Design Support Device, Design Support Method, and Design Support Program

Abstract

To enable optimization of a characteristic such as a product and the like constituting an objective variable in a design process of the product and the like and a design variable with a low load.SOLUTION: A design support device 10 comprises: a data acquisition unit 11 that acquires track record data composed of a design parameter group and a measurement value of a measurement item; a model construction unit 12 that constructs a prediction model predicting the measurement value as a probability distribution and the like on the basis of the design parameter group; a sampling unit 13 that performs sampling of a measurement value group in a prescribed number of points score, using each prediction model; an evaluation value calculation unit 14 that scalarizes a vector with each measurement value included in the measurement value group as an element, and thereby, calculates an evaluation value in each sampling point; an acquisition function evaluation unit 15 that outputs an acquisition function evaluation value on the basis of a distribution of the evaluation value on each sampling point; a design parameter group acquisition unit 16 that acquires a design parameter group by optimization of the acquisition function evaluation value; and an output unit 17 that outputs a design parameter group candidate.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] One aspect of the present disclosure relates to a design support device, a design support method, and a design support program.

Background Art

[0002] Product design utilizing machine learning has been studied. As one field of product design, for example, in the design of functional materials, for example, a model for estimating the properties of a material is constructed by machine learning using learning data consisting of pairs of raw material mixing ratios and properties of experimentally prepared and fabricated materials, and predictions of properties for untested raw material mixing ratios are made. By formulating an experimental plan based on such property predictions, it becomes possible to efficiently optimize design parameters such as material properties and raw material mixing ratios, and the development efficiency is improved. Also, as such an optimization method, Bayesian optimization is known to be effective, and a design device that outputs design values using Bayesian optimization is known (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] On the other hand, in product development such as materials, single-objective optimization may be performed so as to improve one evaluation value calculated by an evaluation function based on measured values of a plurality of properties. In order to optimize one evaluation value and design parameters, it is conceivable to directly predict the evaluation value by machine learning using the one evaluation value as an objective variable. However, due to the non-linearity of the conversion by the evaluation function and other reasons, the learning accuracy may be low, and there is a problem that the optimization efficiency deteriorates. Note that such a problem is not limited to material design but is common to product design in general.

[0005] Therefore, the present invention has been made in view of the above problems, and an object thereof is to enable optimization of one evaluation value composed of target variables and design variables in the production process of products, work in progress, semi-finished products, parts, or prototypes with a lower load and fewer experiment times.

Means for Solving the Problems

[0006] A design support device according to an aspect of the present disclosure is for applying to a method of optimizing design parameters by repeating determination of design parameters and production of a product, work in progress, semi-finished product, part, or prototype based on the determined design parameters in the design of a product, work in progress, semi-finished product, part, or prototype produced based on a group of design parameters composed of a plurality of design parameters. A design support device that obtains a plurality of design parameters such that one evaluation value indicating the characteristics of a product, work in progress, semi-finished product, part, or prototype is improved. The one evaluation value is calculated based on measurement values of a plurality of measurement items, and a data acquisition unit that acquires a plurality of performance data including a group of design parameters and respective measurement values of the plurality of measurement items for a produced product, work in progress, semi-finished product, part, or prototype, and based on the group of design parameters, a prediction model that predicts measurement values of the measurement items as a probability distribution or an approximation or alternative index thereof is constructed based on the performance data. A sampling unit that samples a group of measurement values composed of a plurality of measurement values sampled from a multidimensional probability distribution of measurement values obtained from each prediction model as one sampling point, and a number of measurement values included in the group of measurement values as a dimension number and each measurement value A vector having a value as an element is scalarized by a predetermined operation to calculate an evaluation value of the group of measurement values at each sampling point, and based on the distribution of the evaluation values at each sampling point, a group of design parameters is input, and an acquisition function evaluation unit that outputs an acquisition function evaluation value related to improvement of the evaluation value by a predetermined acquisition function, and a design parameter group acquisition unit that acquires at least one group of design parameters by optimization of the acquisition function evaluation value, and an output unit that outputs the group of design parameters acquired by the design parameter group acquisition unit.

[0007] The design support method according to one aspect of the present disclosure is applicable to a method of optimizing design parameters by repeating determination of design parameters and production of a product, work-in-progress, semi-finished product, component, or prototype based on the determined design parameters in the design of a product, work-in-progress, semi-finished product, component, or prototype made based on a group of design parameters composed of a plurality of design parameters. The design support method is a design support method in a design support device for obtaining a plurality of design parameters such that one evaluation value indicating the characteristics of a product, work-in-progress, semi-finished product, component, or prototype is improved. The one evaluation value is calculated based on measurement values of a plurality of measurement items. The design support method includes a data acquisition step of acquiring a plurality of pieces of performance data including a group of design parameters and measurement values of the plurality of measurement items for a manufactured product, work-in-progress, semi-finished product, component, or prototype; a model construction step of constructing a prediction model for predicting measurement values of measurement items as a probability distribution or an approximation or alternative index thereof based on the performance data, based on the group of design parameters; a sampling step of sampling a group of measurement values composed of a plurality of measurement values sampled from a multi-dimensional probability distribution of measurement values obtained from each prediction model as one sampling point, with a predetermined number of sampling points; an evaluation value calculation step of calculating an evaluation value of the group of measurement values of each sampling point by scalarizing a vector having the number of measurement values included in the group of measurement values as the number of dimensions and the value of each measurement value as an element by a predetermined operation; an acquisition function evaluation step of inputting the group of design parameters based on the distribution of evaluation values of each sampling point and outputting an acquisition function evaluation value regarding improvement of the evaluation value by a predetermined acquisition function; a design parameter group acquisition step of obtaining at least one group of design parameters by optimizing the acquisition function evaluation value; and an output step of outputting the group of design parameters obtained in the design parameter group acquisition step.

