Physical property prediction device, physical property prediction method and physical property prediction program

Abstract

To provide a physical property predication device, a physical property predication method and a physical property predication program capable of appropriately and efficiently developing materials having desired physical properties.SOLUTION: A physical property prediction device 1 comprises: an actual measurement data acquisition unit 21 that acquires actual measurement data including a first manufacturing condition of a sample and an actual value obtained by actually measuring physical properties of a sample manufactured based on the first manufacturing condition; a virtual data acquisition unit 22 that acquires virtual data including a second manufacturing condition of a sample and a virtual value obtained by assuming physical properties of the sample corresponding to the second manufacturing condition; and a model generation unit 23 that generates a prediction model of the physical properties based on the actual measurement data acquired by the actual measurement data acquisition unit 21 and the virtual data acquired by the virtual data acquisition unit 22, the prediction model having a first uncertainty of the physical properties with respect to the actual measurement data and a second uncertainty of the physical properties with respect to the virtual data that is larger than the first uncertainty of the physical properties.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] This disclosure relates to a physical property prediction device, a physical property prediction method, and a physical property prediction program. [Background technology]

[0002] In recent years, a field of technology called materials informatics has been attracting attention, which utilizes information science techniques such as machine learning to improve the efficiency of materials development. In materials informatics, it is necessary to find the appropriate combination of numerous parameters related to the mixing ratio of raw materials and process conditions in order to manufacture materials with desired physical properties. When there are many parameters, the number of parameter combinations becomes enormous. When the number of parameter combinations becomes enormous, it becomes difficult to comprehensively prototype materials. When it is difficult to comprehensively prototype materials, a method is adopted in which the physical properties of a certain number of prototyped materials are measured, and a predictive model of the physical properties is constructed based on the measured data obtained from the measurements. By using a predictive model, it is possible to search for the optimal parameters even if the number of prototyped materials is limited.

[0003] Supervised machine learning methods are sometimes used to build predictive models. However, with supervised machine learning methods, it becomes difficult to build highly accurate predictive models when the amount of measured data is insufficient or when the relationship between parameters and physical properties is complex. To build highly accurate predictive models even when the amount of measured data is insufficient or the relationship between parameters and physical properties is complex, sequential experimental design methods such as Bayesian optimization are used. In Bayesian optimization, Bayesian estimation is used to build predictive models. In Bayesian estimation, in addition to the predicted values ​​of physical properties, the uncertainty of the predicted values ​​is calculated. By calculating the uncertainty, it is possible to calculate the probability that the physical properties will be at a specific value and the expected value of the physical properties. According to Bayesian estimation, the optimal parameters can be found based on limited measured data. In other words, according to Bayesian estimation, an experimental plan for the next prototype of a material can be made based on limited measured data. The predictive model is updated by making the next prototype of the material, measuring the physical properties of the prototyped material, and adding the measurement results to the measured data. In Bayesian optimization, this series of cycles consisting of prototyping, measuring physical properties, and updating the predictive model is repeated. By repeating this series of cycles, it is possible to minimize the number of prototypes required to obtain the desired physical properties.

[0004] However, if the amount of measured data is small or unavailable, it becomes difficult to construct an initial predictive model in Bayesian optimization. This difficulty in constructing an initial predictive model makes it difficult to plan sequential experiments. Specifically, the number of measured data points required to construct a predictive model is usually around 5 to 10. If the number of parameters is large, the number of measured data points required to construct the predictive model may increase further. However, if the monetary and time costs required for prototyping materials and measuring their properties are high, or if the development budget and development period are limited, the number of prototypes that can be made will be limited. A limited number of prototypes will result in an insufficient amount of measured data, making it difficult to construct an initial predictive model and apply Bayesian optimization.

[0005] Patent Document 1 discloses a technique for sequentially designing experiments by applying machine learning to measured data. Patent Document 2 discloses a technique for pre-training material properties by determining the material properties corresponding to the parameters through simulations such as quantum chemical calculations. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] International Publication No. 2021 / 200281 [Patent Document 2] International Publication No. 2020 / 188971 [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] However, the technology disclosed in Patent Document 1 is difficult to implement if a certain amount of measured data is not available in the initial stages. In particular, if the number of prototypes and physical property measurements is limited, it is difficult to obtain the number of measured data necessary to create an initial predictive model.

[0008] Furthermore, the technology disclosed in Patent Document 2 can only be implemented in certain fields where simulation methods are established. In other words, the technology disclosed in Patent Document 2 is difficult to implement in many fields where the application of simulation is difficult.

[0009] This disclosure has been made in consideration of the above points, and aims to provide a material property prediction device, a material property prediction method, and a material property prediction program that can appropriately and efficiently develop materials having desired physical properties. [Means for solving the problem]

[0010] The physical property prediction device described herein is A measurement data acquisition unit acquires measurement data including first manufacturing conditions for a sample and measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions. A virtual data acquisition unit acquires virtual data including second manufacturing conditions for the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing conditions. A model generation unit generates a predictive model of the physical properties based on the measured data acquired by the measured data acquisition unit and the virtual data acquired by the virtual data acquisition unit, such that the model has a first uncertainty of the physical properties with respect to the measured data and a second uncertainty of the physical properties that is greater than the first uncertainty of the physical properties with respect to the virtual data. It is equipped with.

[0011] In the physical property prediction apparatus according to this disclosure, the virtual data acquisition unit may acquire as the virtual value the physical property value of the sample that is predicted when the value of at least one of the plurality of parameters included in the second manufacturing conditions is extremely small or large.

[0012] In the physical property prediction device according to this disclosure, the virtual data acquisition unit may acquire the physical property values ​​of a substance whose raw material or process is similar to that of the sample as the virtual values.

[0013] In the physical property prediction device according to this disclosure, the first uncertainty of the physical property may be 0.

[0014] The physical property prediction device according to this disclosure further comprises an input screen generation unit that generates an input screen for inputting the measured data and the virtual data, The measured data acquisition unit acquires the measured data entered into the input screen generated by the input screen generation unit, The virtual data acquisition unit may acquire the virtual data entered into the input screen generated by the input screen generation unit.

