Trial production condition proposal system and trial production condition proposal method

Abstract

The present invention provides a technology with which it is possible to accurately propose favorable trial production conditions for materials even if the parameter space to be searched is wide with respect to the computational resources used. This trial production condition proposal system 1 is a system for proposing material trial production conditions to material developers, and is provided with a regression model construction processing unit 112 and a trial production condition proposal processing unit 113. The regression model construction processing unit 112 executes a regression model construction process on property measurement data representing actual measurement results of material properties. The trial production condition proposal processing unit 113 performs an optimization process to search for optimal trial production conditions for a material using the constructed regression model, and executes a trial production condition proposal process on the basis of the result of the optimization process.

Description

Prototype condition proposal system and method

[0001] The present invention relates to a system and method for proposing material prototype conditions to material developers.

[0002] In the field of materials science, which conducts research and development of materials, a technique called materials informatics (MI) is now widely used, which utilizes information technologies such as statistical analysis and machine learning to efficiently predict the physical properties and structure of materials. Regarding the research and development of materials using materials informatics, for example, a technique described in Patent Document 1 is known. Patent Document 1 discloses a system that estimates the preparation conditions for a material having optimal physical properties and structure from a data set including the preparation conditions for each of a plurality of materials used as samples and material information representing the physical properties and structure of each material.

[0003] International Publication No. 2021 / 044913

[0004] In materials development using materials informatics, the properties of the material to be developed are typically measured through various evaluation tests to obtain data representing the measurement results (hereinafter referred to as "measured property data"). This measured property data is then input into a computer to construct various trained machine learning models. These trained machine learning models are then used to estimate the conditions for prototyping the material (hereinafter referred to as "prototyping conditions").

[0005] The above-mentioned characteristic measurement data typically contains an extremely large number of explanatory variables. When the number of explanatory variables is extremely large, the number of combinations of explanatory variables also becomes enormous. Therefore, when attempting to estimate favorable prototyping conditions for a material based on materials informatics, it is often necessary to search a very wide parameter space. However, there is generally a limit to the computational resources that can be devoted to executing such processing on a computer. For this reason, when estimating favorable prototyping conditions for a material using realistic computational resources in a realistic computational time, the parameter space to be searched may be too wide for the available computational resources, and it may be impossible to estimate favorable prototyping conditions in a realistic computational time.

[0006] Furthermore, the number of the above-mentioned measured property data (hereinafter also referred to as the "sample number") is often small, and the distribution of explanatory variables contained in each measured property data is often biased. Therefore, in the parameter space searched to estimate favorable prototyping conditions for a material, when a region far from the distribution region of explanatory variables contained in the measured property data (hereinafter also referred to as the "training data") input into a computer to build a machine learning model is searched, the estimation accuracy is lower than when a region close to the distribution region of the explanatory variables is searched. For this reason, the region of the parameter space to be searched that can be searched accurately is limited to the vicinity of the distribution region of the explanatory variables contained in the training data. Therefore, even if the entire parameter space to be searched is thoroughly searched, the entire parameter space cannot be searched with uniform accuracy, and favorable prototyping conditions for the material may not be estimated.

[0007] In view of the above-mentioned problems, the present invention aims to provide a technology that can accurately propose favorable prototyping conditions for materials even when the parameter space to be searched is broad compared to the computational resources used.

[0008] A prototype condition proposal system according to the present invention proposes prototype conditions for a material to a material developer and includes a regression model construction processing unit and a prototype condition proposal processing unit. The regression model construction processing unit executes a regression model construction process for property measurement data representing actual measurement results of the material's properties. The prototype condition proposal processing unit executes an optimization process to search for optimal prototype conditions for the material using the constructed regression model, and executes a prototype condition proposal processing based on the results of the optimization process. A prototype condition proposal method according to the present invention is a method for proposing prototype conditions for a material to a material developer using a computer. This prototype condition proposal method causes a computer to execute a regression model construction process and a prototype condition proposal processing. The regression model construction process represents a process of constructing a regression model for property measurement data representing actual measurement results of the material's properties. The prototype condition proposal processing represents a process of performing an optimization process to search for optimal prototype conditions for the material using the constructed regression model, and proposing prototype conditions for the material based on the results of the optimization process.

[0009] Other problems and solutions disclosed in this application will be made clear by the description of the preferred embodiment and drawings.

[0010] According to the present invention, even if the parameter space to be searched is broad compared to the computational resources used, it is possible to accurately propose favorable prototyping conditions for materials.

[0011] A schematic diagram showing an overview of a prototype condition proposal system according to one embodiment of the present invention. A diagram showing functional blocks of a prototype condition proposal system according to one embodiment of the present invention. A flowchart showing the overall processing flow of a prototype condition proposal system according to one embodiment of the present invention. A flowchart showing details of characteristic measurement data preprocessing. A flowchart showing details of regression model construction processing. A flowchart showing details of prototype condition proposal processing. A diagram showing the outline of a focused search.

[0012] This embodiment will be described in detail below. FIG. 1 is a schematic diagram illustrating an overview of a prototype condition proposal system according to one embodiment of the present invention. The prototype condition proposal system 1 shown in FIG. 1 optimizes various prototype conditions that should be considered when prototyping materials, such as material composition and firing conditions, and proposes the results to a material developer, who is a user of the system. The prototype condition optimization is performed using various machine learning algorithms based on actual property measurement data representing the actual measurement results of the properties of the material. In addition, various regression models, such as Gaussian Process Regression (GPR), Linear Regression, Regression Tree (including cases using an ensemble method), Neural Network Regression, Support Vector Regression (SVR), Logistic Regression, and LASSO Regression (Least Absolute Shrinkage and Selection Operator Regression), are used as prediction models.

[0013] As shown in FIG. 1 , when a user of the prototype conditions proposal system 1 prototypes a material under the prototype conditions proposed by the prototype conditions proposal system 1 and measures the properties of the resulting material, new property measurement data representing the evaluation values ​​of the properties of the resulting material is generated. The prototype conditions proposal system 1 then learns this property measurement data as new learning data, and proposes more optimized prototype conditions to the user. Each time the user repeats this cycle, the prototype conditions proposal system 1 of this embodiment can propose prototype conditions with more favorable predicted property values. The prototype conditions proposal system 1 may also include a function for prototyping a material and a function for measuring the properties of the material to be prototyped, or may be configured integrally with these functions.

