Layout design support device, layout design support system, layout design support method and program

Abstract

Provide appropriate layout information. [Solution] The layout design support device 100 comprises an estimation means and an output means. The estimation means is a trained model for inferring layout information, which includes layout items used for designing the layout of a facility for manufacturing a product, from condition information, which includes condition items representing the conditions for manufacturing a product. The estimation means applies the condition information to the trained model generated by a neural network to estimate the layout information. The output means outputs the layout information estimated by the estimation means.

Description

[Technical Field]

[0001] This disclosure relates to a layout design support device, a layout design support system, a layout design support method, and a program. [Background technology]

[0002] Technologies that support the creation of facility layouts are known. For example, the facility layout system disclosed in Patent Document 1 generates genes using a tree structure that represents the arrangement of units, and determines the arrangement of units using genetic programming on the generated genes. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2001-22815 [Overview of the project] [Problems that the invention aims to solve]

[0004] For example, in block layout design for a factory, detailed constraints are imposed, such as the unit of parts delivery for the products to be manufactured in the factory, the frequency of parts delivery, and the planned production volume. Furthermore, in an actual factory, these constraints change the size of the units corresponding to workplaces. Therefore, it is necessary to design an appropriate layout based on various actual results, including productivity, quality, and delivery time. Genetic programming, such as that described in Patent Document 1, does not take these detailed constraints into account, and may present unrealistic and inappropriate layout information.

[0005] This disclosure has been made in view of the above issues and aims to provide a layout design support device, a layout design support system, a layout design support method, and a program that present appropriate layout information. [Means for solving the problem]

[0006] To achieve the above objectives, the layout design support device according to this disclosure comprises an estimation means and an output means. The estimation means is a trained model for inferring layout information, which includes layout items used for designing the layout of a facility for manufacturing a product, from condition information, which includes condition items representing the conditions for manufacturing a product. The trained model is generated by a neural network, and the condition information is applied to the trained model to estimate the layout information. The output means outputs the layout information estimated by the estimation means. [Effects of the Invention]

[0007] According to this disclosure, a trained model for inferring layout information used for designing the manufacturing layout from condition information, which includes condition items representing the conditions for manufacturing a product, is applied to a trained model generated by a neural network to estimate the layout information and output the estimated layout information. Therefore, appropriate layout information can be presented. [Brief explanation of the drawing]

[0008] [Figure 1] Diagram showing the configuration of the layout design support system according to the embodiment. [Figure 2] Figure 1 shows the functional configuration of the learning unit. [Figure 3] Figure 1 shows the functional configuration of the inference unit. [Figure 4] Figure 1 shows an example of the hardware configuration of the layout design support device. [Figure 5] Figure 1 shows an overview of the neural network that constitutes the trained model of the layout design support system. [Figure 6] A diagram showing an example of condition information and layout information related to the embodiment. [Figure 7] A diagram showing an example of a layout drawing related to the embodiment. [Figure 8] A diagram illustrating the method for estimating the best factory according to the embodiment. [Figure 9]Flowchart of the learning process executed by the learning unit shown in FIG. 1 [Figure 10] Flowchart showing the layout design support process executed by the inference unit shown in FIG. 1 [Figure 11] Flowchart of the inference process shown in FIG. 10

Mode for Carrying Out the Invention

[0009] (Embodiment) Hereinafter, a layout design support device, a layout design support system, a layout design support method, and a program according to an embodiment of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals.

[0010] When condition information indicating the manufacturing conditions of a product is input, the layout design support system according to the embodiment of the present disclosure estimates a factory in which the evaluation index is maximized when the product is manufactured under those conditions, and estimates the block layout of the manufacturing facilities in the estimated factory. Here, the condition information indicates various conditions in the manufacturing process, such as, for example, the unit of parts procurement, the frequency of parts procurement, the planned production quantity, etc. when manufacturing the product. The evaluation index is an index for evaluating the quality of the manufacturing process, such as the defect rate, the delay rate, the productivity, etc.