[0015] In the physical property prediction apparatus according to the present disclosure, the apparatus further comprises a communication unit that transmits the input screen generated by the input screen generation unit to a user's terminal device, and receives the actually measured data and the virtual data input to the input screen from the terminal device, the actually measured data acquisition unit acquires the actually measured data received by the communication unit, and the virtual data acquisition unit may acquire the virtual data received by the communication unit.

[0016] In the physical property prediction apparatus according to the present disclosure, the apparatus further comprises a virtual data complementing unit that complements the virtual data, and the virtual data acquisition unit may acquire the virtual data complemented by the virtual data complementing unit.

[0017] In the physical property prediction apparatus according to the present disclosure, when some parameters having extremely small or large values among a plurality of parameters included in the second manufacturing condition, and the virtual value predicted based on the some parameters are input to the input screen, the virtual data complementing unit may complement the second manufacturing condition in the virtual data by calculating values of parameters other than the some parameters among the plurality of parameters.

[0018] In the physical property prediction apparatus according to the present disclosure, the apparatus further comprises an uncertainty setting unit that sets a first uncertainty of the physical property and a second uncertainty of the physical property, and the model generation unit may generate the prediction model further based on the first uncertainty of the physical property and the second uncertainty of the physical property set by the uncertainty setting unit.

[0019] In the physical property prediction apparatus according to the present disclosure, the apparatus further comprises a setting screen generation unit that generates a setting screen for setting the first uncertainty of the physical property and the second uncertainty of the physical property, and the uncertainty setting unit may set the first uncertainty of the physical property and the second uncertainty of the physical property in accordance with an input operation performed on the setting screen generated by the setting screen generation unit.

[0020] In the physical property prediction device according to this disclosure, the model generation unit generates the prediction model using Bayesian estimation, the first uncertainty may be a first variance, and the second uncertainty may be a second variance that is greater than the first variance.

[0021] In the physical property prediction device according to this disclosure, the model generation unit may generate the prediction model according to the following equation.

number

number

[0022] The physical property prediction device according to this disclosure may further include a manufacturing condition calculation unit that calculates a third manufacturing condition for the sample based on the prediction model generated by the model generation unit.

[0023] In the physical property prediction apparatus according to this disclosure, the third manufacturing condition may include at least one of the manufacturing conditions that result in the maximum uncertainty of the physical property in the prediction model and the manufacturing conditions that result in the minimum uncertainty of the physical property in the prediction model.

[0024] In the physical property prediction apparatus according to this disclosure, the measured data acquisition unit further acquires second measured data including the third manufacturing conditions calculated by the manufacturing condition calculation unit and second measured values ​​obtained by actually measuring the physical properties of the sample manufactured based on the third manufacturing conditions. The model generation unit may update the prediction model based on the second measured data acquired by the measured data acquisition unit.

[0025] In the physical property prediction device according to this disclosure, the model generation unit does not need to update the prediction model if the second measured value is the set optimal value.

[0026] The method for predicting physical properties according to this disclosure is: A step of acquiring measured data including first manufacturing conditions for a sample and measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions, A step of acquiring virtual data including second manufacturing conditions for the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing conditions, A step of generating a predictive model of the physical properties based on the acquired actual measurement data and virtual data, such that the model has a first uncertainty of the physical properties with respect to the actual measurement data and a second uncertainty of the physical properties that is greater than the first uncertainty of the physical properties with respect to the virtual data, It is equipped with.

[0027] The material property prediction program described herein is On the computer, A procedure for obtaining measured data including first manufacturing conditions for a sample and measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions. A procedure for acquiring virtual data including second manufacturing conditions for the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing conditions, and A procedure for generating a predictive model of the physical properties based on the acquired actual measurement data and virtual data, such that the model has a first uncertainty of the physical properties with respect to the actual measurement data and a second uncertainty of the physical properties with respect to the virtual data that is greater than the first uncertainty of the physical properties. Make it run. [Effects of the Invention]

[0028] According to this disclosure, materials having desired physical properties can be developed appropriately and efficiently. [Brief explanation of the drawing]

[0029] [Figure 1] Figure 1 is a schematic diagram showing a physical property prediction device according to the first embodiment. [Figure 2] Figure 2 is a block diagram showing a physical property prediction device according to the first embodiment. [Figure 3] Figure 3 is a flowchart showing an example of the operation of the physical property prediction device according to the first embodiment. [Figure 4] Figure 4 shows an example of an input screen in an example of operation of a physical property prediction device according to the first embodiment. [Figure 5] Figure 5 shows an example of a prediction model in an example of the operation of the physical property prediction device according to the first embodiment. [Figure 6] Figure 6 shows another example of a prediction model in an example of operation of the physical property prediction device according to the first embodiment. [Figure 7] Figure 7 shows an example of the calculation of a third manufacturing condition based on a prediction model in an example of the operation of a physical property prediction device according to the first embodiment. [Figure 8] Figure 8 is a block diagram showing a physical property prediction device according to the second embodiment. [Figure 9] Figure 9 is a flowchart showing an example of the operation of the physical property prediction device according to the second embodiment. [Figure 10] Figure 10 shows an example of an input screen in an example of operation of a physical property prediction device according to the second embodiment. [Figure 11] Figure 11 is a block diagram showing a physical property prediction device according to the third embodiment. [Figure 12] Figure 12 is a flowchart showing an example of the operation of a physical property prediction device according to the third embodiment. [Figure 13] Figure 13 shows an example of an input screen in an example of operation of a physical property prediction device according to the third embodiment. [Figure 14] Figure 14 is a block diagram showing a physical property prediction device according to the fourth embodiment. [Modes for carrying out the invention]

[0030] Embodiments of this disclosure will now be described with reference to the drawings. In the drawings referenced in the embodiments, the same or similar reference numerals are used to denote identical parts or parts having similar functions, and repeated descriptions may be omitted.