[0014] The prototype conditions proposal system 1 of this embodiment is realized by a single general-purpose computer device, as shown in Fig. 1. In the following description, the prototype conditions proposal system 1 is assumed to be realized by a single general-purpose computer device that includes one or more processor devices, one or more storage devices, one or more input / output devices, and wired or wireless communication lines (none of which are shown) connecting these devices.

[0015] This computer device is installed as a terminal inside a laboratory, for example, and is connected to various other terminals installed inside and outside the laboratory, various terminals owned by each user such as laptop PCs, tablets, smartphones, etc. (hereinafter referred to as "user terminals"), other devices such as server devices, etc., via a communication network such as the Internet 400 or a dedicated line. Note that the computer device and the Internet 400 are connected by wire via well-known communication equipment (not shown), but may also be connected wirelessly.

[0016] Next, various functions of the prototype conditions proposal system 1 will be described with reference to FIG. 2. FIG. 2 is a diagram showing functional blocks of the prototype conditions proposal system 1 according to one embodiment of the present invention. Note that each block described below does not represent a hardware-based configuration, but rather represents a functional block. As shown in FIG. 2, the prototype conditions proposal system 1 of this embodiment is configured to include a control unit 11, a storage unit 12, a user interface unit 13, and a communication unit 14.

[0017] The control unit 11 executes various data processing operations based on user operation input detected by the user interface unit 13, data acquired by the communication unit 14, and programs and data stored in the storage unit 12. The control unit 11 also functions as an interface between the user interface unit 13, the communication unit 14, and the storage unit 12.

[0018] The control unit 11 has functional blocks including a characteristic measurement data preprocessing unit 111, a regression model construction processing unit 112, and a prototype condition proposal processing unit 113. The control unit 11 is configured using processor devices (hereinafter simply referred to as "processors") such as a CPU (Central Processing Unit) and various coprocessors, and can realize these functional blocks by executing a predetermined program. Note that the control unit 11 may be configured using a logic circuit such as an FPGA (Field Programmable Gate Array) instead of a processor. Alternatively, the control unit 11 may be configured by combining a processor and a logic circuit.

[0019] The program executed by the control unit 11 may be installed from a program source. The program source may be, for example, a program distribution computer or a computer-readable recording medium. The program executed by the control unit 11 may also be composed of a device driver, an operating system, various application programs located at higher levels than those, and a library that provides common functions to these programs. Furthermore, two or more programs may be realized as a single program, or one program may be realized as two or more programs.

[0020] The characteristic measurement data preprocessing unit 111 preprocesses the characteristic measurement data in the raw data state immediately after recording. This processing performed by the characteristic measurement data preprocessing unit 111 is referred to as characteristic measurement data preprocessing.

[0021] The regression model construction processing unit 112 executes a process of constructing a regression model for the preprocessed characteristic measurement data. This process executed by the regression model construction processing unit 112 is referred to as a regression model construction process.

[0022] The prototype condition proposal processing unit 113 performs optimization processing for searching for optimal prototype conditions for the material using the regression model constructed by the regression model construction processing unit 112 for the preprocessed characteristic measurement data, and then performs processing to propose prototype conditions for the material to the user based on the results of the optimization processing. This processing performed by the prototype condition proposal processing unit 113 is referred to as the prototype condition proposal processing.

[0023] The specific contents of these processes will be described later.

[0024] The storage unit 12 is configured using a storage device such as a RAM or a flash memory, and stores programs that supply various processing commands to the control unit 11 and data representing various information used in the processing executed by the control unit 11. For example, the storage unit 12 stores characteristic measurement data that has been preprocessed by the characteristic measurement data preprocessing unit 111 (hereinafter referred to as "preprocessed data") and data representing the regression model constructed by the regression model construction processing unit 112. The control unit 11 can implement the functional blocks of the characteristic measurement data preprocessing unit 111, the regression model construction processing unit 112, and the prototype condition proposal processing unit 113 by reading and writing this information from and to the storage unit 12.

[0025] The user interface unit 13 accepts input operations from the user and is responsible for processing related to the user interface, such as image display and audio output. The user interface unit 13 has functional blocks of an input unit 131 and an output unit 132. The input unit 131 detects various operations from the user. The input unit 131 is configured using, for example, a keyboard, a pointing device, a touch panel, etc. The output unit 132 executes screen display, audio output, etc. for the user. The output unit 132 is configured using, for example, a liquid crystal display, a touch screen, etc.

[0026] The communication unit 14 is responsible for communication processing with user terminals owned by each user and other devices such as server devices via the Internet 400. The communication unit 14 is configured using, for example, a network interface card (NIC) or a host bus adapter (HBA).

[0027] In this embodiment, the functions of the prototype conditions proposal system 1 are described as being implemented integrally by a single computer. However, these functions may be implemented by multiple interconnected computers or server devices. Furthermore, the prototype conditions proposal system 1 may be configured to include a general-purpose computer such as a laptop PC and a web browser installed thereon, or may be configured to include a web server and various portable devices.

[0028] Furthermore, the description of each function is an example, and multiple functions may be combined into one function, or one function may be divided into multiple functions.

[0029] Next, the overall processing flow of the prototype conditions proposal system 1 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the overall processing flow of the prototype conditions proposal system 1 according to one embodiment of the present invention. Note that in the following description, the processing may be described with the above-mentioned functions or programs as the subject, but the processing described with the function or program as the subject may also be processing performed by a processor or a device having that processor.

[0030] In step S310, the control unit 11 executes characteristic measurement data preprocessing using the characteristic measurement data preprocessing unit 111. As a result, the characteristic measurement data is preprocessed to become preprocessed data, and subsequent processes can be executed normally. Details of the characteristic measurement data preprocessing performed in step S310 will be described later with reference to the flowchart in FIG. 4. When the characteristic measurement data preprocessing is completed, the control unit 11 proceeds to step S320.

[0031] In step S320, the control unit 11 executes a regression model construction process using the regression model construction processing unit 112. As a result, a regression model is constructed for the preprocessed data. Details of the regression model construction process performed in step S320 will be described later with reference to the flowchart in FIG. 5. When the regression model construction process is completed, the control unit 11 proceeds to step S330.