[0011] As shown in FIG. 1, the layout design support system 1 according to the embodiment includes a layout design support device 100 that outputs layout information of a factory suitable for manufacturing and the layout of manufacturing facilities in the factory according to the input condition information, an input unit 200 that supplies various information including the input conditions to the layout design support device 100, and a display device 300 that displays the layout information output by the layout design support device 100.

[0012] The input unit 200 supplies the input information to the layout design support device 100. The input unit 200 includes, for example, a keyboard operated by a user, a mouse, a communication interface device that receives input information from an external device, etc. The input information includes, for example, condition information, learning information for machine learning described later, etc.

[0013] The display device 300 displays the information output by the layout design support device 100. The display device 300 includes, for example, a liquid crystal display. The display device 300 is an example of a display element according to the present disclosure.

[0014] The layout design support device 100 is configured by a computer. The layout design support device 100 applies the condition information of the manufacturing equipment to be layout-designed to a learned model generated by machine learning to obtain layout information. The layout design support device 100 also has a learning function for generating a learned model.

[0015] The layout design support device 100 includes an arithmetic processing unit 110 that executes data processing, and a storage unit 120 that stores programs, learned models, etc. executed by the arithmetic processing unit 110.

[0016] The arithmetic processing unit 110 includes, for example, a CPU (Central Processing Unit). The arithmetic processing unit 110 functions as an information reception unit 111 that receives various information via an input unit 200, a control unit 112 that controls the processing of the layout design support device 100, a learning unit 113 that generates a learned model, an inference unit 114 that estimates the optimal layout of a factory and manufacturing equipment using the learned model, and an output unit 115 that outputs layout information indicating the estimated factory and layout, by executing the program stored in the program storage unit 121 of the storage unit 120.

[0017] The storage unit 120 includes, for example, a non-volatile semiconductor memory such as a flash memory or an EPROM (Erasable Programmable Read Only Memory). The storage unit 120 may also include a non-volatile memory including a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, and a DVD (Digital Versatile Disc).

[0018] The memory unit 120 includes a program memory unit 121 for storing programs, an information memory unit 122 for storing various types of information, and a trained model memory unit 123 for storing trained machine learning models.

[0019] The program storage unit 121 stores the program to be executed by the arithmetic processing unit 110. The program includes programs for executing learning processes, layout design support processes, drawing creation processes, etc., which will be described later. The drawing creation program can utilize a CAD program.

[0020] The information storage unit 122 stores the learning information acquired by the learning unit 113 and the inference information acquired by the inference unit 114. The learning information includes training data for applying machine learning processing to the machine learning model. This training data includes training data that includes multiple pairs of condition information, a factory suitable for manufacturing a product under the conditions indicated by that condition information, and layout information indicating the layout of the manufacturing equipment within that factory. The inference information indicates the manufacturing conditions for the manufacturing equipment for which the layout is to be designed. The information items are the same as those of the input information.

[0021] The trained model storage unit 123 stores the trained machine learning model. In this embodiment, the machine learning model is a neural network. The learning unit 113 generates a trained model by applying machine learning using the training information to the machine learning model. Note that the learning may be performed using other learning methods such as "unsupervised learning method," "reinforcement learning method," or "semi-supervised learning method."

[0022] As illustrated in Figure 5, the neural network that constitutes the machine learning model has an input layer composed of multiple neurons X, a hidden layer composed of neurons Y that receive signals output from each neuron X in the input layer, and an output layer composed of neurons Z that receive signals output from each neuron Y in the hidden layer and output layout information.

[0023] The information receiving unit 111 of the arithmetic processing unit 110 receives the input information. The user inputs condition information, including layout design condition items, via the input unit 200. As illustrated in Figure 6, the condition items include product name, parts delivery unit, parts delivery frequency, planned production quantity, finished product shipment unit, and finished product shipment frequency. The information receiving unit 111 receives the input condition information and stores the received condition information in the information storage unit 122 of the storage unit 120. The control unit 112 is responsible for controlling the operation of the layout design support device 100. The control unit 112 also executes the drawing creation program stored in the program storage unit 121 and creates a drawing of the layout indicated by the layout information.

