Engineering survey method, engineering survey program, engineering survey device and creation method of learned model

Abstract

To provide an engineering survey method, an engineering survey program, an engineering survey device capable of executing engineering processing with high survey efficiency, and a creation method of a learned model.SOLUTION: An engineering survey method uses a parameter set having multiple parameters as sets, and comprises the steps of: after setting one parameter out of the multiple parameters, updating upper limit values and lower limit values of other parameters according to the setting of the parameter; sequentially setting other parameters in the range of the updated upper limit values and lower limit values; creating a preset number of parameter sets; predicting characteristics of products created by the multiple parameter set, using a learned model learned using the parameter sets as explanatory variables, and the characteristics as objective variables; and then selecting the parameter set having the characteristics suitable for a preset condition, out of the predicted characteristics.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a design exploration method, a design exploration program, a design exploration device, and a method for generating a trained model. [Background technology]

[0002] Conventionally, in product design, design exploration is performed to predict characteristics by performing analysis using the finite element method (FEM) or the like and set optimal design values ​​(see, for example, Patent Document 1). Figures 7 and 8 are diagrams for explaining conventional design exploration. For example, as shown in Figure 7, when two through holes 301 and 302 are formed in a plate material 300, the formation positions and diameters of the through holes are set by performing the above-mentioned analysis using the FEM or the like. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2017-054203 Summary of the Invention [Problem to be solved by the invention]

[0004] In the above-mentioned design exploration, it is preferable to analyze various patterns to obtain desirable design values. In this case, for example, the formation positions and diameters of the through holes are randomly generated and then analyzed. However, as shown in FIG. 8, patterns in which through holes 303 and 304 interfere with each other and connect the holes must be eliminated from the design. However, when patterns are generated randomly, it is necessary to determine whether or not to eliminate the pattern after the pattern generation. This determination of whether or not to eliminate the pattern reduces the exploration efficiency.

[0005] The present invention has been made in consideration of the above, and aims to provide a design exploration method, a design exploration program, a design exploration device, and a method for generating a trained model that can execute design processing with high exploration efficiency. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the object, the design exploration method of the present invention is a design exploration method that uses a parameter set consisting of a plurality of parameters, and after setting one of the plurality of parameters, updates the upper and lower limit values ​​of the other parameters each time in accordance with the setting of the parameter, and sequentially sets the other parameters within the range of the updated upper and lower limit values, creates a predetermined number of the parameter sets, predicts the characteristics of each product manufactured using the plurality of parameter sets using a trained model that has been trained using the parameter sets as explanatory variables and the characteristics as target variables, and selects a parameter set from the predicted characteristics that shows characteristics that are suitable for predetermined conditions.

[0007] In addition, in the design exploration method according to the present invention, the upper and lower limit values ​​of the parameters are calculated according to a relational expression between the parameters that is set in advance.

[0008] In addition, the design exploration program of the present invention is a design exploration program that uses a parameter set consisting of a plurality of parameters, and is characterized in that it causes a computer to execute the following steps: after setting one of the plurality of parameters, update the upper and lower limit values ​​of the other parameters each time to correspond to the setting of the parameter, and sequentially set the other parameters within the range of the updated upper and lower limit values; create a predetermined number of the parameter sets; predict the characteristics of each product manufactured using the plurality of parameter sets using a trained model that has been trained with the parameter sets as explanatory variables and the characteristics as target variables; and select from the predicted characteristics a parameter set that shows characteristics that are suitable for predetermined conditions.

[0009] Furthermore, the design exploration device according to the present invention is a design exploration device that uses a parameter set consisting of a plurality of parameters, and is characterized by comprising: a setting unit that, after setting one of the plurality of parameters, updates the upper and lower limit values ​​of the other parameters each time in accordance with the setting of the parameter, and sets the parameter sets by sequentially setting the other parameters within the range of the updated upper and lower limit values, and creates a predetermined number of such parameter sets; a prediction unit that predicts the characteristics of each product manufactured using the plurality of parameter sets using a trained model that has been trained with the parameter sets as explanatory variables and the characteristics as objective variables; and a selection unit that selects, from the characteristics predicted by the prediction unit, a parameter set that exhibits characteristics suitable for predetermined conditions.

