress often changes dynamically due to factors such as purchase time, and the traditional labor dispatch system cannot dynamically optimize for such changes. In addition, there are many types of labor dispatch criteria, and production managers often cannot comprehensively consider all labor dispatch criteria and plan the most efficient production process. Therefore, how to automatically dispatch workers with optimal efficiency is one of the important topics for people in this field.

The present invention provides a work dispatching system and a work dispatching method, which can generate the optimal schedule of products based on machine learning technology.

A work dispatching method of the present invention includes: obtaining a first target management parameter set corresponding to a first product, and obtaining a first feature set, a second feature set and a third feature set according to the first target management parameter set, wherein the A target management parameter set is associated with the first machine; the first feature set is input into the production line cycle time model to obtain the work-in-progress processing capacity of the first machine; the second feature set is input into the utilization rate model to obtain the first The utilization rate of the machine; input the third feature set, work-in-progress processing volume and utilization rate into the output model to obtain the output volume of the first machine; based on the work-in-progress processing volume, utilization rate, output volume and the third Calculate the target output of a product corresponding to the output allocation rate of the first machine and the first product; and execute the real-time dispatch process to generate the best schedule based on the output allocation rate, and output the best schedule .

In an embodiment of the present invention, the above-mentioned calculation of the output distribution rate corresponding to the first machine and the first product is based on the work-in-process processing volume, utilization rate, output volume and the target output volume of the first product. The steps include: inputting the first target management parameter set, work-in-progress processing volume, utilization rate and output volume into the reserve volume model to output the reserve capacity of the first machine, and calculating the third machine based on the output volume and reserve capacity. The output capacity of a machine; input the first target management parameter set and the output capacity into the output distribution model to obtain the predicted output distribution rate corresponding to the first machine and the first product; and according to the predicted output The output distribution rate and the target output volume are used to calculate the output distribution rate.

In an embodiment of the present invention, the above-mentioned first target management parameter set is further associated with an upstream machine of the first machine, wherein the first feature set is input into the production line cycle time model to obtain the on-line performance of the first machine. The step of calculating the product throughput includes: inputting the first feature set into the production line cycle time model to output the cycle time of the upstream machine; and calculating the work-in-progress throughput according to the cycle time and the number of work-in-process of the upstream machine.

In an embodiment of the present invention, the above-mentioned labor dispatching method further includes: generating an optimal schedule of the first product according to the output allocation rate based on a genetic algorithm.

In an embodiment of the present invention, the above-mentioned step of generating the optimal schedule of the first product based on the output allocation rate based on the genetic algorithm includes: obtaining the site process of the first product, where the site process includes the corresponding a first site at the first machine; and generating an initial individual corresponding to the site process based on the work-in-process processing volume and at least one cycle time of the first machine, wherein the initial individual includes a first individual corresponding to the first site Candidate output.

In an embodiment of the present invention, the above-mentioned step of generating the optimal schedule of the first product according to the output allocation rate based on the genetic algorithm further includes: updating the first candidate output according to the restriction formula, where the restriction formula The first candidate output quantity is limited to be less than or equal to the sum of the work-in-progress quantity of the first machine and the second candidate output quantity of a site upstream of the first site.

In an embodiment of the present invention, the above restriction formula limits the first candidate output quantity to be less than or equal to the WIP quantity when the cycle time of the first station is greater than or equal to the preset value.

In an embodiment of the present invention, the above-mentioned dispatching method further includes: calculating the difference between the preset value and the cycle time; and calculating the product of the difference and the work-in-progress processing capacity of the first machine, where the limiting formula is in the cycle When the time is less than the preset value, the first candidate output quantity is limited to be less than or equal to the sum of the WIP quantity and the product.

In an embodiment of the present invention, the above-mentioned work dispatching method further includes: obtaining the total candidate output of the first machine based on the first candidate output; obtaining the output capacity and the total candidate output of the first machine The absolute difference of the quantity, where the fitness function of the genetic algorithm is related to the absolute difference.

In an embodiment of the present invention, the above-mentioned initial individual further includes a second candidate output corresponding to an upstream site of the first site, and the method further includes: obtaining the first machine according to the first candidate output. The first total candidate output volume; and obtaining the second total candidate output volume of at least one upstream machine of the first machine based on the second candidate output volume, wherein the fitness function of the genetic algorithm is associated with the first machine The number of work-in-progress, the first total candidate output, and the second total candidate output of the station.

In an embodiment of the present invention, the above-mentioned real-time work dispatch process includes: obtaining the ranking of multiple work orders of the first machine, and selecting the first work order from the plurality of work orders based on the ranking; responding to the first work order If the order contains the queue time, add the first work order to the work order schedule of the first machine; and in response to the first work order not containing the queue time, determine whether the first work order is associated with the first product's best schedule.