[0024] As shown in Figure 2, the learning unit 113 includes a data acquisition unit 1131 that acquires learning information and a model generation unit 1132 that generates a trained model.

[0025] The data acquisition unit 1131 acquires training information. The model generation unit 1132 performs supervised learning based on the training information output from the data acquisition unit 1131 to generate a trained model. The generated trained model is stored in the trained model storage unit 123.

[0026] As shown in Figure 3, the inference unit 114 includes a data acquisition unit 1141 that acquires inference information, and an information inference unit 1142 that inputs the inference information into a trained model to estimate a suitable factory for manufacturing and a suitable layout of the manufacturing equipment within that factory, and outputs layout information. The layout information estimated by the inference unit 114 and the layout drawing generated by the control unit 112 are displayed on the display device 300 and used as a reference for designing the layout of manufacturing equipment for manufacturing products. The information inference unit 1142 is an example of the estimation means according to this disclosure.

[0027] Layout items included in the layout information include the factory name, parts storage area, assembly area, and productivity.

[0028] The output unit 115 shown in Figure 1 outputs the layout information estimated by the information inference unit 1142 to the display device 300. In addition, a layout drawing corresponding to the factory name included in the layout information may also be output to the display device 300 along with the estimated layout information. Further details will be described later. Note that the output unit 115 is an example of an output means according to this disclosure.

[0029] Next, an example of the hardware configuration of the layout design support device 100 will be described with reference to Figure 4. The layout design support device 100 in Figure 4 is implemented using a computer such as a personal computer or a microcontroller.

[0030] The layout design support device 100 comprises a processor 1001 that executes an operation program, which is a program for the operation of the layout design support device 100, which is connected to each other via a bus 1000; a memory 1002 that serves as the main memory area; an interface 1003 that realizes the communication function of the layout design support device 100; and a secondary storage device 1004 that stores the operation program for executing processing.

[0031] The processor 1001 is, for example, a CPU (Central Processing Unit). The various functions of the layout design support device 100 are realized when the processor 1001 reads the operation program stored in the secondary storage device 1004 into the memory 1002 and executes it.

[0032] Memory 1002 is a main memory device, for example, composed of RAM (Random Access Memory). Memory 1002 stores the operational program read by the processor 1001 from the secondary memory device 1004. Memory 1002 also functions as work memory when the processor 1001 executes the operational program.

[0033] Interface 1003 is an I / O (Input / Output) interface such as a serial port, USB (Universal Serial Bus) port, or network interface. Interface 1003 enables the communication function of the layout design support device 100.

[0034] The secondary storage device 1004 is, for example, flash memory, an HDD (Hard Disk Drive), or an SSD (Solid State Drive). The secondary storage device 1004 stores the operation program executed by the processor 1001, learning information, inference information, information received by the information receiving unit 111, and the trained model generated by the model generation unit 1132 in the learning unit 113.

[0035] Next, we will explain the overview of the neural network that makes up the trained model, referring to Figure 5. As illustrated, a neural network consists of an input layer, a hidden layer, and an output layer, each containing multiple neurons. Here, we assume there is one hidden layer, but the number of hidden layers is arbitrary.

[0036] Each neuron in the input layer X1~X n Each of the condition items indicated by the condition information is assigned to a different value. Here, as shown in Figure 6, the condition items indicated by the condition information are assumed to be the product name, parts delivery unit, parts delivery frequency, planned production quantity, finished product shipment unit, and finished product shipment frequency.

[0037] In this case, for example, neuron X1 is assigned the product name "room air conditioner", neuron X2 is assigned the product name "housing air conditioner", neuron X3 is assigned the product name "water heater", and so on. Also, for example, neuron X 11 The component delivery unit is "400 units", Neuron X 12 The component delivery unit is "500 units", Neuron X 13 A parts delivery unit of "600 units" will be allocated to this.

