WO2019030945A1

WO2019030945A1 - Cause estimation method and program

Info

Publication number: WO2019030945A1
Application number: PCT/JP2018/001561
Authority: WO
Inventors: 昭森口
Original assignee: 株式会社日立ソリューションズ
Priority date: 2017-08-07
Filing date: 2018-01-19
Publication date: 2019-02-14
Also published as: JP6802122B2; JP2019032671A

Abstract

A method for estimating the cause of a manufacturing loss comprises: a step of acquiring production performance information relating to a production process with respect to each of a plurality of individual products; a step of calculating a process time in the production process with respect to each of the plurality of individual products on the basis of the production performance information; a step of acquiring information relating to a manufacturing event indicating a status of the production process; a step of associating the manufacturing event with individual products on the basis of the production performance information and the information relating to the manufacturing event; and a step of causing a distribution of the process times of the plurality of individual products and a distribution of the process times of the individual products with which the manufacturing event is associated to be displayed on a display device in a superposed manner.

Description

Cause estimation method and program

The present disclosure relates to a method and program for estimating causes of production loss.

Due to globalization and diversification of customer needs in recent years, the production industry has been increasing the number of products in small quantities. Increase in incidental work time mainly for setup change work such as replacement of molds and tools and parameter setting change of equipment due to increase of small quantity and variety, decrease in accuracy of cycle time due to increase in product types, increase in number of parts Various losses occur in the manufacturing process of the product, such as deterioration of quality. The influence of losses occurring in the manufacturing process appears at manufacturing time.

For this reason, Patent Document 1 describes a display method for specifying the location of loss occurrence by centrally visualizing the manufacturing time of each process. In Patent Document 1, a time axis is provided for each process, a processing time for each product is described by vertical bars on the time axis, and a graph is generated in which vertical bars of processing times of each process are connected by line segments. .

Patent Document 2 describes a display method for analyzing loss factors for overall optimization instead of individual optimization such as planning and quality by displaying process planning, process progress, production results, and quality inspection all at once. ing.

JP, 2015-075795, A Japanese Patent Application Publication No. 2004-178150

However, the analysis method described in Patent Document 1 only displays the treatment time of the individual product on a time axis with a bar for each production process, and repeatedly occurs under specific conditions across days and weeks. It is difficult to detect manufacturing loss.

For example, manufacturing losses that occur at the same time every day, such as a start-up loss immediately after a break or start of operation, are high in the priority of countermeasures due to repeated losses. However, in the analysis method of Patent Document 1, even if the loss occurs repeatedly, only the temporarily generated loss can be seen and the occurrence condition of the loss can not be known, so it is difficult to estimate the cause.

In addition, although manufacturing loss may occur in a chained manner due to another manufacturing loss, the method of Patent Document 1 has no means for correlating a plurality of manufacturing losses, and the visualization of the influence on production efficiency or the cause It is difficult to estimate. For example, when the abnormal stop of the equipment occurs, the worker is in charge of the restoration work of the equipment, and the process which the worker was originally in charge is delayed. As described above, production losses that cause other production losses in a chained manner have to be preferentially addressed, since they greatly impede production efficiency.

Further, in the case of the method described in Patent Document 2, although information is uniformly displayed on the production control process, there is no means to display even the events at the time of production start such as operation start and the relationship between a plurality of production losses.

This indication is made in view of the above-mentioned point, and provides art which can grasp easily the cause of the manufacture loss of a product individual.

In order to solve the above-mentioned subject, it is the cause estimating method of the manufacture loss which a computer performs, and the step of acquiring the production track record information of a production process about each of a plurality of product individuals, and the above-mentioned plurality based on the above-mentioned Calculating the processing time in the production process for each of the individual products, acquiring information on the production event indicating the status of the production process, and based on the production record information and the information on the production event The step of associating the manufacturing event with the individual product, the distribution of the processing time of the plurality of individual products, and the distribution of the processing time of the individual product linked with the manufacturing event are displayed in a superimposed manner Providing a method of estimating the cause of a manufacturing loss including the step of displaying on a device.

In addition, a step of acquiring production result information of a production process for each of a plurality of product items, a step of calculating processing time in the production process for each of the plurality of product items based on the production result information, and the production Acquiring information on a manufacturing event indicating a status of a process; linking the manufacturing event to the product individual based on the production record information and information on the manufacturing event; and A program is provided for causing a computer to execute a step of superimposing the distribution of the processing time and the distribution of the processing time of the individual product associated with the manufacturing event on a display device.

