Production plan support device and production plan support method

Abstract

A production plan support device according to the present disclosure comprises: a model storage unit storing a machine learning model that has learned a correlation between: order details and information relating to product production at the time when a production plan was formulated in the past using the same production equipment; and parameters of the production plan; and an estimation unit that uses the model to estimate parameters of a production plan on the basis of given order details and information relating to product production.

Description

Production planning support device and production planning support method

[0001] This disclosure relates to a production planning support device and a production planning support method.

[0002] Production planning can also be automated using mathematical optimization techniques. In mathematical optimization, constraints are expressed as an objective function, and a solution is sought that minimizes the value of the objective function within the range that satisfies the constraints (for example, Patent Document 1).

[0003] Creating production plans in a factory is a difficult task. Many constraints exist, depending on the order details, the factory's equipment, and the current production status of the products. These constraints include product delivery dates, upper limits on the number of units produced per hour, resource competition when producing multiple types of products (e.g., product B cannot be processed while product A is being processed), and limitations on production capacity. Under these constraints, skilled and knowledgeable personnel spend considerable time and effort developing production plans.

[0004] When formulating a production plan, it is often impossible to satisfy all constraints. In such cases, the acceptable amount of violation and the degree to which each constraint should be adhered to (priority) are determined for some constraints, and a production plan that violates some constraints is formulated. When using mathematical optimization methods in such cases, one approach is to include the amount of constraint violations in the objective function and optimize the objective function to minimize it. By adding weighting coefficients to the amount of violations in the objective function, the priority of each constraint can also be adjusted.

[0005] Japanese Patent Publication No. 2007-141000

[0006] Since the solution obtained varies greatly depending on the weight coefficient values, those in charge need to use their knowledge and experience, taking into account the production status of the product, to set appropriate weight coefficients. While it is possible to directly estimate the production plan using machine learning techniques, this method does not output information such as which constraints were prioritized when formulating the production plan. As a result, the production planning process tends to become a black box, and it becomes difficult to maintain and modify the model. In manufacturing, there is a need for technology that supports the production planning process so that the planning process can be understood.

[0007] The production planning support device disclosed herein solves the above problem by using data from past production planning to generate a machine learning model for estimating the parameters of the mathematical formulas used in production planning (such as the weight coefficients for the amount of constraint violations), and then using this model to estimate the parameters in mathematical optimization.

[0008] One aspect of this disclosure is a production planning support device comprising: a model storage unit that stores a machine learning model that has learned the correlation between order details and product production information and production plan parameters when production plans were previously formulated using the same production equipment; and an estimation unit that uses the model to estimate production plan parameters based on given order details and product production information.

[0009] This is a schematic hardware configuration diagram of a production planning support device according to one embodiment. This is a block diagram showing the schematic functions of a production planning support device according to the first embodiment. This is a block diagram showing the schematic functions of a production planning support device according to the second embodiment. This is a block diagram showing the schematic functions of a production planning support device according to another embodiment. This is a block diagram showing the schematic functions of a production planning support device according to another embodiment.

[0010] The embodiments of this disclosure will be described below with reference to the drawings. [First Embodiment] Figure 1 is a schematic hardware configuration diagram showing the main parts of a production planning support device according to the first embodiment of this disclosure. The production planning support device 1 of this disclosure can be implemented on a computer such as a personal computer attached to a control device that controls industrial machinery, or a personal computer, cell computer, fog computer 6, or cloud server 7 connected to the control device via a wired / wireless network. In this embodiment, an example is shown in which the production planning support device 1 is implemented on a personal computer connected to a control device that controls industrial machinery via a network.

[0011] The CPU 11 in the production planning support device 1 according to this embodiment is a processor that controls the production planning support device 1 as a whole. The CPU 11 reads the system program stored in the ROM 12 via the bus 22 and controls the entire production planning support device 1 according to the system program. The RAM 13 temporarily stores temporary calculation data, display data, and various data input from external sources.

[0012] The non-volatile memory 14 is composed of, for example, a memory backed up by a battery (not shown) or an SSD (Solid State Drive), and its stored state is maintained even when the power to the production planning support device 1 is turned off. The non-volatile memory 14 stores programs and data read from external devices 72 via the interface 15, programs and data input via the input device 71, and processing programs and data acquired from industrial machines 3. The data stored in the non-volatile memory 14 may be expanded into the RAM 13 when executed or used. In addition, various system programs, such as known analysis programs, are pre-written in the ROM 12.