In an embodiment of the present invention, the above-mentioned real-time work dispatch process further includes: in response to determining that the first work order is associated with the best schedule, adding the first work order to the work order schedule of the first machine; and In response to determining that the first work order is not associated with the best schedule, the first work order is added to the work order schedule of the first machine according to the first-in, first-out rule.

In an embodiment of the present invention, the above-mentioned optimal schedule includes the first optimal output corresponding to the first site, and the real-time dispatch process further includes: periodically updating the work-in-progress of the first site. quantity; and update the first optimal output quantity based on the quantity of work in progress.

A work dispatching system of the present invention includes a processor, a storage medium and a transceiver. The storage medium stores the production line cycle time model, utilization rate model, and output volume model. The processor is coupled to the storage medium and the transceiver, wherein the processor is configured to: obtain a first set of target management parameters corresponding to the first product through the transceiver, and obtain a first set of features, a first set of features, and a first set of features based on the first set of target management parameters. two feature sets and a third feature set, wherein the first target management parameter set is associated with the first machine; the first feature set is input into the production line cycle time model to obtain the work-in-process processing volume of the first machine; the second set of target management parameters is associated with the first machine; The feature set is input into the utilization rate model to obtain the utilization rate of the first machine; the third feature set, the work-in-process processing volume and the utilization rate are input into the output model to obtain the output of the first machine; according to the work-in-progress Calculate the output allocation rate corresponding to the first machine and the first product through the processing capacity, utilization rate, output amount and the target output amount of the first product; and execute the real-time dispatch process to generate output according to the output allocation rate Optimum schedule, and output the optimal schedule through the transceiver.

Based on the above, the present invention can generate the best schedule for a specific product based on machine learning algorithms such as genetic learning algorithms, thereby optimizing the production efficiency of the product and the utilization rate of the production machine.

In order to make the content of the present invention easier to understand, the following embodiments are given as examples according to which the present invention can be implemented. In addition, wherever possible, elements/components/steps with the same reference numbers in the drawings and embodiments represent the same or similar parts.

FIG. 1 shows a schematic diagram of a labor dispatch system 100 according to an embodiment of the present invention. The dispatch system 100 may include a processor 110, a storage medium 120, and a transceiver 130.

The processor 110 is, for example, a central processing unit (CPU), or other programmable general-purpose or special-purpose micro control unit (MCU), microprocessor, or digital signal processing unit. Digital signal processor (DSP), programmable controller, application specific integrated circuit (ASIC), graphics processing unit (GPU), image signal processor (ISP) ), image processing unit (IPU), arithmetic logic unit (ALU), complex programmable logic device (CPLD), field programmable gate array (field programmable gate array) , FPGA) or other similar components or a combination of the above components. The processor 110 can be coupled to the storage medium 120 and the transceiver 130, and access and execute multiple modules and various applications stored in the storage medium 120.

The storage medium 120 is, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM), or flash memory. , hard disk drive (HDD), solid state drive (SSD) or similar components or a combination of the above components, and are used to store multiple modules or various application programs that can be executed by the processor 110 . In this embodiment, the storage medium 120 can store multiple modules including a production line cycle time model 121, a utilization rate model 122, an output model 123, a reserve model 124, and an output allocation model 125. The function will be explained later.

The transceiver 130 transmits and receives signals in a wireless or wired manner. Transceiver 130 may also perform, for example, low noise amplification, impedance matching, mixing, up or down frequency conversion, filtering, amplification, and similar operations.

FIG. 2 is a schematic diagram of a machine capability evaluation process according to an embodiment of the present invention, in which each step of FIG. 2 can be implemented by the labor dispatch system 100 shown in FIG. 1 . The dispatching system 100 can be used to generate the optimal production configuration of product (i), where the optimal production configuration records the requirements of each workstation site of product (i) during the factory operation (for example: on a day-to-day basis) The quantity of work-in-progress of this product being processed. The above i is a positive integer representing the index of the product, and , where N is a positive integer representing the number of product categories. For example, N=3 means that the production line can produce three types of products, namely product #1 corresponding to i=1, product #2 corresponding to i=2, and product # corresponding to i=3. 3.

In order to generate the optimal production configuration of product (i), the processor 110 can obtain the management by objectives (MBO) parameter set (i, j) corresponding to the product (i) through the transceiver 130, where the MBO parameter set ( i, j) include various parameter values that affect the execution of target management, such as the type of machine, the number of machines, the uptime of the machine, the hold rate of the product lot (Lot) or the product The super hot lot (SHL) rate of batches, etc., this disclosure is not limited to this. In this embodiment, the MBO parameter set (i, j) can be associated with the machine (j). The above j is a positive integer representing the index of the machine (or machine group), and , where M is a positive integer representing the number of machines. For example, M=3 means that there are three machines used for production on the product, namely machine A corresponding to j=1, machine B corresponding to j=2, and machine B corresponding to j=3. C.