[0038] Similarly, for each neuron Z1 to Z in the output layer k different contents of each layout item indicated by the layout information are assigned. Here, as shown in FIG. 6, assume that the layout items are product name, factory name, part storage area, assembly area, finished product storage area, material flow, quality, productivity, and delivery date. Note that quality represents the defect rate and delivery date represents the delay rate. In this case, for example, the product name "room air conditioner" is assigned to neuron Z1, the product name "housing air conditioner" is assigned to neuron Z2,... Also, for example, the factory name "A" is assigned to neuron Z 11 , the factory name "B" is assigned to neuron Z 12 , the factory name "C" is assigned to neuron Z 13 ,...

[0039] The structure of this layout information has three components as the layout components: part storage, assembly area, and finished product storage, and it is assumed that the material moves from the part storage to the assembly area and then to the finished product storage.

[0040] As shown in FIG. 6, assume that the condition information indicates "room air conditioner" as the product name, "500 units" as the part intake unit, "2 times / day" as the part intake frequency, "1000 units / day" as the planned production quantity, "200 units" as the finished product shipment unit, and "3 times / day" as the finished product shipment frequency.

[0041] In this case, among the neurons X to which the condition item "product name" of the input layer is assigned, the input of neuron X1 to which "room air conditioner" is assigned is set to "1". Similarly, among the neurons X to which the part intake unit of the input layer is assigned, the input of neuron X 12 to which "500 units" is assigned is set to "1". Similarly, among the neurons X to which the same input item is assigned, the input of the neuron X corresponding to the condition information is set to "1", and the input of the other neurons X in the input layer is set to "0".

[0042] Similarly, among the neurons Z in the output layer, if the output of neuron Z1, which is assigned the product name "room air conditioner," is "1," then the product name of the layout item is "room air conditioner," and neuron Z1 is assigned the factory name "Factory A." 11 If the output is "1", the factory name in the layout item is "Factory A". Note that Factory A is an example of the attribute information of the manufacturing plant related to this disclosure.

[0043] Next, we will explain the method for generating training data included in the learning information. Training data is generated from a combination of data for each condition item and data for layout items that yielded favorable results for that condition item, based on past layout performance data.

[0044] The creator of the training data selects the most suitable combination of layout items for each conditional information. An example of a procedure for identifying the optimal factory based on condition information will be explained with reference to Figure 8. In the example in Figure 8, the optimal factory is identified from factories A to D based on five evaluation indicators: quality (representing the defect rate), delivery time (representing the delay rate), productivity, area, and material flow. As shown in Figure 8, 5 points are assigned to quality, 4 points to delivery time, 3 points to productivity, 2 points to area, and 1 point to material flow.

[0045] First, from the perspective of quality, which represents the defect rate, a factory with a small value is desirable, so factory A, which shows the smallest value of 50 among factories A to D, is selected. From the perspective of delivery time, which represents the delay rate, a factory with a small value is also desirable, so factory C is selected. From the perspective of productivity, a factory with a large value is desirable, so factory D is selected. From the perspective of area and material flow, a factory with a small value is desirable, so factory B is selected from the perspective of area, and factory A is selected from the perspective of material flow.

[0046] Factory A scores the best in both quality and material flow, resulting in a score of 5 points plus 1 point, for a total of 6 points. Similarly, Factory B scores 2 points, Factory C scores 4 points, and Factory D scores 3 points. Therefore, the optimal factory is Factory A, which scores the maximum of 6 points. The person responsible for generating the training data identifies the optimal factory from these perspectives and incorporates this information into the layout information that makes up the training data.

[0047] Next, we will explain the process of generating a trained model by applying training to a neural network, referring to Figure 9. Here, we assume that the conditional information and layout information constituting a single training dataset contain data for each item shown in Figure 6.

[0048] The learning process shown in Figure 9 is initiated when the information receiving unit 111 receives an operation from the user to instruct it to run.

[0049] The data acquisition unit 1131 in the learning unit 113 acquires a set of training data representing learning information (step S301). The training data consists of "1" and "0" information on the input layer side that identifies "product name", "parts delivery unit", "parts delivery frequency", "planned production quantity", "finished product shipment unit", and "finished product shipment frequency", and "1" and "0" information on the output layer side that identifies "product name", "factory name", "parts storage area", "assembly area", "finished product storage area", "material flow", "quality", "productivity", and "delivery date".