[0031] [First Embodiment] First, the physical property prediction device according to the first embodiment will be described with reference to Figures 1 to 7. Figure 1 is a conceptual diagram showing the physical property prediction device 1 according to the first embodiment.

[0032] As shown in Figure 1, the physical property prediction device 1 according to the first embodiment generates a prediction model of a sample based on measured data and prediction data. Measured data includes data on the first manufacturing conditions of the sample and measured values. The sample is, for example, a material manufactured based on multiple raw materials. The first manufacturing conditions are, for example, the mixing ratio of the raw materials and process conditions such as temperature. Measured values ​​are the values ​​of the physical properties of the sample actually manufactured based on the first manufacturing conditions that were actually measured. The physical properties of the sample are, for example, conductivity and dielectric constant. Prediction data includes data on the second manufacturing conditions of the sample and virtual values. The second manufacturing conditions are different from the first manufacturing conditions. The second manufacturing conditions are not the actual manufacturing conditions when the sample was manufactured, but rather virtual manufacturing conditions. Virtual values ​​are virtual values ​​of the physical properties of the sample corresponding to the second manufacturing conditions. The prediction model is information that shows the predicted physical properties of the sample based on the sample's manufacturing conditions. In other words, the prediction model is information that shows the correspondence between the sample's manufacturing conditions and its physical properties.

[0033] Furthermore, the physical property prediction device 1 creates a new experimental plan for the sample based on the generated prediction model. A new sample is fabricated according to the newly created experimental plan. Experiments are conducted to measure the physical properties of the fabricated new sample. The measured data obtained from the experiments is added to the physical property prediction device 1. The physical property prediction device 1 updates the prediction model based on the updated measured data and the virtual data.

[0034] In this way, the material property prediction device 1 optimizes the experimental plan by repeatedly generating a prediction model based on measured data and virtual data, creating an experimental plan, and adding measured data.

[0035] Figure 2 is a block diagram showing a physical property prediction device 1 according to the first embodiment. More specifically, as shown in Figure 2, the physical property prediction device 1 comprises a control unit 2, an input interface 3, a display unit 4, and a storage unit 5. The control unit 2, the input interface 3, the display unit 4, and the storage unit 5 are communicated together via a bus 6.

[0036] Input interface 3 receives various instructions and information input operations from the user. Input interface 3 converts the input operations received from the user into electrical signals. Input interface 3 outputs the input operations converted into electrical signals to control unit 2. Input interface 3 may be, for example, a mouse, keyboard, touchpad, touch panel, trackball, switch button, microphone, etc.

[0037] The display unit 4 displays an image corresponding to the image data transmitted from the control unit 2. The display unit 4 may be, for example, a liquid crystal monitor, a CRT (Cathode Ray Tube) monitor, or a touch panel. The display unit 4 may also be equipped with a speaker.

[0038] The storage unit 5 is a non-transient storage device that stores various types of information. The storage unit 5 may be, for example, an HDD (Hard Disk Drive), an optical disk, a magnetic disk, a magneto-optical disk, a CD (Compact Disc), a DVD (Digital Versatile Disc), or a semiconductor memory. The storage unit 5 stores, for example, a physical property prediction program executed by the physical property prediction device 1 and various types of data used to execute this physical property prediction program.

[0039] The control unit 2 controls the operation of the physical property prediction device 1 in response to user input operations using the input interface 3. The control unit 2 includes a measured data acquisition unit 21, a virtual data acquisition unit 22, a model generation unit 23, an input screen generation unit 24, and a manufacturing condition calculation unit 25.

[0040] The functions corresponding to each component 21-25 of the control unit 2 are recorded in the storage unit 5 in the form of a physical property prediction program that can be executed by a computer. The control unit 2 is, for example, a processor. The processor constituting the control unit 2 reads and executes the physical property prediction program from the storage unit 5, thereby executing the functions of the components 21-25 corresponding to the read physical property prediction program. Alternatively, instead of storing the physical property prediction program in the storage unit 5, the physical property prediction program may be directly incorporated into the processor's circuitry. In this case, the processor executes the physical property prediction program incorporated into the circuitry, thereby executing the functions corresponding to each component 21-25. The physical property prediction program may be a program downloaded by the physical property prediction device 1 from a server via a network. Alternatively, the physical property prediction program may be a program imported into the physical property prediction device 1 from a portable storage medium. The control unit 2 may consist of a single processor, or it may consist of a combination of multiple processors. The processor may be, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an Application Specific Integrated Circuit (ASIC), or a programmable logic device. The programmable logic device may be an SPLD (Simple Programmable Logic Device), a CPLD (Complex Programmable Logic Device), or an FPGA (Field Programmable Gate Array).

[0041] The measurement data acquisition unit 21 acquires measurement data that includes the first manufacturing conditions of the sample and the measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions.

[0042] The virtual data acquisition unit 22 acquires virtual data that includes the second manufacturing conditions of the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing conditions.

[0043] The virtual data acquisition unit 22 may acquire, as virtual values, the physical properties of the sample that are predicted when the values ​​of some of the multiple parameters included in the second manufacturing conditions are extremely small or large.

[0044] An extremely small parameter value is when the parameter value is at or below the lower limit of the predetermined parameter search range. An extremely large parameter value is when the parameter value is at or above the upper limit of the predetermined parameter search range.

[0045] The virtual data acquisition unit 22 may acquire, as virtual values, the physical properties of a substance whose raw materials or processes are similar to those of the sample, instead of the predicted physical properties when the values ​​of some parameters are extremely small or large.

[0046] The model generation unit 23 generates a prediction model of material properties based on the measured data acquired by the measured data acquisition unit 21 and the virtual data acquired by the virtual data acquisition unit 22. The prediction model may be a probabilistic prediction model that shows the predicted material properties in terms of probabilities. The model generation unit 23 generates a prediction model that has a first uncertainty of material properties with respect to the measured data and a second uncertainty of material properties that is greater than the first uncertainty of material properties with respect to the virtual data. The first uncertainty of material properties is, for example, 0.

