Material property generation apparatus, material property generation program, and material property generation method

Abstract

The optimal material properties are obtained from a set that includes not only the material properties of existing materials, but also hypothetical material properties within the range that the engineer in question is estimated to be able to manufacture. [Solution] The material property generation apparatus (4) comprises an acquisition unit (403) that acquires an input which includes a plurality of data that form the material properties, which is an input to a function model (413) generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment, and a generation unit (material property generation unit 404) that generates an evaluation property which is an evaluation material property corresponding to the input using the function model.

Description

[Technical Field]

[0001] This disclosure relates to a material property generation apparatus, a material property generation program, and a material property generation method. [Background technology]

[0002] A method for predicting motor characteristics (torque, efficiency, etc.) is known, which involves using computer-based electromagnetic field analysis simulations (for example, Patent Document 1). Patent Document 1 describes a method for reducing the cost and time required to fabricate and evaluate motors by processing electromagnetic steel sheets. This method uses numerical values ​​calculated by simulation to select the optimal electromagnetic steel sheet from among candidates provided by manufacturers, in order to maximize the torque performance of the motor. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2013-93454 [Overview of the project] [Problems that the invention aims to solve]

[0004] However, the method described in Patent Document 1 involves selecting from existing electrical steel sheets that have already been manufactured by manufacturers, so the material properties obtained are limited to those of the existing sheets. Therefore, the method described in Patent Document 1 cannot obtain material properties different from those of existing electrical steel sheets (for example, hypothetical material properties within the range that the engineer is presumed to be able to manufacture). The material properties of existing electrical steel sheets do not necessarily realize the best motor characteristics, and by exploring solutions that include hypothetical material properties within the range that the engineer is presumed to be able to manufacture, it is possible to obtain better motor characteristics.

[0005] One aspect of this disclosure aims to obtain optimal material properties from a set that includes not only the material properties of existing materials, but also hypothetical material properties within the range that the engineer in question is presumed to be able to manufacture. [Means for solving the problem]

[0006] To solve the above problems, a material property generation apparatus according to one aspect of the present disclosure includes: an acquisition unit that acquires an input to a function model generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment, the input including a plurality of data that form the material properties; a generation unit that uses the function model to generate an evaluation property which is an evaluation material property corresponding to the input acquired by the acquisition unit; an evaluation unit that acquires an evaluation index that can evaluate the characteristics of the electrical equipment calculated based on the evaluation property generated by the generation unit; and an output control unit that determines at least one of the evaluation properties from the evaluation property generated by the generation unit based on the evaluation index.

[0007] To solve the above problems, a material property generation method according to one aspect of the present disclosure is a material property generation method performed by a material property generation device, comprising: an acquisition step of acquiring an input to a function model generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment, the input including a plurality of data that form the material properties; a generation step of generating an evaluation property which is an evaluation material property corresponding to the input acquired in the acquisition step using the function model; an evaluation step of acquiring an evaluation index which is calculated based on the evaluation property generated in the generation step and which is capable of evaluating the characteristics of the electrical equipment; and a determination step of determining at least one evaluation property from the evaluation property generated in the generation step based on the evaluation index.

[0008] Each aspect of the material property generation apparatus according to this disclosure may be implemented by a computer, in which case a material property generation program that implements the material property generation apparatus by a computer by operating the computer as each part (software element) of the material property generation apparatus, and a computer-readable recording medium on which the program is recorded also fall within the scope of this disclosure. [Effects of the Invention]

