Processing device, processing method and storage medium

Abstract

The present invention provides a processing device, a processing method, a program, and a storage medium that can provide users with data useful for making a production plan. The processing device involved in the implementation method refers to a data set, which includes past plan data and actual operation data in a time series that represents the relationship between the time in the past plan and the target production number. The processing device uses the new plan data in the time series that represents the relationship between the time in the new plan and the target production number and each distance between multiple past plan data to calculate multiple first evaluation values. The processing device inputs a new plan image that represents the relationship between the time in the new plan and the target production number into a first model, and uses the classification result of the new plan image output from the first model to calculate multiple second evaluation values ​​for multiple actual operation data. The processing device uses multiple first evaluation values ​​and multiple second evaluation values ​​to extract at least one of the multiple actual operation data.

Description

Technical Field

[0001] Embodiments of the present invention relate to a processing device, a processing method, and a storage medium. Background Art

[0002] When manufacturing a product, a production plan is prepared in advance. It is desirable to develop a technology that can provide users with data useful for preparing the production plan.

[0003] Patent Document 1: Japanese Patent Application Publication No. 2020-149466 Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a processing device, a processing method, and a storage medium that can provide users with data useful for making a production plan.

[0005] The processing device according to the embodiment refers to a plurality of data sets including time-series past plan data and actual operation data representing actual operations relative to the past plan, wherein the time-series past plan data represents the relationship between the time in the past plan and the target production number. The processing device uses each distance between the time-series new plan data representing the relationship between the time in the new plan and the target production number and the plurality of past plan data to calculate a plurality of first evaluation values ​​representing the evaluation of the predicted actual operations for the new plan data for each of the plurality of actual operation data. The processing device inputs a new plan image representing the relationship between the time in the new plan and the target production number into a first model, and uses the classification result of the new plan image output from the first model to calculate a plurality of second evaluation values ​​representing the evaluation of the predicted actual operations for the new plan data for each of the plurality of actual operation data. The processing device extracts at least one of the plurality of actual operation data using the plurality of first evaluation values ​​and the plurality of second evaluation values. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a schematic diagram showing a processing system involved in an embodiment.

[0007] Figure 2 It is a schematic diagram for explaining the processing of the processing system involved in the embodiment.

[0008] Figure 3 It is a schematic diagram for explaining the processing of the processing system involved in the embodiment.

[0009] Figure 4 It is a schematic diagram for explaining the processing of the processing system involved in the embodiment.

[0010] Figure 5 It is a schematic diagram showing the processing result of the processing system involved in the embodiment.

[0011] Figure 6 It is a schematic diagram showing the processing result of the processing system involved in the embodiment.

[0012] Figure 7 It is a schematic diagram showing the processing result of the processing system involved in the embodiment.

[0013] Figure 8 It is a flowchart showing a processing method of the processing system involved in the embodiment.

[0014] Fig. 9 It is a schematic diagram for explaining the advantages of the embodiment.

[0015] Fig.10 This is a flowchart showing a processing method of a processing system according to a modification example of the embodiment.

[0016] Fig.11 It is a schematic diagram showing the hardware structure.

[0017] Description of reference numerals:

[0018] 1: processing device, 2: storage device, 3: input device, 4: output device, 10: processing system, 90: computer, 91: CPU, 92: ROM, 93: RAM, 94: storage device, 95: input interface, 95a: input device, 96: output interface, 96a: output device, 97: communication interface, 97a: server, 98: system bus, 100: processing result, 110: past plan, 111: past plan data, 120: actual work, 121: actual work data, 130: sorting (rank), 131: ranking, 141: first evaluation value, 200: processing result, 210: past plan, 211: past plan image, 220: actual work, 221: actual work data, 230: sorting, 231: ranking, 241: second evaluation value, 300: Judgment result, 310: past plan, 311: past plan data, 320: actual operation, 321: actual operation data, 330: sorting, 331: ranking, 340: evaluation, 341: score, 400: extraction result, 410: past plan data, 420: actual operation data, 500: prediction result, 510: new plan data, 520: first prediction data, A: product, A1: past plan data, A2: actual operation data, B1: past plan data, B2: actual operation data, C1: past plan data, C2: actual operation data, Ch1: chart, Ch2: chart, Ch3: chart, D1: past plan data, D2: actual operation data, N1: new plan data, PM0: processing method, PM1: processing method, a1: past plan image, n1: new plan image. DETAILED DESCRIPTION

[0019] Hereinafter, each embodiment of the present invention will be described with reference to the drawings. In the present specification and each drawing, the same reference numerals are given to the same elements as those already described, and detailed descriptions thereof will be appropriately omitted.

[0020] Figure 1 It is a schematic diagram showing a processing system involved in an embodiment.

[0021] Figure 2 to Figure 4 It is a schematic diagram for explaining the processing of the processing system involved in the embodiment.

[0022] The processing system 10 according to the embodiment provides the user with data that can be used in predicting actual work related to a new production plan. The processing system 10 includes a processing device 1 , a storage device 2 , an input device 3 , and an output device 4 .