[0010] In addition, the method for generating a trained model according to the present invention is a method for generating a trained model that inputs a parameter set consisting of a group of multiple parameters used in product design and outputs the characteristics of a product manufactured using the parameter set, and is characterized in that after setting one of the multiple parameters, the upper and lower limit values ​​of the other parameters are updated each time to correspond to the setting of the parameter, the other parameters are sequentially set within the range of the updated upper and lower limit values, a predetermined number of the parameter sets are created, and a trained model is generated by learning the parameter sets as explanatory variables and the characteristics of the product manufactured using the corresponding parameter sets as objective variables. [Effects of the Invention]

[0011] According to the present invention, it is possible to perform design processing with high search efficiency. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a block diagram showing the configuration of a design exploration device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram for explaining an example of a design object. [Figure 3] FIG. 3 is a flowchart showing the flow of a design exploration process according to an embodiment of the present invention. [Figure 4] FIG. 4 is a flowchart showing the flow of parameter setting processing in the design exploration processing according to one embodiment of the present invention. [Figure 5] FIG. 5 is a diagram for explaining the parameter set setting process. [Figure 6] FIG. 6 is a flowchart showing the flow of the trained model generation process executed by the learning device. [Figure 7] FIG. 7 is a diagram (part 1) for explaining conventional design exploration. [Figure 8] FIG. 8 is a diagram (part 2) for explaining conventional design exploration. DETAILED DESCRIPTION OF THE INVENTION

[0013] In the following description, an exploration and design device will be described as a form for carrying out the present invention (hereinafter referred to as an "embodiment"). The present invention is not limited to this embodiment. Furthermore, in the drawings, identical parts are given the same reference numerals. Furthermore, it should be noted that the drawings are schematic, and the relationship between the thickness and width of each component, the ratio of each component, etc., may differ from reality. Furthermore, the drawings may include parts with different dimensions and ratios.

[0014] (Embodiment) 1 is a block diagram showing the configuration of a design exploration device according to one embodiment of the present invention. The design exploration device 1 uses multiple parameters to predict the characteristics of each parameter and selects the parameter that exhibits the optimal characteristics. The design exploration device 1 includes a setting unit 11, a prediction unit 12, a selection unit 13, an input unit 14, an output unit 15, a control unit 16, and a storage unit 17. The design exploration device 1 is also connected to a learning device 2 so that they can communicate with each other.

[0015] The setting unit 11 sets a plurality of parameters used in the design exploration. 2 is a diagram illustrating an example of a design object. When forming two through holes (a first through hole 101 and a second through hole 102) in a plate material 100, the setting unit 11 performs the above-mentioned FEM analysis or the like to set the formation positions and diameters of the through holes having optimal characteristics. Examples of the characteristics include weight, rigidity, and a combination thereof. The setting unit 11 sets parameters that are a set of a center position P1 and a diameter R1 of the first through hole 101 in the plate material 100, and a center position P2 and a diameter R2 of the second through hole 102. Note that the "diameter" here will be described as the radius of the hole opening.

[0016] The prediction unit 12 performs an analysis on each parameter set by the setting unit 11 and predicts the characteristics of the product corresponding to the parameter. In this embodiment, the prediction unit 12 predicts the characteristics using a trained model (machine learning model) generated by the learning device 2.

[0017] The selection unit 13 extracts optimal characteristics from the characteristics predicted by the prediction unit 12 and selects parameters corresponding to the characteristics. The selection unit 13 extracts characteristics corresponding to, for example, preset conditions and selects the parameters. Examples of the preset conditions include conditions related to weight, conditions related to rigidity, or a combination of both required for a given product.

[0018] The input unit 14 receives input of various signals related to the operation of the design exploration device 1. The input unit 14 is configured using a keyboard, a mouse, a switch, a touch panel, and the like.

[0019] The output unit 15 displays images and outputs sounds and lights under the control of the control unit 16. The output unit 15 is configured using a display, a speaker, a light source, and the like.

[0020] The control unit 16 controls the operation processing of each component part of the design exploration device 1. For example, when an instruction to start the design exploration processing is input via the input unit 14, the control unit 16 causes each unit to execute the processing. In addition, the control unit 16 causes the output unit 15 to output information regarding the result selected by the selection unit 13, for example.

[0021] The setting unit 11, the prediction unit 12, the selection unit 13 and the control unit 16 are each configured using a processor such as a CPU (Central Processing Unit) or various arithmetic circuits that perform specific functions such as an ASIC (Application Specific Integrated Circuit).