[0050] The model generation unit 1132 uses the training data output by the data acquisition unit 1131 to learn the relationship between the combination of "product name," "parts delivery unit," "parts delivery frequency," "planned production quantity," "finished product shipment unit," and "finished product shipment frequency," and the combination of "product name," "factory name," "parts storage area," "assembly area," "finished product storage area," "material flow," "quality," "productivity," and "delivery date," and uses weights W to select the optimal factory layout information. 11 ~W nm , V 11 ~V mk Adjust (step S302).

[0051] The model generation unit 1132 sequentially executes the learning process on multiple training data sets, and when the learning completion condition is met, the trained model is completed. The model generation unit 1132 saves the trained model to the trained model storage unit 123 (step S303) and terminates the training process.

[0052] Next, referring to Figure 10, we will explain a layout design support process that generates layout information using a trained model to assist in layout design.

[0053] The layout designer considers the product to be manufactured and the requirements for the manufacturing process, and sets the condition information. The layout designer inputs the set condition information into the layout design support device 100 via the input unit 200. Here, it is assumed that the condition information set as shown in Figure 6 is the product name "room air conditioner", the parts delivery unit "500 units", the parts delivery frequency "2 times / day", the planned production quantity "1000 units / day", the finished product shipment unit "200 units", and the finished product shipment frequency "3 times / day".

[0054] When the layout design support device 100 starts the layout design support process, the information receiving unit 111 determines whether or not it has acquired inference information (step S101).

[0055] If it is determined that inference information has been acquired (Step S101: Yes), the control unit 112 inputs the inference information to the inference unit 114 (Step S102). Then, it executes the inference process (Step S103). If it is determined that inference information has not been acquired (Step S101: No), the process in Step S101 is repeated.

[0056] In the inference process (step S103), the process shown in Figure 11 is performed. The data acquisition unit 1141 acquires inference information, specifically condition information (step S103a). The information inference unit 1142 applies the condition information to the trained model and obtains its output (step S103b). In the example in Figure 6, the information inference unit 1142 supplies "1" to neuron X representing the product name "room air conditioner", parts delivery unit "500 units", parts delivery frequency "2 times / day", planned production quantity "1000 units / day", finished product shipment unit "200 units", and finished product shipment frequency "3 times / day", and supplies "0" to other neuron X.

[0057] The neurons in the trained model that output "1" are assigned to the recommended layout information, which in other words represents the product name, factory name, parts storage area, assembly area, finished product storage area, material flow, quality, productivity, and delivery date.

[0058] The control unit 112 executes the drawing creation program stored in the program storage unit 121, and as illustrated in Figure 7, adjusts the size, shape, and arrangement of each area on the drawing so that the sizes of the components of the planned layout—in this example, the parts storage area, assembly area, and finished product storage area—match the estimated area sizes, and the lengths of the movement paths of materials match the estimated lengths, thereby forming the drawing. The arrow AL represents an example of a monorail path. The sum of the lengths of the three arrows AL equals the length of the monorail path.

[0059] The information inference unit 1142 outputs the estimated layout information (step S103c). The process then proceeds to step S104, as shown in Figure 10.

[0060] The output unit 115 outputs the layout information output by the information inference unit 1142 to the display device 300. The display device 300 displays the layout information and drawing output from the output unit 115 (step S104). Specifically, the display device 300 displays the product name, factory name, parts storage area, assembly area, finished product storage area, material flow, quality, productivity, and delivery date output from the output unit 115, as illustrated in Figure 6. It also displays the layout drawing as illustrated in Figure 7. In addition, a link may be provided to factory A shown in Figure 6, and when the user clicks that link, the layout drawing shown in Figure 7 may be displayed. The layout design support process is then terminated.

[0061] As described above, the layout design support system 1 according to the embodiment can estimate and present layout information used for layout design, which is estimated from condition information including condition items such as product name, parts delivery unit, and parts delivery frequency, using a trained model.