[0009] According to one aspect of this disclosure, it is possible to obtain optimal material properties from a set that includes not only the material properties of existing materials, but also hypothetical material properties within a range that is difficult for even the engineer to easily estimate, but which the engineer can estimate to be manufacturable based on the results. Verification and experiments are necessary to determine whether the estimated hypothetical material can actually be manufactured, but it is possible to significantly increase the possibility of achieving even better properties than those obtained by applying existing materials to electrical equipment. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows an example of an IPMSM configuration applicable to each aspect of this disclosure. [Figure 2] This block diagram shows an example configuration of a model generation apparatus and a material property generation apparatus according to Embodiment 1. [Figure 3] This is a schematic diagram illustrating an example of the data flow when optimizing component parameters and manufacturing parameters. [Figure 4] This flowchart shows an example of the processing performed by the model generation apparatus and material property generation apparatus according to Embodiment 1. [Figure 5] This figure shows a comparison between the analysis results obtained using the method described in Embodiment 1 and the analysis results obtained for existing electrical steel sheets. [Figure 6] This is a block diagram showing an example configuration of a material property generation apparatus according to Embodiment 2. [Figure 7] This graph shows an example of the relationship between B50 and W10 / 400. [Figure 8]This is a diagram illustrating an example of a method for identifying a point within a triangle. [Figure 9] This is a flowchart illustrating an example of processing executed by the material property generation apparatus according to the second embodiment. [Figure 10] This is a diagram showing a comparison between the analysis result obtained by the method described in the second embodiment and the analysis result of an existing electrical steel sheet. [Figure 11] This is a block diagram illustrating a configuration example of a model generation apparatus and a material property generation apparatus according to the third embodiment. [Figure 12] This is a schematic diagram illustrating an example of the flow of each data when optimizing parameters related to low-dimensional vectors. [Figure 13] This is a flowchart illustrating an example of processing executed by the material property generation apparatus according to the third embodiment. [Figure 14] This is a diagram showing a comparison between the analysis result obtained by the method described in the third embodiment and the analysis result of an existing electrical steel sheet. [Figure 15] This is a block diagram illustrating a configuration example of the material property generation apparatus according to the fourth embodiment. [Figure 16] This is a graph showing an example of magnetization characteristics. [Figure 17] This is a graph showing an example of iron loss characteristics. [Figure 18] This is a flowchart illustrating an example of processing executed by the material property generation apparatus according to the fourth embodiment. [Figure 19] This is a diagram showing a comparison between the analysis result obtained by the method described in the fourth embodiment and the analysis result of an existing electrical steel sheet. [Mode for Carrying Out the Invention]

[0011] Hereinafter, the present disclosure will be described in detail with reference to the drawings, taking as an example a case where the present disclosure is applied to a core used in a motor, with an electric device being a motor (electric motor). In the description of the drawings, the same or substantially the same components are denoted by the same reference numerals, and repeated descriptions thereof will not be given.

[0012] The electrical characteristics of a motor (e.g., torque, efficiency) are mainly determined by the magnetic material used in the core (e.g., electrical steel sheet), the shape of the motor, operating conditions, and manufacturing method. Therefore, selecting the optimal magnetic material to match the motor design greatly contributes to improving the motor's characteristics. However, if a material is selected from existing magnetic materials, as in the conventional technology described above, the resulting material properties (e.g., magnetization properties and iron loss properties) will be limited to those existing properties, and there is a possibility that other properties more suitable for the motor design exist. Therefore, the inventor conceived the idea of ​​finding material properties of magnetic materials that can improve motor characteristics without being limited by the material properties of existing materials, by repeatedly evaluating the motor's characteristics using electromagnetic field analysis simulations while changing the material properties, since the material properties of magnetic materials such as electrical steel sheets can be adjusted by the manufacturing method.

[0013] It is well known that for magnetic materials used in motor cores, lower iron loss and higher magnetic flux density are preferable. Therefore, when optimizing material properties for the purpose of improving motor characteristics without any particular constraints, zero iron loss and infinite magnetic flux density would be optimal. However, such material properties, which are practically impossible to manufacture, are not a meaningful solution. Similarly, material properties that deviate significantly from the material properties of existing materials, and which are considered practically unmanufacturable given the current level of technology, are also not meaningful solutions. Therefore, the inventors, based on the understanding that material properties within a range that does not deviate significantly from existing materials are highly manufacturable, considered that by optimizing various parameters that allow for adjustment of material properties (described later) based on the material properties of existing magnetic materials, it is possible to find material properties that are actually highly manufacturable, that is, hypothetical material properties within a range that is estimated to be manufacturable by the engineer in question. In this way, by generating material properties based on the material properties of existing materials and tailored to motor design, it is possible to obtain a meaningful optimal solution that serves as a guideline for considering materials to be applied to motors and is also relatively highly manufacturable.

[0014] Based on the above-mentioned concept, this disclosure provides a configuration in which predetermined data is input to a generation model generated based on information about existing properties, which are the material properties of existing materials such as the material properties of existing magnetic materials. Evaluation properties, which are material properties for evaluation corresponding to the input, are generated, and the characteristics of electrical equipment using materials with evaluation properties are evaluated by electromagnetic field analysis simulation (numerical analysis), while the input to the generation model is updated (repeated) until predetermined termination conditions are met. This generation model is a model generated based on information about multiple existing properties used in electrical equipment. The reason for preparing information about multiple existing properties is that if only a single existing property is used, it is difficult to limit the range of evaluation properties generated by the generation model, and there is a possibility that the optimal solution will be an evaluation property that exceeds the range that the engineer can estimate to be feasible when looking at the results obtained, or that is not realistically feasible. Such "generation models" include, for example, "trained models," "characteristic set models," "reconstruction models," and "function models." Embodiments using each of these four models will be explained using the case of generating material properties of a motor as an example.