[0008] A design support program according to an aspect of the present disclosure causes a computer to optimize design parameters by repeating determination of design parameters and production of a product, work in process, semi-finished product, part, or prototype based on the determined design parameters in the design of a product, work in process, semi-finished product, part, or prototype made based on a group of design parameters. The design support program is a design support program for functioning as a design support device that obtains a plurality of design parameters such that a single evaluation value indicating characteristics of a product, work in process, semi-finished product, part, or prototype is improved. The single evaluation value is calculated based on measurement values of a plurality of measurement items. The design support program includes a data acquisition function that acquires a plurality of pieces of performance data including a group of design parameters and respective measurement values of the plurality of measurement items for a manufactured product, work in process, semi-finished product, part, or prototype; a model construction function that constructs a prediction model that predicts measurement values of measurement items as probability distributions or approximations or alternative indicators thereof based on the performance data, based on the group of design parameters; a sampling function that samples a group of measurement values, which is a plurality of measurement values sampled from a multidimensional probability distribution of measurement values obtained from each prediction model, as a single sampling point, with a predetermined number of sampling points; an evaluation value calculation function that calculates an evaluation value of the group of measurement values of each sampling point by scalarizing a vector having the number of measurement values included in the group of measurement values as the number of dimensions and the value of each measurement value as an element by a predetermined operation; an acquisition function evaluation function that inputs the group of design parameters and outputs an acquisition function evaluation value regarding improvement of the evaluation value by a predetermined acquisition function based on the distribution of the evaluation values of each sampling point; a design parameter group acquisition function that acquires at least one group of design parameters by optimizing the acquisition function evaluation value; and an output function that outputs the group of design parameters acquired by the design parameter group acquisition function.

[0009] According to such an aspect, a prediction model for predicting measurement values of a plurality of measurement items for calculating and evaluating an evaluation value is constructed for each measurement item based on performance data. Since this prediction model predicts the measurement value of a measurement item as a probability distribution or its approximation or alternative index, based on the multidimensional distribution of the measurement values obtained from the prediction model of each measurement item, it is possible to sample an arbitrary number of measurement value groups. By performing an operation on a vector having a measurement value group at each sampling point as an element using a predetermined evaluation formula, it is possible to obtain an evaluation value regarding each sampling point expressed by a scalar value. Then, based on the distribution of the evaluation values of each sampling point, by optimizing the acquisition function evaluation value output using a predetermined acquisition function, it is possible to obtain a group of design parameters suitable for the next experiment or the like. Therefore, compared with the method of directly learning the evaluation value to construct the acquisition function, it is possible to obtain a more accurate prediction model regarding the measurement value, and by optimizing the acquisition function evaluation value regarding the evaluation value of the measurement value group obtained by the prediction model, as a result, the suitability of the group of design parameters adopted in the experiment or the like is improved, so it is possible to reduce the number of experiments.

[0010] In the design support device according to another aspect, the evaluation value calculation unit may calculate the evaluation value by a predetermined evaluation formula that performs a predetermined operation based on a plurality of measurement values included in the measurement value group.

[0011] According to such an aspect, an arbitrary evaluation value for evaluating a manufactured product or the like can be calculated and set by an evaluation formula.

[0012] In the design support device according to another aspect, the evaluation value calculation unit may calculate the characteristic value as the evaluation value by a theoretical formula that calculates a characteristic value indicating the characteristics of a product, work in process, semi-finished product, part, or prototype based on a plurality of measurement values included in the measurement value group.

[0013] According to such an aspect, the characteristic value for evaluating a manufactured product or the like is calculated by a theoretical formula for calculating the characteristic value. Then, the calculated characteristic value can be used as an evaluation value for evaluating a product or the like.

[0014] In the design support device according to another aspect, when the measured value sampled by the sampling unit using the prediction model of the measurement item is not included in a given domain related to the measurement item, the sampling unit may replace the sampled measured value with a predetermined value preset for the measurement item.

[0015] In the design support device of the above aspect, due to sampling from the probability distribution based on the prediction model, it may happen that the sampled measured value is a value outside the definition domain of the measurement item. Thus, when a measured value not included in the domain is sampled, the measured value is replaced with a predetermined value and included in the measured value group of the sampling point. Therefore, the accuracy of the evaluation value calculated based on the measured value group is improved.

[0016] In the design support device according to another aspect, the acquisition function evaluation unit may output the acquisition function evaluation value by any one of the acquisition functions of LCB (Lower Confidence Bound), UCB (Upper Confidence Bound), EI (Expected Improvement), and PI (Probability of Improvement).

[0017] According to such an aspect, an acquisition function evaluation value suitable for evaluating a group of design parameters suitable for improving the evaluation value is output.

[0018] In the design support device according to another aspect, the design parameter group acquisition unit may acquire one group of design parameters that optimize the acquisition function evaluation value.

[0019] According to such an aspect, a group of design parameters capable of improving the evaluation value related to the measurement item can be obtained.

[0020] In the design support device according to another aspect, the design parameter group acquisition unit may acquire a plurality of groups of design parameters by a predetermined algorithm.

[0021] According to such an aspect, a plurality of groups of design parameters to be subjected to the following experiments can be easily obtained.

[0022] In the design support device according to another aspect, the prediction model is a regression model or a classification model that takes a group of design parameters as an input and outputs a probability distribution of measurement values, and the model construction unit may construct the prediction model by machine learning using the performance data.

[0023] According to such an aspect, since the prediction model is constructed as a predetermined regression model or classification model, a prediction model capable of obtaining the probability distribution of the measurement values of the measurement items, or an approximation or alternative index thereof, can be obtained.