The processor 110 can obtain the site process D(i) of the product (i) through the transceiver 130 . The site process D(i) is a vector that records the sites or machines that the processing process of product (i) needs to pass through. For example, if site process D(i)=[A B C A], it means that product (i) will first be processed by machine A at the first site, then processed by machine B at the second site, and then processed by machine B at the second site. Machine C at the third station processes, and finally machine A at the fourth station completes the entire processing process.

The processor 110 may select a feature set for predicting the cycle time (CT) of product (i) at each site from the MBO parameter set (i, j). Processor 110 may input the feature set to line cycle time model 121 to predict cycle time CT(i) for product (i). CT(i) may be a vector, and each element in cycle time CT(i) corresponds to multiple sites of product (i) respectively. For example, if the site process D(i) of product (i)=[ABCA], then the cycle time CT(i) can be equal to [2 2 4 3]. Assume that the variable ct(i,k) represents the k-th element in the cycle time CT(i) (ie: k is a positive integer representing the index of the site and ), taking ct(i,1)=2 as an example, ct(i,1)=2 means that the cycle time that product (i) needs to spend at the first site - machine A is 2 (for example: 2 hours). Table 1 is an example of features in the feature set used to predict cycle time CT(i), where the feature is a key performance indicator (KPI). In one embodiment, the processor 110 may train the line cycle time model 121 based on training data including the characteristics in Table 1 based on a supervised machine learning algorithm. Table 1 index Category Features Remarks 1 time IS_HOLIDAY Is it a holiday? 2 MONTH_EARLY First month: 1st~10th 3 MONTH_MID Mid-month: 11th~20th 4 MONTH_LATE End of month: 21st~31st 5 equipment M_NUM Total number of machines 6 TOOL_NUM Number of machines on the day 7 AVAIL_TIME_NENG Load time rate (excluding: ENG) 8 C_AVAIL_TIME_NENG Variation value of the machine 9 EQP_UTIL Utilization rate 10 C_EQP_UTIL Variation value of utilization rate 11 EQP_AVAIL_RATE Road width/number of machines 12 TC The time interval between Lot’s continuous output 13 C_TC Variation value of TC 14 materials LOT_SIZE Total number of pieces/Total number of batches 15 C_LOT_SIZE Variation value of LOT_SIZE 16 RUN_LOT_SIZE Avg(wafer quantity) 17 C_RUN_LOT_SIZE Variation value of RUN_LOT_SIZE 18 WIP_BOH 07:30~08:30WIP quantity 19 ENG_LOT_RATE Proportion of ENG Lot 20 SHOT_LOT_RATE SHOT Lot Ratio twenty one HOT_LOT_RATE HOT Lot ratio twenty two BACKUP_BY_RATE Proportion supported by other factories twenty three BACKUP_FOR_RATE Proportion of supporting other factories twenty four REWORK_LOT_RATE REWORK Lot ratio 25 QLIMIT_RATE QTIME Lot ratio 26 SAMPLING_RATE Proportion tested by measuring machine 27 CHANGE_RECIPE Replacement formula ratio 28 BATCH_SIZE Average number of batches 29 Comprehensive NUM_UTIL M_NUM*EQP_UTIL

After predicting the cycle time CT(i) of the product (i), the processor 110 can calculate the work-in-progress throughput (j) corresponding to the machine (j) according to the cycle time CT(i) based on equation (1). ). The work-in-progress processing capacity (j) represents the quantity of work-in-progress of various products that the machine (j) can process in a unit time (for example: within 1 day). The index k representing the site is a positive integer, Represents the (k-1)th site in the site process D(i) of product (i) (ie: the upstream site of the k-th site), The (k-1)th station in the site process D(i) representing product (i) is responsible for the machine (j), represents corresponding to cycle time, and represents corresponding to The amount of work-in-progress inventory of product (i) held by the site. In one embodiment, the processor 110 may obtain from the MBO parameter set (i,j) ,and Can be dynamically updated in response to events such as "work in progress". …(1)