[0032] In step S330, the control unit 11 executes a regression model evaluation process. In this regression model evaluation process, the generalization performance, which is an index representing the predictive accuracy of a regression model, is evaluated for each of the multiple regression models constructed as a result of the processes up to step S320. This evaluation is performed, for example, by performing cross-validation between the regression model and other regression models. The evaluation results are visualized using graphs such as scatter plots and box-and-whisker plots. This allows the user to receive suggestions for prototyping conditions based on regression models with good generalization performance. When the regression model evaluation process is complete, the control unit 11 proceeds to step S340.

[0033] In step S340, the control unit 11 executes a prototype condition proposal process using the prototype condition proposal processing unit 113. In this prototype condition proposal process, the user of the prototype condition proposal system 1 can appropriately modify the prototype conditions for the material proposed by the prototype condition proposal system 1 to further improve the desired properties. The control unit 11 calculates predicted values ​​for the material properties when the prototype is produced under the prototype conditions modified by the user by applying the selected regression model, and presents the results to the user. In other words, the user can interactively modify the prototype conditions while checking the predicted values. In this way, the prototype condition proposal system 1 is a system that can incorporate the knowledge of the user, the developer of the material, into the prototype conditions for the material proposed to the user. Details of the prototype condition proposal process performed in step S340 will be described later with reference to the flowchart of FIG. 6. Upon completing the prototype condition proposal process, the control unit 11 temporarily terminates the process shown in the flowchart of FIG. 3.

[0034] The prototype condition proposal system 1 of this embodiment executes the processes of steps S310 to S340 in FIG. 3 to propose favorable prototype conditions to the user. That is, in step S310, the prototype condition proposal system 1 of this embodiment automatically performs the necessary preprocessing on the characteristic measurement data in the raw data state. Therefore, the prototype condition proposal system 1 of this embodiment can execute the processes of steps S320 to S340 without manually performing complex preprocessing on the characteristic measurement data. Furthermore, the prototype condition proposal system 1 of this embodiment allows the user to select a regression model with good generalization performance. Therefore, the prototype condition proposal system 1 can propose to the user prototype conditions that predict favorable characteristic values.

[0035] As described above in relation to FIG. 1 , a user of the prototype condition proposal system 1 can prototype a material under the prototype conditions proposed by the prototype condition proposal system 1, measure the properties of the resulting product, and have the prototype condition proposal system 1 learn the data representing the actual measurement results as new property measurement data. The user can then have the prototype condition proposal system 1 execute steps S310 to S340 of FIG. 3 again. In this case, the prototype condition proposal system 1 can propose more optimized prototype conditions to the user. That is, the prototype condition proposal system 1 of this embodiment can propose prototype conditions with better predicted property values ​​each time it repeats steps S310 to S340 of FIG. 3 for the same prototype object.

[0036] FIG. 4 is a flowchart showing the details of the characteristic measurement data preprocessing.

[0037] In step S410, the control unit 11 causes the characteristic measurement data preprocessing unit 111 to accept input of characteristic measurement data from the user via the input unit 131 or the communication unit 14. The characteristic measurement data input to the prototype condition proposal system 1 may be, for example, categorical data, continuous data, or discrete data. The specific data format of the characteristic measurement data input to the prototype condition proposal system 1 can be determined as appropriate. When the control unit 11 completes the processing in step S410, the process proceeds to step S420.

[0038] In step S420, the control unit 11 sets the type of variables for the characteristic measurement data input received from the user in step S410, using the characteristic measurement data preprocessing unit 111. Here, either an explanatory variable or a response variable is set. The explanatory variable is a variable that serves as the basis for calculating the predicted value of a characteristic. In this embodiment, the composition of the material that constitutes the prototype conditions, the firing conditions, etc., correspond to the explanatory variables. The response variable is a variable that represents the characteristic value of the prototype material that is the target of prediction. As a specific example of the processing in step S420, the explanatory variables may be set as defaults, and a setting operation may be received from the user who wishes to change the response variable. When the processing in step S420 is completed, the control unit 11 proceeds to step S430.

[0039] In step S430, the control unit 11, via the characteristic measurement data preprocessing unit 111, determines whether or not there is an abnormal value in the characteristic measurement data for which either one of the explanatory variables or the objective variable was set in step S420. This abnormal value determination process is performed, for example, by first displaying the characteristic measurement data in a histogram and determining whether or not there is an outlier outside the range of the mean ±2σ, and then determining whether or not there was a data input error or a malfunction of the evaluation tester when generating the characteristic measurement data for which an outlier was determined to exist. If it is determined that the characteristic measurement data contains an abnormal value, the control unit 11 deletes the abnormal value and sets it as a missing value, and proceeds to step S440. Furthermore, if the type of abnormal value contained in the characteristic measurement data is one of the objective variables in a case where the explanatory variables completely overlap and the characteristics have different levels, the control unit 11 deletes the sample corresponding to the abnormal value, and proceeds to step S440. This is because, when all explanatory variables are overlapped and levels with different characteristics are treated as abnormal values, the level itself must be deleted, unlike when one explanatory variable has an abnormal value due to an input error, etc. On the other hand, when it is determined that the characteristic measurement data does not contain any abnormal values, the control unit 11 proceeds directly to step S440.

[0040] In the case where all the explanatory variables overlap and there are different levels of characteristics, as described above, it may be desirable to retain both explanatory variables because the overlap is meaningful. In such a case, the prototype conditions proposal system 1 of this embodiment can omit the process in step S430.

[0041] In step S440, the control unit 11 determines whether the characteristic measurement data contains missing values ​​using the characteristic measurement data preprocessing unit 111. This is because the characteristic measurement data may contain missing values ​​in advance. If it is determined that the preprocessed data contains missing values, the control unit 11 complements the missing values ​​and proceeds to step S450. The missing value complementation process is performed, for example, by using the mean, median, minimum, or maximum value of the characteristic measurement data excluding outliers as a value to complement the missing value. Alternatively, the missing value complementation may be performed using linear interpolation. For example, when the characteristic measurement data preprocessing unit 111 complements a missing value in step S440, the characteristic measurement data preprocessing unit 111 may display the complemented value in red to make it easier to identify. Alternatively, for example, in step S440, the characteristic measurement data preprocessing unit 111 may delete the level itself without complementing the missing value. Furthermore, in the prototype condition proposal system 1 of this embodiment, the characteristic measurement data preprocessing unit 111 can delete the explanatory variables themselves when the proportion of missing values ​​is high, for example, when 50% or more of the data is missing. On the other hand, if it is determined that the characteristic measurement data does not contain missing values, the control unit 11 proceeds directly to step S450.