[0022] The storage unit 17 stores programs (such as a design exploration program described later) for the control unit 16 to execute various operations. The storage unit 17 is configured using a volatile memory or a nonvolatile memory, or a combination of these. For example, the storage unit 17 is configured using a RAM (Random Access Memory), a ROM (Read Only Memory), etc.

[0023] The learning device 2 is configured using a processor such as a CPU, a processor such as an ASIC or other various arithmetic circuits that execute specific functions, a memory that stores various programs, and the like.

[0024] The learning device 2 is a server device that provides a machine learning function that trains a machine learning model that outputs product characteristics, and is an example of a computer that provides the above-mentioned machine learning function. The learning device 2 can provide the above-mentioned machine learning function by executing software that realizes the above-mentioned machine learning function. For example, the learning device 2 can be realized as a server that provides the above-mentioned machine learning function on-premises. Alternatively, the learning device 2 can be realized as a PaaS (Platform as a Service) or SaaS (Software as a Service) application, thereby providing the above-mentioned machine learning function as a cloud service.

[0025] The learning device 2 can be communicatively connected to the design exploration device 1 via a network NW. For example, the network NW can be any type of communication network, whether wired or wireless, such as the Internet or a local area network (LAN). Note that while FIG. 1 shows an example in which one design exploration device 1 is connected to one learning device 2, any number of design exploration devices 1 can be connected.

[0026] The learning device 2 generates a trained model by performing machine learning using the learning data. The learning device 2 can use a known learning method. The "trained model" referred to here refers to a trained machine learning model, and in this embodiment, a machine learning model that performs characteristic analysis to output characteristics using a combination of parameters such as the diameter of a through hole as input is taken as an example. The model used for machine learning can be realized, for example, by a neural network.

[0027] A parameter set consisting of parameters such as the diameter of the through-hole and the product characteristics configured by that parameter set can be used as a training sample for such a machine learning model. That is, the parameter set included in the training sample is used as the explanatory variable of the machine learning model, and the characteristics are used as the target variable of the machine learning model, and the machine learning model is trained according to any machine learning algorithm, such as deep learning. This results in a trained model. The training samples may be made up of parameters (numerical values) or product design images such as those shown in FIG.

[0028] For example, data related to a machine learning model may include hyperparameters related to the layer structure of the machine learning model, such as neurons and synapses in the input layer, hidden layer, and output layer to which a parameter set is input, as well as parameters related to the objective function, such as weights and biases of each layer.

[0029] Next, a design exploration method according to this embodiment will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of design exploration processing according to one embodiment of the present invention. When a design exploration instruction is received, the control unit 16 executes the design exploration processing. Hereinafter, in a design in which two through holes with circular openings are formed in a plate material 100 as shown in Fig. 2, the design exploration device 1 selects a combination of through hole diameters (radii R1, R2) that satisfy predetermined mechanical property conditions.

[0030] First, the setting unit 11 executes the setting of parameter sets (step S1). The setting unit 11 sets a plurality of parameter sets each of which is a set of a plurality of parameters including the diameters (diameters R1, R2) of the through holes.

[0031] FIG. 4 is a flowchart showing the flow of parameter setting processing in the design exploration processing according to an embodiment of the present invention. FIG. 5 is a diagram for explaining the parameter setting processing. In an example where two through holes (first through hole 201 and second through hole 202) are formed in the plate material 200 shown in FIG. 5, the setting unit 11 sets the horizontal direction as the X direction and the vertical direction as the Y direction along the sides of the rectangular plate material 200. As parameters constituting the parameter set, the center position P1 and diameter R1 of the first through hole 201, the center position P2 and diameter R2 of the second through hole 202, the distance D1 between the center positions P1 and P, and the angle θ formed by the line segment connecting the center positions P1 and P2 and the X direction are set as a set. The center position is expressed in XY orthogonal coordinates with one corner of the plate material 200 being (0, 0). At this time, the maximum value in the X direction of the plate material 200 is X0, and the maximum value in the Y direction is Y0 (<X0). Also, the coordinates of the center position P1 of the first through hole 201 are (X1, Y1), and the coordinates of the center position P2 of the second through hole 202 are (X2, Y2).