[0047] The input screen generation unit 24 generates an input screen for inputting actual measured data and virtual data. The input screen generation unit 24 displays the generated input screen on the display unit 4. The actual measured data acquisition unit 21 acquires the actual measured data that has been input into the input screen generated by the input screen generation unit 24. More specifically, the actual measured data acquisition unit 21 acquires the actual measured data that has been input to the input screen displayed on the display unit 4 using the input interface 3. The virtual data acquisition unit 22 acquires the virtual data that has been input into the input screen generated by the input screen generation unit 24. More specifically, the virtual data acquisition unit 22 acquires the virtual data that has been input to the input screen displayed on the display unit 4 using the input interface 3.

[0048] The manufacturing condition calculation unit 25 calculates a third manufacturing condition for the sample based on the predictive model generated by the model generation unit 23.

[0049] The third manufacturing condition may include at least one of the manufacturing conditions that result in the maximum uncertainty of the physical properties in the prediction model and the manufacturing conditions that result in the minimum uncertainty of the physical properties in the prediction model.

[0050] The measurement data acquisition unit 21 further acquires second measurement data, which includes third manufacturing conditions calculated by the manufacturing condition calculation unit 25 and second measurement values ​​obtained by measuring the physical properties of a sample manufactured based on the third manufacturing conditions.

[0051] The model generation unit 23 updates the prediction model based on the second measured data acquired by the measured data acquisition unit 21. The model generation unit 23 does not update the prediction model if the second measured value is the set optimal value.

[0052] The model generation unit 23 may generate a predictive model, i.e., a stochastic predictive model, using Bayesian estimation. The first uncertainty may be the first variance, and the second uncertainty may be a second variance that is greater than the first variance.

[0053] When generating a predictive model using Bayesian estimation, the model generation unit 23 may generate the predictive model according to the following equation.

[0054]

number

[0055]

number

[0056] JPEG0007921019000006.jpg89170

[0057] In generating a predictive model according to equations (1) and (2), the model generation unit 23 calculates the variance σ of the likelihood corresponding to the measured data. 2 Set it to a small value, and the variance σ of the likelihood corresponding to the virtual data 2 The value is set to a large value. The model generation unit 23 generates the variance σ of the likelihood corresponding to the measured data. 2 You can also set it to 0.

[0058] JPEG0007921019000007.jpg19170

[0059] [Example of operation] Next, an example of the operation of the physical property prediction device 1 having the above configuration will be described. Figure 3 is a flowchart showing an example of the operation of the physical property prediction device 1 according to the first embodiment. First, as a premise for Figure 3, the user creates virtual data. The user may create virtual data by prediction based on experience. For example, when each of the multiple parameters included in the second manufacturing conditions is extremely small or large, the physical properties of the sample may be predicted based on experience. In this case, the user uses the multiple parameters that take extremely small or large values ​​as the second manufacturing conditions and uses the predicted physical properties of the sample as virtual values ​​to create virtual data. Alternatively, the user may create virtual data using known physical properties of a substance whose raw materials or processes are similar to those of the sample. In this case, the user uses the process of the substance as the second manufacturing conditions and uses the known physical properties of the substance as virtual values ​​to create virtual data.

[0060] Furthermore, the user creates actual measurement data. Specifically, the user obtains actual values ​​by measuring the physical properties of a sample manufactured based on the first manufacturing conditions. The user then creates actual measurement data by associating the first manufacturing conditions with the actual measurement values. In creating the actual measurement data, the user may set approximately 1 to 5 combinations of values ​​for multiple parameters that constitute the first manufacturing conditions. The user may then create the actual measurement data by measuring the physical properties of a sample manufactured according to the set combinations of parameter values. In this case, the user may randomly set the values ​​of multiple parameters of the first manufacturing conditions within a predetermined range. Alternatively, the user may set the values ​​of multiple parameters of the first manufacturing conditions using a method that uniformly selects parameter values ​​within a predetermined range, such as Latin square sampling.

[0061] After virtual data and actual measurement data are created, the input screen generation unit 24 generates an input screen in response to user input operations using the input interface 3. The input screen generation unit 24 then displays the generated input screen on the display unit 4 (step S1).

[0062] Figure 4 shows an example of the input screen SC in an example of the operation of the physical property prediction device 1 according to the first embodiment. In the example shown in Figure 4, the input screen SC has an input field F1 for inputting the first to fifth parameters, which are an example of the first manufacturing conditions. The input screen SC also has an input field F2 for inputting the first and second physical properties, which are examples of measured and virtual values. The input screen SC also has an input field F3 for inputting the type of data, i.e., either virtual data or measured data. After entering data in input fields F1 to F3, pressing the execute button B confirms the data entered in input fields F1 to F3.

[0063] After the input screen is displayed on the display unit 4, as shown in Figure 3, the user inputs virtual data into the input screen displayed on the display unit 4 by performing input operations using the input interface 3 (step S2).

[0064] After inputting virtual data, the user inputs actual measured data into the input screen displayed on the display unit 4 using input operations with the input interface 3 (step S3). Note that the order of steps S2 and S3 may be reversed.

[0065] After virtual data and actual data are input, the model generation unit 23 generates a predictive model based on the input virtual data and actual data (step S4). In the example shown in Figure 3, the model generation unit 23 generates a predictive model using Bayesian estimation according to the above-described equations (1) and (2). At this time, the model generation unit 23 calculates the variance σ of the likelihood corresponding to the actual data. 2 Set to 0. Also, the model generation unit 23 calculates the variance σ of the likelihood corresponding to the virtual data. 2 Set to a value greater than 0. The model generation unit 23 also sets the prior distribution p(A) such that it has a high probability density when the value of the physical property θ changes smoothly.

[0066] The model generation unit 23 may display the generated prediction model on the display unit 4.