[0042] In step S450, if the explanatory variables are categorical values ​​rather than continuous values, the control unit 11 causes the characteristic measurement data preprocessing unit 111 to perform encoding processing on the explanatory variables to convert them into numerical data. The characteristic measurement data preprocessing unit 111 executes this encoding processing by, for example, referring to records in a table that represents the correspondence between categorical data and numerical data, which is stored in the storage unit 12. When the control unit 11 completes the processing in step S450, it proceeds to step S460.

[0043] In step S460, the control unit 11, via the characteristic measurement data preprocessing unit 111, determines whether the characteristic measurement data contains redundant explanatory variables. This determination is based on whether a combination of explanatory variables with a correlation coefficient of a predetermined number or more, for example, 0.8 or more, can be extracted. If it is determined that the characteristic measurement data contains redundant explanatory variables, the control unit 11 deletes one of the redundant explanatory variables and proceeds to step S470. Note that in the prototype condition proposal system 1 of this embodiment, combinations of explanatory variables with a correlation coefficient of 0.8 or more are visualized to the user via the output unit 132, and the user can select the explanatory variable to be deleted via the input unit 131. On the other hand, if it is determined that the characteristic measurement data does not contain redundant explanatory variables, the control unit 11 proceeds directly to step S470.

[0044] In step S470, the control unit 11 performs standardization processing on the characteristic measurement data using the characteristic measurement data preprocessing unit 111, as necessary. This standardization processing converts the scale of the characteristic measurement data so that the mean = 0 and the standard deviation (variance) = 1. Upon completing the processing in step S470, the control unit 11 stores the preprocessed data, which is the characteristic measurement data preprocessed in steps S410 to S470 of FIG. 4, in the storage unit 12, and ends the characteristic measurement data preprocessing shown in the flowchart of FIG. 4. Note that the control unit 11 may also perform normalization processing on the characteristic measurement data using the characteristic measurement data preprocessing unit 111, as necessary. In such a case, the control unit 11 may perform the subsequent processing without performing the standardization processing in S470.

[0045] FIG. 5 is a flowchart showing the details of the regression model construction process.

[0046] In step S510, the control unit 11 selects the implementation conditions of cross-validation for evaluating the regression model constructed for the preprocessed data by the regression model construction processing unit 112. The prototyping condition proposal system 1 of this embodiment evaluates each regression model using K-fold cross validation. Furthermore, K=10 is set as the default implementation condition. In this case, the prototyping condition proposal system 1 evaluates the regression model using 10-fold cross validation. The prototyping condition proposal system 1 of this embodiment also allows the user to select the implementation conditions of cross-validation. That is, the regression model construction processing unit 112 can accept the implementation conditions of cross-validation from the user via the input unit 131 or the communication unit 14. After completing the processing in step S510, the control unit 11 proceeds to step S520.

[0047] In step S520, the control unit 11 causes the regression model construction processing unit 112 to select candidate regression models to be used as predictive models in the search for prototyping conditions. At this time, the regression model construction processing unit 112 selects, as candidates, regression models selected by a user via the input unit 131 or the communication unit 14. In the prototyping condition proposal system 1 of this embodiment, the user can select multiple candidate regression models from various regression models, such as Gaussian process regression, the aforementioned linear regression, regression tree (including the ensemble method), neural network regression, support vector regression, logistic regression, and LASSO regression. Upon completing the processing in step S520, the control unit 11 proceeds to step S530.

[0048] In step S530, the control unit 11 executes a process of calculating a weighting standard, which is a standard for weighting the property measurement data, using the regression model construction processing unit 112. The weighting standard is calculated based on the difference between the objective variable included in the property measurement data and the target property representing the target value of the material property, or a statistic representing the rarity of the explanatory variables included in the property measurement data. Note that a specific example of a statistic representing the rarity of an explanatory variable is the occurrence probability of an explanatory variable that satisfies a predetermined condition. When the control unit 11 completes the process in step S530, it proceeds to step S540.

[0049] In step S540, the control unit 11 causes the regression model construction processing unit 112 to execute a process of weighting the characteristic measurement data based on the weighting criteria calculated in step S530. In the prototype condition proposal system 1 of this embodiment, the regression model construction processing unit 112 can also directly weight the regression model based on the weighting criteria calculated in step S530. The process executed by the regression model construction processing unit 112 in step S540 based on the results of step S530 is, specifically, one of the following processes: a process of setting a loss function, which is a function serving as an index for learning; an oversampling process in which rare or important characteristic measurement data is amplified and added as training data; or an undersampling process in which redundant or unimportant characteristic measurement data is deleted from the training data. By executing these processes, the weights of the training data can be appropriately adjusted even when the number of training data used to construct the machine learning model is small or when the distribution of the training data is biased. As a result, estimation accuracy is improved, enabling accurate proposal of prototype conditions for good materials. When the process in step S540 is completed, the control unit 11 proceeds to step S550.

[0050] In step S550, the control unit 11 causes the regression model construction processing unit 112 to search for and set optimal hyperparameters for each regression model of each method selected as a candidate in step S520. In the prototype condition proposal system 1 of this embodiment, the regression model construction processing unit 112 automatically searches for all parameters for each regression model and automatically sets the parameters that will maximize the generalization performance of the regression model when set as hyperparameters during regression model construction. When the control unit 11 completes the processing in step S550, the control unit 11 proceeds to step S560.

[0051] In step S560, the control unit 11 causes the regression model construction processing unit 112 to perform a process of creating a regression model for each method in which optimal hyperparameters have been set. After creating the regression models for each method, the regression model construction processing unit 112 selects the regression model with the highest generalization performance from all the created regression models and performs a process of determining the final regression model. Upon completing the process in step S560, the control unit 11 ends the regression model construction process shown in the flowchart of FIG. 5.

[0052] FIG. 6 is a flowchart showing the details of the prototype condition proposal process.

