Mathematical optimization system, mathematical optimization method, and mathematical optimization program
The mathematical optimization system addresses input data inconsistencies in large-scale problems by dividing them into subproblems and using optimization algorithms to detect and correct issues, enhancing efficiency and reducing costs.
Patent Information
- Application Number
- PCT/JP2026/017414
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-16
- Filing Date
- 2026-05-01
- Publication Date
- 2026-11-19
Smart Images

Figure JP2026017414_19112026_PF_FP_ABST
Abstract
Description
Mathematical optimization system, mathematical optimization method, and mathematical optimization program
[0001] The present disclosure relates to a mathematical optimization system, a mathematical optimization method, and a mathematical optimization program.
[0002] Japanese Unexamined Patent Application Publication No. 2005-284959 describes a mode of reducing evaluation and judgment time by, in the assignment of manufacturing processes to a plurality of products, dividing the product manufacturing process into certain units ("constraint unit units") based on input information, extracting infeasible combinations according to constraint conditions for each "constraint unit unit", and removing the infeasible combinations from all combinations of manufacturing process assignments.
[0003] Japanese Unexamined Patent Application Publication No. 2024-58363 describes a mode of executing a multi-stage optimization process that generates an optimization problem for a production logistics plan based on acquired input information, and solves the generated optimization problem by gradually refining the granularity among time, space, or processes.
[0004] Japanese Unexamined Patent Application Publication No. 2007-183817 describes a mode in a scheduling apparatus for performing production scheduling of a production object consisting of a plurality of processes, in which the whole is modeled as an aggregate of subproblems, various problem settings are made for each subproblem to realize process scheduling from various perspectives, input data interpretation such as error checking of input data and conversion into a predetermined data structure is performed, and when there is an error in the input data, a message is output or re-input is prompted.
[0005] Japanese Unexamined Patent Application Publication No. 2002-236721 describes a mode of simultaneously optimizing all decision aspects included in the problem in a single model by decomposing the problem by item, then solving the production inventory optimization problem for each item independently of others, and alternately repeating this operation and adjustment for eliminating machine interference between items with respect to the solutions.
[0006] An object of the present disclosure is to provide a mathematical optimization system, a mathematical optimization method, and a mathematical optimization program that can efficiently find inconsistencies in input data even for large-scale mathematical optimization problems.
[0007] A mathematical optimization system in the first embodiment includes a processor and solves a mathematical optimization problem, wherein the processor acquires input data for the mathematical optimization problem, generates a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data, solves each of the generated plurality of subproblems by mathematical optimization, determines the feasibility of the subproblem, and outputs information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
[0008] The second embodiment of the mathematical optimization system is the first embodiment of the mathematical optimization system, wherein the mathematical optimization problem is a job shop scheduling problem for multiple jobs, and when the processor generates multiple subproblems, it generates a problem of creating a schedule for a certain job as one of the subproblems.
[0009] In the third embodiment of the mathematical optimization system, the processor generates a problem of creating a schedule for a job as a subproblem, virtually creates the input data relating to the job, and generates the subproblem based on the created virtual input data.
[0010] In the fourth embodiment of the mathematical optimization system, the processor determines the feasibility of the subproblem, and if it is determined to be unfeasible, it obtains the constraints that cause the unfeasibility, and creates inconsistency information of the input data based on the constraints that cause the unfeasibility for the subproblem.
[0011] A fifth embodiment of the mathematical optimization system is a mathematical optimization system of the first embodiment in which the input data includes data for defining decision variables included in a mathematical formula representing the mathematical optimization problem, and data for determining constants included in the mathematical formula.
[0012] The sixth embodiment of the mathematical optimization method is a mathematical optimization method for solving a mathematical optimization problem, wherein a processor in an information processing device performs the following processes: acquiring input data for the mathematical optimization problem; generating a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data; solving each of the generated plurality of subproblems by mathematical optimization; determining the feasibility of the subproblem; and outputting information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
[0013] The seventh embodiment of the mathematical optimization program is a mathematical optimization program for solving a mathematical optimization problem, wherein the program obtains input data for the mathematical optimization problem, generates a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data, solves each of the generated plurality of subproblems by mathematical optimization, determines the feasibility of the subproblem, and outputs information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
[0014] According to this disclosure, even in large-scale mathematical optimization problems, inconsistencies in input data can be efficiently detected.