[0067] Figure 5 shows an example of a prediction model in an example of operation of the physical property prediction device 1 according to the first embodiment. Figure 6 shows another example of a prediction model in an example of operation of the physical property prediction device 1 according to the first embodiment. For example, the model generation unit 23 generates the prediction models shown in Figures 5 and 6. The horizontal axis in Figures 5 and 6 is parameter x, which is an example of the first manufacturing conditions. The vertical axis in Figures 5 and 6 is the physical property value θ. The solid line graph in Figures 5 and 6 shows the average value of the physical property θ corresponding to parameter x. Measured data and virtual data are located on the solid line graph showing the average value. The two dashed line graphs in Figures 5 and 6 show the uncertainty of the physical property θ. Specifically, the two dashed line graphs in Figures 5 and 6 show the upper and lower limits of the 50% confidence interval in the normal distribution, which is the probability distribution of the physical property θ, respectively. As an example, in Figure 5, the normal distribution G when x = 0.4 is shown corresponding to the solid line graph showing the average value and the dashed line graph showing the uncertainty. The hatched region within the normal distribution G corresponds to the 50% confidence interval when x = 0.4.

[0068] As can be seen from Figures 5 and 6, the uncertainty is 0 near the measured data. This is because the variance σ of the likelihood corresponding to the measured data is generated by the model generation unit 23. 2 This is because it was set to 0. On the other hand, the uncertainty is large near the virtual data. This is because the variance σ of the likelihood corresponding to the virtual data is generated by the model generation unit 23. 2 This is because it was set to a value greater than 0.

[0069] After the prediction model is generated, the experimental plan is created as shown in Figure 3 (step S5). In creating the experimental plan, first, the manufacturing condition calculation unit 25 calculates the third manufacturing conditions for the sample based on the prediction model generated by the model generation unit 23. Figure 7 shows an example of the calculation of the third manufacturing conditions based on the prediction model in an example of the operation of the physical property prediction device 1 according to the first embodiment. In the example shown in Figure 7, the manufacturing condition calculation unit 25 calculates the manufacturing conditions when the uncertainty of the physical property in the prediction model is maximum and the manufacturing conditions when the uncertainty of the physical property in the prediction model is minimum as the third manufacturing conditions. More specifically, the manufacturing condition calculation unit 25 calculates the value x1 of parameter x when the uncertainty of physical property θ is maximum and the value x2 of parameter x when the uncertainty of physical property θ is minimum as the third manufacturing conditions. Here, when the uncertainty of the physical property is maximum, there is a low probability that very good physical properties can be obtained. Also, when the uncertainty of the physical property is minimum, there is a high probability that good physical properties can be obtained. Therefore, by setting the value x1 of parameter x when the uncertainty of the physical property θ is maximum and the value x2 of parameter x when it is minimum as the third manufacturing condition, the number of sample prototypes required to obtain the desired physical properties can be reduced. Next, in creating the experimental plan, the user creates a specific plan for manufacturing the sample using the calculated third manufacturing condition.

[0070] After the experimental plan is created, the user conducts the experiment (step S6), as shown in Figure 3. In conducting the experiment, the user first manufactures a sample according to the experimental plan, i.e., the third manufacturing conditions. The sample is manufactured using a manufacturing apparatus (not shown). After the sample is manufactured, the user measures the physical properties of the manufactured sample. The physical properties of the sample are measured using measuring instruments (not shown) appropriate to the physical properties.

[0071] If the desired physical properties are obtained as a result of the experiment (Step S7: Yes), the process is terminated. On the other hand, if the desired physical properties are not obtained as a result of the experiment (Step S7: No), but additional experiments are possible (Step S8: Yes), the user inputs the measured data obtained from the experiment into the input screen (Step S3). In response to the input of the measured data obtained from the experiment, the model generation unit 23 updates the prediction model (Step S4).

[0072] As described above, in the first embodiment, the measurement data acquisition unit 21 acquires measurement data including first manufacturing conditions for the sample and measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions. The virtual data acquisition unit 22 acquires virtual data including second manufacturing conditions for the sample and virtual values ​​of the physical properties of the sample corresponding to the second manufacturing conditions. The model generation unit 23 generates a predictive model of the physical properties based on the measurement data acquired by the measurement data acquisition unit 21 and the virtual data acquired by the virtual data acquisition unit 22. Specifically, the model generation unit 23 generates a predictive model that has a first uncertainty of the physical properties with respect to the measurement data and a second uncertainty of the physical properties that is greater than the first uncertainty of the physical properties with respect to the virtual data.

[0073] This allows for the creation of appropriate sequential experimental plans by effectively utilizing virtual data that reflects tacit knowledge such as experience and intuition in materials development, even when actual measurement data is scarce or unavailable. Furthermore, it enables the development of materials with desired properties even when the number of experiments that can be conducted is limited and it is difficult to produce the number of prototypes and measurements necessary to build an initial predictive model. Moreover, even when a sufficient number of experiments can be conducted, the number of prototypes required to obtain the desired properties can be reduced by utilizing tacit knowledge that is not expressed in actual measurement data. Therefore, according to the first embodiment, materials with desired properties can be developed appropriately and efficiently.

[0074] Furthermore, in the first embodiment, the virtual data acquisition unit 22 may acquire, as virtual values, the predicted physical properties of the sample when some of the multiple parameters included in the second manufacturing conditions are extremely small or large.

[0075] This allows for the acquisition of virtual data useful for generating predictive models using a simple method.

[0076] In the first embodiment, the virtual data acquisition unit 22 may acquire the physical properties of a substance whose raw material or process is similar to that of the sample as virtual values.

[0077] This allows for the acquisition of virtual data useful for generating predictive models using a simple method.

[0078] Furthermore, in the first embodiment, the first uncertainty of the physical properties is 0.

[0079] This makes it possible to generate a predictive model that allows for easy exploration of optimal manufacturing conditions.