[0053] In step S610, the control unit 11 causes the prototyping condition proposal processing unit 113 to search for prototyping conditions based on the regression model constructed in step S560 of FIG. 5 . The prototyping condition proposal processing unit 113 performs this processing using optimization (described in detail below). The prototyping condition proposal system 1 of this embodiment is configured to be able to use various optimization processing techniques, such as mathematical optimization (MO), Bayesian optimization (BO), genetic algorithm (GA), Newton's method (NM), and simplex method (SM). As a result, each explanatory variable that optimizes the predicted value of the characteristic (target variable) is proposed to the user as a provisional prototyping condition representing the results of the search processing. Furthermore, the prototyping condition proposal processing unit 113 then performs sensitivity analysis on the provisional prototyping conditions to evaluate the importance of each explanatory variable constituting the provisional prototyping conditions and presents the evaluation results to the user. In the prototype condition proposal system 1 of this embodiment, when the regression model used is Gaussian process regression, it is also possible to select the prototype conditions that maximize the acquisition function. When the control unit 11 completes the process in step S610, the control unit 11 proceeds to step S620.

[0054] In step S620, the control unit 11 receives, from the user, a modification to the provisional prototyping conditions proposed to the user in step S610 via the prototyping condition proposal processing unit 113. When the prototyping condition proposal processing unit 113 receives an input operation from the user via the input unit 131 or the communication unit 14 to modify the values ​​of the explanatory variables constituting the provisional prototyping conditions, the control unit 11 modifies the provisional prototyping conditions according to the modification. At this time, the control unit 11 uses the prototyping condition proposal processing unit 113 to calculate predicted values ​​of material properties when a prototype is produced under the modified provisional prototyping conditions using a regression model and presents the calculation results to the user. Similarly to step S610, the prototyping condition proposal processing unit 113 also performs a sensitivity analysis on the modified provisional prototyping conditions to evaluate the importance of the explanatory variables constituting the modified provisional prototyping conditions and presents the evaluation results. The evaluation results are updated each time the user modifies the provisional prototyping conditions, so that the latest evaluation results are always presented to the user. When the process in step S620 is completed, the control unit 11 proceeds to step S630.

[0055] In step S630, the control unit 11, via the prototyping condition proposal processing unit 113, determines whether the predicted values ​​of the properties calculated in step S620 for the revised provisional prototyping conditions are insufficient as the property values ​​of the material to be prototyped. For example, if the regression model used is Gaussian process regression, this determination is made by calculating an acquisition function (Acquisition Function) for each provisional prototyping condition, which represents the expected value of the improvement in the material properties when prototyped under the prototype conditions, and determining whether the difference between the value of the acquisition function and the maximum value of the acquisition function is within a predetermined range. The acquisition function is calculated based on the predicted value μ of the material properties when prototyped under any prototype conditions and the standard deviation σ, which represents the variance of the predicted value. If it is determined that the predicted property values ​​are insufficient, the process returns to step S620 and accepts a modification instruction for the prototyping conditions from the user again. If it is determined that the predicted property values ​​are not insufficient, the modified provisional prototyping conditions are sufficient, so the modified provisional prototyping conditions are finalized and proposed to the user as finalized prototyping conditions. In other words, this determination process is repeated until it is determined that the predicted property values ​​of the material related to the provisional prototyping conditions are not insufficient. Upon completion of the process in step S630, the control unit 11 terminates the prototyping condition proposal process shown in the flowchart of FIG. 6.

[0056] In step S610, the prototype conditions proposal processor 113 may perform optimization processing using a sampling method or may perform continuous optimization processing. The sampling method is a method in which multiple prototype level candidates are generated and the prototype level candidate with the best characteristics is selected. In the prototype conditions proposal system 1 of this embodiment, the user can select one of these two execution methods when executing the optimization processing in step S610. Of these, the specific content of the processing executed in step S610 when the prototype conditions proposal processor 113 performs optimization processing using the sampling method is as shown below as steps S611 to S616.

[0057] In step S611, the control unit 11 receives an input operation from the user via the input unit 131 specifying the number of prototype level candidates to be created, and then the control unit 11 performs processing to create the specified number of prototype level candidates via the prototype condition proposal processing unit 113. This processing is referred to as the prototype level candidate creation processing. The prototype level candidate creation processing includes processing to have the computer repeatedly execute each of the processing steps S614 to S616, described below, a specified number of times. As a result of the prototype level candidate creation processing, the number of prototype level candidates specified by the user is created. In the prototype condition proposal system 1 of this embodiment, 10,000 prototype level candidates are created as specified by the user. When the processing in step S611 is completed, the control unit 11 proceeds to step S612.

[0058] In step S612, the control unit 11 causes the prototype condition proposal processing unit 113 to perform a process of calculating predicted values ​​or acquisition functions for all prototype level candidates created in step S611. This process is referred to as the calculation process. As a result of the calculation process, predicted values ​​or acquisition functions are calculated for all prototype level candidates created in step S611. When the control unit 11 completes the process in step S612, the control unit 11 proceeds to step S613.

[0059] In step S613, the control unit 11 causes the prototyping condition proposal processing unit 113 to perform a process of extracting, from all of the prototype level candidates created in step S611, the prototype condition with the best predicted value or acquisition function, based on the predicted value or acquisition function calculated in step S612. This process is referred to as the extraction process. The prototype condition proposal processing unit 113 performs the extraction process in step S613 using the various optimization process techniques described above in connection with step S610. As a result of the extraction process, from all of the prototype level candidates created in step S611, the prototype level candidate with the best predicted value or acquisition function is extracted as the prototype condition. When the process in step S613 is completed, the control unit 11 terminates the optimization process executed in step S610.

[0060] As described above, the prototype level candidate creation process executed in step S611 includes a process of causing a computer to repeatedly execute the processes of steps S614 to S616, which will be described later, a specified number of times.