[0013] Interface 15 is an interface for connecting the CPU 11 of the production planning support device 1 to external devices 72 such as USB devices. From the external devices 72, for example, pre-stored control programs and data related to the operation of each industrial machine 3 can be read. In addition, control programs and setting data edited within the production planning support device 1 can be stored in external storage means via the external devices 72.

[0014] Interface 20 is an interface for connecting the CPU 11 of the production planning support device 1 to a wired or wireless network 5. Industrial machinery 3, fog computers 6, cloud servers 7, etc., are connected to the network 5, and they exchange data with the production planning support device 1.

[0015] The display device 70 displays data, programs, and other data obtained as a result of their execution, which are loaded into memory, and output via the interface 17. The input device 71, consisting of a keyboard and pointing device, transmits commands and data based on operator operations to the CPU 11 via the interface 18.

[0016] Figure 2 is a schematic block diagram showing the functions of the production planning support device 1 according to the first embodiment of this disclosure. Each function of the production planning support device 1 according to this embodiment is realized by the CPU 11 of the production planning support device 1 shown in Figure 1 executing a system program and controlling the operation of each part of the production planning support device 1.

[0017] The production planning support device 1 according to this embodiment includes a data acquisition unit 100, a preprocessing unit 110, a learning unit 120, an estimation unit 130, and an output unit 140. In addition, the RAM 13 to non-volatile memory 14 of the production planning support device 1 is pre-configured with a model storage unit 200, which is an area for storing the model generated by the learning process performed by the learning unit 120.

[0018] The production planning support device 1 according to this embodiment operates in two modes: a learning mode that generates a model that learns the priority of each constraint in the production plan based on the production status of the product by performing machine learning learning processing based on the acquired data, and an estimation mode that estimates the priority of each constraint in the production plan based on the production status of the product by performing machine learning estimation processing based on the acquired data.

[0019] The operation of each function in learning mode is described below. In learning mode, the data acquisition unit 100 acquires data used for learning processing related to machine learning by the learning unit 120. The learning data acquired by the data acquisition unit 100 includes at least the order details from when a production plan was previously formulated using the same production equipment, information related to product production such as product information, material information, and machine information, and parameters of the production plan. The order details from when a production plan was previously formulated include, for example, the customer who ordered the product, the ordered product, whether or not there are options for the product, the number of products, and the delivery date of the product. The product information includes the product name, manufacturing days, materials required for manufacturing, machinery required for manufacturing, and the number of products that can be manufactured per predetermined period. The material information includes the material name, inventory quantity, price, and delivery days if ordered. The machine information includes the machine name, number of units in operation, and operating costs. In addition, the parameters of the production plan include weight coefficients assigned to the amount of constraint violations. The order details, product information, material information, and machine information include information related to constraints such as delivery dates, quantities, required materials, and required machinery. The data acquisition unit 100 may also acquire information related to the constraints of each piece of data. For example, an upper limit on the total operating cost of the machine could be further constraint information.

[0020] The data acquisition unit 100 may acquire data based on the user's operation of the input device 71 in learning mode. Alternatively, it may acquire data according to instructions from other computers such as the fog computer 6 and the cloud server 7 via the network 5. The data acquisition unit 100 may, for example, acquire input data from the input device 71. It may also acquire data from external devices 72. Furthermore, it may acquire data from other computers such as the fog computer 6 and the cloud server 7 via the network 5. The data acquisition unit 100 outputs the acquired data to the preprocessing unit 110.

[0021] In learning mode, the preprocessor 110 creates input data I and output data O based on the data acquired by the data acquisition unit 100, converting them (numerically, sampling, etc.) into a unified format to be handled by the learning unit 120. For example, the preprocessor 110 creates a set of input data I as state data and output data O as label data in a predetermined format used when supervised learning is performed in the learning unit 120. The preprocessor 110 then outputs the created input data I as state data and output data O as label data to the learning unit 120.

[0022] The preprocessing unit 110 creates order content data I1 as a numerical column of sets (orderer, product, product options, quantity, delivery date) from data related to the order details. At this time, the orderer, product, and product options can be individually identified beforehand by assigning a unique number (so-called ID), and then numerically converted using that number. For the delivery date, the date and time can be treated directly as a numerical column, or it can be numerically converted using an internal format of a computer or other device. If multiple product options are specified, the product options may become a numerical column of the numbers of multiple options. If there are multiple order details, the order content data I1 as input data will consist of multiple numerical columns.