[0080] Furthermore, in the first embodiment, the physical property prediction device 1 further includes an input screen generation unit 24 that generates an input screen for inputting measured data and virtual data. The measured data acquisition unit 21 acquires the measured data that has been input into the input screen generated by the input screen generation unit 24. The virtual data acquisition unit 22 acquires the virtual data that has been input into the input screen generated by the input screen generation unit 24.

[0081] This allows for the calculation of predictive models that reflect the user's intentions.

[0082] Furthermore, in the first embodiment, the model generation unit 23 generates a predictive model using Bayesian estimation according to equations (1) and (2). That is, the model generation unit 23 generates a predictive model that has a first variance with respect to the actual data and a second variance with respect to the virtual data that is larger than the first variance.

[0083] This makes it possible to generate a predictive model that can accurately search for optimal manufacturing conditions.

[0084] In the first embodiment, the manufacturing condition calculation unit 25 calculates a third manufacturing condition for the sample based on the predictive model generated by the model generation unit 23.

[0085] This allows for the exploration of optimal manufacturing conditions based on a predictive model.

[0086] Furthermore, in the first embodiment, the third manufacturing condition includes at least one of the manufacturing condition when the uncertainty of the physical property in the prediction model is maximized and the manufacturing condition when the uncertainty of the physical property in the prediction model is minimized.

[0087] This reduces the number of prototypes required to obtain the desired physical properties, enabling the efficient development of materials with the desired properties.

[0088] In the first embodiment, the measurement data acquisition unit 21 further acquires second measurement data, which includes third manufacturing conditions calculated by the manufacturing condition calculation unit 25 and second measurement values ​​obtained by measuring the physical properties of a sample manufactured based on the third manufacturing conditions. The model generation unit 23 updates the prediction model based on the second measurement data acquired by the measurement data acquisition unit 21.

[0089] This allows for the optimization of the experimental design, enabling the appropriate development of materials with desired physical properties.

[0090] Furthermore, in the first embodiment, the model generation unit 23 does not update the prediction model if the second measured value is the set optimal value.

[0091] This reduces the number of sample prototypes required.

[0092] (Second embodiment) Next, we will describe a second embodiment that generates a prediction model based on the set uncertainty, focusing on the differences from the first embodiment. Figure 8 is a block diagram showing the physical property prediction device 1 according to the second embodiment.

[0093] As shown in Figure 8, the control unit 2 of the physical property prediction device 1 according to the second embodiment further includes an uncertainty setting unit 26 and a setting screen generation unit 27, in addition to the configuration of the first embodiment.

[0094] The uncertainty setting unit 26 sets a first uncertainty and a second uncertainty for the material properties. The model generation unit 23 generates a prediction model based on the first uncertainty and second uncertainty for the material properties set by the uncertainty setting unit 26.

[0095] The setting screen generation unit 27 generates a setting screen for setting the first uncertainty and the second uncertainty of the physical properties. The setting screen generation unit 27 displays the generated setting screen on the display unit 4. The uncertainty setting unit 26 sets the first uncertainty and the second uncertainty of the physical properties in response to input operations on the setting screen generated by the setting screen generation unit 27. More specifically, the uncertainty setting unit 26 sets the first uncertainty and the second uncertainty of the physical properties in response to input operations using the input interface 3 on the setting screen displayed on the display unit 4. The setting screen generation unit 27 may display the setting screen at the same time as the input screen displayed by the input screen generation unit 24. Alternatively, the setting screen generation unit 27 may display the setting screen at a different timing than the input screen displayed by the input screen generation unit 24.

[0096] Figure 9 is a flowchart showing an example of operation of the physical property prediction device 1 according to the second embodiment. As shown in Figure 9, in the second embodiment, the setting screen generation unit 27 displays the setting screen simultaneously with the input screen displayed by the input screen generation unit 24 (step S11).

[0097] Figure 10 shows an example of the input screen SC in an example of the operation of a physical property prediction device according to the second embodiment. In the example shown in Figure 10, the setting screen generation unit 27 displays the setting screen SC2 on the input screen SC. The setting screen SC2 has a slider bar S on which the ratio of the weight of the measured data to the weight of the virtual data can be input. By sliding the slider bar S to input the ratio of the weight of the measured data to the weight of the virtual data, the first uncertainty and the second uncertainty corresponding to the input ratio can be set.

[0098] As shown in Figure 9, in the second embodiment, the uncertainty setting unit 26 sets the first uncertainty and the second uncertainty in response to an input operation on the setting screen (step S9).

[0099] The model generation unit 23 generates a predictive model based on the input virtual data and actual data, and the set first uncertainty and second uncertainty (step S4).

[0100] As described above, in the second embodiment, the uncertainty setting unit 26 sets a first uncertainty and a second uncertainty of the material properties. The model generation unit 23 then generates a prediction model based on the first uncertainty and the second uncertainty of the material properties set by the uncertainty setting unit 26.

[0101] This allows for the calculation of predictive models that reflect the user's intentions, thereby improving the degree of freedom in exploring manufacturing conditions.

[0102] In the second embodiment, the setting screen generation unit 27 generates a setting screen for setting the first uncertainty and the second uncertainty of the physical properties. The uncertainty setting unit 26 sets the first uncertainty and the second uncertainty of the physical properties in response to input operations on the setting screen generated by the setting screen generation unit 27.

[0103] This allows for easy configuration of the first and second uncertainties using the settings screen.

[0104] (Third embodiment) Next, we will describe a third embodiment that complements virtual data, focusing on the differences from the first embodiment. Figure 11 is a block diagram showing the physical property prediction device 1 according to the third embodiment.

[0105] As shown in Figure 11, the control unit 2 of the physical property prediction device 1 according to the third embodiment further includes a virtual data completion unit 28 in addition to the configuration of the first embodiment.

[0106] The virtual data completion unit 28 completes the virtual data. When the virtual data entered on the input screen is not complete and some data is missing, the virtual data completion unit 28 completes the virtual data by filling in the missing data.

[0107] The virtual data acquisition unit 22 acquires the virtual data that has been supplemented by the virtual data supplementation unit 28.