[0062] (modified version) In this embodiment, the layout information is output using a trained model learned by the model generation unit 1132 of the layout design support device 100. However, the system is not limited to this; a trained model may be acquired from outside the layout design support device 100, and the layout information may be output to the display device 300 based on this trained model.

[0063] In this embodiment, the blocks to be arranged are defined as three: a parts storage area, an assembly area, and a finished product storage area. However, there may be two or four or more blocks. In that case, the layout information should include the area of ​​each block and information on the movement of objects passing through all the blocks.

[0064] Furthermore, the various data in the storage unit 120 of the layout design support device 100 may be stored on an external cloud-type server.

[0065] Although preferred embodiments have been described in detail above, the invention is not limited to the embodiments described above, and various modifications and substitutions can be made to the embodiments described above without departing from the scope of the claims.

[0066] The various aspects of this disclosure are summarized below as an appendix. (Note 1) A trained model for inferring layout information, which includes layout items used for designing the layout of a facility for manufacturing a product, from condition information, which includes condition items representing the conditions for manufacturing a product, the model comprising estimation means for estimating the layout information by applying the condition information to a trained model generated by a neural network, An output means for outputting the layout information estimated by the estimation means, Equipped with, Layout design support device. (Note 2) The aforementioned condition items include the product name, the unit of parts delivery, and the frequency of parts delivery. The aforementioned layout item includes attribute information of the manufacturing area and the area of ​​the parts storage area. Layout design support device as described in Appendix 1. (Note 3) The layout design support device described in Appendix 1 or 2, The system includes a display device that displays the layout information output by the output means, Layout design support system. (Note 4) A trained model for inferring layout information used for designing the layout of a facility for manufacturing a product from condition information representing the conditions for manufacturing the product, wherein the condition information is applied to a trained model generated by a neural network to estimate the layout information. The estimated layout information is displayed on the display element. Layout design support method. (Note 5) On the computer, A trained model for inferring layout information used for designing the layout of a facility for manufacturing a product from condition information representing the conditions for manufacturing the product, wherein the condition information is applied to a trained model generated by a neural network to estimate the layout information. The estimated layout information is displayed on the display element. A program that executes a process. [Explanation of symbols]

[0067] 1 Layout design support system, 100 Layout design support device, 110 Arithmetic processing unit, 111 Information receiving unit, 112 Control unit, 113 Learning unit, 114 Inference unit, 115 Output unit, 120 Storage unit, 121 Program storage unit, 122 Information storage unit, 123 Trained model storage unit, 200 Input unit, 300 Display device, 1000 Bus, 1001 Processor, 1002 Memory, 1003 Interface, 1004 Secondary storage device, 1131, 1141 Data acquisition unit, 1132 Model generation unit, 1142 Information inference unit, AL Arrow.

Claims

1. A trained model for inferring layout information, which includes layout items used for designing the layout of a facility for manufacturing a product, from condition information, which includes condition items representing the conditions for manufacturing a product, the model comprising estimation means for estimating the layout information by applying the condition information to a trained model generated by a neural network, An output means for outputting the layout information estimated by the estimation means, Equipped with, Layout design support device.

2. The aforementioned condition items include the product name, the unit of parts delivery, and the frequency of parts delivery. The aforementioned layout item includes attribute information of the manufacturing area and the area of ​​the parts storage area. The layout design support device according to claim 1.

3. A layout design support device according to claim 1 or 2, The system includes a display device that displays the layout information output by the output means, Layout design support system.

4. A trained model for inferring layout information used for designing the layout of a facility for manufacturing a product from condition information representing the conditions for manufacturing the product, wherein the condition information is applied to a trained model generated by a neural network to estimate the layout information. The estimated layout information is displayed on the display element. Layout design support method.

5. On the computer, A trained model for inferring layout information used for designing the layout of a facility for manufacturing a product from condition information representing the conditions for manufacturing the product, wherein the condition information is applied to a trained model generated by a neural network to estimate the layout information. The estimated layout information is displayed on the display element. A program that executes a process.

Images