On the other hand, the processor 110 may select a feature set for predicting the activation rate (activation) of the machine (j) from the MBO parameter set (i, j). The processor 110 may input the feature set into the utilization model 122 to predict the utilization rate (j) of the machine (j). Table 2 is an example of features in a feature set used to predict utilization rate (j), where the feature is a key performance indicator. In one embodiment, the processor 110 may train the utilization model 122 based on training data including the characteristics of Table 2 based on a supervised machine learning algorithm. Table 2 index Category Features Remarks 1 time IS_HOLIDAY Is it a holiday? 2 equipment M_NUM Total number of machines 3 AVAIL_TIME_NENG Load time rate (excluding: ENG) 4 TC The time interval between Lot’s continuous output 5 materials WIP_BOH 7:30WIP on the same day 6 CLOSE_WIP_QTY WIP of upcoming stations within one day 7 ENG_LOT_RATE Proportion of ENG Lot 8 SHOT_LOT_RATE SHOT Lot Ratio 9 HOT_LOT_RATE HOT Lot ratio 10 BATCH_SIZE Average number of batches

After predicting the cycle time CT(i) of the product (i) and the utilization rate (j) of the machine (j), the processor 110 may select from the MBO parameter set (i, j) for predicting the machine ( The characteristic set of output (j) of j). The processor 110 may input the feature set and the predicted cycle time CT(i) and utilization rate (j) to the throughput model 123 to predict the throughput (j) of the machine (j). Table 3 is an example of features in a feature set used to predict output (j), where the feature is a key performance indicator. In one embodiment, the processor 110 may train the throughput model 123 based on a supervised machine learning algorithm based on training data including the characteristics, cycle time, and availability of Table 3. table 3 index Category Features Remarks 1 time IS_HOLIDAY Is it a holiday? 2 MONTH_EARLY First month: 1st~10th 3 MONTH_MID Mid-month: 11th~20th 4 MONTH_LATE End of month: 21st~31st 5 equipment M_NUM Total number of machines 6 TOOL_NUM Number of machines on the day 7 AVAIL_TIME_NENG Load time rate (excluding: ENG) 8 C_AVAIL_TIME_NENG Variation value of load time rate 9 EQP_UTIL Utilization rate 10 C_EQP_UTIL Variation value of utilization rate 11 EQP_AVAIL_RATE Road width/number of machines 12 TC The time interval between Lot’s continuous output 13 C_TC Variation value of TC 14 materials LOT_SIZE Total number of pieces/Total number of batches 15 C_LOT_SIZE Variation value of LOT_SIZE 16 RUN_LOT_SIZE Avg(wafer quantity) 17 C_RUN_LOT_SIZE Variation value of RUN_LOT_SIZE 18 WIP_BOH 07:30~08:30WIP quantity 19 CLOSE_WIP_QTY WIP of upcoming stations within one day 20 C_CLOSE_WIP Variation value of WIP twenty one ENG_LOT_RATE Proportion of ENG Lot twenty two SHOT_LOT_RATE SHOT Lot Ratio twenty three HOT_LOT_RATE HOT Lot ratio twenty four BACKUP_BY_RATE Proportion supported by other factories 25 BACKUP_FOR_RATE Proportion of supporting other factories 26 REWORK_LOT_RATE REWORK Lot ratio 27 QLIMIT_RATE QTIME Lot ratio 28 SAMPLING_RATE Proportion tested by measuring machine 29 CHANGE_RECIPE Replacement formula ratio 30 BATCH_SIZE Average number of batches 31 Comprehensive NUM_UTIL M_NUM*EQP_UTIL

The processor 110 may obtain input parameters for predicting the backup capability (j) of the machine (j), where the input parameters may include the HBO parameter set (i, j), the work-in-progress throughput (j), and the utilization rate ( j) and output (j). The processor 110 may input the input parameters (i, j) into the backup capacity model 124 to predict the backup capability (j, q) of the machine (j) to other machines (q). The above q is a positive integer representing the index of the machine (or machine group), and , where M is a positive integer representing the number of machines. Table 4 is an example of features in the feature set used to predict redundancy capability (j, q), where the feature is a key performance indicator, and the feature can be a feature of the supporting machine or a feature of the supported machine. . In one embodiment, the processor 110 may train the reserve capacity model 124 based on a supervised learning algorithm based on training data including the characteristics of Table 4, WIP throughput, utilization rate, and output. Table 4 index Category Features Remarks 1 time IS_HOLIDAY Is it a holiday? 2 MONTH_EARLY First month: 1st~10th 3 MONTH_MID Mid-month: 11th~20th 4 MONTH_LATE End of month: 21st~31st 5 equipment AVAIL_TIME_NENG Load time rate (excluding: ENG) 6 C_AVAIL_TIME_NENG Variation value of load time rate 7 EQP_UTIL Utilization rate 8 C_EQP_UTIL Variation value of utilization rate 9 TOOL_NUM Number of machines on the day 10 M_NUM Total number of machines 11 materials WIP_BOH 07:30~08:30WIP quantity 12 CLOSE_WIP_QTY WIP of upcoming stations within one day 13 MOVE_QTY_INTERNAL Total output of the day (excluding: supported/remanufactured/cross-factory Lot) 14 QLIMIT_RATE QTIME Lot ratio 15 TC The time interval between Lot’s continuous output 16 C_TC Variation value of TC