[0108] The virtual data completion unit 28 may complete the virtual data when some of the multiple parameters included in the second manufacturing conditions that have extremely small or large values, and virtual values ​​predicted based on those parameters, are entered into the input screen. Specifically, the virtual data completion unit 28 may complete the second manufacturing conditions of the virtual data by calculating the values ​​of the parameters other than some of the multiple parameters.

[0109] Figure 12 is a flowchart showing an example of the operation of the physical property prediction device 1 according to the third embodiment. Figure 13 is a diagram showing an example of the input screen in the example of the operation of the physical property prediction device 1 according to the third embodiment.

[0110] As shown in Figure 12, in the third embodiment, the virtual data completion unit 28 completes the virtual data if the virtual data entered on the input screen is not complete (step S10).

[0111] For example, m parameters x₁, x₂, …, x included in the second manufacturing condition m among which, the value of one parameter x₁ is an extremely small or large value x₁ * , the physical property value of the sample can be determined as θ * based on the user's tacit knowledge, and it may be predictable that this is the case. In this case, the user enters, on the input screen, x₁ as the value of the parameter x₁ * , enters θ as the physical property value * , and does not enter values for parameters other than x₁. In the example shown in FIG. 13, on the input screen SC, an extremely small value 0 is entered as the value of the first parameter, 0.0 is entered as the first physical property and the second physical property respectively, and values of parameters other than the first parameter are set to undetermined, that is, ***.

[0112] When values of parameters other than x₁ are not entered, the virtual data is incomplete. In this case, the virtual data complementing unit 28 complements the virtual data by calculating and adding values for parameters x₂, …, x m other than x₁. For example, the virtual data complementing unit 29 divides each predetermined search range of parameters x₂, …, x m into k levels. Then, the virtual data complementing unit 28 combines values selected from each of the divided regions divided into k levels to obtain k m-1 combinations of values of parameters x₂, …, x m and calculates said values. Instead of such a method, the virtual data complementing unit 28 may calculate values of parameters x₂, …, x m by generating (m-1)-dimensional uniform random numbers.

[0113] Further, m parameters x₁, x₂, …, x included in the second manufacturing condition mIn some cases, when the values ​​of two to m-1 parameters are extremely small or extremely large, the user may be able to predict the physical properties of the sample through tacit knowledge. In this case, the user enters the values ​​of two to m-1 parameters and the predicted physical properties into the input screen, and does not enter any parameters other than the two to m-1 parameters. In the example shown in Figure 13, an extremely large value of 100 is entered as the value of the third parameter in the input screen SC, an extremely large value of 100 is entered as the value of the fourth parameter, 100.0 is entered as the first physical property, 133.3 is entered as the second physical property, and the values ​​of parameters other than the third and fourth parameters are set to undetermined.

[0114] If values ​​for parameters other than 2 to m-1 parameters are not entered, the virtual data becomes incomplete. In this case, the virtual data completion unit 28 divides the search area for each parameter other than 2 to m-1 parameters into k levels. The virtual data completion unit 28 then calculates the parameter value by combining the values ​​selected from each of the k divided areas.

[0115] If actual measurement data of a substance similar to the sample is relevant, the user may input that substance's measurement data as virtual data on the input screen.

[0116] Furthermore, if complete virtual data is entered into the input screen, the virtual data completion unit 28 will not complete the virtual data.

[0117] As described above, in the third embodiment, the virtual data completion unit 28 completes the virtual data. The virtual data acquisition unit 22 acquires the virtual data completed by the virtual data completion unit 28.

[0118] This allows for the efficient generation of virtual data.

[0119] In the third embodiment, the virtual data completion unit 28 completes the second manufacturing conditions of virtual data by calculating the values ​​of the parameters other than the selected parameters when some of the multiple parameters included in the second manufacturing conditions, and virtual values ​​predicted based on those selected parameters, are entered into the input screen.

[0120] This allows for the efficient supplementation of virtual data.

[0121] (Fourth embodiment) Next, we will describe a fourth embodiment that provides an input screen to the user's terminal device, focusing on the differences from the first embodiment. Figure 14 is a block diagram of the physical property prediction device 1 according to the fourth embodiment.

[0122] Figure 1 illustrates an example in which the input interface 3 and display unit 4 are provided on the physical property prediction device 1. In contrast, in the example shown in Figure 14, the input interface 3 and display unit 4 are provided on the user's terminal device 9. The physical property prediction device 1 and the terminal device 9 are connected via a network 8 for communication.

[0123] The physical property prediction device 1 includes a communication unit 7. The communication unit 7 is an interface for communicating with a terminal device 9 via a network 8. For example, the communication unit 7 may be a network card and a network adapter. The communication unit 7 transmits the input screen generated by the input screen generation unit 24 to the terminal device 9. The communication unit 7 receives the measured data and virtual data entered on the input screen from the terminal device 9.

[0124] The actual measurement data acquisition unit 21 acquires the actual measurement data received by the communication unit 7. The virtual data acquisition unit 22 acquires the virtual data received by the communication unit 7.

[0125] The model generation unit 23 may transmit the prediction model it has generated to the terminal device 9 via the communication unit 7. The prediction model transmitted to the terminal device 9 may be displayed on the display unit 4 of the terminal device 9.

[0126] The manufacturing condition calculation unit 25 may transmit the manufacturing conditions calculated by the manufacturing condition calculation unit 25 to the terminal device 9 via the communication unit 7. The manufacturing conditions transmitted to the terminal device 9 may be displayed on the display unit 4 of the terminal device 9.

[0127] According to the fourth embodiment, a user using the terminal device 9 can search for optimal manufacturing conditions through an input screen provided by the physical property prediction device 1, i.e., the server.