The present specification includes the disclosure content of Japanese Patent Application No. 2017-152739 on which the priority of the present application is based.

According to the present disclosure, it is possible to easily grasp the cause of the production loss of the individual product. Problems, configurations, and effects other than the above are clarified by the description of the embodiments below.

It is a figure which shows the structure of the estimation system which estimates the cause of the manufacturing loss of a present Example. It is a figure which shows the structure of manufacture information DB. It is a figure which shows the example of a manufacture performance table. It is a figure which shows the example of an installation error information table. It is a figure which shows the example of an installation operation information table. It is a figure which shows the example of a manufacture event list table. It is a figure which shows the example of a manufacture event time table. It is a figure which shows the example of a manufacture event period table. It is a figure which shows the structure of calculation result DB. It is a figure which shows the example of the manufacture performance table with an event. It is a figure which shows the example of a log normal distribution information table. It is a figure which shows the example of an event correlation table. It is a figure which shows the sequence of an acquisition process of manufacture information. It is a figure which shows the sequence of a log normal distribution approximation process. It is a figure which shows the example of the approximated lognormal distribution. It is a figure which shows the sequence of the process which provides event information with respect to manufacture performance. It is a figure which shows the sequence of the process which calculates | requires the correlation of a manufacture loss and a manufacture event. It is an example of GUI which an arithmetic unit outputs to a display. It is an example of GUI which shows the cause estimation result of a manufacturing loss. It is an example of the detailed screen which shows the correlation of a manufacture loss and a manufacture event.

Hereinafter, embodiments of the present disclosure will be described based on the drawings. Note that the embodiments of the present disclosure are not limited to the embodiments described later, and various modifications can be made within the scope of the technical idea thereof. The same reference numerals are given to the corresponding parts of the respective drawings used for describing the respective embodiments to be described later, and the overlapping description will be omitted.

<Example>
FIG. 1 is a diagram showing the configuration of an estimation system for estimating the cause of manufacturing loss according to the present embodiment. As shown in FIG. 1, the estimation system includes a production planning system 101, a manufacturing execution system 102, a manufacturing / inspection facility 103, a facility abnormality detection system 104, an arithmetic device 110, and a display device 120.

The production planning system 101 is a system for formulating types of products to be manufactured and the number thereof, work schedules of workers and equipment, raw materials and parts necessary for manufacturing the products, and has been introduced in many manufacturing industries. Is a well known system.

The manufacturing execution system 102 is a system that collects information on work instructions and manufacturing results for equipment and workers. The equipment abnormality detection system 104 is a system that monitors the state of equipment through a sensor or the like and detects an abnormality that leads to equipment failure.

The arithmetic unit 110 acquires a data acquisition unit 111 that acquires data related to the manufacture of individual products, a manufacturing event processing unit 112 that generates a manufacturing event from the data acquired by the data acquisition unit 111, and associates a manufacturing result with a manufacturing event. Statistics processing unit, and statistical processing unit 113 for estimating the cause of manufacturing loss, manufacturing information DB 114 for storing the manufacturing data acquired by data acquiring unit 111, operation result DB 115 for storing the processing result of statistical processing unit 113 .

The arithmetic unit 110 is provided with a recording unit (not shown), and the recording unit is recorded with mathematical expressions for determining parameters μ, σ and variance v of log normal distribution described later. In addition, various programs are recorded in the recording unit, and the arithmetic device 110 can function as the data acquisition unit 111, the manufacturing event processing unit 112, and the statistical processing unit 113 by executing the programs. The arithmetic device 110 is connected to the display device 120.

The data acquisition unit 111 acquires data (production record information and information on manufacturing events) relating to the manufacture of individual products from the production planning system 101, the manufacturing execution system 102, the manufacturing and inspection facility 103, and the equipment abnormality detection system 104. The result of the cause estimation process stored in the calculation result DB 115 is output to the display device 120.

FIG. 2 is a diagram showing the configuration of the manufacturing information DB 114. As shown in FIG. The manufacturing information DB 114 includes a manufacturing result table 201, an equipment error information table 202, a equipment operation information table 203, a manufacturing event list table 204, a manufacturing event time table 205, and a manufacturing event period table 206. The data held by each table will be described below.