[0015] This is a block diagram showing an example of the schematic configuration of a mathematical optimization system. This is a block diagram showing an example of the server hardware configuration. This is a block diagram showing an example of the server's functional configuration. This is a block diagram showing an example of the server's output unit's functional configuration. This is a flowchart showing an example of mathematical optimization processing. This is a diagram showing an example of product type data in the embodiment. This is a diagram showing an example of worker data in the embodiment. This is a diagram showing an example of work sequence data in the embodiment. This is a diagram showing an example of worker skill data in the embodiment. This is a diagram showing an example of the correspondence between product type and work set and task in the embodiment. This is a diagram showing an example of application information virtually created in the embodiment. This is a flowchart showing an example of mathematical optimization processing in the embodiment.
[0016] Hereinafter, with reference to the drawings, examples of embodiments for carrying out the technology of this disclosure will be described in detail.
[0017] <Overview of this Embodiment> In large-scale optimization systems that utilize mathematical optimization, including schedulers, it is common to encounter cases where an unexecutable optimization problem is constructed, preventing the desired results from being obtained. When there are no contradictions in the mathematical formula of the optimization problem itself, the main cause of unexecutability is inconsistency in the input data. Even if the cause lies in one of the input data, it is virtually impossible to cover all causes with rule-based validation checks performed before the optimization calculation. The input data itself is enormous and has complex dependencies, making it difficult and computationally expensive to identify the problematic parts from the entire optimization problem. Therefore, it is extremely laborious to initially gather valid input data, or to continuously maintain input data that can be correctly calculated as the situation changes.
[0018] Furthermore, even if a large-scale optimization problem is unfeasible, its IIS (Irreducable Infeasible Set: the smallest set of mutually contradictory equations in the optimization problem) is almost always much smaller than that of the original optimization problem. Therefore, before attempting to solve the large-scale overall optimization problem, solving numerous very small-scale optimization problems constructed by extracting only a portion of the input data allows for the efficient discovery of data inconsistencies that would make the problem unfeasible.
[0019] While there are conventional methods that divide a large optimization problem into smaller optimization problems and solve them in one go, these assume that the input data is correct and are not aimed at efficiently detecting data inconsistencies and reducing data preparation costs. For example, conventional methods divide the original problem into multiple subproblems to speed up computation, but the size of the subproblems cannot be reduced to very small in order to ensure that the sought local solutions do not deviate too far from the global optimal.
[0020] In this embodiment, it is possible to efficiently detect inconsistencies in input data, even in large-scale mathematical optimization problems.
[0021] <System Configuration> Figure 1 is an explanatory diagram showing an example configuration of the mathematical optimization system 2. In this embodiment, the mathematical optimization system 2 that provides the user with an optimal or near-optimal solution through mathematical optimization will be described. The mathematical optimization system 2 includes a server 10 and a user terminal 90. The server 10 and the user terminal 90 are connected via a network N.
[0022] Server 10 is, for example, a server computer capable of various information processing and information transmission / reception. Note that the device corresponding to Server 10 is not limited to a server computer; for example, it may be a personal computer or the like. In this embodiment, Server 10 functions as a device that provides the user with an optimal or near-optimal solution through mathematical optimization.
[0023] The user terminal 90 is a general-purpose computer such as a personal computer. In this embodiment, the user terminal 90 functions as a device that presents the user with information about inconsistencies in the input data.
[0024] In this embodiment, the server 10 provides the user with the optimal or near-optimal solution through mathematical optimization. However, the local user terminal 90 may also provide the user with the optimal or near-optimal solution through mathematical optimization. In other words, the distinction between the two is merely for convenience, and a single computer may perform the entire process.
[0025] <Server Configuration According to This Embodiment> Referring to Figure 2, the hardware configuration of the server 10 according to this embodiment will be described. An example of the server 10 is a personal computer or a server computer. As shown in Figure 1, the server 10 includes a CPU (Central Processing Unit) 20, a memory 21 as a temporary storage area, and a non-volatile storage unit 22. The server 10 also includes a display 23 such as a liquid crystal display, input devices 24 such as a keyboard and mouse, and a network interface 25 connected to a network. The CPU 20, memory 21, storage unit 22, display 23, input devices 24, and network interface 25 are connected to a bus 27. The CPU 20 is an example of a processor according to the disclosed technology.