[0128] While several embodiments of this disclosure have been described, these embodiments are presented as examples and are not intended to limit the scope of the disclosure. These embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of this disclosure, as well as in the claims of this disclosure and its equivalents. [Explanation of Symbols]

[0129] 1. Physical property prediction device 21 Measurement data acquisition unit 22 Virtual Data Acquisition Unit 23 Model Generation Unit 24 Input screen generation unit 25 Manufacturing condition calculation department 26 Uncertainty Setting Section 27 Setting screen generation section 28 Virtual Data Completion Unit

Claims

1. A measurement data acquisition unit acquires measurement data including first manufacturing conditions for a sample and measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions. A virtual data acquisition unit acquires virtual data including second manufacturing conditions for the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing conditions. A model generation unit generates a predictive model of the physical properties based on the measured data acquired by the measured data acquisition unit and the virtual data acquired by the virtual data acquisition unit, such that the model has a first uncertainty of the physical properties with respect to the measured data and a second uncertainty of the physical properties that is greater than the first uncertainty of the physical properties with respect to the virtual data. Equipped with, The material property prediction device comprises a model generation unit that generates the prediction model using Bayesian estimation, wherein the first uncertainty is a first variance and the second uncertainty is a second variance that is greater than the first variance.

2. The physical property prediction apparatus according to claim 1, wherein the virtual data acquisition unit acquires the predicted physical property values ​​of the sample as virtual values ​​when the value of at least one of the plurality of parameters included in the second manufacturing conditions is extremely small or large.

3. The physical property prediction device according to claim 1, wherein the virtual data acquisition unit acquires the physical property values ​​of a substance whose raw material or process is similar to that of the sample as the virtual values.

4. The physical property prediction device according to claim 1, wherein the first uncertainty of the aforementioned physical property is 0.

5. The system further includes an input screen generation unit that generates an input screen for inputting the measured data and the virtual data, The measured data acquisition unit acquires the measured data entered into the input screen generated by the input screen generation unit, The physical property prediction apparatus according to claim 1, wherein the virtual data acquisition unit acquires the virtual data input to the input screen generated by the input screen generation unit.

6. The system further comprises a communication unit that transmits the input screen generated by the input screen generation unit to the user's terminal device and receives the measured data and virtual data entered on the input screen from the terminal device. The measured data acquisition unit acquires the measured data received by the communication unit, The physical property prediction apparatus according to claim 5, wherein the virtual data acquisition unit acquires the virtual data received by the communication unit.

7. The system further includes a virtual data completion unit that completes the aforementioned virtual data, The physical property prediction apparatus according to claim 5, wherein the virtual data acquisition unit acquires the virtual data that has been supplemented by the virtual data supplementation unit.

8. The physical property prediction device according to claim 7, wherein the virtual data completion unit completes the second manufacturing conditions of the virtual data by calculating the values ​​of the parameters other than the partial parameters among the plurality of parameters when a subset of parameters having extremely small or large values ​​and the virtual values ​​predicted based on the partial parameters are input to the input screen.

9. The system further includes an uncertainty setting unit for setting a first uncertainty and a second uncertainty of the aforementioned physical properties, The material property prediction apparatus according to claim 1, wherein the model generation unit generates the prediction model based on the first uncertainty and the second uncertainty of the material property set by the uncertainty setting unit.

10. The system further comprises a setting screen generation unit that generates a setting screen for setting a first uncertainty and a second uncertainty of the aforementioned physical properties, The physical property prediction device according to claim 9, wherein the uncertainty setting unit sets the first uncertainty of the physical property and the second uncertainty of the physical property in accordance with the input operation to the setting screen generated by the setting screen generation unit.

11. The physical property prediction apparatus according to claim 1, wherein the model generation unit generates the prediction model according to the following equation. [Math 1] [Math 2]

12. The physical property prediction apparatus according to any one of claims 1 to 11, further comprising a manufacturing condition calculation unit that calculates third manufacturing conditions for the sample based on the prediction model generated by the model generation unit.

13. The physical property prediction apparatus according to claim 12, wherein the third manufacturing condition includes at least one of the manufacturing condition when the uncertainty of the physical property in the prediction model is maximized and the manufacturing condition when the uncertainty of the physical property in the prediction model is minimized.

14. The measurement data acquisition unit further acquires second measurement data, which includes the third manufacturing conditions calculated by the manufacturing condition calculation unit and second measurement values ​​obtained by measuring the physical properties of the sample manufactured based on the third manufacturing conditions. The physical property prediction apparatus according to claim 12, wherein the model generation unit updates the prediction model based on the second measured data acquired by the measured data acquisition unit.

15. The physical property prediction device according to claim 14, wherein the model generation unit does not update the prediction model if the second measured value is the set optimal value.

16. A computer, Obtain measured data including the first manufacturing conditions for the sample and the measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions. Virtual data is obtained that includes the second manufacturing conditions for the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing conditions. Based on the acquired actual measurement data and virtual data, a predictive model for the physical properties is generated such that the actual measurement data has a first uncertainty for the physical properties, and the virtual data has a second uncertainty for the physical properties that is greater than the first uncertainty for the physical properties. Equipped with, The method for predicting material properties involves generating the prediction model using Bayesian estimation, wherein the first uncertainty is a first variance, and the second uncertainty is a second variance that is greater than the first variance.

17. The method for predicting physical properties according to claim 16, wherein the computer displays an input screen for inputting the measured data and the virtual data, and the acquisition of the measured data and the acquisition of the virtual data are performed in accordance with input operations on the input screen.

18. On the computer, A procedure for obtaining measured data including first manufacturing conditions for a sample and measured values ​​of the physical properties of the sample manufactured based on the first manufacturing conditions. A procedure for acquiring virtual data including a second manufacturing condition for the sample and virtual values ​​representing the physical properties of the sample corresponding to the second manufacturing condition, and A procedure for generating a predictive model of the physical properties based on the acquired actual measurement data and virtual data, such that the model has a first uncertainty of the physical properties with respect to the actual measurement data and a second uncertainty of the physical properties with respect to the virtual data that is greater than the first uncertainty of the physical properties. Make it run, A material property prediction program in which the procedure for generating the prediction model is performed using Bayesian estimation, the first uncertainty being the first variance, and the second uncertainty being the second variance which is greater than the first variance.

Images