After predicting the output (j) and backup capacity (j, q) of the machine (j), the processor 110 can calculate the output (j) and the backup capacity (j) based on equation (2). The output capacity (j) of the computer station (j). …(2)

FIG. 3 is a schematic diagram for evaluating the output allocation rate according to an embodiment of the present invention, in which each step of FIG. 3 can be implemented by the labor dispatch system 100 shown in FIG. 1 . After predicting the output capacity (j), the processor 110 may select a feature set from the MBO parameter set (i, j) for generating the predicted output allocation rate (i, j), where the predicted output allocation rate The rate (i,j) represents the ratio of the output capacity (j) of the machine (j) to the product (i). The processor 110 may input the feature set and the throughput capacity (j) into the throughput allocation model 125 to generate a predicted throughput allocation rate (i,j) of the machine (j) to the product (i). Table 5 is an example of features in the feature set used to predict the output allocation rate (i, j), where the feature is a key performance indicator. In one embodiment, the processor 110 may train the throughput allocation model 125 based on a supervised machine learning algorithm based on training data including the characteristics of Table 5 and the throughput capacity (j). table 5 index Category Features Remarks 1 time MONTH_EARLY First month: 1st~10th 2 MONTH_MID Mid-month: 11th~20th 3 MONTH_LATE End of month: 21st~31st 4 equipment M_NUM Total number of machines 5 MOVE_QTY_TG The output of the machine on the day 6 materials WIP_BOH_TG 7:30WIP on the same day 7 WIP_BOH 7:30 WIP quantity on the same day 8 CLOSE_WIP_QTY_TG WIP of upcoming stations within one day 9 CLOSE_WIP_QTY Number of WIPs arriving at the station within a day 10 Comprehensive MOVE_RATIO_F Machine WIP>MOVE on the day: the characteristic value is the WIP ratio; WIP≤MOVE on the machine on the day: (1) Machine cycle time ≥ preset value (such as day): the characteristic value is the WIP ratio; (2) Machine cycle time <Default value: Characteristic value = WIP + work in process processing volume * (1-cycle time)

After obtaining the predicted output allocation rate (i, j), the processor 110 can calculate the predicted output allocation rate (i, j) of the product (i) based on the machine (j) and the target of the machine (j). Output (j) The output distribution rate (i,j) of computer station (j) to product (i) is shown in equation (3). For example, "output allocation rate (i, j) = 50%" means that machine (j) is expected to use 50% of the output capacity (j) of the in-process product (i). In one embodiment, the target output volume (j) can be customized by the production manager. The processor 110 can obtain the target throughput (j) through the transceiver 130 . The output allocation rate (i,j) can be a function of both the predicted output allocation rate (i,j) and the target output (j). The processor 110 can determine the output of the machine (j) based on the average output of the machine (j) in the past, the priority of the product (i), the load of the machine downstream of the machine (j), the date of the production line operation, and the current status of the product (i). The output distribution rate (i, j) is calculated using parameters such as product inventory or the in-process incoming goods quantity of product (i). The present invention does not limit the calculation method of the output distribution rate (i, j). …(3)

After determining the output allocation rate (i, j) of the machine (j) to the product (i), the processor 110 can determine the output allocation rate (i, j) based on a genetic algorithm (GA). ) generates the optimal production configuration (i) of product (i), and outputs the optimal production configuration (i) through the transceiver 130 for reference by production managers. The optimal production configuration (i) is a vector, and each element in the optimal production configuration (i) is used to indicate the candidate output of product (i) at the corresponding site. For example, assume that the site process D(i) of product (i) = [A B C A], if the optimal production configuration (i) = [100 100 50 25], represents the first site of product (i) (i.e. : Machine A) needs to produce 100 in-process or finished products of product (i) within unit time (for example: within 1 day), and the second station of product (i) (ie: machine B) needs to produce 100 pieces of work-in-process or finished product of product (i) within unit time. To produce 100 work-in-progress or finished products of product (i), the third station of product (i) (ie: machine C) needs to produce 50 work-in-progress or finished products of product (i) in unit time, and the product The fourth station of (i) (i.e., machine A) needs to produce 25 in-process or finished products of product (i) per unit time.