[0015] Furthermore, since this disclosure is based on the material properties of existing materials, the calculated optimal material properties may be the same as the material properties of one of the multiple existing materials. Therefore, we do not rule out the possibility that the material properties generated by the method described below may be the same as the material properties of existing materials.

[0016] First, let's explain an example of a motor with reference to Figure 1. Figure 1 shows an example where the motor is an IPMSM (Interior Permanent Magnet Synchronous Motor).

[0017] Figure 1 shows an example of the configuration of the IPMSM10. As shown in Figure 1, the IPMSM10 comprises a rotor 11 and a stator 12. The rotor 11 comprises a rotor core 111 and a plurality of permanent magnets 131, 132, and 133. The stator 12 comprises a stator core 121 having an annular core back 121a (yoke) and a plurality of teeth 121b protruding radially inward from the core back 121a, and windings (not shown) wound around each tooth. The rotor 11 is arranged coaxially with the fixed stator 12 and rotates around the axis of rotation.

[0018] The IPMSM10 shown in Figure 1 is an IPMSM in which permanent magnets 131 and 132, which serve as the magnetic poles of the rotor 11, are arranged in a V-shape, and the number of poles is 8. In Figure 1, one of the four regions obtained by dividing the cross-section that appears when the IPMSM10 is cut perpendicular to the rotation axis of the IPMSM10 is shown. Therefore, in Figure 1, the region of the IPMSM10 that constitutes two poles of the rotor 11 is shown. In Figure 1, the case in which there are 3 permanent magnets per pole is illustrated as an example, but the number of permanent magnets per pole is not particularly limited. The number of permanent magnets per pole may be 1, 2, or 4 or more. Also, although the number of poles is explained as 8, this is just an example, and the number of poles is not limited to 8. Furthermore, the number of teeth 121b on the stator 12 is not limited to that shown in Figure 1. The rotor core 111 and stator core 121 are constructed by laminating multiple electromagnetic steel sheets.

[0019] Furthermore, the type of motor is not limited to IPMSMs, but may also include other motors such as SPM motors (Surface Permanent Magnet Motors), induction motors (IM), electrically excited synchronous motors (EESM), synchronous reluctance motors (SynRM), and switched reluctance motors (SRM). Motors other than rotating electric machines, such as linear motors, may also be included. Moreover, the electrical equipment covered is not limited to motors. For example, generators and transformers may also be included.

[0020] Furthermore, while this disclosure uses a stator core as an example of a component of the electrical equipment covered, the materials of components that affect the characteristics of the electrical equipment, such as rotor cores, permanent magnets, and windings, may also be covered. In addition, while this disclosure uses electromagnetic steel sheets as an example of core materials, soft magnetic materials such as amorphous, permalloy, ferrite, and nanocrystalline materials may also be covered. Permanent magnets may include neodymium magnets, ferrite magnets, samarium-cobalt magnets, and alnico magnets. Winding materials may include copper and aluminum. The type of electromagnetic steel sheet is not particularly limited, but it may include non-oriented electromagnetic steel sheets used in rotating electrical machines such as motors and generators, as well as grain-oriented electromagnetic steel sheets used in transformers, for example. In other words, the target is any electrical equipment that has components whose material properties affect the characteristics of the electrical equipment, and the material properties of the generated material may also be those of such components, and are not limited to the equipment and component materials (electromagnetic steel sheets) described herein.

[0021] The embodiments described below for each generation model will be explained using the existing material as electrical steel sheet and the material properties as magnetization properties and iron loss properties. However, this disclosure is not limited thereto. The material properties only need to include at least one of magnetization properties or iron loss properties, and may also include other material properties. Alternatively, the material properties may include at least one other material property other than these.

[0022] [Embodiment 1] (Example configuration of model generation device 1a and material property generation device 1b) Referring to Figure 2, the model generation apparatus 1a and material property generation apparatus 1b according to Embodiment 1 will be described. Figure 2 is a block diagram showing an example of the configuration of the model generation apparatus 1a and the material property generation apparatus 1b.