[0061] In step S614, the control unit 11, via the prototype condition proposal processing unit 113, performs a process of selecting a reference prototype level from prototype level candidates whose characteristics have been measured. This process is referred to as the selection process. In the selection process, measured characteristic data is used as prototype level candidates whose characteristics have been measured. That is, in the selection process, first, measured characteristic data with good characteristics is extracted from a plurality of measured characteristic data. Next, a selection probability is calculated so that the extracted measured characteristic data with good characteristics are selected as prototype conditions with priority. Furthermore, measured characteristic data is randomly selected according to the calculated selection probability. A combination of materials in this selected measured characteristic data is set as a prototype level candidate. In this way, in the selection process, the reference prototype level is randomly selected according to the selection probability calculated so that, among the measured characteristic data that are prototype level candidates, those with good characteristics are selected as prototype conditions with priority. This makes it easier to select prototype level candidates with good characteristics as prototype conditions. The selection process may also include a process of weighting the selection probability of the reference prototype level. In this case, weighting increases the likelihood of selecting a prototype level that is preferable as the reference. As a result of the selection process, the combination of raw materials for the reference prototype level candidate is determined. When the process in step S614 is completed, the control unit 11 proceeds to step S615.

[0062] In step S615, the control unit 11 performs a process in which the prototype condition proposal processing unit 113 varies the prototype level selected in step S614 using random numbers within the range in which the material can be prototyped. This process is referred to as the variation process. In this variation process, the raw material composition ratio for the prototype level candidate is first randomly set while observing the total value constraint. Next, the average particle size and maximum firing temperature for the prototype level candidate are calculated based on the raw material combination determined in step S614 and the raw material composition ratio determined in step S615. After that, other explanatory variables are set according to the distribution of the explanatory variables contained in each characteristic measurement data. As a result of the variation process, values ​​of various explanatory variables, including the material composition ratio, for the reference prototype level candidate are determined. After completing the process in step S615, the control unit 11 proceeds to step S616.

[0063] In step S616, the control unit 11 causes the prototype condition proposal processing unit 113 to perform a process of saving the prototype level candidates that have been subjected to the variation process in step S615. This process is referred to as the saving process. As a result of the saving process, the prototype level candidates are saved by being written to the memory unit 12. When the control unit 11 completes the process in step S616, it returns to step S614 and again executes the processes of steps S614 to S616 described above. Each of the processes of steps S614 to S616 is repeatedly executed a specified number of times. In the prototype condition proposal system 1 of this embodiment, 10,000 times is specified as the number of times that each of the processes of steps S614 to S616 is repeatedly executed. Therefore, each of the processes of steps S614 to S616 is repeatedly executed 10,000 times. As a result, in the prototype level candidate creation process performed in step S611, prototype level candidates that are necessary and sufficient for accurately extracting prototype conditions with good characteristics are created. When the control unit 11 has repeatedly executed each of the processes in steps S614 to S616 10,000 times, it ends the prototype level candidate generation process in step S611 and proceeds to step S612.

[0064] The prototype level candidate creation process executed in step S611 may further include a determination process for determining whether or not there is a prototype level to be used as a reference, and a reference generation process for generating a reference using random numbers if it is determined in the determination process that there is no reference. In this case, even if there is no reference, by generating the reference using random numbers, it is possible to create prototype level candidates in the same way as when there is a reference.

[0065] On the other hand, when the prototype condition proposal processing unit 113 performs continuous optimization processing, the specific content of the processing executed in step S610 is as shown below as steps S617 to S619.

[0066] In step S617, the control unit 11, through the prototyping condition proposal processing unit 113, performs a process of selecting a prototype level to be used as an initial value from prototype level candidates whose characteristics have been actually measured. This process is referred to as the selection process. In the selection process, measured characteristic data is used as prototype level candidates whose characteristics have been actually measured. That is, in the selection process, measured characteristic data to be used as the initial value of the prototype level is selected from multiple measured characteristic data. In the selection process, the prototype level to be used as the initial value is selected randomly. This prevents bias in the search area within the parameter space when searching for prototype conditions in step S610. Furthermore, the selection process may include a process of weighting the selection probability of the prototype level to be used as the initial value. In this case, weighting in this manner makes it easier to select a prototype level that is preferable as the initial value. When the control unit 11 completes the process in step S617, the control unit 11 proceeds to step S618.

[0067] In step S618, the control unit 11 causes the prototype conditions proposal processing unit 113 to perform a process of setting a penalty for the selected prototype level if it deviates from the predetermined constraints. This process is referred to as the penalty setting process. As a method of setting a penalty, if a penalty is added in step S619, a very large negative number is assigned as the evaluation value for the penalty. Furthermore, if a penalty is multiplied in step S619, 0 is assigned as the evaluation value for the penalty. In the prototype conditions proposal system 1 of this embodiment, if the selected prototype level deviates from the predetermined constraints, 0 is assigned as the evaluation value for the penalty in step S618 by default. Then, by multiplying by 0 in step S619, the prototype level is ignored. Furthermore, in the prototype conditions proposal system 1 of this embodiment, a constraint is set in advance that "the material can be prototyped," or in other words, "only parameter ranges and combinations that allow prototyping of the material are allowed." This imposes constraints from a material science perspective, allowing the prototype condition proposal system 1 to propose only those prototyping conditions that can actually be used to prototype a material. Other specific examples of constraints include "focusing the search on existing property measurement data close to the target physical properties" and "focusing the search on prototyping conditions considered effective from a material science perspective." The number of pre-set constraints may be one or more. That is, a single constraint may be set, or multiple constraints may be set in combination. The prototype condition proposal processing unit 113 may also automatically adjust the pre-set constraints and their internal parameters (e.g., weights for performing a focused search) to optimize the predicted values ​​and acquisition functions. After completing the processing in step S618, the control unit 11 proceeds to step S619.

[0068] In step S619, the control unit 11 causes the prototype conditions proposal processing unit 113 to perform a process of optimizing the prototype conditions so that the predicted value or acquisition function to which the penalty set in step S618 is added or multiplied becomes the best. The prototype conditions proposal processing unit 113 performs this process in step S619 using the various optimization processing techniques described above in connection with step S610. The prototype conditions proposal processing unit 113 also repeatedly performs this process in step S619 a number of times specified by the user. In the prototype conditions proposal system 1 of this embodiment, 10,000 times is specified as the number of times to repeatedly perform the process in step S619. Therefore, the optimization process is repeated 10,000 times in step S619. When the process in step S619 is completed, the control unit 11 terminates the optimization process performed in step S610.

[0069] As in the case of the sampling method, the optimization process executed in step S610 may further include a determination process for determining whether or not there is a reference prototype level, and a reference generation process for generating a reference using random numbers if it is determined in the determination process that there is no reference. In this case, even if there is no reference, by generating a reference using random numbers, it is possible to create prototype level candidates in the same way as when there is a reference.