Specifically, in order to generate the optimal production configuration (i) for product (i), processor 110 may calculate the total output (i, j) of station (j) for product (i) according to equation (4). Then, the processor 110 can process the work-in-progress (j) of the machine (j) according to the cycle time of the machine (j). And the total output (i,j) generates the initial individual (i) of the genetic algorithm, where the initial individual (i) corresponds to the site process D(i) of the product (i). …(4)

The gene of the initial individual (i) can contain the candidate output amount (i,k) of the corresponding site of the product (i), as shown in equation (5), where the candidate output amount (i,k) represents the product (i ) candidate output of the k-th site. For example, assuming that the site process D(i)=[ABCA] of product (i), the initial individual (i) can contain four genes arranged in sequence, as shown in Table 6. Taking gene #1 of the initial individual (i) as an example, gene #1 records that the first site of product (i) is machine A, and the candidate output of machine A is 50. …(5) Table 6 Gene #1 #2 #3 #4 machine A B C A Candidate output 50 80 40 60

The processor 110 may generate a plurality of initial individuals (i) to form an initial population in the same manner as the initial individual (i), and perform crossover or mutation on the initial population to generate a new population. . The processor 110 can continuously update the group based on methods such as roulette wheel selection, competition selection or rank based wheel selection until the best individual whose fitness meets the conditions is found ( i) until. The best individual (i) can include the best output (i,k) of the corresponding site of product (i), where the best output (i,k) represents the k-th site of product (i) optimal output. The processor 110 can extract each gene in the best individual (i) as the best production configuration of the product (i). For example, assuming that the site process D(i)=[ABCA] of product (i), the best individual (i) can contain four genes arranged in sequence, as shown in Table 7. Taking gene #1 of the best individual (i) as an example, gene #1 records that the first site of product (i) is machine A, and the optimal output of machine A is 100 pieces. The processor 110 can extract each gene in Table 7 to obtain the optimal production configuration [100 100 50 25] for product (i). Table 7 Gene (site) #1 #2 #3 #4 site machine A B C A optimal output 100 100 50 25

In one embodiment, in the process of the processor 110 calculating the best individual (i) according to the genetic algorithm, the update of the genes (i.e., candidate output) on each individual of the genetic algorithm needs to meet specific requirements. Limiting formulas, as shown in Equations (6), Equations (7) and Equations (8), in which the preset value can be customized by the production manager. For example, the production manager can set the preset value as the normal operating time of the machine ( uptime) or set to 1 day. …(6) …(7) …(8)

In one embodiment, the fitness function of the genetic algorithm can be associated with function (9) and function (10). When individual information is brought into function (9), the smaller the function value, the higher the efficiency of machine use, and the greater the fitness of the individual. When individual information is brought into function (10), the smaller the function value, the closer the predicted work-in-progress quantity (j) of machine (j) is to the safe water level (j) of machine (j), and represents the individual's adaptation. The greater the degree. The safety water level (j) can be customized by production management personnel. In one embodiment, the fitness function of the genetic algorithm can be expressed as equation (11). …(9) (10)

After generating the optimal production configuration (i) of product (i), the processor 110 can add the work order containing the optimal production configuration (i) to the queue of the real-time dispatch process to wait for the production line to execute the work order. . FIG. 4 illustrates a flow chart of a real-time labor dispatch process according to an embodiment of the present invention, wherein the real-time labor dispatch process can be implemented by the labor dispatch system 100 shown in FIG. 1 . Before starting step S401, the processor 110 sets the variable y to 1, where y represents the ordering of multiple work orders to be queued into the sequence of the immediate dispatch process. In step S401, the processor 110 may determine whether the sorting y is less than or equal to a constant Y, where the constant Y is the total number of work orders to be queued into the sequence of the real-time dispatch process. If the sorting y is less than or equal to the constant Y, then enter step S402. If the sorting y is greater than the constant Y, the processor 110 resets the sorting y to 1 and proceeds to step S404.

In step S402, the processor 110 may select the work order corresponding to sorting y from the Y work orders, and determine whether the work order includes the queue time. Queue time is customized by production managers to ensure that work orders can be executed before the specified time. Therefore, tickets that include queue time have higher priority than regular tickets or AI tickets. Accordingly, if the work order includes queuing time, step S403 is entered, and the processor 110 can add the work order to the work order schedule of the machine. In one embodiment, the processor 110 may add the work order to the work order schedule of the machine according to the first-in, first-out rule. On the other hand, if the work order does not include queuing time, the processor 110 increases the value of sorting y by 1 and re-executes step S401

In step S404, the processor 110 may determine whether the sorting y is less than or equal to the constant Y. If the sorting y is less than or equal to the constant Y, then enter step S405. If the sorting y is greater than the constant Y, the processor 110 resets the sorting y to 1 and proceeds to step S407.