[0023] The preprocessing unit 110 creates product data I2 as a numerical sequence of sets (product, manufacturing days, required materials, required machinery, and number of units that can be manufactured per predetermined period) from data related to product information, for example. At this time, required materials and required machinery can be assigned numbers that allow for individual identification of materials and machinery in advance, similar to the ordering party and product type, and then numerically converted using these numbers. For products manufactured using multiple materials and multiple machines, required materials and required machinery may be numerical sequences of multiple numbers. If there is multiple product information, the product data I2 as input data will be multiple numerical sequences.

[0024] The preprocessing unit 110 creates material data I3, which is a numerical sequence of sets (material, inventory quantity, price, and delivery time if ordered) from data related to material information. If there are multiple pieces of material information, the material data I3 as input data will consist of multiple numerical sequences.

[0025] The preprocessing unit 110 creates machine data I4, which is a numerical sequence of sets (machine, number of units in operation, operating cost) from data related to machine information, for example. If there are multiple pieces of machine information data, the machine data I4 used as input data will consist of multiple numerical sequences.

[0026] The preprocessing unit 110 may further create a numerical sequence from the constraint information data and use it as constraint data I5. Constraint data I5 may be, for example, a numerical sequence of lower and upper limits for each data point.

[0027] The production plan parameters acquired by the data acquisition unit include, for example, a weighting coefficient assigned to the amount of constraint violation. This weighting coefficient w i (i = 1 to k) are the amount of violation x related to each constraint in the objective function y shown in the following equation 1. erri These are the weighting coefficients (i = 1 to k). The preprocessing unit 110 creates a parameter O1, which is a numerical sequence of production plan parameters, from these production plan parameters.

[0028]

[0029] When creating a production plan, we consider equation 1 and the respective order details, product information, material information, and machine information as variables, and create constraint equations that define the relationships between these variables. Then, we adjust the values ​​of the variables related to the respective order details, product information, material information, and machine information to satisfy these constraint equations as much as possible. Finally, we find the solution that minimizes the value of the objective function (known mathematical optimization). At this time, the amount that violates the constraint defined in the constraint equation is the violation amount x related to the constraint. erri (i = 1 to k). In this disclosure, instead of directly estimating the production plan, this weighting coefficient w i Please note that we will be creating a machine learning model to estimate (i = 1 to k).

[0030] The learning unit 120 generates a model that learns the correlation between the input data I and output data O created by the preprocessing unit 110. The generated model is then stored in the model storage unit 200. The model generated by the learning unit 120 is a model that estimates the output data O created from production plan parameters based on the input data I = (order details data I1, product data I2, material data I3, machine data I4, constraint data I5) created from order details, product information, material information, machine information, and constraint information. This model can use, for example, a known regression equation, SVM (Support Vector Machine), multilayer neural network, Bayesian network, or Gaussian mixture model.

[0031] Next, we will explain the operation of each function in estimation mode. In estimation mode, the parameters of the production plan are estimated using the model generated by the learning unit 120 in learning mode.

[0032] In the estimation mode of the production planning support device 1, the data acquisition unit 100 acquires data used for estimation processing related to machine learning by the estimation unit 130. The estimation data acquired by the data acquisition unit 100 includes at least the order details, product information, material information, and machine information from when production plans were previously formulated for the same production equipment. The order details from when production plans were previously formulated include, for example, the customer who ordered the product, the ordered product, whether or not there are options for the product, the number of products, and the delivery date of the product. The product information includes the product name, manufacturing days, materials required for manufacturing, machinery required for manufacturing, and the number of products that can be manufactured per predetermined period. The material information includes the material name, inventory quantity, price, and delivery days if ordered. The machine information includes the machine name, number of units in operation, and operating costs. In addition to these, the data acquisition unit 100 may also acquire information related to the constraints of each of these data.

[0033] The data acquisition unit 100 may acquire respective data based on the user's operation of the input device 71 in the estimation mode. Alternatively, the data may be acquired in accordance with instructions from other computers such as the fog computer 6 and the cloud server 7 via the network 5. For example, the data acquisition unit 100 may acquire data input from the input device 71. Alternatively, the data may be acquired from the external device 72. Further, the data may be acquired from other computers such as the fog computer 6 and the cloud server 7 via the network 5. The data acquisition unit 100 outputs the acquired data to the preprocessing unit 110.