[0032] Here, in this example, it is explained that the following relational expressions are set between the parameters. X1 - R1 > 0 ···(1) X1 + R < X0 ···(2) Y1 - R1 > 0 ···(3) Y1 + R1 < Y0 ···(4) Y1 + R2 + (R1 + R2 + D1)sinθ < Y0 ···(5) X1 + R2 + (R1 + R2 + D1)cosθ < X0 ···(6)

[0033] The setting unit 11 first sets X1, which is the coordinate of the center position P1 of the first through hole 201 (step S11). The setting unit 11 sets X1 within the range of the lower limit value and the upper limit value that satisfy, for example, the following conditions. Here, the subscript min indicates the minimum value, and the subscript max indicates the maximum value. 〔Lower limit value of X1〕 Condition 1-1: X1 > R1 (from the above formula (1)) 〔Upper limit value of X1〕 Condition 1-2: X1 < X0 - R2 min -(R1 min + R2 min + D1 min )cosθ min

[0034] After that, the setting unit 11 sets the upper limit value and the lower limit value of R1, R2, Y1, D1, and θ (step S12). The setting unit 11 sets the settable range of each parameter according to the determination of X1. The setting unit 11 sets the values that satisfy the conditions set for each parameter as the upper limit value and the lower limit value.

[0035] 〔Lower limit value of R1〕 No restriction 〔Upper limit value of R1〕<00002uiality 2-1: R1 < X1 and R1 < X0 - X1 (from the above formulas (1), (2)) Condition 2-2: R1 < Y1 and R1 < Y0 - Y1 (from the above formulas (3), (4)) Condition 2-3: R1 < (Y0 - R2 min -(R2 <00o0007>+ D1 min )sinθ min ) / (1 + sinθ min ) Condition 2-4: R1 < (Y0 - R2 min -(R2 [[ID=5uiality 2-3, from the above formula (1), Y1 min ​​​​​​​​​​​​=R1, so R1 is entered into Y1 in the above formula (5) to perform the calculation. Also, under condition 2-4, from the above formula (3), X1 min =R1, so R1 is entered as X1 in the above formula (6) to perform the calculation. The upper limit of R1 is set to the minimum value among the above conditions 3-1 to 3-4.

[0036] [Y1 lower limit] Condition 3-1: Y1>R1 (from equation (3) above) [Y1 upper limit] Condition 3-2: Y1 <Y0-R2 min -(R1 min +R2 min +D1 min ) sinθ min

[0037] [R2 lower limit] Condition 4-1: R2<(Y0-Y1 min -(R1 min +D1 min ) sinθ min ) / (1+sinθ min ) [Upper limit of R2] Condition 4-2: R2<(X0-X1 min -(R1 min +D1 min )cosθ min ) / (1+cosθ min )

[0038] [D1 lower limit] Condition 5-1: D1<(Y0-Y1 min -R2 min ) / sinθ min -R1 min -R2 min [D1 upper limit] Condition 5-2: D1<(X0-X1 min -R2 min ) / cosθ min -R1 min -R2 min

[0039] [θ lower limit] Condition 6-1:θ <Arccos((X0-X1 min -R2 min ) / (R1 min +R2 min +D1 min )) [Upper limit of θ] Condition 6-2:θ <Arcsin((Y0-Y1 min -R2 min ) / (R1 min +R2 min +D1 min ))

[0040] After setting the upper and lower limit values, the setting unit 11 sets D1 within the range of the set upper and lower limit values ​​(step S13).

[0041] After setting D1, the setting unit 11 sets the upper and lower limit values ​​of R1, R2, Y1 and θ (step S14). Hereinafter, the upper and lower limit values ​​are set according to the above conditions.

[0042] After setting the upper and lower limit values, the setting unit 11 sets θ within the range between the set upper and lower limit values ​​(step S15).

[0043] After setting θ, the setting unit 11 sets the upper and lower limits of R1, R2 and Y1 (step S16).

[0044] After setting the upper and lower limit values, the setting unit 11 sets R2 within the range of the set upper and lower limit values ​​(step S17).

[0045] After setting R2, the setting unit 11 sets the upper and lower limits of R1 and Y1 (step S18).

[0046] After setting the upper and lower limit values, the setting unit 11 sets R1 within the range of the set upper and lower limit values ​​(step S19).

[0047] After setting R1, the setting unit 11 sets the upper and lower limits of Y1 (step S20).

[0048] After setting the upper limit value and the lower limit value, the setting unit 11 sets Y1 within the range of the set upper limit value and the lower limit value (step S21).