[0023] The storage device 2 stores a data set of past plan data and actual operation data for the past plan data. The past plan data is time series data indicating the relationship between the time in the past plan and the target production number at each time for a certain product. The actual operation data is time series data indicating the relationship between the time and the actual production number (actual operation production number) at each time for the product. For example, the past plan data indicates the relationship between the date and the target production number of the product at each date. The actual operation data indicates the relationship between the date and the actual operation production number of the product at each date.

[0024] A plurality of data sets are stored in the storage device 2. The products to be planned and actually worked on may be the same or different between the data sets.

[0025] Figure 2 (a)~ Figure 2 (d) is an example of past planning data. Figure 2 (a)~ Figure 2 The past plan data A1 to D1 of (d) respectively represent plans for the production of products A to D. Figure 2 (e)~ Figure 2 (h) is an example of actual operation data. Figure 2 (e)~

[0026] Figure 2 The actual operation data A2 to D2 of (h) represent the Figure 2 (a)~ Figure 2 (d) The actual operation of the past planning data A1 to D1. Figure 2 (a)~ Figure 2 In (h), the horizontal axis represents time and the vertical axis represents the number of productions.

[0027] include Figure 2 (a) The past planning data A1 and Figure 2 The data set of the actual operation data A2 of (e) is stored in the storage device 2. Similarly, Figure 2 (b) The past planning data B1 and Figure 2 (f) The actual operation data B2 data set, Figure 2 (c) The past planning data C1 and Figure 2 (g) a data set of actual operation data C2, and Figure 2 (d) The past planning data D1 and Figure 2 The data set of the actual operation data D2 (h) is stored in the storage device 2 .

[0028] The processing device 1 obtains new plan data. The new plan data is time series data indicating the relationship between the time in the new plan and the target production number at each time for any product. For example, the new plan data indicates the relationship between the future date and the target production number of the product at each date. The product to be the subject of the new plan may be the same as or different from the product to be the subject of the past plan and actual operation.

[0029] The input device 3 is used for the user to input data to the processing device 1. The output device 4 outputs the data sent from the processing device 1 to the user. For example, the user uses the input device 3 to input new plan data to the processing device 1. The new plan data is stored in the storage device 2, and the processing device 1 can obtain the new plan data from the storage device 2.

[0030] Figure 3 (a) is an example of new planning data. Figure 3 The new plan data N1 of (a) represents the plan for the production of product A. Figure 3 In (a), the horizontal axis represents time and the vertical axis represents the number of productions.

[0031] The processing device 1 extracts actual work data that can be used as a prediction of actual work for a new plan from a plurality of data sets. The processing device 1 executes the following first and second processes for the extraction.

[0032] In the first process, the processing device 1 calculates the distances between the new plan data and the plurality of past plan data, and uses the calculated plurality of distances to calculate a plurality of first evaluation values ​​representing the evaluation as a prediction of the actual work relative to the new plan data for each of the plurality of actual work data.

[0033] For example, the processing device 1 calculates Figure 3 (a) The new plan data N1 and Figure 2 (a)~ Figure 2 (d) The distance between the multiple past plan data A1 to D1. The more similar the patterns of changes in the target production number with respect to time are, the smaller the distance is. The smaller the distance is, the more suitable the actual operation data relative to the past plan data for which the distance is obtained is as a prediction of the actual operation relative to the new plan data.

[0034] The processing device 1 arranges the plurality of actual operation data in ascending order of the plurality of distances to the plurality of past plan data to generate a ranking. The smaller the distance to the past plan data, the higher the actual operation data corresponding to the past plan data is in the ranking. The processing device 1 calculates the first evaluation value of each actual operation data according to the order in the ranking.

[0035] As the distance, Euclidean distance, Manhattan distance, Chebyshev distance, or distance calculated by dynamic time warping (DTW) or the like can be used.

[0036] In the second process, the processing device 1 uses the new plan image and the first model. The new plan image is image data showing the relationship between time and target production quantity in the new plan. Figure 3 (b) is an example of the new planned image. Figure 3 The new plan image n1 of (b) is Figure 3 (a) is the image data corresponding to the new plan data N1.

[0037] In this example, an image of a graph in which the horizontal axis represents time and the vertical axis represents the number of production is used as the new plan image n1. The content shown in the image can be anything as long as it represents the relationship between time and the target number of production. For example, the image can also be a distribution graph showing the relationship between time and the number of production.

[0038] The user inputs a new planned image to the processing device 1 using the input device 3. Alternatively, the processing device 1 may generate a new planned image based on new planning data.

[0039] The first model outputs a classification result based on the input of the plan image. The first model includes, for example, a model learned by a neural network, a random forest, a model learned by a decision tree, or a model learned by a support vector machine. Preferably, the first model includes a convolutional neural network (CNN). A plurality of actual operation data are classified into any one of the categories. The classification result includes the probability that the plan image is classified into each of the plurality of classes.

[0040] The first model is learned in advance using a past plan image. The past plan image is image data showing the relationship between time and target production quantity in past plans for a certain product.

[0041] Figure 4 (a)~ Figure 4 (d) is an example of a past plan image. Figure 4 (a)~ Figure 4 The past planning images a1 to d1 of (d) correspond to Figure 2 (a)~ Figure 2 The past plan data A1 to D1 of (d) represent plans for production of products A to D. In this example, an image of a graph in which the horizontal axis represents time and the vertical axis represents the number of productions is used as a past plan image. Figure 2 (e)~ Figure 2 The actual work data A2 to D2 of (h) represent actual work with respect to the past plans represented by the past plan images a1 to d1, respectively.