[0034] In the estimation mode, the preprocessing unit 110 creates input data I that is converted (digitized, sampled, etc.) into a unified format handled by the estimation unit 130, based on the data acquired by the data acquisition unit 100. For example, the preprocessing unit 110 creates the input data I as state data in a predetermined format for using a supervised learning model in the estimation unit 130. Then, the preprocessing unit 110 outputs the created input data I as state data to the estimation unit 130.

[0035] The estimation unit 130 executes estimation processing for production plan parameters (weighting coefficients assigned to the amount of constraint violation) using the model stored in the model storage unit 200, based on the estimation-purpose input data I received from the preprocessing unit 110. The estimation processing performed by the estimation unit 130 may be a known estimation process based on supervised learning. The estimation unit 130 outputs the estimation result of the production plan parameters to the output unit 140.

[0036] The output unit 140 outputs the production plan parameters as the result of the estimation processing performed by the estimation unit 130. The output unit 140 may, for example, display and output the production plan parameters to the display device 70. Further, if necessary, the parameters may be transmitted and output to an external computer such as the fog computer 6 or the cloud server 7 via the network 5.

[0037] A user utilizing the production planning support device 1, which has the above configuration, first inputs the order details, product information, material information, machine information, constraint information, and the values ​​of the production plan parameters set by the skilled worker at the time, as a single dataset into the production planning support device 1, which operates in learning mode. The user inputs the number of datasets necessary to generate a model with sufficiently high estimation accuracy, and then instructs the production planning support device 1 to generate a model. As a result, a model is generated for estimating the values ​​of the production plan parameters based on the order details, product information, material information, machine information, and constraint information.

[0038] From this point forward, when the user creates a new production plan, they operate the production planning support device 1 in estimation mode. They then input the order details, product information, material information, machine information, and constraint information that they will refer to when creating the production plan into the production planning support device 1. Based on this input data, the production planning support device 1 estimates and outputs the weighting coefficients to be assigned to the amount of constraint violations, which are parameters of the production plan.

[0039] The user then sets weighting coefficients for the output constraint violations against the objective function and performs mathematical optimization considering other constraints. This mathematical optimization can be done manually or using dedicated mathematical optimization application software. By referring to the weighting coefficients assigned to the output constraint violations, the user can learn which constraints should be prioritized in similar situations. Furthermore, if the weighting coefficients assigned to the output constraint violations appear to be abnormal values, the user can consider retraining the model.

[0040] The production planning support device 1 according to this embodiment, having the above configuration, can estimate the parameters of the objective function used in production planning using a model created based on information from past production planning. Unlike estimating the production plan itself, it is possible to confirm the objective function that forms the basis for formulating the production plan, thus clarifying the flow of production planning. This not only supports production planning but also assists in the learning of new employees, and makes it easier to maintain and evaluate the model.

[0041] [Second Embodiment] The following describes a production planning support device according to a second embodiment. The production planning support device 1 according to this embodiment has the same hardware configuration as the production planning support device 1 according to the first embodiment.

[0042] Figure 3 is a schematic block diagram showing the functions of the production planning support device 1 according to the second embodiment of this disclosure. Each function of the production planning support device 1 according to this embodiment is realized, similar to the production planning support device 1 according to the first embodiment, by the CPU 11 of the production planning support device 1 shown in Figure 1 executing a system program and controlling the operation of each part of the production planning support device 1.

[0043] The production planning support device 1 according to this embodiment includes a data acquisition unit 100, a preprocessing unit 110, a learning unit 120, an estimation unit 130, and an output unit 140. Furthermore, the RAM 13 to non-volatile memory 14 of the production planning support device 1 is pre-prepared with a model storage unit 200, which is an area for storing the model generated by the learning process performed by the learning unit 120. In addition, the RAM 13 to non-volatile memory 14 of the production planning support device 1 is pre-prepared with a history data storage unit 210, which is an area for storing the input data I used when estimation was performed in the past and the actual data related to the production plan formulated at that time.

[0044] The following describes the operation of each function in the learning mode of the production plan support apparatus 1 according to the present embodiment. In the learning mode, the data acquisition unit 100 according to the present embodiment acquires, as learning data, order details, product information, material information, machine information, constraint information, production plan parameters, and performance information related to the created production plan. The performance information related to the created production plan may include, for example, information related to the production plan created at that time, parameters of the production plan estimated at that time, and information related to the quality of the production state.