[0024] In the design support device according to another aspect, the prediction model may be a machine learning model that predicts the probability distribution of measurement values, or an approximation or alternative index thereof, using any one of the posterior distribution of prediction values based on Bayesian theory, the distribution of prediction values of predictors constituting an ensemble, the theoretical formulas of prediction intervals and confidence intervals of a regression model, Monte Carlo dropout, and the distribution of predictions of a plurality of predictors constructed under different conditions.

[0025] According to such an aspect, a prediction model capable of predicting the probability distribution of the measurement values of the measurement items based on the group of design parameters, or an approximation or alternative index thereof, is constructed.

Effects of the Invention

[0026] According to one aspect of the present disclosure, optimization of one evaluation value and design variables constituted by an objective variable in the production process of a product, work in process, semi-finished product, component, or prototype can be achieved with a lower load using a smaller number of experiments.

Brief Description of the Drawings

[0027]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Mode for Carrying Out the Invention

[0028] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are denoted by the same reference numerals, and duplicate descriptions are omitted.

[0029] FIG. 1 is a diagram showing an outline of a material design process, which is an example of a design process of a product, work-in-progress, semi-finished product, component, or prototype to which the design support apparatus according to the embodiment is applied. Hereinafter, "product, work-in-progress, semi-finished product, component, or prototype" will be referred to as "product, etc.". The design support apparatus 10 of the present embodiment can be applied to the design process of any product, etc. that requires optimization of one evaluation value indicating the characteristics of the product, etc. The design support apparatus 10 can be applied to a method of optimizing the design parameters and evaluation values of a product, etc. by repeating the determination of design parameters and the production of a product, work-in-progress, semi-finished product, component, or prototype based on the determined design parameters. Specifically, the design support apparatus 10 can be applied to, in addition to the development and design of materials, for example, the design of products such as automobiles and drugs, the optimization of the molecular structure of drugs, and the like. In the present embodiment, as described above, the design support process by the design support apparatus 10 will be described by taking an example of material design as an example of the design of a product, etc.

[0030] As shown in FIG. 1, the design support process by the design support apparatus 10 is applied to the production of materials and experiments in a plant, laboratory A, etc. That is, based on the set design parameter group x, materials are produced in a plant, laboratory A, etc., and based on the produced materials, measurement values y of a plurality of measurement items indicating the characteristics of the materials are acquired. Note that the production of materials and experiments in the plant and laboratory A may be simulations. In this case, the design support apparatus 10 provides a design parameter group x for executing the next simulation.

[0031] The design support apparatus 10 performs optimization of one evaluation value and design parameters based on the performance data including the design parameter group x and the measurement values y of a plurality of measurement items of the material produced based on the design parameter group x. Specifically, the design support apparatus 10 outputs a design parameter group x that may obtain more suitable characteristics for the next production and experiment based on the design parameter group x and the measurement values y regarding the produced material. One evaluation value is a value serving as an index for evaluating a product or the like, and is calculated as a scalar value by a predetermined calculation on a measurement value group composed of measurement values of a plurality of measurement items.

[0032] The evaluation value may be a value calculated by a predetermined evaluation formula that performs a predetermined calculation based on a plurality of measurement values included in the measurement value group. The evaluation formula in this case may be arbitrarily set, and for example, it may be independently set by the industry to which the product or the like to be produced belongs, the company that produces the product or the like and the department of the company, and an individual involved in research and development or the like of the product or the like.

[0033] Further, the evaluation value may be a characteristic value indicating a characteristic regarding a product or the like. The evaluation formula in this case may be a theoretical formula for calculating the characteristic value. For example, when the evaluation value of a product or the like is the transmission loss of an electrical signal, the evaluation formula is a formula that multiplies the product of the frequency, the square root of the relative dielectric constant, and the dielectric tangent by a coefficient.

[0034] For example, the design support device 10 of the present embodiment is applied for the purpose of tuning a plurality of design variables to improve a single evaluation value in the design of a material product. As an example of the design of a material product, when a certain material is produced by mixing a plurality of compositions, the design support device 10 uses a group of design parameters such as the blending amount of each composition as design variables, and uses, as an objective variable, an evaluation value calculated based on a plurality of measured values measured for the produced material, for tuning the group of design parameters to improve a single evaluation value.

[0035] FIG. 2 is a block diagram showing an example of the functional configuration of the design support device 10 according to the embodiment. The design support device 10 is a device that obtains a plurality of design parameters so as to improve a single evaluation value indicating the characteristics of a material in the design of a material produced based on a group of design parameters composed of a plurality of design parameters. As shown in FIG. 2, the design support device 10 may include a functional unit configured in the processor 101, a design parameter storage unit 21, and a measured value storage unit 22. Each functional unit will be described later.

[0036] FIG. 3 is a diagram showing an example of the hardware configuration of a computer 100 constituting the design support device 10 according to the embodiment. Note that the computer 100 may constitute the design support device 10.

[0037] As an example, the computer 100 includes, as hardware components, a processor 101, a main storage device 102, an auxiliary storage device 103, and a communication control device 104. The computer 100 constituting the design support device 10 may further include an input device 105 such as a keyboard, a touch panel, a mouse, etc., which are input devices, and an output device 106 such as a display.

[0038] Processor 101 is an arithmetic unit that executes an operating system and application programs. Examples of processors include a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit), but the type of processor 101 is not limited to these. For example, processor 101 may be a combination of sensors and dedicated circuits. The dedicated circuit may be a programmable circuit such as an FPGA (Field-Programmable Gate Array) or other types of circuits.

[0039] Main storage device 102 is a device that stores programs for realizing the design support device 10 etc. and arithmetic results output from processor 101. Main storage device 102 is constituted by at least one of, for example, a ROM (Read Only Memory) and a RAM (Random Access Memory).

[0040] Auxiliary storage device 103 is a device that can generally store a larger amount of data than main storage device 102. Auxiliary storage device 103 is constituted by a non-volatile storage medium such as a hard disk or a flash memory. Auxiliary storage device 103 stores a design support program P1 for causing computer 100 to function as a design support device 10 etc. and various types of data.