[0026] The storage unit 22 is implemented by an HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory, etc. The storage unit 22, as a storage medium, stores the information processing program 30. The CPU 20 reads the information processing program 30 from the storage unit 22, expands it into memory 21, and executes the expanded information processing program 30. The information processing program 30 is an example of a mathematical optimization program.
[0027] Furthermore, the memory unit 22 stores problem data 32, which is a problem example of an optimization problem that the server 10 aims to solve, in a data format that the server 10 can process. The problem data 32 includes a set of elements that constitute the optimization problem to be solved, and which consists of multiple elements. The problem data 32 also includes constraints.
[0028] The elements that make up the set of elements to be solved refer to, for example, in the case of a job shop scheduling problem, the worker, the work date and time, and the type of product being worked on. Specific examples of the set of elements to be solved will be explained in the examples described later.
[0029] Incidentally, the server 10 according to this embodiment has a function to present the user with information about inconsistencies in the input data before performing the optimization calculation in the example of the optimization problem to be solved.
[0030] Next, with reference to Figure 3, the functional configuration of the server 10 according to this embodiment will be described. As shown in Figure 3, the server 10 includes an acquisition unit 40, a first determination unit 42, a generation unit 44, a second determination unit 46, an output unit 48, and a calculation unit 50. The CPU 20 executes the information processing program 30, thereby enabling the acquisition unit 40, the first determination unit 42, the generation unit 44, the second determination unit 46, the output unit 48, and the calculation unit 50 to function.
[0031] The acquisition unit 40 acquires input data for a mathematical optimization problem. The input data includes data for defining decision variables included in the mathematical formula representing the mathematical optimization problem, and data for determining constants included in the formula.
[0032] The first determination unit 42 determines whether there are any problems with the format of the acquired input data, based on a predetermined input data format.
[0033] The generation unit 44 generates multiple subproblems by extracting a part of the mathematical optimization problem based on the input data. Specifically, if the mathematical optimization problem is a job shop scheduling problem for multiple jobs, the generation unit 44 generates the problem of creating a schedule for a particular job as a subproblem when generating multiple subproblems.
[0034] More specifically, when the generation unit 44 generates a subproblem for creating a schedule for a job, it virtually creates input data related to the job and generates the subproblem based on the created virtual input data.
[0035] The second determination unit 46 solves each of the generated subproblems by mathematical optimization and determines the feasibility of the subproblem.
[0036] The output unit 48 outputs information based on the feasibility determination results for each of the multiple subproblems as inconsistency information for the input data.
[0037] Specifically, if the output unit 48 determines that a subproblem is unexecutable, it obtains the constraints that caused the unexecutability and creates inconsistency information for the input data based on the constraints that caused the unexecutability of the subproblem.
[0038] More specifically, as shown in Figure 4, the output unit 48 includes a setting unit 48A, a selection unit 48B, an execution unit 48C, and an extraction unit 48D.
[0039] The setting unit 48A first sets the set of elements to be solved included in the problem data 32 as the provisional element set. The setting unit 48A also adds 1 to the variable C, which is used to count the number of times the execution unit 48C (described later) has determined that the provisional theorem model is executable. Furthermore, if the execution unit 48C (described later) determines that the provisional theorem model is not executable, the setting unit 48A sets the set of elements obtained by subtracting the set of excluded elements (described later) from the provisional element set used when formulating the provisional theorem model as the provisional element set. In other words, the provisional element set is either the set of elements to be solved, or a set of elements that is a part of the set of elements to be solved.
[0040] The selection unit 48B selects an exclusion element set, which is a subset of the provisional element set, from the provisional element set. In this embodiment, the exclusion element set is defined as an element set from which, when its elements are excluded, the provisional theoretical model generated using the provisional element set becomes executable. Furthermore, if the selection unit 48B repeatedly performs the process of selecting an exclusion element set from the provisional element set, it will select an exclusion element set that is different from the exclusion element sets selected so far. In this embodiment, the selection unit 48B randomly selects an exclusion element set from the provisional element set.