[0045] In the learning mode, the data acquisition unit 100 according to the present embodiment further acquires, as one information set, order details, product information, material information, machine information, constraint information, production plan parameters, and performance information related to the created production plan at a plurality of time points. For example, the data acquisition unit 100 acquires, as one information set, information when a production plan was created a predetermined number of times previously, in addition to the information when a production plan was created at a certain time. The data acquisition unit 100 outputs the acquired data to the preprocessing unit 110.

[0046] In the learning mode, the preprocessing unit 110 according to the present embodiment creates input data I and output data O that are converted (digitized, sampled, etc.) into a unified format handled by the learning unit 120, based on the data acquired by the data acquisition unit 100. For example, the preprocessing unit 110 creates a set of input data I as state data in a predetermined format used when performing supervised learning in the learning unit 120, and output data O as label data. Then, the preprocessing unit 110 outputs the created input data I as state data and the output data O as label data to the learning unit 120.

[0047] The input data I created by the preprocessing unit 110 according to the present embodiment includes order detail data I 0-1 , product data I 0-2 , material data I 0-3 , machine data I 0-4 , constraint data I 0-5 . In addition to this, it further includes order detail data I created based on information obtained when creating production plans up to a predetermined number of times previously j-1 , product data Ij-2 Material data I j-3 Machine data I j-4 Constraint data I j-5 Actual Data I j-6 (where j is 1 to a predetermined number of times) is included. Actual data I j-6 For example, information related to the production plan included in the actual performance information, the parameters of the estimated production plan, and information related to the quality of the production status can be converted into numerical columns in a predetermined order to create the output data O. The output data O is the same as in the first embodiment.

[0048] The learning unit 120 in this embodiment generates a model that learns the correlation between the input data I and output data O created by the preprocessing unit 110. The generated model is then stored in the model storage unit 200. The model generated by the learning unit 120 is a model that estimates the output data O created from the parameters of the production plan based on the input data I created from the order details, product information, material information, machine information, constraint information, and actual information from production planning up to a predetermined number of times prior, in addition to the order details, product information, material information, machine information, constraint information. For example, this model can use known regression equations, SVM (Support Vector Machine), multilayer neural networks, Bayesian networks, Gaussian mixture models, recurrent neural networks, etc.

[0049] Next, the operation of each function of the production planning support device 1 according to this embodiment in estimation mode will be described. In estimation mode, the parameters of the production plan are estimated using the model generated by the learning unit 120 in learning mode.

[0050] In this embodiment, the data acquisition unit 100 acquires order details, product information, material information, machine information, and constraint information as estimation data in estimation mode. The data acquisition unit 100 then outputs the acquired estimation data to the preprocessing unit 110.

[0051] Furthermore, the data acquisition unit 100 in this embodiment acquires actual performance information related to the production plan formulated based on the estimation results after the estimation processing by the estimation unit 130 has been performed. The data acquisition unit 100 outputs the acquired actual performance information to the preprocessing unit 110.

[0052] In this embodiment, the preprocessor 110, in estimation mode, creates input data I by converting (digitizing, sampling, etc.) the data acquired by the data acquisition unit 100 into a unified format to be handled by the estimation unit 130. For example, the preprocessor 110 creates input data I as state data in a predetermined format for use in the estimation unit 130 with a supervised learning model. The preprocessor 110 then outputs the created input data I as state data to the estimation unit 130.

[0053] Furthermore, the preprocessing unit 110 in this embodiment stores the created input data I in the history data storage unit 210. When actual information related to the production plan is later input, the preprocessing unit 110 creates actual data I6 based on the actual information and stores the created actual data I6 in association with the input data I stored in the history data storage unit 210. In this way, the history data storage unit 210 stores data related to past production plans as history. In the initial stage when history data has not yet been stored, a predetermined number of data can be obtained from the data used for learning and stored in the history data storage unit 210.

[0054] In this embodiment, when the estimation unit 130 receives the input data I for estimation from the preprocessing unit 110, it refers to the history data storage unit 210 to retrieve the order details I up to a predetermined number of times prior. j-1 Product Data I j-2 Material data I j-3 Machine data I j-4 Constraint data I j-5 Actual Data I j-6The number of times j (where j is 1 to a predetermined number of times) is obtained and added to the input data I. Then, based on the input data I to which the history has been added, the estimation process for production plan parameters (weight coefficients to be assigned to the amount of constraint violations) using the model stored in the model storage unit 200 is executed. The estimation process performed by the estimation unit 130 may be a known supervised learning estimation process.

[0055] The operation of the output unit 140 is the same as that of the output unit 140 in the first embodiment.