FIG. 3 is a diagram showing an example of the manufacturing result table 201. As shown in FIG. The manufacturing results table 201 includes product numbers for identifying individual products or lots, process IDs for uniquely identifying manufacturing processes, start times and completion times of the individual products in processes identified by the process IDs, start times Stores the processing time which is the difference between the and the completion time.

Here, in the item of “processing time”, for example, the processing time for each of a plurality of individual products in the production process, which is calculated based on the start time and the completion time by the arithmetic device 110, is stored. Moreover, although only process1 is described as process ID in FIG. 3, the data regarding another process ID are also recorded.

FIG. 4 is a diagram showing an example of the equipment error information table 202. As shown in FIG. The facility error information table 202 stores a facility ID uniquely identifying a facility in which an error has occurred, an error occurrence time, and error information representing the content of the error. In the example of FIG. 4, it is recorded that a tool damage error has occurred in the facility whose facility ID is mac_a at March 10, 2017 10:10 in the first row.

FIG. 5 is a diagram showing an example of the equipment operation information table 203. As shown in FIG. The facility operation information table 203 includes a facility ID uniquely identifying the facility, a facility state such as operation or stop, an event causing a stop or operation such as maintenance and production, and a duration of the facility state (a facility state start date and time And end date and time) and. For example, in the first row of the facility operation information table 203, it is recorded that the facility whose facility ID is mac_a is operating from 9:00 to 12:00 on March 2, 2017, and a product is produced. ing.

FIG. 6 is a diagram showing an example of the manufacturing event list table 204. As shown in FIG. The manufacturing event list table 204 is a table for registering a manufacturing event, and includes an event ID uniquely indicating the event and an event name. In FIG. 6, an event ID that includes “t” in the ID name string is an event that occurs with a temporal width, and the other event IDs are events that occur at a temporary point.

FIG. 7 is a diagram showing an example of the manufacturing event time table 205. As shown in FIG. The manufacturing event time table 205 stores the occurrence date and time of a manufacturing event (manufacturing event whose event ID does not include “t”) among manufacturing events registered in the manufacturing event list table 204. It is a table. In the manufacturing event time table 205, manufacturing events are arranged in order of occurrence date and time.

FIG. 8 is a diagram showing an example of the manufacturing event period table 206. As shown in FIG. The manufacturing event period table 206 is a table for storing the start date and the end date of a manufacturing event (manufacturing event having an event ID including "t") among manufacturing events registered in the manufacturing event list table 204. It is.

FIG. 9 is a diagram showing the configuration of the calculation result DB 115. As shown in FIG. The calculation result DB 115 includes an event-added manufacturing result table 901, a log normal distribution information table 902, and an event correlation table 903. The data held by each table will be described below.

FIG. 10 is a diagram showing an example of the manufacturing result table with event 901. The manufacturing result table with event 901 is the manufacturing result table 201 with information on whether or not a manufacturing event has occurred, true for the column of the manufacturing event that has occurred, and false for the column of the manufacturing event that has not occurred. Is stored. The method of determining the occurrence of the manufacturing event will be described below.

First, based on at least one of the occurrence date and time and the end date and time of the manufacturing event, the arithmetic device 110 determines a period in which the manufacturing event is considered to be influenced. The period is, for example, a predetermined period determined based on experience for each process, and is a period during which a manufacturing event is considered to affect the processing time of a product individual. The computing device 110 links a product individual and a manufacturing event in which the determined period and the manufacturing period overlap, and sets the column of the manufacturing event to true.

For example, if the above period in step 1 is 600 seconds before and after the occurrence date of a production event, the start time of the product individual with product number 0001 and the occurrence date of eve_1 are both 9:00, and the production event is affected. The column of eve_1 becomes true because the possible period of time overlaps with the period of manufacture of the individual product.

Further, the arithmetic unit 110 determines, for each product individual, a time difference indicating temporal perspective between the generated manufacturing event and the manufacturing time of the product individual linked to the manufacturing event, and stores the time difference in the manufacturing result table with event 901 . The time difference is determined based on at least one of the occurrence date and time and the end date and time of the manufacturing event, and at least one of the start time and the completion time of the individual product linked to the manufacturing event.