[0041] The execution unit 48C generates a provisional theoretical model, which is a mathematical model formulated with the set of elements obtained by removing the set of excluded elements selected by the selection unit 48B from the provisional set of elements, and the constraint conditions. During this generation, conditions relating to the elements of the set of excluded elements are also removed from the constraint conditions. The execution unit 48C then determines whether the generated provisional theoretical model is executable or not. Specifically, for example, the execution unit 48C determines whether the provisional theoretical model is executable or not by using software called a solver for solving optimization problems to determine whether an executable solution can be obtained from the provisional theoretical model. For solving this optimization problem, known algorithms such as branch and bound, local search, tabu search, simulated annealing, quantum annealing, genetic algorithms, and swarm intelligence (e.g., particle swarm optimization and ant colony optimization) are used.
[0042] The setting unit 48A repeatedly performs the process of adding 1 to a variable C, which is used to count the number of times the execution unit 48C has determined that the provisional theoretical model is executable, until a predetermined condition (hereinafter referred to as the "termination condition") is met. Furthermore, if the execution unit 48C determines that the provisional theoretical model is not executable, the setting unit 48A repeatedly performs the process of setting the set of elements obtained by removing the set of excluded elements from the set of provisional elements used when formulating the provisional theoretical model as the set of provisional elements, until the termination condition is met.
[0043] Further, the selection unit 48B repeatedly executes the process of selecting an exclusion element set from the provisional element set until an end condition is satisfied. Further, the execution unit 48C generates a provisional mathematical model, which is a mathematical model in which an element set obtained by excluding the exclusion element set selected by the selection unit 48B from the provisional element set and constraint conditions are formulated, and performs a process of determining whether the generated provisional mathematical model is executable. The execution unit 48C repeatedly executes this process until the end condition is satisfied. As described above, when repeatedly executing the process of selecting an exclusion element set from the provisional element set, the selection unit 48B sets the exclusion element set to a set different from the exclusion element sets selected so far. That is, the execution unit 48C executes the process of determining whether the provisional mathematical model is executable a plurality of times while changing the exclusion element set.
[0044] Examples of the end condition for the above repetitive processing include a condition that the number of consecutive times the provisional mathematical model is determined to be executable is equal to or greater than a threshold, or a condition that the number of executions of the repetitive processing exceeds an upper limit value.
[0045] When the execution unit 48C determines that the provisional mathematical model is not executable, the extraction unit 48D extracts a list of conflicting constraint conditions (hereinafter referred to as "conflicting constraints") from that provisional mathematical model. In the present embodiment, when the execution unit 48C has determined that the provisional mathematical model is executable at least once and has determined that the provisional mathematical model is not executable at least once, the extraction unit 48D extracts a list of conflicting constraints from the provisional mathematical model determined to be non-executable. For the extraction of this list of conflicting constraints, for example, a solver function is used.
[0046] The output unit 48 creates and outputs inconsistency information including the list of conflicting constraints extracted as described above.
[0047] When there is no non-executable subproblem, the calculation unit 50 uses software called a solver for solving an optimization problem to perform a solution process on the problem example of the optimization problem to be solved.
[0048] Next, the operation of the server 10 according to this embodiment will be explained with reference to Figure 5. The CPU 20 executes the information processing program 30, thereby executing the mathematical optimization process shown in Figure 5. The mathematical optimization process shown in Figure 5 is executed, for example, when a user inputs an instruction to start execution.
[0049] In step S100 of Figure 5, the acquisition unit 40 acquires the input data for the mathematical optimization problem.
[0050] In step S102, the first determination unit 42 determines whether there are any problems with the format of the acquired input data, based on a predetermined input data format.
[0051] In step S104, the generation unit 44 generates multiple subproblems by extracting a part of the mathematical optimization problem based on the input data.
[0052] In step S106, the second determination unit 46 solves the target subproblem by mathematical optimization and determines the feasibility of the target subproblem.
[0053] In step S108, the second determination unit 46 determines whether the process in step S106 has been performed for all of the generated subproblems. If the process in step S106 has been performed for all of the generated subproblems, the process proceeds to step S110. On the other hand, if there are any subproblems for which the process in step S106 has not been performed, the process returns to step S106 and is repeated for those subproblems.