[0023] As shown in Figure 2, the model generation device 1a comprises a control unit 10a that centrally controls all parts of the model generation device 1a, and a storage unit 11a that stores various data used by the model generation device 1a. The model generation device 1a also comprises a communication unit 12a for the model generation device 1a to communicate with other devices, an input unit 13a that receives various data inputs to the model generation device 1a, and an output unit 14a for the model generation device 1a to output various data. The control unit 10a also includes a learning data acquisition unit 101a and a learning unit 102a. The storage unit 11a stores the learning data 111a.

[0024] The model generation device 1a may be configured as a general-purpose computer. When the model generation device 1a is configured as a general-purpose computer, the control unit 10a may be configured to include a processor, memory, etc. The processor can execute various programs by reading programs from the storage unit 11a and loading the read programs into memory. The processor is not particularly limited, but can be implemented as, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an MPU (Micro Processor Unit), or an FPGA (Field-Programmable Gate Array). The number of processors provided is not particularly limited; there may be one or more.

[0025] The storage unit 11a may consist of an HDD (Hard Disk Drive), an SSD (Solid State Drive), or the like. Furthermore, the storage unit 11a does not necessarily have to be located in the model generation device 1a. For example, the storage unit 11a may be located on an external server. Such an external server may be a physical server or a cloud server. When the storage unit 11a is located on an external server, the model generation device 1a can access the storage unit 11a via a communication network such as a wired or wireless LAN (Local Area Network). Thus, Embodiment 1 includes both cases where the storage unit 11a is physically connected to the model generation device 1a and cases where the storage unit 11a is not physically connected to the model generation device 1a.

[0026] The input unit 13a is typically a keyboard, mouse, etc., but is not limited to these; any configuration that can input various types of data is acceptable. For example, the input unit 13a may be a microphone, touch panel, etc. The output unit 14a is typically a display, but is not limited to this; any configuration that can output various types of data is acceptable. For example, the output unit 14a may be a speaker, warning light, etc.

[0027] The training data 111a stored in the memory unit 11a will now be described. The training data 111a is labeled data used in machine learning. Machine learning is a method that learns the features contained in input data and generates a "model" that estimates the result corresponding to newly input data. The trained model 111b, which will be described later, is a trained model generated by machine learning.

[0028] The learning data 111a is generated by associating the input data with the correct data and stored in the memory unit 11a. Examples of "input data" here include "component parameters" and "manufacturing parameters". The "component parameters" and "manufacturing parameters" are explained below. Component parameters are parameters that indicate the various components that make up the electrical steel sheet, and are often expressed in mass %. Examples of elements to be considered as component parameters include Si, Mn, Al, C, S, N, P, Sn, Sb, Cu, Ni, Cr, Mg, Ca, Ce, La, Nd, etc. Manufacturing parameters are parameters that indicate the manufacturing conditions when manufacturing the electrical steel sheet. Information regarding the crystal structure may also be included as manufacturing parameters. In the manufacturing process of electrical steel sheets, hot rolling, hot-rolled sheet annealing, cold rolling, finish annealing, and coating application are performed. In the manufacturing process of motor cores, strain relief annealing may be performed after the core is formed. The parameters that indicate these manufacturing conditions may include the following parameters. (1) Temperature at each pass in the hot rolling process, cooling rate after passing through the final pass, temperature during winding, etc. (2) Annealing temperature and holding time in the hot-rolled sheet annealing process, reduction ratio in the cold-rolling process, etc. (3) Annealing temperature, holding time, type of coating to be applied in the finish annealing process, etc. (4) Annealing temperature, holding time, etc. in the stress relief annealing process

[0029] Thus, the input data in Embodiment 1 consists of component parameters that include at least some of the various components constituting the electrical steel sheet, and manufacturing parameters that include at least some of the manufacturing conditions when manufacturing the electrical steel sheet. The types and number of component parameters and manufacturing parameters that constitute the input data are appropriately selected according to the required accuracy and the degree of influence on accuracy. These parameters may be measured values ​​obtained by actual measurements, or, in addition to measured values, calculated values ​​obtained by simulations may be used.