For example, in the case of an event having no time width, the difference between the occurrence date and time of the start of the product individual is taken as the time difference. For example, in the case of the product individual of the product number 0033, since the start time of manufacture is 13:05 and the occurrence time of the manufacture event is 13:00, a time difference of -300 seconds is recorded. Here, when the start time of manufacture is in the past rather than the manufacture event, it is considered as a positive time difference, and when the start time of manufacture is future than the manufacture event, it is considered as a negative time difference.

Also, in the case of an event having a temporal width, that is, an event that has a time interval from occurrence to end, for example, the minimum value of the difference between any time point of the time interval and the start time of the product individual Time difference is assumed. For example, if the start time or completion time of the product individual is within the occurrence period of the event, the time difference is zero.

FIG. 11 is a diagram showing an example of the lognormal distribution information table 902. As shown in FIG. The log normal distribution information table 902 stores the results of calculating the parameters, variances, and log likelihoods of the probability density function corresponding to the log normal distribution by approximating the distribution of the processing time of each step with the log normal distribution. There is. In the example shown in FIG. 11, μ and σ in the probability density function described below are stored. μ and σ are parameters characterizing the probability density function, and e represents the number of Napiers.

Users can, for example, use the variance to assess the amount of manufacturing loss that a process involves. The process including the production loss has a large variation in processing time, that is, dispersion.

FIG. 12 is a diagram showing an example of the event correlation table 903. The event correlation table 903 stores a process ID, an event ID, an event log likelihood, a log likelihood difference, and a time difference (seconds). The event log likelihood indicates the log likelihood when it is assumed that the distribution of the population based on the processing time of the product individual linked to the manufacturing event follows the determined log normal distribution. The above-mentioned population extracts only the processing time of the product individual linked to the specific manufacturing event from the first population of the processing time of one process so that it becomes the same number as the number of elements of the first population. It is a second population obtained by randomly duplicating elements of processing time.

In other words, the processing time of the individual product linked to the event Y (eve_1, eve_2 ...) is extracted from the first population of the entire processing time of process X (X = 1, 2 ...), and the sample number is as described above The log likelihood in the case where it is assumed that the set obtained by duplicating in the second population is the second population, and the distribution of the second population follows the lognormal distribution obtained by fitting to the distribution of the first population is It is an event log likelihood.

The log likelihood difference is the difference between the log likelihood stored in the log normal distribution information table 902 and the event log likelihood stored in the event correlation table 903. That is, the difference between the log likelihood assuming that the first population conforms to the log normal distribution and the log likelihood assuming that the second population obeys the same log normal distribution is the log likelihood Degree difference.

FIG. 13 is a diagram showing a sequence of acquisition processing of manufacturing information. Each step of the sequence will be described below.
(Step 1301)
First, the data acquisition unit 111 acquires information on the manufacturing results of individual products from the manufacturing execution system 102 or the manufacturing / inspection facility 103, and stores the information in the manufacturing results table 201. The data acquisition unit 111 acquires facility error information from the manufacturing / inspection facility 103 or the facility abnormality detection system 104, and stores the facility error information in the facility error information table 202. Further, the data acquisition unit 111 acquires facility operation information from the production planning system 101, the manufacturing execution system 102 or the manufacturing / inspection facility 103, and stores the facility operation information table 203.
(Step 1302)
The data acquisition unit 111 calculates the processing time in the production process for each of the plurality of product individuals based on the start time and completion time of the production process of the plurality of product individuals acquired in step 1301 and (production performance information). , And stored in the manufacturing result table 201. In the case of the example described in FIG. 3, the processing time of the process 1 of the product number 0001 is 600 seconds which is the difference between the start time 09:00:00 and the completion time 09:10:00.
(Step 1303)
The manufacturing event processing unit 112 converts the information on the manufacturing event included in the equipment error information and the equipment operation information acquired in step 1301 into an event ID registered in the manufacturing event list table 204. The manufacturing event processing unit 112 stores, in the manufacturing event time table 205, the occurrence date and time of a manufacturing event having no time width and the event ID of the manufacturing event. Further, the manufacturing event processing unit 112 stores, in the manufacturing event period table 206, the event ID, the occurrence date and time, and the end date and time for a production event having a temporal width.