[0041] Communication control device 104 is a device that executes data communication with other computers via a communication network. Communication control device 104 is constituted by, for example, a network card or a wireless communication module.

[0042] Each functional element of the design support device 10 is realized by causing the processor 101 to read the corresponding program P1 onto the processor 101 or the main memory device 102 and execute the program. The program P1 includes code for realizing each functional element of the corresponding server. The processor 101 operates the communication control device 104 according to the program P1, and reads and writes data in the main memory device 102 or the auxiliary storage device 103. Through such processing, each functional element of the corresponding server is realized.

[0043] The program P1 may be provided after being fixedly recorded on a tangible recording medium such as a CD-ROM, a DVD-ROM, or a semiconductor memory. Alternatively, at least one of these programs may be provided via a communication network as a data signal superimposed on a carrier wave.

[0044] Referring again to FIG. 2, the design support device 10 includes a data acquisition unit 11, a model construction unit 12, a sampling unit 13, an evaluation value calculation unit 14, an acquisition function evaluation unit 15, a design parameter group acquisition unit 16, and an output unit 17. As shown in FIG. 2, the design parameter storage unit 21 and the measurement value storage unit 22 may be configured as part of the design support device 10, or may be configured as other devices accessible from the design support device 10.

[0045] The data acquisition unit 11 acquires a plurality of performance data regarding the fabricated materials. The performance data consists of pairs of a design parameter group and the respective measured values of a plurality of measurement items. The design parameter storage unit 21 is a storage means for storing the design parameter group in the performance data, and may be configured, for example, in the main memory device 102 and the auxiliary storage device 103. The measurement value storage unit 22 is a storage means for storing the measured values in the performance data.

[0046] FIG. 4 is a diagram showing an example of the design parameter group stored in the design parameter storage unit 21. As shown in FIG. 4, the design parameter storage unit 21 stores the design parameter group x in the material fabrication from the first time (t = 1) to the (T - 1)th time (t = T - 1)t stores it. The design parameter group x may include, for example, the blending amount of raw material A, the blending amount of raw material B, and design parameter D, etc., and can constitute vector data with a dimensionality corresponding to the number D of design parameters. The design parameters may be, in addition to those exemplified, for example, non-vector data such as molecular structures and images. Also, when dealing with the problem of selecting an optimal molecule from a plurality of molecule types, the design parameters may be data indicating options among the plurality of molecules.

[0047] FIG. 5 is a diagram showing an example of the measured value y stored in the measurement value storage unit 22. As shown in FIG. 5, the measurement value storage unit 22 stores the measured values y of a plurality of measurement items k (k = 1 to K) indicating the characteristics of the materials produced in the material production from the first time (t = 1) to the (T - 1)th time (t = T - 1). k,t stores it. The measurement item k may include, for example, relative permittivity, dielectric loss tangent, and measurement item K. The design parameter group x t and the measured value y k,t form a pair to constitute performance data.

[0048] Based on the performance data in the material production from the first time (t = 1) to the (T - 1)th time (t = T - 1), the design support device 10 obtains the design parameter group x for the Tth time such that one evaluation value calculated based on the measured values of each measurement item is improved. T to obtain.

[0049] The model construction unit 12 constructs a prediction model based on the performance data. The prediction model is a model that predicts the measured value y of the measurement item k as the target variable based on the design parameter group x k as a probability distribution or its approximation or alternative index. The model constituting the prediction model may be any model that can predict the measured value y k as a probability distribution or its approximation or alternative index, and its type is not limited.

[0050] For example, the prediction model takes the design parameter x as an input and the measured value y kIt may also be a regression model that outputs a probability distribution. When the prediction model is a regression model, the prediction model may be constituted by any one of regression models such as linear regression, PLS regression, Gaussian process regression, random forest, and neural network. The model construction unit 12 may construct a prediction model by a well-known machine learning method using the performance data. The model construction unit 12 may also construct a prediction model by a machine learning method that applies the performance data to the prediction model and updates the parameters of the prediction model.

[0051] The probability distribution obtained from the prediction model is not limited to a specific probability distribution such as a Gaussian distribution, but may be a probability distribution according to the characteristics of the measurement item, for example, a beta distribution or a discrete probability distribution.

[0052] In a prediction model constructed as Gaussian process regression, the probability distribution of the measured value is predicted by inputting the design parameter group x in the performance data constituting the explanatory variable of the training data, the measured value y constituting the objective variable, and the design parameter x of the prediction target into the model.

[0053] Furthermore, the prediction model may be a machine learning model that predicts the probability distribution of the measured value or its approximation or alternative index using any one of the posterior distribution of the predicted value based on Bayesian theory, the distribution of the predicted values of the predictors constituting the ensemble, the theoretical formulas of the prediction interval and confidence interval of the regression model, the distribution obtained by Monte Carlo dropout, and the distribution of the predictions of a plurality of predictors constructed under different conditions.

[0054] The probability distribution of the measured values, or the prediction of its alternative index, can be obtained by a model-specific method. The probability distribution of the measured values, or its approximation or alternative index, can be based on the posterior distribution of the predicted values if it is Gaussian process regression or Bayesian neural network, based on the distribution of the predictions of the predictors that make up the ensemble if it is random forest, based on the prediction intervals and confidence intervals if it is linear regression, and based on Monte Carlo dropout if it is a neural network. However, the calculation method of the distribution of the measured values or its alternative index for each machine learning model is not limited to the above methods.

[0055] In addition, any model can also be extended to a model that can predict the probability distribution of the measured values or its alternative index. For example, a model that uses the distribution of the predicted values of each model obtained by constructing a plurality of data sets by the bootstrap method or the like and constructing a prediction model for each of them as an alternative index of the probability distribution of the measured values can be cited as an example. However, the method of extending the machine learning model to a model that can predict the probability distribution of the measured values or its alternative index is not limited to the above method.