In step S405, the processor 110 may select a work order corresponding to sorting y from the Y work orders, and determine whether the work order is associated with the optimal production configuration. If the work order is associated with the optimal production configuration, it means that the work order is an artificial intelligence work order, and the work order has a higher priority than a general work order. Accordingly, if the work order is associated with the optimal production configuration, step S406 is entered, and the processor 110 can add the work order to the work order schedule of the machine. In one embodiment, the processor 110 may add the work order to the work order schedule of the machine according to the first-in, first-out rule. On the other hand, if the work order is not related to the optimal production configuration, the processor 110 may determine that the work order is a general work order. The processor 110 may increase the value of sort y by 1 and re-execute step S404.

In step S407, the processor 110 may determine whether the sorting y is less than or equal to the constant Y. If the sorting y is less than or equal to the constant Y, then enter step S408. If the sorting y is greater than the constant Y, the processor 110 ends the immediate dispatch process. After completing the real-time dispatch process, the machine's final work order schedule is the best schedule.

In step S408, the work order can be added to the work order schedule of the machine. In one embodiment, the processor 110 may add the work order to the work order schedule of the machine according to the first-in, first-out rule.

In one embodiment, the optimal production configuration (i) of product (i) may include the optimal output volume (i,k) corresponding to the k-th site. The processor 110 may periodically update the work-in-progress quantity (i) of the k-th site, and then update the optimal output quantity (i, k) based on the work-in-progress quantity (i) based on the processes of FIGS. 2 and 3 . In this way, the optimal production configuration (i) of product (i) can be dynamically updated to adapt to the inventory status of work-in-progress that may change at any time.

FIG. 5 illustrates a flow chart of a labor dispatching method according to an embodiment of the present invention, wherein the labor dispatching method can be implemented by the labor dispatching system shown in FIG. 1 . In step S501, a first target management parameter set corresponding to the first product is obtained, and a first feature set, a second feature set and a third feature set are obtained according to the first target management parameter set, wherein the first target management parameter set Associated with the first machine. In step S502, the first feature set is input into the production line cycle time model to obtain the work-in-progress throughput of the first machine. In step S503, the second feature set is input into the utilization rate model to obtain the utilization rate of the first machine. In step S504, the third feature set, the work-in-progress throughput, and the utilization rate are input into the output model to obtain the output of the first machine. In step S505, the output distribution rate corresponding to the first machine and the first product is calculated based on the work-in-progress processing volume, utilization rate, output volume, and the target output volume of the first product. In step S506, the real-time dispatch process is executed to generate an optimal schedule based on the output allocation rate, and the optimal schedule is output.

To sum up, the work dispatching system of the present invention can predict indicators such as the machine's work-in-process processing volume, cycle time, and total output based on machine learning algorithms and production management parameters, and use genetic algorithms and indicators to provide Plan the best schedule for a specific product. When executing the real-time dispatch process, the dispatch system can determine the priority of the work order based on factors such as the queuing time of the work order and whether the work order is an artificial intelligence work order, and determine the work order schedule based on the priority.

100: Work dispatch system 110: Processor 120:Storage media 121:Production line cycle time model 122: Utilization rate model 123:Output model 124: Reserve capacity model 125:Output distribution model 130:Transceiver S401, S402, S403, S404, S405, S406, S407, S408, S501, S502, S503, S504, S505, S506: Steps

Figure 1 is a schematic diagram of a labor dispatch system according to an embodiment of the present invention. FIG. 2 is a schematic diagram illustrating a process for evaluating machine capabilities according to an embodiment of the present invention. FIG. 3 illustrates a schematic diagram for evaluating the output allocation rate according to an embodiment of the present invention. FIG. 4 illustrates a flow chart of the real-time dispatch process according to an embodiment of the present invention. FIG. 5 illustrates a flow chart of a method for dispatching workers according to an embodiment of the present invention.

S501, S502, S503, S504, S505, S506: steps

Claims (14)