For example, in the case of the example illustrated in FIG. 4, an error occurs in the facility ID of mac_a at 2017/03/02 02 10:10:00. And in the example shown in FIG. 6, since the event ID of the equipment error of mac_a is eve_6, the record of event ID: eve_6, occurrence time: 2017/03/02 10:10:00 is the manufacturing event time table 205 Stored in

Moreover, in the case of the example shown by FIG. 5, the installation of mac_b is stopped by maintenance from 2017/03/02 02 10:10:00 to 2017/03/02 10:20:00. And in the example shown in FIG. 6, since event ID showing that mac_b is under maintenance is eve_t_6, event ID: eve_t_6, start date and time: 2017/03/02 10:10:00, end date and time The records of 2017/03/02 10:20:00 are stored in the manufacturing event period table 206.

FIG. 14 is a diagram showing a sequence of log normal distribution approximation processing. Each step of the sequence will be described below.
(Step 1401)
The statistical processing unit 113 sets, as a population, a set of processing times of one step among the processing times of the plurality of product individuals calculated in step 1302, and performs all the processing of fitting the histogram of the population with the lognormal distribution. Run through the steps. Then, the statistical processing unit 113 obtains parameters μ and σ characterizing the lognormal distribution for each process. Further, the statistical processing unit 113 calculates the log likelihood in the case where it is assumed that the histogram of the above-mentioned population conforms to the fitted log normal distribution. Note that the fitting with the lognormal distribution is performed using, for example, the least squares method or the maximum likelihood estimation method.

FIG. 15 is a diagram showing an example of the approximated lognormal distribution. In FIG. 15, the histogram of the processing time for one process and the approximated log normal distribution are drawn together. In FIG. 15, the vertical axis represents a ratio, and indicates the number of samples of processing time included in a specific section with respect to the number of samples of total processing time. Thus, the sum of the values across all the sections of the histogram is one.
(Step 1402)
The statistical processing unit 113 obtains the variance v from the parameters μ and σ calculated in step 1401 according to the following equation, and stores the parameters μ, σ, the variance v, and the log likelihood in the log normal distribution information table 902.

FIG. 16 is a diagram showing a sequence of processing for adding event information to a manufacturing result. Each step of the sequence will be described below.
(Step 1601)
The manufacturing event processing unit 112 repeats the processing of the subsequent steps 1602 and 1603 for all the events described in the manufacturing event time table 205 and the manufacturing event period table 206.
(Step 1602)
First, the manufacturing event processing unit 112 searches for a product individual manufactured at the same time as the manufacturing event described in the manufacturing event time table 205 and the manufacturing event period table 206. The manufacturing event processing unit 112 searches for a product individual manufactured at the same time as the manufacturing event based on at least one of the occurrence date and end time of the manufacturing event and at least one of the start time and the completion time of the product individual. Do.

Specifically, the manufacturing event processing unit 112 searches for a product individual whose manufacturing period overlaps with a predetermined period considered to be influenced by the manufacturing event. The predetermined period is a predetermined time width defined for each production process centered on at least one of the occurrence date and the end date of the production event. The predetermined time width is determined, for example, by the experience of the process manager, because the influence of the manufacturing event on the processing time differs depending on the product to be manufactured or the manufacturing process. Further, “a predetermined period and a manufacturing period overlap” means that, for example, at least one of the start time and the completion time of the individual product is included in the predetermined period.

For example, the event: eve_1 described in FIG. 7 occurs at 2017/03/02 09:00:00, and the step of the product number 0001 described in FIG. 3: The start time at process 1 is 09:00: It is 00. That is, the occurrence time of the event: eve_1 and the start time of the process of the product number 0001: process 1 are the same. Therefore, since the organization affected by the manufacturing event overlaps with the manufacturing period, the search by the manufacturing event processing unit 112 is hit.

In addition, the event described in FIG. 8: eve_t_6 is generated at 2017/03/02 02 10:10:00 and then ended at 2017/03/02 10:20:00 and is described in FIG. Step of the product number 0011: The end time of process 1 is included in the event occurrence period. Therefore, since the organization affected by the manufacturing event overlaps with the manufacturing period, the search by the manufacturing event processing unit 112 is hit.
(Step 1603)
Subsequently, the manufacturing event processing unit 112 adds manufacturing event information used for the search to the information on the manufacturing results of the product unit hit in the search in step 1601, and merges the information into the manufacturing result table with event 901. Store. That is, the manufacturing event processing unit 112 links the manufacturing event to the individual product based on the production record information and the information on the manufacturing event.