[0042] The first model uses the past plan images as input data and uses the values ​​representing the categories of the actual work data as labels for learning. Figure 4 The past plan image a1 of (a) is used as input data, and the representation includes Figure 2 The values ​​of the categories including the actual operation data A2 of (e) are learned as labels. Thus, when a planned image is input, the first model learns to classify the image into which category.

[0043] The category is set according to the change pattern of the target production number with respect to time. For example, regardless of the target production number and the production period, the same category is set for past plans with similar change patterns of the target production number with respect to time.

[0044] The processing device 1 inputs the new plan image to the first model and obtains the classification result output from the first model. The processing device 1 uses the classification result to calculate a second evaluation value for each actual operation data. The second evaluation value represents an evaluation of the actual operation data as a prediction of the actual operation relative to the new plan data represented by the new plan image.

[0045] For example, the processing device 1 arranges the categories in descending order of probability to generate a ranking. The higher the probability, the higher the possibility that the actual operation data classified into the category is similar to the actual operation of the new plan, and the more suitable it is as a prediction of the actual operation. The processing device 1 calculates the second evaluation value of the actual operation data classified into each category according to the order in the ranking.

[0046] The processing device 1 calculates the score of each actual operation data using the first evaluation value and the second evaluation value. For example, the processing device 1 uses the sum of the first evaluation value and the second evaluation value, the product of the first evaluation value and the second evaluation value, or the average of the first evaluation value and the second evaluation value as the score. The processing device 1 may also use the weighted sum of the first evaluation value and the second evaluation value as the score.

[0047] For example, the processing device 1 calculates the weight for each actual operation data. The weight for the actual operation data can be set based on the time difference between the start time of the new plan and the start time of the past plan that is the basis of the actual operation data. It can be considered that the smaller the time difference is, the more suitable the actual operation data is as a prediction of the actual operation relative to the new plan. The smaller the time difference is, the larger the weight is set by the processing device 1.

[0048] The weight of the actual operation data may also be set based on the difference between the total amount of the target production number in the new plan and the total amount of the target production number in the past plan which is the basis of the actual operation data. It can be considered that the smaller the difference in the production number is, the higher the eligibility of the actual operation data is. The smaller the difference in the production number is, the larger the weight is set by the processing device 1. The total amount of the production number can be calculated by accumulating the production number at each time during the production period from the start time to the end time.

[0049] Alternatively, the weight for the actual operation data may be set based on the difference between the length of the production period in the new plan and the length of the production period in the past plan that is the basis of the actual operation data. It can be considered that the smaller the difference in the production period, the higher the eligibility of the actual operation data. The smaller the difference in the production period, the larger the weight is set by the processing device 1.

[0050] The processing device 1 extracts one actual operation data from a plurality of actual operation data based on the score. In the case where a higher score indicates that the actual operation data is more qualified, the processing device 1 extracts the actual operation data with the highest score. The processing device 1 may extract two or more actual operation data in descending order of the scores. The processing device 1 stores the extracted actual operation data in the storage device 2. The processing device 1 may also output the extracted actual operation data to the output device 4.

[0051] The processing device 1 can correct the production period and the number of productions in the extracted actual operation data and generate the first prediction data. The production period and the number of productions can be corrected by the following method.

[0052] The processing device 1 calculates the ratio of the production period of the actual operation data to the production period of the past plan which is the basis of the actual operation data. The processing device 1 multiplies the production period of the new plan by the ratio to calculate the production period of the first forecast data. The processing device 1 sets the start time of the new plan as the start time of the first forecast data. The processing device 1 sets the time obtained by adding the calculated production period to the start time of the new plan as the end time of the first forecast data.

[0053] The processing device 1 calculates the ratio of the total actual operation production number of the actual operation data to the total target production number of the past plan which is the basis of the actual operation data. The processing device 1 multiplies the target production number at each time of the new plan by the ratio to calculate the production number of the first prediction data.

[0054] The processing device 1 stores the first prediction data in the storage device 2. The processing device 1 may also output the first prediction data to the output device 4.

[0055] Figure 5 (a) Figure 5 (b) Figure 6 , Figure 7 (a) and Figure 7 (b) is a schematic diagram showing the processing result of the processing system involved in the embodiment.

[0056] The processing device 1 can be as follows Figure 5 (a) Figure 5 (b) Figure 6 , Figure 7 (a) and Figure 7 The processing result is output as shown in (b) of FIG. The output method is any method such as display on a monitor, projection on a projector, printing on a printer, etc.

[0057] Figure 5 The processing result 100 of (a) represents the result obtained by the first processing. In the processing result 100, one or more past plan data 111 representing the past plan are displayed in the column of the past plan 110. One or more actual operation data 121 representing the actual operation relative to the past plan are displayed in the column of the actual operation 120. The past plan data 111 and the actual operation data 121 are displayed in a chart in a manner that is easy for the user to understand. The number of past plan data 111 displayed in the column of the past plan 110 and the number of actual operation data 121 displayed in the column of the actual operation 120 can be appropriately set by the user. The rank 131 of each actual operation data when each actual operation data is ranked based on the first evaluation value is displayed in the column of the sort 130. The first evaluation value 141 after each actual operation data is multiplied by the weight is displayed in the column of the evaluation 140.