[0070] As described above, the optimization process executed in step S610 is performed by a sampling method or continuously. When the optimization process is performed by a sampling method, the prototype level candidate with the best properties is selected from multiple prototype level candidates. Therefore, the prototype condition proposal system 1 can estimate the best prototype conditions for the material and propose them as provisional prototype conditions. Furthermore, when the optimization process is performed continuously, the prototype level is more refined than when the optimization process is performed one-off. Therefore, the prototype condition proposal system 1 can estimate the best prototype conditions for the material and propose them as provisional prototype conditions. FIG. 7 shows how the prototype condition proposal processing unit 113 searches for provisional prototype conditions to be proposed by performing the optimization process in step S610 in this manner. Note that FIG. 7 also shows, as a comparative example, how provisional prototype conditions to be proposed are searched for using conventional technology. FIG. 7 shows predicted data for the coefficient of thermal expansion (CTE) and bending strength, which are properties of a ceramic composite material made from a certain number of raw materials, plotted on a scatter plot. As described above, conventional techniques only thoroughly search the entire, broad parameter space, resulting in a sparse search near the target region within the parameter space. In contrast, the prototype condition proposal system 1 of this embodiment performs optimization processing using a sampling method, as shown in steps S611 to S616, or performs continuous optimization processing, as shown in steps S617 to S619, thereby enabling a focused search near the target region within the parameter space. In other words, the prototype condition proposal system 1 of this embodiment increases the influence of characteristic measurement data with good characteristics as a prototype level candidate, thereby focusedly searching for prototype conditions close to the characteristics of the characteristic measurement data, as shown in FIG. 7, and efficiently finding many prototype conditions that can satisfy the desired characteristics.

[0071] As described above, when a user prototypes a material based on the prototype conditions proposed in step S630 of FIG. 6 and data representing the actual measurement results of the properties of the resulting product is input as new property measurement data, the prototype conditions proposal system 1 of this embodiment proposes more optimized prototype conditions to the user based on the newly input property measurement data. In this case, the control unit 11 of the prototype conditions proposal system 1 determines whether the newly input property measurement data contains missing values. If it is determined that the newly input property measurement data contains missing values, the control unit 11 performs property measurement data preprocessing on the newly input property measurement data to fill in the missing values, starting from step S440 of FIG. 4. On the other hand, if it is determined that the newly input property measurement data does not contain missing values, it is not necessary to perform property measurement data preprocessing on the newly input property measurement data. Therefore, in such a case, the control unit 11 determines whether the regression model needs to be updated. If it is determined that the regression model needs to be updated, the control unit 11 executes a regression model construction process for the newly input characteristic measurement data, starting from step S510 in Fig. 5, in order to construct a new regression model. On the other hand, if it is determined that the regression model does not need to be updated, the control unit 11 omits the regression model construction process and the regression model evaluation process, and performs a prototyping condition proposal process for the newly input characteristic measurement data, starting from step S610 in Fig. 6, using the previously constructed regression model. Note that even if the newly input characteristic measurement data contains missing values, if the regression model does not need to be updated, the control unit 11 similarly omits the regression model construction process and the regression model evaluation process.

[0072] According to the embodiment of the present invention described above, the following advantageous effects are achieved.

[0073] (1) The prototype condition proposal system 1 is a system that proposes prototype conditions for a material to a material developer, and includes a regression model construction processing unit 112 and a prototype condition proposal processing unit 113. The regression model construction processing unit 112 executes a regression model construction process (FIG. 5) for characteristic measurement data representing actual measurement results of the material's characteristics (step S320). The prototype condition proposal processing unit 113 executes an optimization process (step S610) to search for optimal prototype conditions for the material using the constructed regression model, and executes a prototype condition proposal process (FIG. 6) based on the results of the optimization process (step S610) (step S340). This configuration allows for accurate proposal of favorable prototype conditions for the material, even if the parameter space to be searched is too broad for the computational resources used.

[0074] (2) The optimization process (step S610) includes a prototype level candidate creation process (step S611) for creating a specified number of prototype level candidates, a calculation process (step S612) for calculating predicted values ​​or acquisition functions for all the created prototype level candidates, and an extraction process (step S613) for extracting the prototype conditions with the best predicted values ​​or acquisition functions. As a result, the prototype level candidate with the best predicted value or acquisition function is extracted as the prototype conditions from all the created prototype level candidates. As a result, the prototype condition proposal system 1 can estimate the best prototype conditions for the material.

[0075] (3) The prototype level candidate creation process (step S611) includes a process of causing a computer to repeatedly execute the following steps a designated number of times: a selection process (step S614) for selecting a reference prototype level from prototype level candidates whose characteristics have been measured; a variation process (step S615) for varying the selected prototype level using random numbers within the range in which the material can be prototyped; and a storage process (step S616) for storing the prototype level candidate that has been subjected to the variation process (step S615). As a result, the prototype level candidate creation process (step S611) creates prototype level candidates that are necessary and sufficient to accurately extract prototype conditions with good characteristics.

[0076] (4) The reference prototype level is selected randomly in the selection process (step S614). This makes it easier to select prototype level candidates with good characteristics as prototype conditions.

[0077] (5) The selection process (step S614) may include a process (not shown) for weighting the selection probability of the reference prototype level. In this case, weighting increases the likelihood of selecting a preferred prototype level as the reference.

[0078] (6) The prototype level candidate creation process (step S611) may further include a determination process (not shown) for determining whether or not there is a prototype level to be used as a reference, and a reference generation process (not shown) for generating a reference using random numbers if it is determined that there is no reference in the determination process (not shown). In this case, even when there is no reference, by generating the reference using random numbers, it is possible to create prototype level candidates in the same way as when there is a reference.

[0079] (7) The optimization process (step S610) includes a selection process (step S617) for selecting an initial prototype level from prototype level candidates whose characteristics have already been measured, a penalty setting process (step S618) for setting a penalty for deviation from a predetermined constraint condition for the selected prototype level, and a process (step S619) for optimizing the prototype conditions so that the predicted value or acquisition function obtained by adding or multiplying the set penalty is optimal. This allows the prototype level to be more refined than when the optimization process is performed one-off. As a result, the prototype condition proposal system 1 can estimate the best prototype conditions for a material.