For example, in the case of the search of eve_1 illustrated in step 1602, since the occurrence time of eve_1 and the process time of the process of product number 0001 overlap: as shown in FIG. 10, the record eve_1 of the product number 0001 as shown in FIG. Stores true. Further, in the case of the search of eve_t_6 exemplified in the step 1601, since the start time of the process of the product 0011: process1 overlaps, true is stored in the eve_t_6 of the record of the product number 0011 as well.

FIG. 17 is a diagram showing a sequence of processing for determining the correlation between the manufacturing loss and the manufacturing event. Each step of the sequence will be described below.
(Step 1701)
The subsequent processing from step 1702 to step 1705 is repeatedly performed for all the events described in the manufacturing event list table 204. Hereinafter, manufacturing events will be described as event i (0 <i <total number of events + 1).
(Step 1702)
The processes from step 1703 to step 1705 are repeatedly executed for all process IDs. Hereinafter, the j-th process ID is described as process j (0 <j <total number of processes + 1).
(Step 1703)
The statistical processing unit 113 acquires, for the process j, the manufacturing results (the manufacturing results at the time of event) of the individual product whose flag of event i is true. Specifically, the statistical processing unit 113 acquires, for the process j, the processing time of the individual product whose event i flag is true.

Subsequently, the statistical processing unit 113 determines that the histogram of the processing time set of the product individual linked to the manufacturing event is represented by the log normal distribution parameter of process j stored in the log normal distribution information table 902. Log likelihood (event log likelihood) is calculated assuming that the normal distribution is followed. At this time, since the log likelihood comparison is performed in step 1704, the number of processing times for process j related to the product individual associated with event i is equal to the number of processing times for all product individual related to process j And randomly duplicate the processing time.
(Step 1704)
The statistical processing unit 113 obtains the difference between the event log likelihood calculated in step 1703 and the log likelihood of process j stored in the log normal distribution information table 902, and stores the difference in the event correlation table 903.

The statistical processing unit 113 calculates an average time difference (an average value of time differences of each product individual) between a product individual having a true event i flag of process j and the event i, and stores the calculated average time difference in the event correlation table 903. The statistical processing unit 113 may store, instead of the average value of the time differences, an intermediate value, a maximum value, or a minimum value of the time differences in the event correlation table 903.

An output example of the manufacturing loss cause estimation result obtained by the cause estimating method of the present embodiment will be described below.

FIG. 18 is an example of a GUI that the arithmetic device 110 outputs to the display device 120. In the GUI, inefficient processes in which production losses have occurred are displayed in order. The arithmetic unit 110 refers to, for example, the variances stored in the lognormal distribution information table 902, and displays the inefficient processes on the display unit 120 in descending order of variance of processing time. A cause analysis button is attached to each process, and when the button is pressed, the screen changes to a screen displaying the cause estimation result.

FIG. 19 is an example of a GUI showing the cause estimation result of the manufacturing loss. The arithmetic device 110 extracts, for example, a manufacturing event having a log likelihood difference equal to or larger than a predetermined value from the event correlation table 903 for the process to which the cause analysis button pressed on the screen of FIG. 18 belongs. Then, the arithmetic device 110 acquires the log likelihood difference and the average time difference associated with the extracted manufacturing event, arranges the manufacturing event names in the order of the magnitude of the average time difference, and displays the same on the display device 120. In the example shown in FIG. 19, manufacturing events with small average time differences are displayed in order from the left.

Between the manufacturing event name and the process name displayed on the GUI, an arrow from the manufacturing event name to the process name and the log likelihood difference are displayed. Each arrow is displayed changing in thickness based on the magnitude of the log likelihood difference. Specifically, thicker arrows are displayed for manufacturing events with larger log likelihood differences. That is, the production event in which the bold arrow is displayed is a production event in which the distribution of the processing time is largely deviated, which is a production event that causes a large production loss. Also, each manufacturing event name is displayed from top to bottom in descending order of log likelihood difference.

As described above, the GUI indicating the cause estimation result of the manufacturing loss determines the two-dimensional arrangement of the manufacturing event name and the process name according to the magnitude of the log likelihood difference and the average time difference. Moreover, the process name and each manufacturing event name are connected by the arrow, and the thickness of the arrow is changed. By displaying in this manner, the user can visually understand the magnitude of the influence of the manufacturing event on the processing time of the process and the strength of the correlation between the manufacturing event and the manufacturing loss.