A method of dispatching workers, including: Obtain a first target management parameter set corresponding to the first product, and obtain a first feature set, a second feature set and a third feature set according to the first target management parameter set, wherein the first target management parameter set is associated with at the first machine; Input the first feature set into the production line cycle time model to obtain the work-in-progress throughput of the first machine; Input the second feature set into the utilization rate model to obtain the utilization rate of the first machine; Input the third feature set, the work-in-progress throughput and the utilization rate into the output model to obtain the output of the first machine; Calculate the output distribution rate corresponding to the first machine and the first product based on the work-in-process processing volume, the utilization rate, the output volume, and the target output volume of the first product. ;as well as Execute a real-time dispatch process to generate an optimal schedule based on the output allocation rate, and output the optimal schedule. The labor dispatch method according to claim 1, wherein the number of workers corresponding to the first product is calculated based on the work-in-progress processing volume, the utilization rate, the output volume and the target output volume of the first product. The step of allocating the output volume of the machine and the first product includes: The first target management parameter set, the work-in-progress processing volume, the utilization rate and the output volume are input into the reserve volume model to output the reserve capacity of the first machine, and according to the The output volume and the backup capacity are used to calculate the output capacity of the first machine; Input the first target management parameter set and the output capacity into an output allocation model to obtain a predicted output allocation rate corresponding to the first machine and the first product; and The output allocation rate is calculated based on the predicted output allocation rate and the target output. The method of dispatching work according to claim 1, wherein the first target management parameter set is further associated with an upstream machine of the first machine, and the first feature set is input to the production line cycle time The step of modeling to obtain the work-in-progress throughput of the first machine includes: Input the first feature set into the production line cycle time model to output the cycle time of the upstream machine; and The WIP throughput is calculated based on the cycle time and the WIP quantity of the upstream machine. The method of dispatching workers as described in request item 1 further includes: The optimal schedule of the first product is generated based on the output allocation rate based on a genetic algorithm. The work dispatching method according to claim 4, wherein the step of generating the optimal schedule of the first product according to the output allocation rate based on the genetic algorithm includes: Obtain a site process for the first product, wherein the site process includes a first site corresponding to the first machine; and An initial individual corresponding to the site process is generated according to the work-in-progress throughput and at least one cycle time of the first machine, wherein the initial individual includes a first candidate product corresponding to the first site. Output. The work dispatching method of claim 5, wherein the step of generating the optimal schedule of the first product based on the output allocation rate based on the genetic algorithm further includes: The first candidate output quantity is updated according to the restriction formula, wherein the restriction formula limits the first candidate output quantity to be less than or equal to the work-in-progress quantity of the first machine and the quantity of work-in-progress of the first site. The sum of the output quantities of the second candidates at the upstream site. The labor dispatch method of claim 6, wherein the restriction formula limits the first candidate output to be less than or equal to the Quantity of work in progress. The method of dispatching workers as described in request item 7 further includes: Calculate the difference between the preset value and the cycle time; and Calculate the product of the difference and the work-in-progress throughput of the first machine, where The restriction formula limits the first candidate output quantity to be less than or equal to the sum of the work-in-progress quantity and the product when the cycle time is less than the preset value. The method of dispatching workers as described in request item 4 further includes: Obtain the total candidate output of the first machine according to the first candidate output; Obtain the absolute difference between the output capacity of the first machine and the total candidate output, where The fitness function of the genetic algorithm is related to the absolute difference. The method of dispatching work as described in claim 4, wherein the initial individual further includes a second candidate output corresponding to an upstream site of the first site, wherein the method further includes: Obtain the first total candidate output of the first machine according to the first candidate output; and The second total candidate output of at least one upstream machine of the first machine is obtained according to the second candidate output, wherein The fitness function of the genetic algorithm is related to the work-in-progress quantity of the first machine, the first total candidate output volume, and the second total candidate output volume. The method of dispatching workers as described in request item 1, wherein the instant dispatch process includes: Obtain the ranking of multiple work orders of the first machine, and select a first work order from the plurality of work orders based on the ranking; In response to the first work order including the queue time, adding the first work order to the work order schedule of the first machine; and In response to the first work order not including the queue time, it is determined whether the first work order is associated with the optimal schedule of the first product. The method of dispatching workers as described in request item 11, wherein the instant dispatch process further includes: In response to determining that the first work order is associated with the optimal schedule, add the first work order to the work order schedule of the first machine; and In response to determining that the first work order is not associated with the optimal schedule, the first work order is added to the work order schedule of the first machine according to a first-in, first-out rule. The method of dispatching workers as described in claim 11, wherein the optimal schedule includes the first optimal output corresponding to the first site, and the real-time dispatch process further includes: Periodically updating the work-in-progress quantity at the first site; and The first optimal output quantity is updated according to the quantity of work in progress. A labor dispatch system including: transceiver; Storage media to store the production line cycle time model, utilization rate model and output volume model; A processor coupled the storage medium and the transceiver, wherein the processor is configured to perform: Obtain a first target management parameter set corresponding to the first product through the transceiver, and obtain a first feature set, a second feature set and a third feature set according to the first target management parameter set, wherein the first The target management parameter set is associated with the first machine; Input the first feature set into the production line cycle time model to obtain the work-in-progress throughput of the first machine; Input the second feature set into the utilization rate model to obtain the utilization rate of the first machine; Input the third feature set, the work-in-progress throughput, and the utilization rate into the output model to obtain the output of the first machine; Calculate the output distribution rate corresponding to the first machine and the first product based on the work-in-process processing volume, the utilization rate, the output volume, and the target output volume of the first product. ;as well as A real-time dispatch process is executed to generate an optimal schedule according to the output allocation rate, and the optimal schedule is output through the transceiver.

Images