[0054] In step S110, the output unit 48 determines whether or not there are any unexecutable subproblems among the multiple subproblems. If there are any unexecutable subproblems among the multiple subproblems, the process proceeds to step S112. On the other hand, if there are no unexecutable subproblems, the process proceeds to step S114.
[0055] In step S112, if the output unit 48 determines that a certain subproblem is unexecutable, it obtains the constraints that caused the unexecutability of that subproblem. Based on the constraints that caused the unexecutability of that subproblem, the output unit 48 creates inconsistency information of the input data, displays it to the user terminal 90, and terminates the mathematical optimization process.
[0056] In step S114, the calculation unit 50 uses software called a solver to solve the optimization problem to perform the problem-solving process on an example of the optimization problem to be solved. The calculation unit 50 displays the result of the problem-solving process on the example of the optimization problem to be solved on the user terminal 90 and terminates the mathematical optimization process.
[0057] Next, we will describe some examples.
[0058] <Examples> Examples describe an application of the disclosed technology to a job shop scheduling problem. Specifically, the examples address an optimization problem in which a plan is formulated to complete a job shop scheduling problem as quickly as possible by efficiently assigning tasks related to the production of various product types to be manufactured within a given period to multiple workers.
[0059] The problem data 32 relating to the embodiment includes the variety data shown in Figure 6. As shown in Figure 6, the variety data associates the start time of work for each variety to be manufactured with the delivery time for that variety.
[0060] Furthermore, the problem data 32 in the embodiment includes worker data as shown in Figure 7. As shown in Figure 7, in the worker data, the start time and end time of work for each worker are associated with that worker.
[0061] Furthermore, the problem data 32 relating to the embodiment includes the work sequence data shown in Figure 8. As shown in Figure 8, the work sequence data includes, for each product type, the work required for manufacturing that product type, constraints on the work sequence, the time required for each work, and the waiting time required after each work. In the example in Figure 8, the arrows represent constraints on the work sequence, the time written near the arrows represents the waiting time after the work is completed, and the time written to the right of the work represents the time required for each work. Also, in the example in Figure 8, product type 1 is shown to have a waiting time of 0 minutes with no lower limit and no upper limit after the completion of work 1, followed by work 2, and then work 3 is shown to have a waiting time of 20 minutes with no lower limit and no upper limit after work 2. Also, in the example in Figure 8, product type 1 is shown to have a waiting time of 8 minutes for work 1, a waiting time of 5 minutes for work 2, and a waiting time of 3 minutes for work 3.
[0062] Furthermore, in the example in Figure 8, for variety 2, after the completion of task 1, a waiting period with a minimum of 0 minutes and no upper limit is observed before task 2 is performed, and after task 2, a waiting period with a minimum of 0 minutes and no upper limit is observed before task 4 is performed. Furthermore, in the example in Figure 8, for variety 2, after the completion of task 1, task 3 is performed without any waiting period, and after task 2, a waiting period with no minimum and an upper limit of 5 minutes is observed before task 4 is performed. Furthermore, in the example in Figure 8, for variety 2, the required time for task 1 is 7 minutes, the required time for task 2 is 10 minutes, the required time for task 3 is 3 minutes, and the required time for task 4 is 5 minutes.
[0063] Furthermore, the problem data 32 relating to the embodiment includes data on the skills of the workers shown in Figure 9. As shown in Figure 9, the data on the skills of the workers includes information indicating whether each worker is capable of performing the work for each product type and work pair. In the example in Figure 9, a circle indicates that the worker is capable of performing the work for that product type, and a blank space indicates that the worker is not capable of performing the work for that product type. In the example in Figure 9, worker A is capable of performing work 1 for product type 1, but not for work 2 and work 3, and is capable of performing work 1 and work 4 for product type 2, but not for work 2 and work 3.
[0064] Furthermore, as shown in Figure 10, in this embodiment, a pair of variety and work is treated as a task. In the example in Figure 10, the pair of variety 1 and work 1 is shown as task 1.
[0065] In the examples, the above data will be represented by the symbols shown below.