In the example of FIG. 19, it can be seen that three production events of product type switching, inspection equipment: defect detected in ins 2, and manufacturing equipment: error generated in mac 2 are largely associated with variation in process time of process 1 . Further, it can be seen that the product type switching, ins2_defect, mac2_error, and so on have affected the production of the individual product for a long time in the order. From the above, it is suggested that, for example, the quality is not stable immediately after the product type switching, a defect occurs, and the process 1 process may be delayed due to the defect repair.

In the GUI shown in FIG. 19, when any arrow is pressed by the user, the screen transitions to a screen that displays in detail the correlation between the variation of the processing time in the process and the manufacturing event. The above screen will be described below.

FIG. 20 is an example of a detail screen showing the correlation between the manufacturing loss and the manufacturing event. In the above-mentioned detailed screen, a histogram of the processing time of a plurality of product individuals not limited to the occurrence of a manufacturing event (normal time histogram) and a histogram of the processing time of a product individual linked to a manufacturing event are displayed superimposed There is. As a result, the user can visually recognize which manufacturing event has occurred and how long the normal processing time is extended.

<Modification 1>
In the example, a log normal distribution is given as an example of the distribution curve fitted to the processing time histogram. However, the processing time histogram may be fitted using a distribution curve (probability distribution function) other than the lognormal distribution. For example, fitting using a general probability distribution such as a normal distribution, a Poisson distribution, or a gamma distribution is possible. Further, in the present specification, the log likelihood is used as an example of the evaluation method of the relationship between the manufacturing loss and the manufacturing event, but the evaluation may be performed using the likelihood.

<Modification 2>
Also, in the embodiment, the log-normal distribution with respect to the processing time histogram of all product individuals in one process (both product individuals linked with manufacturing events and product individuals not linked with manufacturing events) The curve was fitted. The fitting of the distribution curve may be performed on the histogram of the population from which the samples of the processing time at which the outliers occur are removed. For detection of outliers, a general method for detecting outliers can be used, such as a method in which a value with a predetermined difference from the average value of the treatment time is taken as an outlier, Tukey-Kramer test, Sumirnov-Grubbs test, or the like.

[Summary]
As described above, in the estimation system of the present disclosure, the histogram of the processing time of the entire product individual manufactured in a certain process, and the processing time of the individual product associated with the manufacturing event among the entire product individual manufactured in the above process And the histogram of the above are superimposed and displayed on the display device 120. Therefore, the user can visually understand the influence of the production event on the processing time of the individual product.

In addition, the estimation system of the present disclosure includes, for example, the likelihood (or log likelihood) of the probability distribution function fitted to the histogram of the processing time of the entire product individual manufactured in a certain process, and the product individual manufactured in the above process Calculate the event likelihood (or event log likelihood) when assuming that the histogram of the processing time of the product individual linked to the manufacturing event in the whole follows the above probability distribution function, and the difference (the likelihood Calculate the degree difference or log likelihood difference). The user can understand the degree of influence of a certain manufacturing event on the processing time of a product individual by confirming the above difference in likelihood (or difference in log likelihood).

Also, the estimation system of the present disclosure, for example, determines a time difference indicating temporal perspective of the manufacturing event and the product individual linked to the manufacturing event for each product individual and calculates the average. By confirming the average time difference, the user can understand how long the production event affects the processing time of production of the product individual.

In addition, the estimation system of the present disclosure, for example, arranges one or more manufacturing events together with information indicating the difference between the likelihood and the event likelihood and displays them on the display device 120 in order of the average magnitude of the time difference. The user can at first glance understand the temporal influence and degree of the production event on the production processing time of the product individual. Furthermore, the user can obtain auxiliary knowledge that recognizes the relationship between a plurality of manufacturing events.

In addition, the estimation system of the present disclosure links, for example, a production event with a production event whose production period overlaps with a predetermined period during which the production event is considered to have affected the production of the production product. That is, the estimation system of the present disclosure effectively extracts the processing time that production loss is considered to have occurred due to a manufacturing event from a set of overall processing times, and what manufacturing event affected the processing time in the production process Can be effectively shown to the user.

The predetermined period is, for example, a predetermined time width defined for each production step centered on at least one of the occurrence date and time and the end date and time of the production event. Since the processing time of a product event varies depending on the type of production process, setting a predetermined period as described above makes it possible to properly extract a sample of processing time affected by the manufacturing event.