[0056] In this embodiment, the data acquisition unit 100 acquires, in learning mode, order details, product information, material information, machine information, constraint information, production plan parameters, and actual information related to the formulated production plan from multiple time periods as a single information set. For example, the data acquisition unit 100 acquires, in addition to the information from when a production plan was formulated at a certain time, the information from when production plans were formulated up to a predetermined number of times prior, as a single information set. The data acquisition unit 100 outputs the acquired data to the preprocessing unit 110.

[0057] The production planning support device 1 according to this embodiment, equipped with the above configuration, can estimate the parameters of the objective function used in production planning using a model created based on information from past production planning. When estimating these parameters, it refers to actual performance information from production planning up to a predetermined number of times prior. Depending on the data used for training, it becomes possible to perform flexible estimations that take history into account, such as prioritizing customers who have previously missed delivery deadlines in the current plan.

[0058] [Other Embodiments] In the embodiments described above, a production planning support device 1 equipped with a learning unit 120 and an estimation unit 130 was described. However, by preparing a model storage unit 200 that stores already learned models, for example, a production planning support device 1 equipped only with an estimation unit 130, as illustrated in Figures 4 and 5, can also be configured. In this case, the model storage unit 200 can also be configured on an external computer such as a fog computer 6 or a cloud server 7.

[0059] While embodiments of this disclosure have been described in detail above, this disclosure is not limited to the individual embodiments described above. These embodiments can be added, replaced, modified, partially deleted, etc., in any way that does not depart from the spirit of the invention or from the idea and intent of this disclosure derived from the claims and their equivalents. For example, the order of operations and processes in the embodiments described above are shown as examples only and are not limited thereto. The same applies when numerical values ​​or mathematical formulas are used in the description of the embodiments described above.

[0060] The following are annotations relating to embodiments of the present disclosure. (Annotation 1) A production planning support device (1) according to one aspect of the present disclosure includes a model storage unit (200) that stores a machine learning model which has learned the correlation between the order details and product production information and the parameters of the production plan when a production plan was previously formulated with the same production equipment, and an estimation unit (130) that uses the model to estimate the parameters of the production plan based on the given order details and product production information.

[0061] (Note 2) The production planning support device (1) according to another aspect of the present disclosure further comprises a learning unit (120) that generates a machine learning model that learns the correlation between the order details and product production information and the parameters of the production plan when a production plan was previously formulated with the same production equipment. (Note 3) The parameters of the production plan estimated by the production planning support device (1) according to another aspect of the present disclosure are weight coefficients for the amount of violation of each constraint defined in the mathematical optimization process for formulating the production plan. (Note 4) The model used by the production planning support device (1) according to another aspect of the present disclosure is trained using data from production plans formulated up to a predetermined number of times prior.

[0062] (Note 5) A production planning support method according to one aspect of the present disclosure includes the step of a computer estimating production planning parameters based on given order details and product production information, using a machine learning model that has learned the correlation between order details and product production information and production planning parameters when production planning has been formulated in the past with the same production equipment.

[0063] 1 Production planning support device 3 Industrial machinery 5 Network 6 Fog computer 7 Cloud server 11 CPU 12 ROM 13 RAM 14 Non-volatile memory 15, 17, 18, 20 Interface 22 Bus 70 Display device 71 Input device 72 External device 100 Data acquisition unit 110 Preprocessing unit 120 Learning unit 130 Estimation unit 140 Output unit 200 Model storage unit 210 History data storage unit

Claims

1. A production planning support device comprising: a model storage unit that stores a machine learning model that has learned the correlation between order details and product production information from past production planning using the same production equipment and the parameters of the production plan; and an estimation unit that uses the model to estimate the parameters of the production plan based on the given order details and product production information.

2. The production planning support device according to claim 1, further comprising a learning unit that generates a machine learning model that learns the correlation between the order details and product production information from when production plans were previously formulated using the same production equipment, and the parameters of the production plan.

3. The production planning support device according to claim 1, wherein the parameters of the production plan are weight coefficients for the amount of violation of each constraint defined in the mathematical optimization process for formulating the production plan.

4. The production planning support device according to claim 1, wherein the model uses data from production planning up to a predetermined number of times prior to the planning stage for training.

5. A production planning support method that performs the step of estimating production plan parameters based on given order details and product production information, using a machine learning model that has learned the correlation between order details and product production information and production plan parameters when a production plan was previously formulated using the same production equipment.

Images