[0066]
[0067] Furthermore, the problem data 32 relating to the embodiment includes decision variables, constraints, and an objective function in order to formulate it as a mixed-integer linear programming problem. The decision variables are as follows: i and a i,w It is defined as follows.
[0068]
[0069] The constraints are shown by equations (1) to (8) below. Equation (1) represents the condition that each task must be started after the start time. Equation (2) represents the condition that each task must be completed by the deadline. Equation (3) represents the condition that the worker's start time must be observed. Equation (4) represents the condition that the worker's end time must be observed. Equation (5) represents the condition that each task must be assigned to one worker. Equation (6) represents the condition that each worker can only be assigned one task at a time. Equation (7) represents the condition that the order constraints between tasks must be observed. Equation (8) represents the condition that each task must only be assigned to a worker who is capable of performing that task.
[0070] …(6) M is a sufficiently large constant. o is a 0-1 variable necessary for the formulation. …(7)
[0071] The objective function is to complete the last task to be completed as early as possible, and is shown by equations (9) and (10) below.
[0072]
[0073] In this embodiment, the calculation unit 50 solves this problem example using an optimization algorithm such as the branch-and-bound method. This problem example can also be solved using a quantum computer.
[0074] This example is impossible to execute. This is because the time difference constraints in the work sequence data shown in Figure 8 above are contradictory. Specifically, if we set the time when work 1 is completed in product type 2 to 0 (the unit is minutes, but this is omitted), there is no suitable value for the start time of work 4 (there is no start time for work 4 that satisfies both the conditions of path A and path B below). And relatively simple data inconsistencies like this, where the cause of impossibility is contained within a single product type, are prone to occur.
[0075] Path A: Task 1 → 2 → 4 The earliest time to start Task 2 is "0". The earliest time to finish Task 2 is "10". The earliest time to start Task 4 is "10". Path B: Task 1 → 3 → 4 The latest time to start Task 3 is "0". The latest time to finish Task 3 is "3". The latest time to start Task 4 is "8".
[0076] The above is a cause of impossibility that is confined to variety 2 only; therefore, any calculation that includes even one instance of variety 2 will be impossible for the same reason. Consequently, by performing a very small-scale calculation, such as one instance of variety 2 in Figure 6, it can be determined that the calculation is impossible, and the time required to search for the cause of impossibility is also short. On the other hand, in actual calculations, a very large number of varieties are calculated at once, and feasibility can only be determined after performing a large-scale calculation, which increases computational costs and makes the causes of impossibility more complex. For searching for the cause of impossibility, methods described in patent documents (WO2024 / 018693), etc., are used.
[0077] Therefore, this section presents an example of using mathematical optimization during data checking to identify the cause of execution failure that is confined to each product category. Assume that the application information shown in Figure 6 is application information that changes each time, and the rest is master information that changes infrequently, and consider the cases of creating new master information or updating it. Furthermore, during the update process, as shown in Figure 11, a very small optimization problem is created as a subproblem by virtually creating application information corresponding to Figure 6 for each product category, and the process of solving this subproblem is repeated. Figure 11 shows an example of virtually creating application information for each product category, with a start time of 8:00 and a delivery time of 18:00. This makes it possible to detect complex data inconsistencies that are difficult to identify with normal validation checks in advance. Figure 11 shows an example of setting values by referring to the worker's shift and relaxing the conditions as much as possible.
[0078] Specifically, the processing routine shown in Figure 12 is executed as a mathematical optimization process.
[0079] In step S200, the acquisition unit 40 reads the update master information as input data for the mathematical optimization problem.
[0080] In step S202, the first determination unit 42 performs a validation check to determine whether there are any problems with the format of the acquired update master information based on a predetermined master information format.
[0081] In step S204, the generation unit 44 sets all product types on the work information sheet as a set of unchecked product types, and sets the execution failure reason message to an empty string.
[0082] In step S206, the generation unit 44 determines whether the set of unchecked varieties is an empty set. If the set of unchecked varieties is an empty set, the process proceeds to step S220. On the other hand, if the set of unchecked varieties is not an empty set, the process proceeds to step S208.
[0083] In step S208, the generation unit 44 selects one variety i from the set of unchecked varieties.
[0084] In step S210, the generation unit 44 generates virtual application information consisting only of variety i.