Further, the estimation system of the present disclosure, for example, when fitting a predetermined probability distribution function to the distribution of the processing time of a plurality of product individuals, a predetermined outlier detection method from the distribution of the processing times of the plurality of product individuals. The predetermined probability distribution function is fitted to the distribution from which the outliers have been removed by. As described above, removing outliers can improve the estimation accuracy of the cause of manufacturing loss.

The present invention is not limited to the embodiments described above, but includes various modifications. For example, the embodiments described above are described in detail in order to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the configurations described. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. In addition, with respect to a part of the configuration of each embodiment, it is possible to add, delete, and replace other configurations. In addition, each of the configurations, functions, processing units, processing means, etc. described above may be realized by hardware, for example, by designing part or all of them with an integrated circuit. Further, each configuration, function, etc. described above may be realized by software by the processor interpreting and executing a program that realizes each function. Information such as a program, a table, and a file for realizing each function can be placed in a memory, a recording device such as a hard disk or SSD, or a recording medium such as an IC card, an SD card, or a DVD.

All publications and patent documents cited herein are incorporated herein by reference in their entirety.

101: production planning system, 102: production execution system, 103: production / inspection equipment, 104: equipment abnormality detection system, 110: arithmetic device, 111: data acquisition unit, 112: production event processing unit, 113: statistical processing unit, 114: manufacturing information DB, 115: calculation result DB, 120: display device, 201: manufacturing results table, 202: equipment error information table, 203: equipment operation information table, 204: manufacturing event list table, 205: manufacturing event time table , 206: manufacturing event period table, 901: manufacturing result table with event, 902: log normal distribution information table, 903: event correlation table

Claims

A computer-implemented method for estimating the cause of manufacturing loss,
Acquiring production result information of a production process for each of a plurality of product items;
Calculating the processing time in the production process for each of the plurality of product items based on the production result information;
Obtaining information on a manufacturing event indicating the status of the manufacturing process;
Associating the production event with the individual product based on the production record information and information on the production event;
Superimposing the distribution of the processing time of the plurality of product individuals and the distribution of the processing time of the product individual to which the manufacturing event is linked, and displaying the result on a display device;
Method of estimating the cause of manufacturing loss including.
The step of calculating the processing time is a step of calculating the processing time based on the start time and the completion time included in the production record information,
The step of associating the production event with the individual product is based on at least one of an occurrence date and an end date included in the information on the production event and at least one of the start time and the completion time of the individual product. , Associating the production event with the individual product;
A method of estimating the cause of manufacturing loss according to claim 1.
Fitting a predetermined probability distribution function to the distribution of the processing times of the plurality of product individuals;
Calculating the likelihood on the assumption that the distribution of the processing time follows a probability distribution function determined by fitting;
Calculating an event likelihood when it is assumed that the distribution of the processing time of the product individual linked to the manufacturing event follows the probability distribution function obtained by fitting;
Calculating the difference between the likelihood and the event likelihood;
The method for estimating the cause according to claim 1, further comprising
The manufacturing event and the manufacturing event are based on at least one of an occurrence date and an end date of the manufacturing event and at least one of an opening time and a completion time of the individual product linked to the manufacturing event. Determining a time difference indicating temporal perspective with the linked product individual for each product individual and calculating an average thereof;
The method for estimating the cause according to claim 3, further comprising
Displaying one or more production events on a display device in order of information indicating the difference between the likelihood and the event likelihood and in the order of the average magnitude of the time difference;
The method for estimating the cause according to claim 4, further comprising
The step of associating the production event with the individual product comprises
Linking the product event with a product individual whose manufacturing period overlaps with a predetermined period that is considered to have affected the manufacture of the product individual;
The method for estimating the cause according to claim 1.
The step of fitting a predetermined probability distribution function to the distribution of the processing time of the plurality of product individuals is:
Fitting the predetermined probability distribution function to a distribution in which outliers are removed by a predetermined outlier detection method from the distribution of the processing time of the plurality of product individuals;
The method for estimating the cause according to claim 3.
Acquiring production result information of a production process for each of a plurality of product items;
Calculating the processing time in the production process for each of the plurality of product items based on the production result information;
Obtaining information on a manufacturing event indicating the status of the manufacturing process;
Associating the production event with the individual product based on the production record information and information on the production event;
Superimposing the distribution of the processing time of the plurality of product individuals and the distribution of the processing time of the product individual to which the manufacturing event is linked, and displaying the result on a display device;
A program to make a computer run.