[0056] In addition, the model construction unit 12 may tune the hyperparameters of the prediction model by a well-known hyperparameter tuning method. That is, the model construction unit 12 may update the hyperparameters of the prediction model by maximum likelihood estimation using the vector representing the design parameter group x, which is the explanatory variable in the performance data, and the measured value y, which is the objective variable.

[0057] In addition, the prediction model may be constructed by a classification model. When the prediction model is a classification model, the model construction unit 12 can construct the prediction model by a machine learning method that can evaluate a well-known probability distribution using the performance data.

[0058] In this way, by constructing the prediction model by the model construction unit 12 using a predetermined regression model or classification model, it becomes possible to obtain the probability distribution of the measured values of the measurement items based on any design parameter group x.

[0059] Further, the prediction model may be a single-task model that predicts the measured value of one measurement item as a probability distribution or an approximation or alternative index thereof, or a multi-task model that predicts the measured values of a plurality of measurement items as a probability distribution or an approximation or alternative index thereof. In this way, by constructing a prediction model using a multi-task model or a single-task model appropriately configured according to the nature of the measurement item, the accuracy of predicting the measured value by the prediction model can be improved.

[0060] The sampling unit 13 samples a plurality of groups of measured values sampled from the multi-dimensional probability distribution of the measured values obtained from each prediction model as one sampling point, and samples a predetermined number of groups of measured values.

[0061] Specifically, for example, when the measured values y k (k = 1 to K) follow a multi-dimensional normal distribution based on the prediction model and are uncorrelated with each other, they are expressed as follows. y k ~N(m k (x),σ k (x) 2 ) In such a case, the sampling unit 13 samples a plurality of measured values y k from the probability distribution of the measured value y that is the target variable in the prediction model, based on one group of design parameters x. k (k = 1 to K).

[0062] More specifically, the sampling unit 13 samples groups of measured values y k,n (k = 1 to K, n = 1 to N) as the n-th (n: 1 to N) sampling point. The group of measured values y n of the n-th sampling point constitutes a vector with the number of measured values included in the group of measured values as the number of dimensions and each measured value as an element, as expressed below. y n =[y 1,n ,y 2,n ,..,y k,n ,..,y K,n ​

[0063] Then, the sampling unit 13 acquires a set of measurement value groups Y corresponding to N sampling points, which is a predetermined number of points. Y = [y1, y2,.., y n ,.., y N The measurement value groups y1, y2,.., y n ,.., y N each constitute a vector.

[0064] In the above example, it is assumed that the measurement value y as the target variable in the prediction model follows a normal distribution and is independent of each other. However, there may be a correlation, and it is not limited to being a normal distribution. It may also follow other probability distributions. k That is, it is assumed that the measurement value y as the target variable in the prediction model follows a normal distribution and is independent of each other. However, there may be a correlation, and it is not limited to being a normal distribution. It may also follow other probability distributions.

[0065] Also, when the measurement value sampled by the sampling unit 13 using the prediction model of the measurement item is not included in the given domain regarding the measurement item, the sampling unit 13 may replace the sampled measurement value with a predetermined value preset for the measurement item. Specifically, for example, when a certain measurement item is a physical property value that cannot take negative values, due to sampling from the probability distribution based on the prediction model, it may happen that the sampled measurement value is negative. In such a case, the sampling unit 13 replaces the sampled negative measurement value with the predetermined value "0" preset for the measurement item. By including the thus-replaced predetermined value in the set of measurement value groups, the accuracy of the evaluation value calculated later is improved.

[0066] Also, regarding the sampling by the sampling unit 13, an example of sampling for each measurement item has been described. However, it is not limited to such an example. The sampling unit 13 may, for example, sample the set of measurement value groups at once based on the multivariate normal distribution defined by the prediction model of each measurement item.

[0067] Also, based on the sampling points from the standard normal distribution sampled and stored in advance, the sampling unit 13 samples the set of measurement value groups y​k,n It may also be possible to obtain a sample. Specifically, the sampling unit 13 pre-samples sampling points y_std k,n (n = 1 to N) from a standard normal distribution which is a normal distribution with a mean of 0 and a variance of 1, and according to the following conversion formula, a sample of the measured value group y k,n can be obtained. y k,n = y_std k,n * σ k (x) + m k (x)

[0068] In the design support device 10 implemented on a computer, the sampling unit 13 may perform N-point sampling in a batch, or may perform sampling corresponding to each of a plurality of design parameter groups x in a batch.

[0069] The evaluation value calculation unit 14 calculates the evaluation value of the measured value group corresponding to one design parameter group x and one sampling point. Specifically, as described above, the measured value group of one sampling point constitutes a vector with the number of measured values included in the measured value group as the dimension number and each measured value as an element. Therefore, the evaluation value calculation unit 14 calculates the evaluation value of each sampling point by scalarizing the vector representing the measured value group by a predetermined operation.

[0070] The evaluation value calculation unit 14 may calculate the evaluation value by a predetermined evaluation formula that performs a predetermined operation based on a plurality of measured values included in the measured value group. The evaluation formula can be arbitrarily set, and for example, it may be independently set by the industry to which the manufactured product etc. belongs, the company that manufactures the product etc. and the department of the company, and the individuals involved in research and development etc. of the product etc.

[0071] The evaluation value calculation unit 14 may be a theoretical formula for calculating a characteristic value indicating a characteristic related to a product etc. As an example, when the evaluation value is the transmission loss of an electrical signal, the evaluation value calculation unit 14 calculates the evaluation value v n by the following formula (1). Evaluation value (transmission loss) v n∝ Frequency × √Relative permittivity × Dielectric loss tangent ···(1) That is, the evaluation value v n is calculated as a scalar value obtained by multiplying the product of the frequency, the square root of the relative permittivity, and the dielectric loss tangent by a coefficient.

[0072] The evaluation value calculation unit 14 calculates the evaluation value v n for each of the measurement value groups y n at the first to Nth sampling points, thereby obtaining an evaluation value set V n consisting of the first to Nth evaluation values v. V = [v1, v2,.., v n ,.., v N As described above, the evaluation values v1, v2,.., v n ,.., v N each constitute a scalar value.

[0073] In the design support device 10 implemented on a computer, the evaluation value calculation unit 14 may calculate the evaluation values v1, v2,.., v n ,.., v N corresponding to each of the N sampling points in a batch.

[0074] The acquisition function evaluation unit 15 takes the design parameter group as input based on the distribution of the evaluation values at each sampling point, and outputs an acquisition function evaluation value regarding the improvement of the evaluation value by a predetermined acquisition function.

[0075] The acquisition function evaluation unit 15 may output an acquisition function evaluation value using a well-known acquisition function such as LCB (Lower Confidence Bound), for example. LCB is used when minimizing the output of the function, and a suitable design parameter group x can be obtained by minimizing the value of LCB.

[0076] When constructing the acquisition function with LCB, the acquisition function evaluation unit 15 defines and constructs the acquisition function evaluation value A(x) as in the following formula (2). A(x) = mv(x) - aσv(x) ···(2) ​The acquisition function evaluation unit 15 evaluates and obtains the mean mv(x) and the standard deviation σv(x) based on the distribution of the evaluation values v included in the evaluation value set V, and outputs an acquisition function evaluation value using the acquisition function shown in Equation (2). a in Equation (2) is an arbitrary parameter. Equation (2) of the acquisition function represents the lower limit of the confidence interval when it is assumed that the measured value of v in the next experiment follows a normal distribution with the design parameter group x as a parameter. n Based on the distribution of v, the mean mv(x) and the standard deviation σv(x) are evaluated and obtained, and an acquisition function evaluation value is output by the acquisition function shown in Equation (2). a in Equation (2) is an arbitrary parameter. The acquisition function of Equation (2) represents the lower limit of the confidence interval when it is assumed that the measured value of v in the next experiment follows a normal distribution with the design parameter group x as a parameter. n when assuming that the measured value of v follows a normal distribution.

[0077] Also, the acquisition function evaluation unit 15 may output an acquisition function evaluation value A(x) by well-known functions such as UCB (Upper Confidence Bound), EI (Expected Improvement), and PI (Probability of Improvement).

[0078] The design parameter group acquisition unit 16 obtains at least one design parameter group by optimizing the acquisition function evaluation value A(x) output by the acquisition function evaluation unit 15.

[0079] As an example, the design parameter group acquisition unit 16 may obtain at least one design parameter group x that optimizes the output of the acquisition function. Specifically, the design parameter group acquisition unit 16 performs optimization with the acquisition function evaluation value A(x) output by the acquisition function evaluation unit 15 as an objective variable, and obtains the design parameter group x as an optimal solution.

[0080] Also, as an example, the design parameter group acquisition unit 16 may obtain a plurality of design parameter groups by a predetermined algorithm. Specifically, the design parameter group acquisition unit 16 may obtain a plurality of design parameter groups by applying a batch Bayesian optimization method to the acquisition function. The batch Bayesian optimization method may be, for example, a method such as Local Penalization, but the method is not limited.

[0081] The output unit 17 outputs the design parameter group acquired by the design parameter group acquisition unit 16. That is, the output unit 17 outputs the design parameter group obtained based on the performance data in the material production from the first time (t = 1) to the (T - 1)-th time (t = T - 1) as the design parameter group x T for the production of the T-th material.

[0082] Also, when the design parameter group acquisition unit 16 acquires a plurality of design parameter groups, the output unit 17 outputs the acquired design parameter groups as the design parameter groups for the production of materials for N times after the next time of the (T - 1)-th time. The design parameter groups for the production of materials for multiple times may be used for simultaneous experiments and material production.

[0083] The output mode is not limited, but the output unit 17 outputs the design parameter group candidates, for example, by causing them to be displayed on a predetermined display device or stored in a predetermined storage means.

[0084] FIG. 6 is a flowchart showing the measurement items in material design and the process of optimizing the design parameter group.

[0085] In step S1, a design parameter group is acquired. The design parameter group acquired here is for the initial material production (experiment) and may be an arbitrarily set design parameter group or a design parameter group set based on experiments that have already been conducted.

[0086] In step S2, material production is performed. In step S3, the measured values of the measurement items of the produced material are acquired. The pair of the design parameter group as the production condition in step S2 and the measured value of each measurement item acquired in step S3 constitutes the performance data.

[0087] In step S4, it is determined whether a predetermined end condition is satisfied. The predetermined end condition is a condition for optimizing the design parameter group and the evaluation value, and may be arbitrarily set. The end condition for optimization may be, for example, reaching a predetermined number of times of production (experiment) and acquisition of measurement values, reaching a target value of the evaluation value, and convergence of optimization. When it is determined that the predetermined end condition is satisfied, the optimization process ends. When it is not determined that the predetermined end condition is satisfied, the process proceeds to step S5.

[0088] In step S5, the design support process by the design support device 10 is performed. The design support process is a process of outputting a design parameter group for the next material production. Then, the process returns to step S1 again.

[0089] Note that in the first cycle of the processing cycle composed of steps S1 to S5, when a plurality of pairs of the design parameter group and the measured values of the measurement items are obtained as initial data, the processing of steps S1 to S4 is omitted. When the initial data cannot be obtained, in step S1, a design parameter group obtained by an arbitrary method such as an experimental design method and a random search is acquired. In the second and subsequent cycles of the processing cycle, in step S1, the design parameter group output in step S5 is acquired.