[0049] In this way, the setting unit 11 sets a parameter set consisting of the parameters X1, R1, R2, Y1, D1, and θ. According to this setting method, parameters are set so that two through holes are formed in a plate material and the through holes do not interfere with each other. For example, as shown in Table 1 below, an upper limit and a lower limit are set for each parameter, and each parameter is set. In Table 1, values ​​shown in hatching indicate that they have already been set. As shown in Table 1, the parameters X1, R1, R2, Y1, D1, and θ are set sequentially in the above order. [Table 1]

[0050] 3, the prediction unit 12 performs characteristic analysis based on the parameter set set by the setting unit 11 (step S2). The prediction unit 12 uses the trained model generated by the calculation device 2 to predict the characteristics of a product having a shape represented by the parameter set.

[0051] Here, the trained model generated by the learning device 2 will be described with reference to Fig. 6. Fig. 6 is a flowchart showing the flow of the trained model generation process executed by the learning device.

[0052] First, the learning device 2 sets a parameter set (step S101). The learning device 2 sets a plurality of parameter sets, each of which is a combination of a plurality of parameters including the diameters (radii R1 and R2) of the through holes. Specifically, the learning device 2 sets a parameter set, each of which is a combination of parameters X1, R1, R2, Y1, D1, and θ, according to the flow of the parameter setting process shown in FIG. 4. Note that the setting of the parameter set is not limited to the parameter setting process shown in FIG. 4, and the parameter set may be created in advance, for example. Furthermore, the parameter set does not need to include all of the parameters X1, R1, R2, Y1, D1, and θ, and may be, for example, a combination of parameters X1, R1, R2, and Y1.

[0053] After the parameter set is set, the learning device 2 performs characteristic analysis based on the set parameter set (step S102). The prediction unit 12 predicts the characteristics of a product having a shape represented by the parameter set, for example, using FEM analysis. If the parameter set is for a product that has already been manufactured, actual measured values ​​of the characteristics may be combined. Furthermore, the analysis process may be performed not only by the learning device 2 but also by an external device.

[0054] After the characteristic analysis, the control unit 16 determines whether analysis using another parameter set is necessary (step S103). The control unit 16 determines whether analysis is necessary according to conditions such as the number of parameter sets (number of analyses) set in advance. The number of analyses here is set based on the number of parameter sets required as learning data. If the control unit 16 determines that the number of generated parameter sets is less than the set number and analysis using another parameter set is necessary (step S103: No), the control unit 16 returns to step S101 and generates a new parameter set. If the control unit 16 determines that the number of generated parameter sets has reached the set number and analysis using another parameter set is not necessary (step S103: Yes), the control unit 16 proceeds to step S104. In each of the processes in steps S102 and S103, a required number of parameter sets may be generated in step S101, and after the required number of parameter sets are obtained, characteristic analysis may be performed on all of the parameter sets.

[0055] In step S104, the learning device 2 generates a trained model by learning using the parameter set as explanatory variables and the characteristics as objective variables.

[0056] Then, the learning device 2 stores the generated trained model in a memory (step S105). The learning device 2 also outputs the generated trained model to the design exploration device 1.

[0057] The prediction unit 12 uses the trained model acquired from the learning device 2 to acquire the estimated characteristics.

[0058] Returning to FIG. 3, after the characteristic analysis, the control unit 16 determines whether or not analysis using another parameter set is necessary (step S3). The control unit 16 determines whether or not analysis is necessary according to conditions such as the number of parameter sets (number of analyses) that are set in advance. If the control unit 16 determines that analysis using another parameter set is necessary (step S3: Yes), the control unit 16 returns to step S1 and generates a new parameter set. If the control unit 16 determines that analysis using another parameter set is not necessary (step S3: No), the control unit 16 proceeds to step S4. In each of the processes in steps S1 to S3, a required number of parameter sets may be generated in step S1, and after the required number of parameter sets are obtained, characteristic analysis may be performed on all of the parameter sets. Furthermore, it is preferable that the parameter sets set in step S1 include parameter sets that are made up of a combination of parameters that are different from the parameter sets used as the learning data.

[0059] In step S4, the selection unit 13 refers to the analysis results, extracts analysis results that satisfy preset conditions, and selects a parameter set corresponding to the extracted analysis results as the optimum parameter set.

[0060] Then, the control unit 16 outputs the design exploration result including the parameter set selected by the selection unit 13 to the output unit 15. The output unit 15 may display the design exploration result on a display or output it to an external terminal. The control unit 16 may also store the design exploration result in the storage unit 17.

[0061] Users such as designers can refer to the design exploration results to design a product having desired characteristics.