[0058] Figure 5 The processing result 200 of (b) represents the result obtained by the second processing. In the processing result 200, one or more past plan images 211 representing the past plan are displayed in the column of the past plan 210. One or more actual operation data 221 representing the actual operation relative to the past plan are displayed in the column of the actual operation 220. The number of past plan images 211 displayed in the column of the past plan 210 and the number of actual operation data 221 displayed in the column of the actual operation 220 can be appropriately set by the user. Instead of the actual operation data, an actual operation image representing the actual operation relative to the past plan can also be displayed. In the column of the sorting 230, the rank 231 of each actual operation data when the actual operation data is ranked based on the second evaluation value is displayed. In the column of the evaluation 240, the second evaluation value 241 after the weight is multiplied for each actual operation data is displayed.

[0059] The processing device 1 adds the weighted first evaluation value and the second evaluation value to calculate a score. Figure 6Indicates the judgment result obtained based on the processing results of the first processing and the second processing. In the judgment result 300, one or more past plan data 311 representing the past plan are displayed in the column of the past plan 310. The past plan image can also be displayed instead of the past plan data. In the column of the actual operation 320, one or more actual operation data 321 representing the actual operation relative to the past plan are displayed. The actual operation image can also be displayed instead of the actual operation data. The number of past plan data 311 displayed in the column of the past plan 310 and the number of actual operation data 321 displayed in the column of the actual operation 320 can be appropriately set by the user. In the column of the evaluation 340, a score 341 as a weighted sum of the first evaluation value and the second evaluation value is displayed for each actual operation data. In the column of the sort 330, the ranking 331 of each actual operation data when ranked based on the score 341 is displayed.

[0060] For example, the processing device 1 extracts the actual work data having the highest score and the past plan data which is the basis of the actual work data. Figure 7 (a) shows an extraction result 400. The extraction result 400 includes the extracted actual operation data 420 and the past plan data 410 which is the basis of the actual operation data 420. The past plan data 410 and the actual operation data 420 specifically show the production time, the production quantity, and the like.

[0061] The processing device 1 generates first prediction data based on the extraction result. Figure 7 (b) shows the forecast result 500. The forecast result 500 includes new plan data 510 and first forecast data 520. The production time, production quantity, etc. are specifically shown in the new plan data 510 and the first forecast data 520. The new plan data 510 is data prepared by the user and used in the first process and the second process. The production period and production quantity in the first forecast data 520 are corrected using the past plan data 410, the actual operation data 420, and the new plan data 510.

[0062] User confirms Figure 5 (a) shows a first row of one or more actual operation data arranged in a plurality of actual operation data, and Figure 5 The second row of one or more actual work data in which a plurality of actual work data are arranged as shown in (b) can easily understand which actual work data the processing device 1 evaluates and how. Figure 6 The judgment result can easily understand which actual operation data is extracted and what the score of each actual operation data is. Figure 7 The extraction result of (a) can easily grasp the data used for the first prediction data. The user confirms Figure 7 The prediction result of (b) can easily grasp the prediction result of the processing device 1. For example, the user Figure 7 (a) and Figure 7 By comparing (b) with FIG. 1 , it is possible to easily determine whether the data used for the first prediction data is appropriate.

[0063] Figure 8 It is a flowchart showing a processing method of the processing system involved in the embodiment.

[0064] In the processing method PM0, the processing device 1 obtains new plan data and a new plan image (step S1). The processing device 1 performs a first process (step S2). Thus, a first evaluation value is obtained. The processing device 1 weights the first evaluation value (step S3). The processing device 1 performs a second process (step S4). Thus, a second evaluation value is obtained. The processing device 1 weights the second evaluation value (step S5). The processing device 1 uses the weighted first evaluation value and the second evaluation value to calculate a score (step S6). The processing device 1 extracts actual operation data based on the score (step S7). The processing device 1 generates first prediction data (step S8). The processing device 1 outputs the data obtained by the process (step S9).

[0065] In the processing method PM0, the order of processing can be changed appropriately. For example, steps S4 and S5 can be executed before steps S2 and S3. Steps S4 and S5 can also be executed in parallel with steps S2 and S3.

[0066] Advantages of the embodiment will be described.

[0067] For the production of products, it is hoped that limited resources can be used more effectively and flexibly. Through the effective and flexible use of resources, more production numbers and more varieties can be achieved with fewer resources. As a result, for example, sales opportunities can be increased and profits can be improved. On the other hand, the production number will change over time. For example, the production number changes according to changes in product demand and the increase in varieties caused by the production of new products. If the production number changes, the necessary resources will also change. In order to cope with changes in production numbers, it is necessary to plan the relationship between future time and the target production number and determine the securing and allocation of resources. At this time, it is hoped that the securing and allocation of resources will be determined while referring to the actual manufacturing capacity. Manufacturing capacity is an indicator that represents the production number relative to the resources that can be allocated. By referring to the manufacturing capacity, it is possible to reduce excess or shortage of resources when producing according to the plan and utilize resources more effectively and flexibly.

[0068] In the past, manufacturing capacity was calculated based on pre-set rules or optimization techniques. However, there are many uncertain factors in actual manufacturing capacity. Examples of uncertain factors include the sudden absence of operators, the compatibility of products and machines, aging of machines, and differences in operator skills. As a result, it is difficult to produce according to a predetermined new plan. In the past, various responses such as changes in personnel allocation and production adjustments of other products were made based on skills and experience to address these factors. In other words, in order to produce products according to a new plan predetermined by previous methods, human knowledge, skills, experience, etc. are required. In addition, research and execution of responses also take time. Therefore, there is a desire for a technology that can reduce dependence on human knowledge, skills, experience, etc. and reduce the burden of responses.

[0069] In order to solve this problem, the implementation method provides the user with data used to estimate the actual manufacturing capacity with higher accuracy. By referring to the data, the actual manufacturing capacity can be estimated with higher accuracy. Even if uncertain factors arise, a new plan with a high probability of being executable can be created.

[0070] Specifically, the processing system involved in the embodiment extracts one or more actual operation data that can be used in the prediction of the new plan from a plurality of past actual operation data. The extracted actual operation data represents the actual operation based on the past plan similar to the new plan. When the past plan is similar to the new plan, the actual operation based on the past plan is also likely to be similar to the actual operation based on the new plan. The actual operation data is the result of uncertain elements generated for the past plan, and represents the actual manufacturing capacity. By referring to the extracted actual operation data, the actual manufacturing capacity can be estimated with higher accuracy.

[0071] In an embodiment, actual operation data is extracted using the first evaluation value and the second evaluation value. The first evaluation value is calculated using the distance between the new plan data and the past plan data. The second evaluation value is calculated using the classification result of the new plan image. By using the distance and the classification result of the image, a past plan that is more similar to the new plan can be extracted. As a result, actual operation data useful in predicting actual operation relative to the new plan can be extracted.

[0072] Fig. 9 This is a schematic diagram for explaining the advantages of the embodiment.

[0073] exist Fig. 9In the figure, multiple graphs Ch1 to Ch3 showing the relationship between time and production number are shown. In graphs Ch1 to Ch3, the horizontal axis represents time and the vertical axis represents production number. The solid line represents the new plan. The dotted line represents the past plan. By using the distance, it is possible to extract the past plan with a smaller difference in the target production number at each time from the new plan, as shown in graphs Ch1 and Ch2. On the other hand, in the extraction using the distance, regardless of whether the form of the change in the target production number with respect to time is similar or not, it is not easy to extract the past plan with a larger difference in the target production number from the new plan. For example, the distance between the new plan represented by graph Ch3 and the past plan is greater than the distance between the new plan and the past plan represented by graph Ch2. Therefore, the new plan in graph Ch3 is judged to be dissimilar to the past plan in graph Ch3.

[0074] When the classification result of the image is used, even if the difference between the production number of the actual operation data in the classified category and the target production number of the new plan is large, the new plan is mainly classified based on the form of the change in the production number relative to time. For example, even if the difference between the target production number of the new plan and the target production number of the past plan is large as shown in graph Ch3, the new plan is classified into the category of the actual operation data based on the past plan. By using both the distance and the classification result of the image, it is possible to extract actual operation data that is more useful for predicting the actual operation.

[0075] When calculating the distance between the new plan data and the past plan data, it is preferable to use the DTW method. According to the DTW method, the change of the target production number with respect to time can be compared between the new plan data and the past plan data regardless of the length of the production period in each past plan. Therefore, more useful actual operation data can be extracted.

[0076] The first model preferably includes a CNN. According to the CNN, a new plan processed as an image can be more appropriately classified. For example, even in the case where there is a slight difference between the new plan and the past plan used for learning, or in the case where there are outliers in the past plan, the new plan can be more appropriately classified. That is, the new plan can be classified into a category based on the actual operation data of the past plan that looks similar to the new plan.

[0077] The time of one or both of the past plan data and the actual operation data may be corrected. The processing device 1 shifts the time of the actual operation data to the past by the time required for the production of the product. The time required for production is stored in the past plan as, for example, the lead time. Alternatively, the processing device 1 may shift the time of the past plan data to the future by the lead time.

[0078] (Variation Example)

[0079] The processing device 1 may further perform a third process. In the third process, the processing device 1 generates second prediction data based on production data. The production data is time series data indicating the relationship between time and production parameters. The production parameter is at least one selected from the group consisting of the operation rate of the equipment, the occurrence rate of the malfunction, the maintenance rate, and the occurrence rate of the defective products.

[0080] When making a plan, the change of production parameters with respect to time is also predicted. The processing device 1 inputs the production data predicted when making a new plan into the second model. The second model outputs the second prediction data indicating the relationship between time and production quantity.

[0081] The second model includes a neural network. Preferably, the second model includes a recurrent neural network (RNN). More preferably, the RNN includes a Long Short Term Memory (LSTM) structure.

[0082] The second model is learned in advance. Past production data and actual operation data predicted when making past plans are used in learning the second model. Past production data is time series data representing the relationship between time and production parameters in past plans. The second model is learned using past production data as input data and actual operation data as teaching data. Thus, the second model is learned based on production data so as to be able to predict actual operation data.

[0083] The processing device 1 stores the second prediction data in the storage device 2. The processing device 1 may also output the second prediction data to the output device 4.

[0084] Fig.10 This is a flowchart showing a processing method of a processing system according to a modification example of the embodiment.

[0085] In the processing method PM1, the processing device 1 obtains new planning data, new planning images, and production data (step S11). Figure 8 The represented processing method PM0 similarly performs the first processing (step S2) and weights the first evaluation value (step S3). The processing device 1 performs the second processing (step S4) and weights the second evaluation value (step S5). The processing device 1 calculates the score (step S6), extracts the actual operation data (step S7), and generates the first prediction data (step S8). The processing device 1 performs the third processing (step S12). In the third processing, the processing device 1 obtains the second prediction data output from the second model. The processing device 1 outputs the data obtained by the processing (step S9).

[0086] According to the modified example, the user can estimate the manufacturing capacity by referring to both the first prediction data and the second prediction data. In addition, even if there is no past plan similar to the new plan, the user can use the second prediction data in the prediction of the actual operation. According to the modified example, the convenience of the user can be improved.

[0087] Fig.11 It is a schematic diagram showing the hardware structure.

[0088] The processing system 10 according to the embodiment can be Fig.11 The hardware structure shown is implemented. Fig.11 The computer 90 shown includes a CPU 91 , a ROM 92 , a RAM 93 , a storage device 94 , an input interface 95 , an output interface 96 , and a communication interface 97 .

[0089] The ROM 92 stores a program for controlling the operation of the computer 90. The ROM 92 stores a program necessary for the computer 90 to implement the above-mentioned processing. The RAM 93 functions as a storage area for developing the program stored in the ROM 92.

[0090] The CPU 91 includes a processing circuit. The CPU 91 uses the RAM 93 as a working memory to execute a program stored in at least one of the ROM 92 and the storage device 94. During the execution of the program, the CPU 91 controls each component via the system bus 98 and executes various processes.

[0091] The storage device 94 stores data necessary for executing the program and data obtained by executing the program.

[0092] An input interface (I / F) 95 connects the computer 90 to an input device 95a. The input I / F 95 is, for example, a serial bus interface such as USB. The CPU 91 can read various data from the input device 95a via the input I / F 95.

[0093] The output interface (I / F) 96 connects the computer 90 to the output device 96a. The output I / F 96 is, for example, a video output interface such as a digital visual interface (DVI) or a high-definition multimedia interface (HDMI (registered trademark)). The CPU 91 can send data to the output device 96a via the output I / F 96, so that the output device 96a outputs the data. For example, the output device 96a outputs Figure 5 (a) Figure 5 (b) Figure 6 , Figure 7 (a), and Figure 7At least any one of the processing results shown in (b).

[0094] The communication interface (I / F) 97 connects a server 97a outside the computer 90 to the computer 90. The communication I / F 97 is, for example, a network card such as a LAN card. The CPU 91 can read various data from the server 97a via the communication I / F 97.

[0095] The storage device 94 includes one or more selected from a hard disk drive (HDD) and a solid state drive (SSD). The input device 95a includes one or more selected from a mouse, a keyboard, a microphone (sound input), and a touch pad. The output device 96a includes one or more selected from a monitor, a projector, and a printer. A device having the functions of both the input device 95a and the output device 96a, such as a touch panel, may also be used.

[0096] The computer 90 functions as a processing device 1. The storage device 94 and the server 97a function as a storage device 2. The input device 95a is used as an input device 3 of the processing system 10. The output device 96a is used as an output device 4 of the processing system 10.

[0097] By using the processing device or the processing method described above, data useful for making a production plan can be provided to the user. In addition, by using a program for operating a computer as a processing device, the same effect can be obtained.

[0098] The various data processing mentioned above can also be recorded as a program that can be executed by a computer on a disk (floppy disk, hard disk, etc.), an optical disk (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, etc.), a semiconductor memory, or other non-transitory computer-readable storage medium.

[0099] For example, information recorded in a recording medium can be read by a computer (or embedded system). In a recording medium, the recording format (storage format) is arbitrary. For example, a computer reads a program from a recording medium, and a CPU executes instructions recorded in the program based on the program. In a computer, the acquisition (or reading) of a program can be performed through a network.

[0100] The implementation method may include the following technical solutions.

[0101] (Technical Solution 1)

[0102] A processing device refers to a plurality of data sets including time-series past plan data and actual operation data representing actual operations relative to the past plans, wherein the time-series past plan data represents the relationship between time in the past plans and a target production number,

[0103] Using each distance between the new plan data of the time series representing the relationship between the time in the new plan and the target production number and the plurality of past plan data, a plurality of first evaluation values ​​representing the evaluation of the predicted actual operation relative to the new plan data are calculated for each of the plurality of actual operation data,

[0104] A new plan image representing the relationship between the time in the new plan and the target production number is input into the first model, and a plurality of second evaluation values ​​representing the evaluation of the prediction of the actual operation relative to the new plan data are calculated for each of the plurality of actual operation data using the classification result of the new plan image output from the first model.

[0105] At least one actual work data among the plurality of actual work data is extracted using the plurality of first evaluation values ​​and the plurality of second evaluation values.

[0106] (Technical Solution 2)

[0107] According to the processing device described in technical solution 1, the multiple actual operation data are time series data representing the relationship between time and the actual operation production number.

[0108] (Technical Solution 3)

[0109] According to the processing device described in technical solution 2, the first prediction data of the time series representing the relationship between time and the production number is generated using the new plan data and the extracted actual operation data.

[0110] (Technical Solution 4)

[0111] According to the processing device described in technical solution 3, wherein the ratio of the total amount of the actual operation production numbers in the extracted actual operation data to the total amount of the target production numbers in the past planning data that is the basis of the extracted actual operation data is calculated,

[0112] The production number in the first forecast data is calculated by multiplying the target production number in the new plan data by the ratio.

[0113] (Technical Solution 5)

[0114] A processing device according to claim 3 or 4, wherein a ratio of a production period in the extracted actual operation data to a production period in the past plan data which is a basis for the extracted actual operation data is calculated.

[0115] The production period in the first forecast data is calculated by multiplying the production period in the new plan data by the ratio.

[0116] (Technical Solution 6)

[0117] According to the processing device described in any one of technical solutions 1 to 5, the distances are calculated using a dynamic time warping method.

[0118] (Technical Solution 7)

[0119] A processing device according to any one of technical solutions 1 to 6, wherein the first model includes a convolutional neural network, uses a past plan image representing the relationship between the time in the past plan and the target production number as input data, and uses values ​​representing the categories of each of the multiple actual operation data as labels for learning.

[0120] (Technical Solution 8)

[0121] According to the processing device described in any one of technical solutions 1 to 7, the classification result includes the probability of the multiple categories into which the multiple actual operation data are classified,

[0122] The plurality of first evaluation values ​​are calculated based on the rank of the plurality of distances when they are arranged in order from short to long values,

[0123] The plurality of second evaluation values ​​are calculated based on the ranking when the plurality of probabilities are arranged in order from high to low values.

[0124] (Technical Solution 9)

[0125] According to the processing device described in any one of technical solutions 1 to 8, wherein the plurality of first evaluation values ​​are multiplied by a plurality of weights corresponding to the plurality of actual operation data, respectively.

[0126] multiplying the plurality of second evaluation values ​​by the plurality of weights respectively,

[0127] extracting at least one of the plurality of actual operation data using the plurality of weighted first evaluation values ​​and the plurality of weighted second evaluation values,

[0128] The plurality of weights are set based on dates of the plurality of data sets.

[0129] (Technical Solution 10)

[0130] A processing device according to any one of technical solutions 1 to 9, wherein production data in the new plan representing the relationship between time and one or more production parameters selected from the group consisting of equipment operation rate, occurrence rate of malfunctions, maintenance rate and occurrence rate of defective products are input into a second model, and second prediction data of a time series representing the relationship between time and production number are obtained from the output of the second model.

[0131] Several embodiments of the present invention are illustrated above, but these embodiments are only illustrative and are not intended to limit the scope of the invention. These new embodiments can be implemented in various other ways, and various omissions, substitutions, and changes can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and purpose of the invention and are included in the invention described in the technical solution and its equivalent scope. In addition, the above-mentioned embodiments can be implemented by comparing data sets with each other.

Claims

1. A processing device, comprising a processing circuit, wherein: The processing circuit performs the following processing: Referring to a plurality of data sets including time-series past plan data and actual operation data, the time-series past plan data indicating the relationship between time and target production numbers in past plans, and the actual operation data indicating actual operations relative to the past plans, Using each distance between the new plan data of the time series and the plurality of past plan data, a plurality of first evaluation values ​​representing evaluation are calculated for the plurality of actual operation data, the evaluation being a prediction of the actual operation relative to the new plan data, the new plan data of the time series representing the relationship between the time in the new plan and the target production number, A new plan image representing the relationship between the time in the new plan and the target production number is input into the first model, and a plurality of second evaluation values ​​representing the evaluation of the prediction of the actual operation relative to the new plan data are calculated for each of the plurality of actual operation data using the classification result of the new plan image output from the first model. extracting at least one actual operation data from among the plurality of actual operation data using the plurality of first evaluation values ​​and the plurality of second evaluation values, The first model includes a convolutional neural network, which uses a past plan image representing the relationship between the time in the past plan and the target production number as input data, and uses values ​​representing the categories of each of the plurality of actual operation data as labels for learning, The processing circuit inputs the production data in the new plan, which represents the relationship between production parameters and time, into the second model, and obtains second prediction data of the time series output from the second model, which represents the relationship between time and production number, wherein the production parameters are one or more selected from the group consisting of equipment operation rate, occurrence rate of malfunctions, maintenance rate and occurrence rate of defective products.

2. The processing device according to claim 1, wherein: The plurality of actual operation data are time series data indicating the relationship between time and the actual operation production number.

3. The processing device according to claim 2, wherein: The first time-series forecast data indicating the relationship between time and the number of productions is generated using the new plan data and the extracted actual operation data.

4. The processing device according to claim 3, wherein: calculating a ratio of the total amount of the actual operation production numbers in the extracted actual operation data to the total amount of the target production numbers in the past plan data which is the basis of the extracted actual operation data, The production number in the first forecast data is calculated by multiplying the target production number in the new plan data by the ratio.

5. The processing device according to claim 3 or 4, wherein: calculating a ratio of the production period in the extracted actual operation data to the production period in the past plan data which is a basis for the extracted actual operation data, The production period in the first forecast data is calculated by multiplying the production period in the new plan data by the ratio.

6. The processing device according to any one of claims 1 to 4, wherein: The distances are calculated using dynamic time warping.

7. The processing device according to any one of claims 1 to 4, wherein: The classification result includes probabilities of a plurality of categories into which the plurality of actual operation data are classified, The plurality of first evaluation values ​​are calculated based on the rank of the plurality of distances when they are arranged in order from short to long values, The plurality of second evaluation values ​​are calculated based on the ranking when the plurality of probabilities are arranged in order from high to low values.

8. The processing device according to any one of claims 1 to 4, wherein: The plurality of first evaluation values ​​are multiplied by a plurality of weights corresponding to the plurality of actual operation data, respectively. multiplying the plurality of second evaluation values ​​by the plurality of weights respectively, extracting at least one actual operation data from among the plurality of actual operation data using the plurality of weighted first evaluation values ​​and the plurality of weighted second evaluation values, The plurality of weights are set based on dates of the plurality of data sets.

9. A processing device comprising a processing circuit, wherein: The processing circuit performs the following processing: When new plan data of a time series indicating the relationship between time in a new plan and a target production number are received, one or more of a plurality of actual operation data are arranged and displayed in a first rank based on respective distances between the new plan data and a plurality of past plan data of a time series indicating the relationship between time in past plans and a target production number, wherein the plurality of actual operation data respectively indicate actual operations relative to the plurality of past plan data, inputting a new plan image showing the relationship between the time and the target production quantity in the new plan into the first model, arranging one or more of the plurality of actual operation data based on the output result from the first model to display a second ranking, The first model includes a convolutional neural network, which uses a past plan image representing the relationship between the time in the past plan and the target production number as input data, and uses values ​​representing the categories of each of the plurality of actual operation data as labels for learning, The processing circuit inputs the production data in the new plan, which represents the relationship between production parameters and time, into the second model, and obtains second prediction data of the time series output from the second model, which represents the relationship between time and production number, wherein the production parameters are one or more selected from the group consisting of equipment operation rate, occurrence rate of malfunctions, maintenance rate and occurrence rate of defective products.

10. A processing method, causing a processing circuit to perform the following processing: Referring to a plurality of data sets including time-series past plan data and actual operation data, the time-series past plan data indicating the relationship between time and target production numbers in past plans, and the actual operation data indicating actual operations relative to the past plans, Using each distance between the new plan data of the time series and the plurality of past plan data, a plurality of first evaluation values ​​representing evaluations are calculated for the plurality of actual operation data, the evaluations being predictions of actual operations relative to the new plan data, the new plan data of the time series representing the relationship between the time in the new plan and the target production number, A new plan image representing the relationship between the time in the new plan and the target production number is input into the first model, and a plurality of second evaluation values ​​representing the evaluation of the prediction of the actual operation relative to the new plan data are calculated for each of the plurality of actual operation data using the classification result of the new plan image output from the first model. extracting at least one actual operation data from among the plurality of actual operation data using the plurality of first evaluation values ​​and the plurality of second evaluation values, The first model includes a convolutional neural network, which uses a past plan image representing the relationship between the time in the past plan and the target production number as input data, and uses values ​​representing the categories of each of the plurality of actual operation data as labels for learning, The processing circuit is configured to input the production data in the new plan, which represents the relationship between the production parameters and time, into the second model, and obtain the second prediction data of the time series output from the second model, which represents the relationship between time and the production number, wherein the production parameters are one or more selected from the group consisting of the equipment operation rate, the occurrence rate of malfunctions, the maintenance rate, and the occurrence rate of defective products.

11. A storage medium, wherein: A program is stored to cause the processing circuit to execute the following processing: Referring to a plurality of data sets including time-series past plan data and actual operation data, the time-series past plan data indicating the relationship between time and target production numbers in past plans, and the actual operation data indicating actual operations relative to the past plans, Using each distance between the new plan data of the time series and the plurality of past plan data, a plurality of first evaluation values ​​representing evaluations are calculated for the plurality of actual operation data, the evaluations being predictions of actual operations relative to the new plan data, the new plan data of the time series representing the relationship between the time in the new plan and the target production number, A new plan image representing the relationship between the time in the new plan and the target production number is input into the first model, and a plurality of second evaluation values ​​representing the evaluation of the prediction of the actual operation relative to the new plan data are calculated for each of the plurality of actual operation data using the classification result of the new plan image output from the first model. extracting at least one actual operation data from among the plurality of actual operation data using the plurality of first evaluation values ​​and the plurality of second evaluation values, The first model includes a convolutional neural network, which uses a past plan image representing the relationship between the time in the past plan and the target production number as input data, and uses values ​​representing the categories of each of the plurality of actual operation data as labels for learning, The processing circuit is configured to input the production data in the new plan, which represents the relationship between the production parameters and time, into the second model, and obtain the second prediction data of the time series output from the second model, which represents the relationship between time and the production number, wherein the production parameters are one or more selected from the group consisting of the equipment operation rate, the occurrence rate of malfunctions, the maintenance rate, and the occurrence rate of defective products.

Images