[0090] FIG. 7 is a flowchart showing an example of the content of the design support method in the design support device 10 according to the embodiment, and shows the processing of step S5 in FIG. 6. The design support method is executed by the design support program P1 being read into the processor 101 and the program being executed to realize each functional unit 11 to 17.

[0091] In step S11, the data acquisition unit 11 acquires a plurality of performance data regarding the produced material. The performance data consists of pairs of the design parameter group and the respective measured values of the measurement items.

[0092] In step S12, the model construction unit 12 constructs a prediction model based on the performance data.

[0093] In step S13, the sampling unit 13 samples a plurality of measurement value groups sampled from the multidimensional probability distribution of the measurement values obtained from each prediction model based on one design parameter group x as one sampling point based on the prediction model, and performs sampling of the measurement value groups of a predetermined number of points.

[0094] In step S14, the evaluation value calculation unit 14 calculates the evaluation value of each sampling point by scalarizing the vector having the measurement values of each measurement item of the measurement value group by a predetermined operation.

[0095] In step S15, the acquisition function evaluation unit 15 outputs a predetermined acquisition function evaluation value based on the distribution of the evaluation values of each sampling point.

[0096] In step S16, the design parameter group acquisition unit 16 acquires at least one design parameter group by optimizing the acquisition function evaluation value obtained by the acquisition function evaluation unit 15 in step S15.

[0097] In step S17, the output unit 17 outputs the design parameter group acquired by the design parameter group acquisition unit 16 in step S16 as the design parameter group for the next material production (step S1).

[0098] Next, a design support program for causing a computer to function as the design support apparatus 10 of the present embodiment will be described. FIG. 8 is a diagram showing the configuration of the design support program.

[0099] The design support program P1 includes a main module m10, a data acquisition module m11, a model construction module m12, a sampling module m13, an evaluation value calculation module m14, an acquisition function evaluation module m15, a design parameter group acquisition module m16, and an output module m17 that comprehensively control the design support process in the design support device 10. And, each module m11 to m17 realizes each function for the data acquisition unit 11, the model construction unit 12, the sampling unit 13, the evaluation value calculation unit 14, the acquisition function evaluation unit 15, the design parameter group acquisition unit 16, and the output unit 17.

[0100] Note that the design support program P1 may be transmitted via a transmission medium such as a communication line, or may be stored in the recording medium M1 as shown in FIG. 8.

[0101] According to the design support device 10, the design support method, and the design support program P1 of the present embodiment described above, a prediction model for predicting measurement values of a plurality of measurement items for calculation and evaluation of evaluation values is constructed for each measurement item based on actual performance data. Since this prediction model predicts the measurement value of the measurement item as a probability distribution or its approximation or alternative index, based on the multidimensional distribution of the measurement values obtained from the prediction model of each measurement item, an arbitrary number of measurement value groups can be sampled. By performing an operation on the vector having the measurement value group at each sampling point as an element using a predetermined evaluation formula, an evaluation value regarding each sampling point expressed by a scalar value can be obtained. Then, based on the distribution of the evaluation values of each sampling point, by optimizing the acquisition function evaluation value output using a predetermined acquisition function, a design parameter group suitable for the next experiment or the like can be obtained. Therefore, compared with the method of directly learning the evaluation value to construct the acquisition function, a more accurate prediction model regarding the measurement value can be obtained, and by optimizing the acquisition function evaluation value regarding the evaluation value of the measurement value group obtained by the prediction model, as a result, the suitability of the design parameter group adopted in the experiment or the like is improved, so that the number of experiments can be reduced.

[0102] The present invention has been described in detail based on its embodiments. However, the present invention is not limited to the above embodiments. The present invention can be variously modified without departing from its gist.

Explanation of Signs

[0103] P1... Design support program, m10... Main module, m11... Data acquisition module, m12... Model construction module, m13... Sampling module, m14... Evaluation value calculation module, m15... Acquisition function evaluation module, m16... Design parameter group acquisition module, m17... Output module, 10... Design support device, 11... Data acquisition unit, 12... Model construction unit, 13... Sampling unit, 14... Evaluation value calculation unit, 15... Acquisition function evaluation unit, 16... Design parameter group acquisition unit, 17... Output unit, 21... Design parameter storage unit, 22... Measured value storage unit.

Claims

1. In the design of a product, work-in-progress, semi-finished product, component, or prototype produced based on a group of design parameters consisting of a plurality of design parameters, for the purpose of applying a method of optimizing the design parameters by repeating the determination of the design parameters and the production of the product, work-in-progress, semi-finished product, component, or prototype based on the determined design parameters, a design support device for obtaining the plurality of design parameters such that one evaluation value indicating the characteristics of the product, work-in-progress, semi-finished product, component, or prototype is improved, comprising: The one evaluation value is calculated based on measurement values of a plurality of measurement items. A data acquisition unit that acquires a plurality of performance data consisting of the group of design parameters and respective measurement values of the plurality of measurement items for the manufactured product, work-in-progress, semi-finished product, component, or prototype. A model construction unit that constructs a prediction model for predicting the measurement value of the measurement item as a probability distribution, an approximation thereof, or an alternative index based on the group of design parameters, based on the performance data. A sampling unit that samples a group of measurement values consisting of a plurality of measurement values sampled from a multi-dimensional probability distribution of measurement values obtained from each prediction model as one sampling point, with a predetermined number of points. An evaluation value calculation unit that calculates an evaluation value of the group of measurement values of each sampling point by scalarizing a vector having the number of measurement values included in the group of measurement values as the number of dimensions and the value of each measurement value as an element by a predetermined operation. An acquisition function evaluation unit that takes the group of design parameters as an input and outputs an acquisition function evaluation value related to the improvement of the evaluation value by a predetermined acquisition function based on the distribution of the evaluation values of each sampling point. A design parameter group acquisition unit that acquires at least one group of design parameters by optimizing the acquisition function evaluation value. An output unit that outputs the group of design parameters acquired by the design parameter group acquisition unit. A design support device comprising the above.