[0080] (8) The initial prototype level is selected randomly in the selection process (step S617). This makes it possible to search for prototype conditions (step S610) in a non-uniform area within the parameter space.

[0081] (9) The selection process (step S617) may include a process (not shown) for weighting the selection probability of the initial prototype level. In this case, weighting may be performed in this manner, making it easier to select a desirable prototype level as the initial value.

[0082] (10) The optimization process (step S610) may further include a determination process (not shown) for determining whether or not there is a reference prototype level, and a reference generation process (not shown) for generating a reference using random numbers if it is determined that there is no reference in the determination process (not shown). In this case, even when there is no reference, by generating the reference using random numbers, it is possible to create prototype level candidates in the same way as when there is a reference.

[0083] (11) The constraint is that the material can be prototyped. Since the constraint is imposed in this way based on materials science, the prototype condition suggestion system 1 can suggest only the prototype conditions that can actually be used to prototype the material.

[0084] The present invention is not limited to the above-described embodiment, and can be implemented using any components without departing from the spirit of the present invention.

[0085] The above-described embodiments and modifications are merely examples, and the present invention is not limited to these details as long as the features of the invention are not impaired. Furthermore, although various embodiments and modifications have been described above, the present invention is not limited to these details. Other aspects that can be considered within the scope of the technical idea of ​​the present invention are also included within the scope of the present invention.

[0086] 1: Prototype condition proposal system 11: Control unit 12: Storage unit 13: User interface unit 14: Communication unit 111: Characteristics measurement data preprocessing unit 112: Regression model construction processing unit 113: Prototype condition proposal processing unit 131: Input unit 132: Output unit 400: Internet

Claims

1. A prototype condition suggestion system that suggests prototype conditions for a material to a material developer, a regression model construction processing unit that executes a regression model construction process on characteristic measurement data representing actual measurement results of the characteristics of the material; a prototype condition proposal processing unit that performs an optimization process for searching for optimal prototype conditions for the material using the constructed regression model, and executes a prototype condition proposal process based on a result of the optimization process; Equipped with The optimization process includes: A prototype level candidate creation process for creating a designated number of prototype level candidates; A calculation process of calculating a predicted value or an acquisition function for all the created prototype level candidates; An extraction process for extracting the prototyping conditions with the best predicted value or acquisition function; Including, The prototype level candidate creation process includes: A selection process for selecting a reference prototype level from prototype level candidates whose characteristics have already been measured; A variation process for varying the selected prototype level by random numbers within a range in which the material can be prototyped; A storage process for storing the prototype level candidates subjected to the variation process; The prototype condition proposal system includes a process of causing a computer to repeatedly execute the above a specified number of times.

2. 2. The prototype condition proposal system according to claim 1, The prototype condition proposing system, wherein the reference prototype level is selected randomly in the selection process.

3. 3. The prototype condition proposal system according to claim 2, The selection process includes a process of weighting the selection probability of the reference prototype level.

4. 2. The prototype condition proposal system according to claim 1, The prototype level candidate creation process includes: A determination process for determining whether or not a reference prototype level is present; a criterion generating process for generating a criterion using random numbers if it is determined that there is no criterion in the determination process; The prototype condition suggestion system further includes:

5. A prototype condition suggestion system for suggesting prototype conditions for a material to a material developer, comprising: a regression model construction processing unit that executes a regression model construction process on characteristic measurement data representing actual measurement results of the characteristics of the material; a prototype condition proposal processing unit that performs an optimization process for searching for optimal prototype conditions for the material using the constructed regression model, and executes a prototype condition proposal process based on a result of the optimization process; Equipped with The optimization process includes: A selection process for selecting a prototype level that will be an initial value from prototype level candidates whose characteristics have already been measured; a penalty setting process for setting a penalty for deviation from a predetermined constraint condition for the selected prototype level; A process of optimizing the prototype conditions so that the predicted value or the acquisition function to which the set penalty is added or multiplied becomes the best. A prototype condition suggestion system including:

6. 6. The prototype conditions proposal system according to claim 5, wherein the prototype level that is the initial value is selected randomly in the selection process.

7. 7. The prototype condition proposal system according to claim 6, The selection process includes a process of weighting the selection probability of the prototype level that is the initial value.

8. 6. The prototype condition proposal system according to claim 5, The optimization process includes: A determination process for determining whether or not a reference prototype level is present; a criterion generating process for generating a criterion using random numbers if it is determined that there is no criterion in the determination process; The prototype condition suggestion system further includes:

9. 6. The prototype conditions suggestion system according to claim 5, wherein the constraint condition is that the material can be prototyped.

10. A method for proposing prototype conditions for a material to a material developer using a computer, comprising the steps of: A regression model construction process for constructing a regression model for actual characteristic measurement data representing actual measurement results of the characteristics of the material; performing an optimization process for searching for optimal prototype conditions for the material using the constructed regression model, and proposing prototype conditions for the material based on the results of the optimization process; causing the computer to execute The optimization process includes: A prototype level candidate creation process for creating a designated number of prototype level candidates; A calculation process of calculating a predicted value or an acquisition function for all the created prototype level candidates; An extraction process for extracting the prototyping conditions with the best predicted value or acquisition function; Including, The prototype level candidate creation process includes: A selection process for selecting a reference prototype level from prototype level candidates whose characteristics have already been measured; A variation process for varying the selected prototype level by random numbers within a range in which the material can be prototyped; A storage process for storing the prototype level candidates subjected to the variation process; a process of causing the computer to repeatedly execute the above a specified number of times.

11. A method for proposing prototype conditions for a material to a material developer using a computer, comprising the steps of: A regression model construction process for constructing a regression model for actual characteristic measurement data representing actual measurement results of the characteristics of the material; performing an optimization process for searching for optimal prototype conditions for the material using the constructed regression model, and proposing prototype conditions for the material based on the results of the optimization process; causing the computer to execute The optimization process includes: A selection process for selecting a prototype level that will be an initial value from prototype level candidates whose characteristics have already been measured; a penalty setting process for setting a penalty for deviation from a predetermined constraint condition for the selected prototype level; A process of optimizing the prototype conditions so that the predicted value or the acquisition function to which the set penalty is added or multiplied becomes the best. A method for proposing prototype conditions, including: