JP7826136B2 - Planning device and planning method - Google Patents

Description

The present invention relates to a planning device and planning method in which a planner modifies an automatically generated plan and adopts the manually revised plan as the final plan.

There are many events in which advance planning is important, such as product manufacturing and the operation and management of large-scale systems. When creating such plans, it is necessary to create a plan that satisfies the planning's objective functions, such as maximizing production volume, maximizing equipment utilization, and minimizing the number of workers, while adhering to constraints such as time, space, equipment, and human resources. Creating such plans by hand takes too much time, so computers are often used.

On the other hand, real-world environments often involve complex multiple constraints and objective functions, making it difficult to accurately define all of them and input them into a computer. In such cases, the accuracy of the plan can be improved by defining constraints and objective functions to the extent possible and combining them with rule-of-thumb logic. However, it is extremely difficult to define constraints and objective functions on a computer that completely simulate the real environment, and rule-of-thumb logic lacks versatility and scalability, so the plan output by the computer is unlikely to be satisfactory to the planner as is.

For this reason, a method has been proposed in which the ambiguous priorities of multiple constraints and objective functions are measured based on a plan output by a computer and then revised by the planner, and these priorities are reflected in the next plan creation.

For example, Patent Document 1 proposes a configuration in which, with the aim of enabling more accurate plan creation by reflecting correction know-how extracted from pre-revision plans and revised plans by the planner, "the plan creation unit 112 uses plan condition data 121 and revision know-how data 126-129 read from the storage device 120 to determine decision variables for the new plan so that the new plan satisfies the constraints and objective function and fits into one of the groups included in the revision know-how data. Here, the revision know-how data includes multiple groups for which anti-patterns 127 indicating trends in the work patterns of multiple pre-revision plans, reference patterns 128 indicating trends in the work patterns of multiple revised plans in which the planner has revised the multiple pre-revision plans, and plan-related information for multiple pre-revision plans and revision probabilities 126, which are statistics of decision variables, have been determined."

Japanese Patent Application Laid-Open No. 2021-071771

According to Patent Document 1, by reflecting the revision know-how extracted from the pre-revision plan and the revised plan prepared by the planner, it is possible to create a more accurate plan.

However, human corrections are not always correct, and there may have been other reasons, for example. Furthermore, while Patent Document 1 extracts correction know-how from the differences between the pre-correction plan and the post-correction plan, it is difficult to extract policy changes, such as changing the priority of each indicator, from the differences.

In light of the above, the present invention aims to provide a planning device and planning method that enables continuous, stable production by automatically updating the planning device at appropriate times, rather than updating it unnecessarily.

According to the present invention, "a planning device in which a manually revised plan obtained by a planner revising an automatically generated plan is made into a final plan, the planning device comprising: a storage device that stores the automatically generated plan created by the automatic planning processing unit and the manually revised plan obtained by the planner revising the automatically generated plan; an evaluation processing unit that evaluates a plurality of predetermined evaluation indexes for the automatically generated plan and the manually revised plan; a detection processing unit that determines the difference between the evaluation result of the automatically generated plan and the evaluation result of the manually revised plan; and a proposal processing unit that, when the difference is greater than a predetermined threshold, recommends that it is time to update the adjustment of the weights of the evaluation indexes to minimize the difference; and an output device that presents the adjustment of the weights of the evaluation indexes recommended by the proposal processing unit to the planner."

The present invention also provides a "planning method in which a manually revised plan obtained by a planner modifying an automatically generated plan is used as a final plan, the method comprising: evaluating a plurality of predetermined evaluation indices for the automatically generated plan and the manually revised plan obtained by the planner modifying the automatically generated plan; determining the difference between the evaluation results of the automatically generated plan and the manually revised plan; and, if the difference is greater than a predetermined threshold, notifying the planner that it is time to update the adjustment of the weights of the evaluation indices to minimize the difference."

This invention makes it possible to achieve sustained, stable production by automatically updating equipment at appropriate times, rather than needlessly updating it.

FIG. 1 is a diagram showing an example of the configuration of a planning support device 10. FIG. 1 is a diagram showing an example of a plan. FIG. 4 is a diagram showing an example of a data structure of plan information. FIG. 10 is a diagram showing an example of a data structure of product information. FIG. 10 is a diagram showing an example of a data structure of an automatically generated plan result. FIG. 10 is a diagram showing an example of a data structure of a manual correction planning result. FIG. 10 is a diagram showing an example of the data structure of a correction log. FIG. 2 is a diagram showing the time-series processing of various functions in the central processing unit of FIG. 1; 9 is a flowchart specifically illustrating various functions shown in FIG. 8 . A diagram summarizing the amount of constraint violations and evaluation index values by month in chronological order. A diagram showing the result after correction. FIG. 10 is a diagram showing an example of a screen that recommends the variance value of the difference between the automatically generated plan and the revised plan.

The following describes an embodiment of the present invention with reference to the drawings. The planning device is not limited to specific plans, and can be applied to a wide variety of plans, such as factory production, maintenance of social infrastructure, or personnel allocation in various facilities. However, the following description focuses on an example of a planning support device that creates so-called production plans, in which products are manufactured by working in a predetermined process.

In an embodiment of the present invention, we will explain how part of the plan creation device is configured as a plan creation support device that leaves it up to the judgment of the planner.

A planning device according to an embodiment of the present invention will be described using Figure 1. The specific configuration for implementing the planning support device 10 will be described assuming a mainframe or personal computer, but it can also be implemented on the cloud.

The plan creation support device 10 has the following hardware configuration: a storage device 120 consisting of a non-volatile storage element such as an SSD (Solid State Drive) or a magnetic medium such as a hard disk drive; a memory 150 consisting of a volatile storage element such as RAM; a central processing unit (arithmetic unit) 110 that reads and executes a program 130 stored in the storage device 120 into the memory 150, thereby providing overall control of the device itself and performing various judgments, calculations, and control processes; an input device 160 that accepts key input and voice input from the planner M; and an output device 140 such as a display that displays processed data; these pieces of hardware are connected to each other via a bus so that they can communicate with each other.

The central processing unit 110 reads the program 130 stored in the storage device 120 into the memory 150 and executes it to perform the functions of an evaluation processing unit 111 that evaluates the automatically generated plan and the manually corrected plan, a detection processing unit 112 that detects situations that require action from the evaluation results, a proposal processing unit 113 that requests the plan creator M to adjust the weights of the evaluation indexes when the difference in evaluation values becomes large, an adjustment processing unit 114 that adjusts the weights of the evaluation indexes, and an automatic plan creation processing unit 115 that creates an automatically generated plan.

With the above configuration, the planning support device 10 according to an embodiment of the present invention evaluates the difference between the automatically generated plan created by the automatic planning processing unit 115 and the manually revised plan manually revised by the planner, and helps enable optimization and adjustment of the functions of the automatic planning processing unit 115.

In addition to being stored in the storage device 120, the program 130 may also be installed by the planning support device 10 from another device via a specified medium into the storage device 120 when necessary, such as during execution. The medium refers to, for example, a storage medium or communication medium that can be attached to or detached from a specified interface of the planning support device 10.

The storage device 120 also stores the data necessary to execute these functions, specifically, constraints and objective functions (planning conditions) 121, plan-related information 122, automatically generated plan results 123, manually revised plan results 124, evaluation results 125, weights for each evaluation 126, and programs 130.

The various data stored in the storage device 120 in Figure 1 can be divided into preset information, continuously updated information, and information generated by the present invention. Of these, the preset information is the constraints and objective functions (planning conditions) 121 and plan-related information 122, the continuously updated information is the automatically generated plan results 123 and manually revised plan results 124, and the information generated by the present invention is the evaluation results 125 and the weight of each evaluation 126.

Next, we will explain each piece of data in detail, but as a prerequisite, we will first explain an example of a plan created by the planning support device 10 using Figure 2. In Figure 2, a production plan is created to execute N processes Lj (j = 1 to N) to produce M products Pi (i = 1 to M), with the aim of obtaining a production plan such as that shown in Gantt chart 20. Note that Gantt chart 20 is an example when N = 4 and M = 3, and also shows the time axis for process L4.

2 also shows planning conditions that the planning support device 10 uses to create a plan as shown in a Gantt chart 20. When creating a production plan, a set of products Pi and a set of processes Li required to produce the products are defined, and the processing time _Tij of each process for each product is given as an explanatory variable.

In the example of FIG. 2 , two constraints are set. The first is a sequence constraint, which requires that in each process, a process for another product cannot begin until the process for the current product is completed. The second is an equipment constraint, which requires that in each process, work cannot be performed on multiple products. Furthermore, an objective function f is set as the planning goal, which minimizes the time it takes for all processes for all products to be completed (total processing time). Therefore, in this case, creating a production plan means finding the start time t _ij (decision function) of each process for each product that satisfies the above two constraints and minimizes the objective function f. Note that the above is merely an example, and other constraints, such as product type combination conditions and delivery date conditions, or objective functions such as minimizing production costs, may also be set. Furthermore, multiple objective functions may also be set.

The above data on constraints, objective functions, decision variables, and explanatory variables is stored as constraints and objective functions (planning conditions) 121 in the storage device 120. Next, we will explain the data structure of the data used by the planning support device 10 to create plans.

Plan-related information 122 is a collection of plan-related data collected about past plans. The contents of plan-related information 122 can be selected at will by plan creator M as information that will affect the quality of the plan. For example, when the plan creator makes revisions to a plan, it is desirable to decide what data to accumulate as plan-related information 122 based on their knowledge of what information they use to determine whether or not revisions are necessary. Here, plan information 122a and product information 122b are given as specific examples of plan-related information 122.

Each record of the plan information 122a illustrated in Figure 3 stores values such as a plan number 2211, plan execution month 2212, plan execution date 2213, and number of operations 2214. The plan number 2211 is identification information that uniquely identifies a plan, and records with the same plan number represent information explaining the plan with that plan number. The plan execution month 2212 is the month the plan is executed, the plan execution date 2213 is the date the plan is executed, and the number of operations 2214 is the total number of operations in the plan (number of processes N x number of products M). This data is collected as information 122a explaining the plan when the month or date the plan is executed or the overall load of the plan is thought to affect the success or failure of the plan. Other data may also include identification information that uniquely identifies the plan creator, the operating status of production equipment, weather, and other data. Furthermore, unless plan creator M considers it information explaining the plan, even the data illustrated in Figure 3 does not need to be included in the plan information 122a. The same applies to the other data described below.

Each record of the product information 122b illustrated in Figure 4 stores values such as plan number 2221, product 2222, color 2223, and orderer 2224. Plan number 2221 is identification information that uniquely identifies the plan information and is the same as plan number 2211. Product 2222 is identification information that uniquely identifies the product that the plan targets, color 2223 is the color of the product, and orderer 2224 is a code that identifies the orderer who placed the product. The data collected as product information 122b is selected as product information that affects the success or failure of the plan. For example, color information is included in this example because the plan includes a painting process and the color of the product affects the painting sequence. Other data such as product type and delivery date may also be included.

Next, the automatically generated plan result 123 and the manually revised plan result 124 will be explained. If necessary, the planner may revise the start time determined by the planning support device 10 based on their own knowledge and know-how, and execute the revised plan. In such cases, the plans before and after the revision are stored as the automatically generated plan result 123 and the manually revised plan result 124, respectively. Therefore, for the plan (before revision) stored in the automatically generated plan result 123, there exists a corresponding plan (after revision) stored in the manually revised plan result 124.

Note that the automatically generated plan result 123 is generated using the automatic planning processing unit 115, but the specific method for creating the automatically generated plan result 123 in the automatic planning processing unit 115 is known, for example, from Patent Document 1, and therefore a detailed explanation will be omitted here. The present invention does not limit the method for creating an automatically generated plan in the automatic planning processing unit 115.

Each record of the automatically generated plan result 123 shown in FIG. 5 stores values such as a plan number 2301, a product 2302, a process 2303, a processing time 2304, and a start time 2305. The plan number 2301 is identification information that uniquely identifies the plan and is the same as the plan number 2211. The product 2302 is identification information that uniquely identifies the product targeted by the plan, the process 2303 is identification information that uniquely identifies the process targeted by the plan, the processing time 2304 is the processing time T _ij of the corresponding process for the corresponding product, and the start time 2305 is the start time t _ij of the corresponding process for the corresponding product. It can be seen that the Gantt chart shown in FIG. 2 can be created from the information of records with the same plan number. The field values of the product 2302, the process 2303, and the processing time 2304 are values given for creating the plan, and the field value of the start time 2305 is a value determined by the planning support device 10.

Each record of the manually corrected plan result 124 shown in Fig. 6 stores values such as a plan number 2401, an operation 2402, a process 2403, a processing time 2404, and a start time 2405. The meaning of each record item is the same as the items in the automatically generated plan result 123 shown in Fig. 5. By comparing the two, it can be seen that, for example, for plan numbers 0601-0800, the start time _t21 of product P2, process L1 has been corrected from 08:10 to 08:20.

It is advisable to store a correction log in addition to the manual correction plan result 124 in the storage device 120. Storing the correction log and presenting it as appropriate is expected to assist the planner in making decisions. The correction log is a record of the actions corrected by the planner for each step. Each record of the correction log 125 shown in Figure 7 stores values such as the plan number 2501, step 2502, product 2503, process 2504, start time before correction 2505, and start time after correction 2506.

The plan number 2501 is identification information that uniquely identifies the plan and is the same as the plan number 2211. The step 2502 is the order of actions revised by the planner in the plan with the corresponding plan number, starting from 1 for each plan number. The product 2302 is identification information that uniquely identifies the product to be revised, the process 2303 is identification information that uniquely identifies the process to be revised, the pre-revised start time 2505 is the start time t _ij before the revision by the planner, i.e., the start time determined by the planning support device 10 when the plan was created, and the revised start time 2506 is the start time t _ij revised by the planner. The revision log 125 stores information indicating the importance of each revision, since the planner is likely to make important revisions, for example, revisions that have a large impact on other decision variables, first.

The contents stored in the storage device 120 have been explained above, but since the evaluation results 125 and the weights 126 of each evaluation index, among the various data stored in the storage device 120, are information generated by the present invention, their specific contents will be explained separately.

Returning to Figure 1, the planning support device 10 in Figure 1 uses the various functions shown in the central processing unit 110 to perform calculations using various data stored in the storage device 120, and as a result, obtains intermediate products and final deliverables. The various functions shown in the central processing unit 110 are processed chronologically as shown in the flow in Figure 8.

As shown in Figure 8, the automatically generated plan 123 and the manually revised plan 124 are input to the central processing unit 110, and appropriate information is displayed on an output device 140 such as a monitor. The planner M can then input appropriate instructions from the input device 160, thereby helping to optimize the functions of the automatic planning processing unit 115.

In the internal processing of the central processing unit 110, the evaluation processing unit 111 first inputs the automatically generated plan 123 and the manually revised plan 124. Here, the automatically generated plan 123 was automatically created by the automatic planning processing unit 115, and the manually revised plan 124 was revised by the plan creator M after looking at past operating results and taking into account his or her own know-how, and the plant, etc. was ultimately operated according to the manually revised plan 124. The evaluation processing unit 111 inputs the evaluation results of the automatically generated plan 123 and the manually revised plan 124 for a certain period in the past. Note that in the following explanation, it is assumed that the evaluation results of the automatically generated plan 123 and the manually revised plan 124 on a monthly basis for approximately six months have been input.

The evaluation processing unit 111 evaluates two plans: the automatically generated plan 123 generated by the automatic planning processing unit 115, and the manually revised plan 124 that was manually revised by the planner M. The evaluation utilizes the constraint C violation checking function and evaluation index value E calculation function built into the automatic planning processing unit 115. The evaluation results are stored and accumulated as 125 in the storage device 120.

The detection processing unit 112 presents the evaluation results 125 to the planner M via a comparative visualization 130 on an output device 140 such as a monitor. The proposal processing unit 113 compares and evaluates the automatically generated plan 123 and the manually revised plan 124, and when the difference in evaluation values becomes large, it presents an update recommendation 131 for adjusting the weights of the evaluation indexes to the planner M via an output device 140 such as a monitor. Depending on the presented content, the planner M can reflect in subsequent processing factors that are deemed to require adjustment and their magnitude by appropriately providing them to the central processing unit 110 via the input device 160.

When the difference in evaluation values becomes large, the adjustment processing unit 114 searches for and determines the weights of the evaluation indexes that will reduce the difference in evaluation values. This process only searches for weight values, and does not change the weight values incorporated into the automatic generation during operation. The weights of the evaluation indexes determined with the approval of the planner M are stored and accumulated in the storage device 120 as 126, and are also reflected in the next plan in the automatic planning processing unit 115. The determined weights of the evaluation indexes are appropriately presented to the planner M via a monitor screen or the like using the comparison visualization unit 130.

Figure 9 is a flowchart specifically illustrating the various functions shown in Figure 8. In the flow of Figure 9, processing steps S100 to S400 are processing in the evaluation processing unit 111 and detection processing unit 112, processing steps S500 and S510 are processing in the proposal processing unit 113, processing steps S520 to S540 are processing in the adjustment processing unit 114, and processing step S600 is processing in the detection processing unit 112.

First, we will explain the processing in the evaluation processing unit 111 and the detection processing unit 112. In the flow of Figure 9, first, in processing step S100, one or more sets of a manually revised plan prepared by the planner M and an automatically generated plan generated by the automatic plan preparation processing unit 115 are prepared.

Next, in processing step S200, the evaluation processing unit 112 accepts the manually corrected plan and the automatically generated plan and calculates the evaluation values of the evaluation indices. Below, the evaluation indices and their evaluation values for the manually corrected plan and the automatically generated plan are explained in detail using examples.

Here, we will assume constraint C and evaluation index value E as evaluation indicators. Constraint C is an item that should be improved, such as "must not be done, should be avoided if possible" (a negative evaluation factor, so to speak), while evaluation index value E is an item that should be recommended, such as "it is desirable to maintain it for as long or as large as possible" (a positive evaluation factor, so to speak). The evaluation value is the numerical representation of these positive and negative evaluation indicators.

In the example shown below, three cases, constraint condition C01 to constraint condition C03, have been set for constraint condition C, as shown in the number # and classification column on the left side of Figure 10, and three cases, evaluation index value E01 to evaluation index value E03, have been set for evaluation index value E.

For example, these are as follows: Constraint C01 is a constraint that in each process, a process for another product cannot begin until the process for the current product is completed; constraint C02 is a constraint that in each process, work cannot be performed on multiple products; and constraint C03 is a constraint that production must be completed by the delivery date set for each product.

Regarding the evaluation index value E, evaluation index value E01 highly evaluates plans with a small optimal inventory days in terms of inventory days, evaluation index value E02 highly evaluates plans with a large number of consecutive productions of the same product in terms of the number of consecutive productions of the same product, and evaluation index value E03 highly evaluates plans with as short a changeover time as possible within a day in terms of the rate of single item production per day.

Looking at the three cases of constraints C01 to C03, the smaller the value of each constraint, the better the plan. Looking at the three cases of evaluation indexes E01 to E03, the smaller the value of each evaluation index, the better the plan.

Returning to Figure 9, in processing step S200, as a result of processing by the evaluation processing unit 111, the plan evaluation shown in Figure 10 is obtained as a result of evaluating two plans: the automatically generated plan 123 generated by the automatic planning processing unit 115, and the manually revised plan 124 manually revised by the plan creator M.

Figure 10 shows a chronological summary of the amount of violation of multiple constraints C and the evaluation results of multiple evaluation index values E for each month. The symbol ● represents the automatically generated plan, and the symbol ▲ represents the manually revised plan. In Figure 10, the amount of violation of multiple constraints C and the evaluation results of multiple evaluation index values E are compared chronologically for each month of the four months from January to April shown on the horizontal axis, with zero violation amount on the vertical axis indicating no violation. These evaluation results are stored and accumulated in the evaluation results 125 of the storage device 120. Note that hereafter, the amount of violation of constraints C and the evaluation results of multiple evaluation index values E may be referred to as evaluation values.

Note that the trends in Figure 10 presented by the detection processing unit 112 in Figure 9 represent various events. For example, it may be detected that in April the difference in evaluation value between the automatically generated plan shown by the thick line and the manually revised plan shown by the thin line has increased, and as shown in Figure 10, the results of test plans generated in April, May, and June by adjusting the weights of the evaluation indicators may also be shown.

Here, testing means that when the difference in evaluation values becomes large, the adjusted evaluation index weights are applied to the automatically generated plan to generate a plan, the newly generated plan is evaluated, and the evaluation results of the plan and the manually revised plan are displayed alongside the automatically generated plan before adjustment to the plan creator M.

Furthermore, the following can be seen from the trends in the amount of violation of constraint C and the fluctuations in evaluation index E shown in Figure 10. In this example, the automatically generated plan generally generates a better plan than the manually revised plan. The amount of violation of constraints C01, C02, and C03 is greater in the manually revised plan than in the automatically generated plan, indicating that a certain degree of violation is tolerated for these constraints. In particular, the amount of violation in the manually revised plan for constraint C03 tends to increase, suggesting that this constraint is not important. The evaluation indexes E01, E02, and E03 are similar between the automatically generated plan and the manually revised plan.

At this time, the detection processing unit 112 performs comparative visualization 130 on the output device 140, and the evaluation results of the automatically generated plan 123 and the manually corrected plan 124 stored in the storage device 120 up to this point are displayed in chronological order in a format such as that shown in Figure 10.

Since some of the events detected by the detection processing unit 112 should be reflected in the automatic plan by the automatic planning processing unit 115, the proposal processing unit 113 compares and evaluates the automatically generated plan and the manually revised plan, and when the difference in evaluation values becomes large, it presents the planner M with a suggestion to adjust the weights of the evaluation indicators. This process corresponds to processing step S300 in the flow of Figure 9.

In the evaluation process of processing step S300, the difference in the evaluation results of the automatically generated plan and the manually revised plan is calculated and stored and accumulated in the storage device 120 along with the evaluation results 125. Here, the difference in the evaluation results, for example, the difference di in the evaluation results of the i-th plan, is calculated using equation (1).

In equation (1), pj is the value of the jth evaluation index in the automatically generated plan, and pj' is the value of the jth evaluation index in the manually revised plan. When comparing the quality of a plan, the balance of priorities between different evaluation indexes has a greater impact than the magnitude of the evaluation index value, so we used cosine similarity instead of Euclidean distance.

In the evaluation process of processing step S400, the detection processing unit 112 uses equation (2) to calculate the variance of the difference between the evaluation results of the automatically generated plan and the manually revised plan. Specifically, the variance value α is calculated from the difference in the evaluation values of the plan set calculated in the preparatory stage and the difference in the evaluation results of the latest plan, in accordance with Hotelling's theory, using equation (2). Here, d0 and σ0 are the mean and standard deviation of the differences in the evaluation values of the plan set calculated in the preparatory stage. Also, dn is the difference in the evaluation results of the latest plan. This completes the processing in the evaluation processing unit 111 and detection processing unit 112.

Next, in the processing by the proposal processing unit 113, first in processing step S500, the variation value α is compared with the threshold value k, and if it is greater than or equal to the threshold value, the process proceeds to processing step S510, and if it is less than the threshold value, the process proceeds to processing step S600. Since the variation value α follows a chi-squared distribution, the threshold value k = 2.71 is set, which is deemed to indicate a change with a probability of 90% or more. In processing step S510, the proposal unit 113 recommends via the output device 140 that it is time to update.

Next, in the processing by the adjustment processing unit 113, first in processing step S520, the weights of the evaluation indexes are adjusted so that they approach the evaluation indexes of the revised plan. Specifically, the following inverse optimization problem is solved to adjust the weights of the evaluation indexes. The calculated weights of the evaluation indexes are stored and accumulated in the storage device 120 as 126 and are reflected in the next processing by the automatic plan creation unit 115.

The calculation formula (objective function) f(x) for the automatically generated plan score can be expressed as equation (3) using the plan value x, the calculation formula fi(x) for the i-th evaluation index, and the weight wi.

The inverse problem E(w) is to find the weight w of the unknown evaluation index from the planning value x ^* obtained by the manual correction planning, and can be expressed by equation (4).

The inverse problem E(w) is solved using a local search method with the initial solution as the currently set weight of the evaluation index. The planned value x is obtained by automatic planning with the weight w of the evaluation index set.

Next, in processing step S530, the automatic planning processing unit 115 solves the plan optimization problem while reflecting the adjusted weights and outputs the result as a test plan. The test planning process here applies the adjusted evaluation index weights to the automatic plan to generate a plan. Then, for that test plan, the automatic planning processing unit 115 uses its constraint condition C violation checking function and evaluation index value E calculation function to calculate an evaluation index value.

In processing step S540, the automatic planning processing unit uses equation (1) to calculate the difference between the test plan generated in processing step S530 and the manual correction plan received in processing step S200.

The calculated test results are displayed in the detection processing unit 112 in a chronological order, with each evaluation index value for the automatically generated plan and the manually revised plan, and the fluctuation value of the difference between the automatically generated plan and the manually revised plan also displayed in a chronological order. Based on this display, the plan creator M can confirm the test results. Note that the adjusted weights may be reflected directly in the device for the self-running generation plan, rather than being used for the test plan.

In processing step S530, the new plan (e.g., test plan) obtained by solving the plan optimization problem while reflecting the adjusted weights should be called a second automatically generated plan to distinguish it from the original automatic plan, and the difference between the evaluation results of the second automatically generated plan and the evaluation results of the manually modified plan should be called a second difference value.

Figure 11 shows the results after correction. The dotted line shows the test results, which show how the weighting of the evaluation values for the automatically generated plan and the manually corrected plan in Figure 10, which were adjusted to reflect the evaluation values from January to the current month, was reflected in the evaluation values from April to June.

In this example, the weights of each evaluation index (constraint C and evaluation index value E) for the automatically generated plan and the manually revised plan before weight adjustment, i.e. when the evaluation values in Figure 10 were obtained, are represented as "1.0." Furthermore, the weights of constraints C01 and C03 and evaluation index values E01 and E03 have been reduced, and the weight of evaluation index value E02 has been increased.

In this example, the test results for April and May are closer to the manually revised plan than the automatically generated plan, indicating an improvement. In particular, the test results for constraint C03 and evaluation index value E03 in April show that they are approaching the manually revised plan. However, the test results for June show that they are deviating from the manually revised plan and have deteriorated.

Figure 12 shows an example screen that recommends the variance value of the difference between the automatically generated plan and the revised plan. It shows an enlarged view of the individual case with constraint C and evaluation index value E in Figure 10, and in response to the increasing deviation between the automatically generated plan before adjustment and the manually revised plan for the previous three months, the plan creator M is notified that the deviation has increased as of April and a weight adjustment is recommended. As a result, the deviation between the automatically generated plan after adjustment and the manually revised plan for the last three months has improved, as shown in the test results.

More specifically, the vertical axis in Figure 12 represents the difference between the manually revised plan and the automatically generated plan, or the difference between the manually revised plan and the test plan, and ranges from 0 to 1, with 0 meaning there is no difference and they match. Furthermore, by converting the variation value to the difference from the manually revised plan using equation (2), the difference where there is a 90% chance of change (= 0.412, variation value = 2.71) and the difference where there is a 99% chance of change (= 0.500, variation value = 6.63) are color-coded.

In April, the difference in evaluation value between the automatically generated plan and the manually revised plan was 0.415 (variation value = 2.80), slightly exceeding the threshold, so the results for test plans generated by adjusting the weights of the evaluation indicators for April, May, and June are also shown. The difference in evaluation value between the test plan generated by adjusting the weights in April and the manually revised plan was 0.163 (variation value = 1.00), below the threshold and showing a significant improvement.

The overall difference from the manually revised plan was difficult to discern from the trend in the amount of violation of constraint C and the fluctuations in evaluation index value E in Figure 10, but this becomes clearer in Figure 12. We can see that the difference from the manually revised plan was particularly large in April. This is thought to be due to the influence of constraint C03 and evaluation index E03. By adjusting the weights of the evaluation index in accordance with the manually revised plan as of April and generating a test plan, we were able to create plans for both April and May that differed less from the original automatically generated plan.

According to the present invention described above, it is possible to evaluate the difference between the automatically generated plan created by the automatic planning processing unit 115 and the manually revised plan revised by the planner, and to assist in optimizing and adjusting the functions of the automatic planning processing unit 115. This makes it possible to achieve sustained, stable production by automatically updating equipment at appropriate times rather than updating it unnecessarily.

In the above explanation, we have shown an example in which the invention is configured as a support device that presents information to the planner at appropriate times and proceeds with processing in response to the planner's instructions, but it is also possible to configure an automatic plan creation device that presents information but proceeds with processing automatically.

10: Planning support device 110: Central processing unit 111: Evaluation processing unit 112: Detection processing unit 113: Proposal processing unit 114: Adjustment processing unit 115: Automatic planning processing unit 121: Constraint conditions and objective function (planning conditions)
122: Plan-related information 123: Automatically generated plan result 124: Manually corrected plan result 125: Evaluation result 126: Evaluation weight 120: Storage device 130: Program 150: Memory 140: Output device 160: Input device

Claims (12)

A planning device in which a planner modifies an automatically generated plan and obtains a manually modified plan as a final plan,
a storage device that stores an automatically generated plan created by an automatic planning processing unit and a manually corrected plan obtained by correcting the automatically generated plan by a planner;
a central processing unit including an evaluation processing unit that evaluates a plurality of predetermined evaluation indexes for the automatically generated plan and the manually corrected plan, a detection processing unit that determines a difference between an evaluation result of the automatically generated plan and an evaluation result of the manually corrected plan, and a proposal processing unit that, when the difference is greater than a predetermined threshold, recommends that it is time to update the adjustment of the weights of the evaluation indexes to minimize the difference;
an output device that presents the adjustment of the weight of the evaluation index recommended by the proposal processing unit to the planner. The plan creation device according to claim 1,
The plan creation device is characterized in that the evaluation processing unit obtains, as the evaluation result, an evaluation value that quantifies a predetermined evaluation index for each of the automatically generated plan and the manually corrected plan. 3. The plan creation device according to claim 2,
The planning device is characterized in that the evaluation indexes include a plurality of sets of constraint conditions that are matters to be improved and evaluation indexes that are matters to be recommended. The plan creation device according to claim 1,
The plan generation device is characterized in that the central processing unit includes an adjustment processing unit that adjusts weights of the evaluation indexes of the automatically generated plan so that the weights approach the evaluation indexes of the manually corrected plan. 5. The plan creation device according to claim 4,
The automatic planning processing unit creates a second automatically generated plan based on the adjusted weights, calculates a second difference value which is the difference between the evaluation result of the second automatically generated plan and the evaluation result of the manually corrected plan, and presents at least the second difference value in chronological order on the output device. The plan creation device according to claim 1,
The planning device is characterized in that the output device presents the evaluation results of the automatically generated plan and the evaluation results of the manually corrected plan in chronological order. The plan creation device according to claim 1,
further comprising an input device for inputting an instruction for adjustment according to the content presented by the output device;
The plan creation device is characterized in that the central processing unit acquires instructions input to the input device and proceeds with processing based on the instructions. A planning method executed by a planning device in which a manually revised plan obtained by a planner revising an automatically generated plan is used as a final plan,
a central processing unit evaluates a plurality of predetermined evaluation indexes for the automatically generated plan and a manually corrected plan obtained by correcting the automatically generated plan by a plan creator, calculates a difference between the evaluation result of the automatically generated plan and the evaluation result of the manually corrected plan, and, if the difference is greater than a predetermined threshold, recommends that it is time to update the adjustment of the weights of the evaluation indexes to minimize the difference;
A planning method , characterized in that an output device presents to a planner the adjustments of weights of evaluation indexes recommended by said central processing unit . 9. The planning method of claim 8, further comprising:
a central processing unit for determining, as the evaluation result, an evaluation value obtained by quantifying a predetermined evaluation index for each of the automatically generated plan and the manually corrected plan; 10. The planning method of claim 9,
The planning method is characterized in that the evaluation indexes include a plurality of sets of constraints that are matters to be improved and evaluation indexes that are matters to be recommended. 9. The planning method of claim 8, further comprising:
a central processing unit that adjusts weights of the evaluation indexes of the automatically generated plan so that the weights approach the evaluation indexes of the manually corrected plan; 12. The planning method of claim 11,
the central processing unit creates a second automatically generated plan based on the adjusted weights, and calculates a second difference value that is a difference between the evaluation result of the second automatically generated plan and the evaluation result of the manually corrected plan;
The planning method is characterized in that the output device presents the second difference value .