2. The evaluation value calculation unit calculates the evaluation value by a predetermined evaluation formula that performs a predetermined operation based on a plurality of measurement values included in the group of measurement values. The design support device according to Claim 1.

3. The evaluation value calculation unit calculates the characteristic value as the evaluation value by a theoretical formula that calculates a characteristic value indicating the characteristics of the product, work-in-progress, semi-finished product, component, or prototype based on a plurality of measurement values included in the group of measurement values. The design support device according to Claim 1.

4. When the measured value sampled by the sampling unit using the prediction model of the measurement item is not included in a given domain related to the measurement item, the sampling unit replaces the sampled measured value with a predetermined value preset for the measurement item. The design support device according to any one of claims 1 to 3.

5. The acquisition function evaluation unit outputs the acquisition function evaluation value by any one of acquisition functions of LCB (Lower Confidence Bound), UCB (Upper Confidence Bound), EI (Expected Improvement), and PI (Probability of Improvement). The design support device according to any one of claims 1 to 4.

6. The design parameter group acquisition unit acquires one group of design parameters that optimizes the acquisition function evaluation value. The design support device according to any one of claims 1 to 5.

7. The design parameter group acquisition unit acquires a plurality of the design parameter groups by a predetermined algorithm. The design support device according to any one of claims 1 to 5.

8. The prediction model is a regression model or a classification model that takes the design parameter group as an input and outputs the probability distribution of the measured value. The model construction unit constructs the prediction model by machine learning using the performance data. The design support device according to any one of claims 1 to 7.

9. The prediction model is a machine learning model that predicts the probability distribution of the measured value or its approximation or alternative index using any one of the posterior distribution of the predicted value based on Bayes' theory, the distribution of the predicted values of the predictors constituting the ensemble, the theoretical formulas of the prediction interval and the confidence interval of the regression model, Monte Carlo dropout, and the distribution of the predictions of a plurality of predictors constructed under different conditions. The design support device according to claim 8.

10. In the design of a product, work-in-progress, semi-finished product, component, or prototype produced based on a set of design parameters consisting of a plurality of design parameters, in order to apply a method for optimizing the design parameters by repeating the determination of the design parameters and the production of the product, work-in-progress, semi-finished product, component, or prototype based on the determined design parameters, a design support method in a design support apparatus for obtaining the plurality of design parameters such that a single evaluation value indicating the characteristics of the product, work-in-progress, semi-finished product, component, or prototype is improved, the single evaluation value is calculated based on measurement values of a plurality of measurement items, the design support method includes: a data acquisition step of acquiring a plurality of pieces of performance data including the set of design parameters and respective measurement values of the plurality of measurement items for the produced product, work-in-progress, semi-finished product, component, or prototype; a model construction step of constructing a prediction model for predicting the measurement values of the measurement items as a probability distribution or an approximation or alternative index thereof based on the set of design parameters, based on the performance data; a sampling step of sampling the measurement value group consisting of a plurality of measurement values sampled from the multidimensional probability distribution of the measurement values obtained from each prediction model as one sampling point, by sampling the measurement value group a predetermined number of times; an evaluation value calculation step of calculating the evaluation value of the measurement value group of each sampling point by scalarizing a vector having the number of measurement values included in the measurement value group as the number of dimensions and the value of each measurement value as an element, by a predetermined operation; an acquisition function evaluation step of inputting the set of design parameters based on the distribution of the evaluation values of each sampling point, and outputting an acquisition function evaluation value regarding the improvement of the evaluation value by a predetermined acquisition function; a design parameter set acquisition step of acquiring at least one set of design parameters by optimizing the acquisition function evaluation value; an output step of outputting the set of design parameters acquired in the design parameter set acquisition step; and having the design support method.

11. A design support program for causing a computer to function as a design support device for obtaining a plurality of design parameters such that a single evaluation value indicating characteristics of a product, work-in-progress, semi-finished product, component, or prototype improves, in order to apply it to a method of optimizing design parameters by repeating determination of design parameters and production of a product, work-in-progress, semi-finished product, component, or prototype based on the determined design parameters, in the design of a product, work-in-progress, semi-finished product, component, or prototype produced based on a group of design parameters consisting of a plurality of design parameters. The single evaluation value is calculated based on measurement values of a plurality of measurement items. The design support program has a data acquisition function for acquiring a plurality of performance data consisting of the group of design parameters and respective measurement values of the plurality of measurement items for the manufactured product, work-in-progress, semi-finished product, component, or prototype. has a model construction function for constructing a prediction model that predicts the measurement values of the measurement items as a probability distribution or an approximation or alternative index thereof based on the group of design parameters, based on the performance data. has a sampling function for sampling a group of measurement values consisting of a plurality of measurement values sampled from a multi-dimensional probability distribution of measurement values obtained from each prediction model as one sampling point, by sampling the group of measurement values by a predetermined number of points. has an evaluation value calculation function for calculating an evaluation value of the group of measurement values of each sampling point by scalarizing a vector having the number of measurement values included in the group of measurement values as the number of dimensions and the value of each measurement value as an element by a predetermined operation. has an acquisition function evaluation function for inputting the group of design parameters and outputting an acquisition function evaluation value regarding improvement of the evaluation value by a predetermined acquisition function based on the distribution of the evaluation values of each sampling point. has a design parameter group acquisition function for obtaining at least one group of design parameters by optimizing the acquisition function evaluation value. has an output function for outputting the group of design parameters obtained by the design parameter group acquisition function. A design support program for realizing the above.

Images