[0085] In step S212, the generation unit 44 extracts master information related only to variety i from the master information for updating.
[0086] In step S214, the second determination unit 46 defines a subproblem of the mathematical optimization problem from the generated virtual application information and the extracted master information, and solves it by mathematical optimization.
[0087] In step S216, the second determination unit 46 determines whether the subproblem defined in step S212 is executable. If the subproblem is executable, the process proceeds to step S206. On the other hand, if the subproblem is not executable, the process proceeds to step S218.
[0088] In step S218, the output unit 48 obtains the constraints that cause the subproblem to be unexecutable, adds the constraints that cause the subproblem to be unexecutable to the unexecutable cause message, and returns to step S206.
[0089] In step S220, the output unit 48 determines whether the execution failure reason message is an empty string. If the execution failure reason message is an empty string, the process proceeds to step S222. On the other hand, if the execution failure reason message is not an empty string, the process proceeds to step S224.
[0090] In step S222, the output unit 48 updates the master information with the update master information and terminates the processing routine.
[0091] In step S224, the output unit 48 converts the execution failure reason message into inconsistency information of the input data.
[0092] In step S226, the output unit 48 displays information about the inconsistency of the input data to the user terminal 90 and terminates the processing routine.
[0093] The processing routine described above prevents updating the master information if even one product type is unexecutable. This ensures that the products in the master information always maintain a state where there are no self-contained causes of execution failure. This allows for quick correction of data inconsistencies and enables calculations to be performed with a certain degree of elimination of causes of execution failure in large-scale calculations involving numerous product types. Furthermore, if an execution failure does occur, it becomes easier to identify the cause.
[0094] As described above, this embodiment allows for the efficient detection of inconsistencies in input data, even in large-scale mathematical optimization problems. Furthermore, by utilizing mathematical optimization in the validation check of input data within a system that solves large-scale mathematical optimization problems, inconsistencies in input data can be efficiently detected, significantly reducing the cost of creating the input data.
[0095] Furthermore, the scope of input data validation checks is expanded, allowing for more efficient preparation of input data. Additionally, by the time the desired optimization calculation is performed, input data with some unusable causes removed can be prepared, reducing the risk of rework in preparing the input data.
[0096] In the above embodiment, for example, the hardware structure of a processing unit that performs various processes, such as each functional unit of the server 10, can be the following types of processors. As mentioned above, these types of processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as a programmable logic device (PLD), such as an FPGA (Field Programmable Gate Array), whose circuit configuration can be changed after manufacturing, and a dedicated electrical circuit, such as an Application Specific Integrated Circuit, which has a circuit configuration specifically designed to perform a particular process.
[0097] A single processing unit may be composed of one of these various processors, or it may be composed of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, multiple processing units may be composed of a single processor.
[0098] Examples of configuring multiple processing units with a single processor include, firstly, a configuration where one or more CPUs and software combine to form a single processor, as exemplified by client and server computers, and this processor functions as multiple processing units. Secondly, a configuration using a processor that realizes the functions of the entire system, including multiple processing units, on a single IC (Integrated Circuit) chip, as exemplified by System-on-a-Chip (SoC). Thus, various processing units are configured, in terms of hardware structure, using one or more of the above-mentioned processors.
[0099] Furthermore, the hardware structure of these various processors can more specifically utilize an electrical circuit (circuitry) that combines circuit elements such as semiconductor elements.
[0100] Furthermore, although the above embodiment describes a configuration in which the information processing program 30 is pre-stored (installed) in the storage unit 22, the invention is not limited to this configuration. The information processing program 30 may be provided in a form recorded on a recording medium such as a CD-ROM (Compact Disc Read Only Memory), DVD-ROM (Digital Versatile Disc Read Only Memory), or USB (Universal Serial Bus) memory. Alternatively, the information processing program 30 may be provided in a form that is downloaded from an external device via a network.
[0101] The program of this application can be provided as a program product. A program product includes all forms of products for providing a program. For example, a program product includes a program provided via a network such as the Internet, and non-temporary computer-readable recording media such as CD-ROMs and DVDs on which the program is stored.
[0102] The following additional information is disclosed regarding the embodiments described above.