[0062] In the above-described embodiment, for a parameter set consisting of multiple parameters for manufacturing a product, after one parameter is set, the upper and lower limit values ​​of the other parameters are updated each time in accordance with the setting of the parameter, and the other parameters are set within the ranges of the updated upper and lower limit values. Therefore, it is possible to efficiently generate a parameter set used for design exploration, such as a parameter set that prevents interference between through holes formed in a plate material. According to this embodiment, efficient setting of parameter sets enables design processing to be performed with high exploration efficiency.

[0063] Furthermore, according to this embodiment, when generating a trained model for characteristic analysis, the parameter set to be used as training data is generated according to the flowchart shown in FIG. 4, so that training data can be created efficiently, and ultimately, a trained model can be generated efficiently.

[0064] The learning program does not necessarily have to be stored in the memory of the learning device 2 from the beginning. For example, the learning program can be stored on a "portable physical medium" such as a flexible disk inserted into a computer, such as a FD, CD-ROM, DVD disk, magneto-optical disk, or IC card. The learning device 2 can then retrieve and execute the learning program from such a portable physical medium. Alternatively, the learning program can be stored on another computer or server device connected to the learning device 2 via a public line, the Internet, a LAN, a WAN, or the like. The learning program stored in this way can be downloaded to the learning device 2 and then executed.

[0065] In the above-described embodiment, an example has been described in which the design exploration device 1 and the learning device 2 are separate devices. However, the design exploration device 1 may have the functions of the learning device 2.

[0066] In this way, the present invention can include various embodiments not described here, and various design changes can be made within the scope that does not deviate from the technical idea specified by the claims.

[0067] As described above, the design exploration method, design exploration program, design exploration device, and learned model generation method according to the present invention are suitable for executing design processing with high exploration efficiency. [Explanation of symbols]

[0068] 1 Design exploration equipment 2 Learning device 11 Setting section 12 Prediction Department 13 Selection section 14 Input section 15 Output section 16 Control Unit 17 Memory section

Claims

1. A design exploration method using a parameter set including a plurality of parameters, after setting one of the plurality of parameters, updating upper and lower limit values ​​of the other parameters in accordance with the setting of the parameter, and sequentially setting the other parameters within the ranges of the updated upper and lower limit values; creating a preset number of the parameter sets; Predicting the characteristics of each product produced using the plurality of parameter sets using a trained model trained with the parameter sets as explanatory variables and the characteristics as objective variables; Selecting a parameter set that exhibits characteristics suitable for predetermined conditions from among the predicted characteristics. A design exploration method comprising:

2. The upper and lower limit values ​​of the parameters are calculated according to a predetermined relational expression between the parameters.

2. The design exploration method of claim 1.

3. A design exploration program using a parameter set including a plurality of parameters, after setting one of the plurality of parameters, updating upper and lower limit values ​​of the other parameters in accordance with the setting of the parameter, and sequentially setting the other parameters within the ranges of the updated upper and lower limit values; creating a preset number of the parameter sets; Predicting the characteristics of each product produced using the plurality of parameter sets using a trained model trained with the parameter sets as explanatory variables and the characteristics as objective variables; Selecting a parameter set that exhibits characteristics suitable for predetermined conditions from among the predicted characteristics. A design exploration program that causes a computer to execute the above steps.

4. A design exploration device using a parameter set including a plurality of parameters, a setting unit that, after setting one parameter among the plurality of parameters, updates upper and lower limit values ​​of other parameters in accordance with the setting of the parameter, and sets the other parameters sequentially within the ranges of the updated upper and lower limit values ​​to set the parameter sets, and creates a preset number of such parameter sets; a prediction unit that predicts the characteristics of each product produced by each of the plurality of parameter sets using a trained model that has been trained using the parameter sets as explanatory variables and the characteristics as objective variables; a selection unit that selects a parameter set that exhibits characteristics suitable for a predetermined condition from among the characteristics predicted by the prediction unit; A design exploration device comprising:

5. A method for generating a trained model that inputs a parameter set consisting of a plurality of parameters used in product design and outputs characteristics of a product manufactured using the parameter set, after setting one of the plurality of parameters, updating upper and lower limit values ​​of the other parameters in accordance with the setting of the parameter, and sequentially setting the other parameters within the ranges of the updated upper and lower limit values; creating a preset number of the parameter sets; A trained model is generated by training the parameter set as an explanatory variable and the characteristics of a product produced by the corresponding parameter set as a target variable. A method for generating a trained model.

Images