[0103] (Note 1) A mathematical optimization system that includes a processor and solves a mathematical optimization problem, wherein the processor: acquires input data for the mathematical optimization problem; generates a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data; solves each of the generated plurality of subproblems by mathematical optimization; determines the feasibility of the subproblem; and outputs information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
[0104] (Note 2) The mathematical optimization system according to Note 1, wherein the mathematical optimization problem is a job shop scheduling problem for multiple jobs, and the processor generates a problem of creating a schedule for a certain job as the subproblem when generating multiple subproblems.
[0105] (Note 3) The mathematical optimization system according to Note 2, wherein the processor generates a problem of creating a schedule for a job as the subproblem, virtually creates the input data relating to the job, and generates the subproblem based on the created virtual input data.
[0106] (Note 4) The mathematical optimization system according to any one of Notes 1 to 3, wherein the processor determines the feasibility of the subproblem, and if it is determined to be unfeasible, it obtains the constraints that cause the unfeasibility, and creates inconsistency information of the input data based on the constraints that cause the unfeasibility for the subproblem.
[0107] (Note 5) The input data is a mathematical optimization system as described in any one of Notes 1 to 4, which includes data for defining decision variables included in the mathematical formula representing the mathematical optimization problem, and data for determining constants included in the mathematical formula.
[0108] (Note 6) A mathematical optimization method for solving a mathematical optimization problem, wherein a processor in an information processing device performs the following processes: acquiring input data for the mathematical optimization problem; generating a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data; solving each of the generated plurality of subproblems by mathematical optimization; determining the feasibility of the subproblems; and outputting information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
[0109] (Note 7) A mathematical optimization program for solving a mathematical optimization problem, wherein the program causes a processor in an information processing device to perform the following processes: acquire input data for the mathematical optimization problem; generate multiple subproblems by extracting a part of the mathematical optimization problem based on the input data; solve each of the generated subproblems by mathematical optimization; determine the feasibility of the subproblems; and output information based on the feasibility determination result for each of the multiple subproblems as inconsistency information of the input data.
[0110] The disclosure of Japanese Patent Application 2025-082647 is incorporated herein by reference in its entirety.
[0111] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually described as being incorporated by reference.
Claims
1. A mathematical optimization system that includes a processor and solves a mathematical optimization problem, wherein the processor acquires input data for the mathematical optimization problem, generates a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data, solves each of the generated plurality of subproblems by mathematical optimization, determines the feasibility of the subproblem, and outputs information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
2. The mathematical optimization system according to claim 1, wherein the mathematical optimization problem is a job shop scheduling problem for multiple jobs, and the processor generates a problem of creating a schedule for a certain job as the subproblem when generating multiple subproblems.
3. The mathematical optimization system according to claim 2, wherein the processor generates a problem of creating a schedule for a job as a subproblem, virtually creates the input data relating to the job, and generates the subproblem based on the created virtual input data.
4. The mathematical optimization system according to claim 1, wherein the processor determines the feasibility of the subproblem, and if it determines that it is not feasible, it obtains the constraints that cause the feasibility, and creates inconsistency information of the input data based on the constraints that cause the feasibility of the subproblem.
5. The mathematical optimization system according to claim 1, wherein the input data includes data for defining decision variables included in a mathematical formula representing the mathematical optimization problem, and data for determining constants included in the mathematical formula.
6. A mathematical optimization method for solving a mathematical optimization problem, wherein a processor in an information processing device performs the following processes: acquiring input data for the mathematical optimization problem; generating a plurality of subproblems by extracting a part of the mathematical optimization problem based on the input data; solving each of the generated plurality of subproblems by mathematical optimization; determining the feasibility of the subproblems; and outputting information based on the determination result of the feasibility of each of the plurality of subproblems as inconsistency information of the input data.
7. A mathematical optimization program for solving a mathematical optimization problem, wherein the program causes a processor in an information processing device to perform the following processes: acquire input data for the mathematical optimization problem; generate multiple subproblems by extracting a part of the mathematical optimization problem based on the input data; solve each of the generated subproblems by mathematical optimization; determine the feasibility of the subproblems; and output information based on the feasibility determination results for each of the multiple subproblems as inconsistency information of the input data.