[0030] Next, an example of "ground truth data" will be explained. Ground truth data is data that shows the material properties of electrical steel sheets manufactured using "component parameters" and "manufacturing parameters" as input data. The material properties of electrical steel sheets show the relationships between multiple variables, and include, for example, magnetization properties, iron loss properties, yield point (YP), tensile strength (TS), elongation (EL), hardness (e.g., Vickers hardness Hv), etc. Here, "magnetization properties" are properties that show the relationship between two variables, magnetic flux density and magnetic field strength, and for example, magnetic field strength measured within a predetermined range with a predetermined step size may be used. The predetermined range of magnetic flux density here may be, for example, 0 or more and 2T or less. The predetermined step size of magnetic flux density may be, for example, 0.1T. In this case, the magnetization properties show discrete properties that do not take continuous values, but are not limited to this. For example, the magnetization properties may be properties shown as continuous values ​​that approximate the measurement results. Note that the level of magnetic flux density to be measured may be coarser than described above. Furthermore, calculated values ​​derived from manufacturing conditions and / or crystal structure may be used as correct answer data. For example, magnetization characteristics obtained using the technology described in Japanese Patent Application Publication No. 2021-15032 may be included as correct answer data.

[0031] Furthermore, "iron loss characteristics" refer to characteristics that show the relationship between two variables, magnetic flux density and iron loss value, for each frequency during sinusoidal excitation. For example, iron loss values ​​measured within a predetermined range with a predetermined step size at 50Hz, 100Hz, 200Hz, 400Hz, and 800Hz may be used. The predetermined range of magnetic flux density here may be, for example, 0 or more and 2T or less. The predetermined step size of magnetic flux density may be, for example, 0.1T. In this case, the iron loss characteristics show discrete characteristics that do not take continuous values, but are not limited to this. For example, the iron loss characteristics may be characteristics shown as continuous values ​​that approximate the measurement results. The levels of magnetic flux density and frequency measured may be coarser than those described above. Also, a representative value of iron loss, such as the Steinmetz coefficient, may be used as the iron loss value.

[0032] Here, we will describe an example of a method for actually measuring and obtaining the magnetization characteristics and iron loss characteristics of existing electrical steel sheets, which will be used as the correct answer data mentioned above. Note that the acquisition method described here may also be adopted in Embodiments 2 to 4 described later. The magnetization characteristics and iron loss characteristics of existing electrical steel sheets can be obtained by known methods described in JIS C 2550-1:2011 "Test methods for electrical steel strips - Part 1: Method for measuring magnetic properties of electrical steel strips using an Epstein tester" (hereinafter referred to as Document 1), JIS C 2550-3:2019 "Test methods for electrical steel strips - Part 3: Method for measuring intermediate frequency magnetic properties" (hereinafter referred to as Document 2), JIS C 2556:2015 "Method for measuring magnetic properties of electrical steel strips using a single-sheet tester" (hereinafter referred to as Document 3), etc. Documents 1 and 2 are measurement methods using Epstein samples, and Document 3 is a measurement method using single-sheet samples. Note that in Document 3, the length of the single-sheet sample is specified as 500 mm or more, but measurements may be taken using other sample dimensions. For example, measurements may be performed using small single-plate samples, as described in the IEEJ Magnetics Workshop materials, MAG-08-180 "Development of a Magnetic Property Measurement Device under High Magnetic Flux Density Using a Single-Plate Magnetic Tester" (2008) and MAG-17-083 "Investigation of a Simple Magnetic Measurement Method Using a Small Single-Plate Magnetic Tester" (2017). As another example, measurements may be performed using an annular sample as described in Japanese Patent Publication No. 2015-114163, or by arranging multiple samples in an annular shape as described in Japanese Patent Publication No. 2007-139717. As yet another example, measurements may be performed using a sample in the shape of a motor stator core as described in Japanese Patent Publication No. 4660110. In order to obtain an appropriate optimal solution, the material properties used in the optimization calculation must be measured using the same method for all steel types, or the results obtained using different measurement methods must be converted to equivalent measurement results using correction coefficients, etc. The conversion method is not particularly limited, but for example, the method described in Annex JC "Method for converting measured values ​​obtained by a single-sheet tester to Epstein equivalent values" of JIS C 2556:2015 "Method for measuring the magnetic properties of electromagnetic steel strips using a single-sheet tester" can be used.Alternatively, information about the measurement method may be treated as a categorical variable and part of the material properties. The training data may be the ground truth data obtained in this way.

[0033] Furthermore, data from multiple electrical steel sheets may be used for both the "input data" and the "correct answer data." The reason for preparing data from multiple electrical steel sheets is as described above.

[0034] The learning data acquisition unit 101a acquires the learning data 111a from the storage unit 11a. The learning data acquisition unit 101a outputs the acquired learning data 111a to the learning unit 102a.

[0035] The learning unit 102a generates a trained model 111b (generative model) for generating material properties different from those of existing electrical steel sheets by machine learning using the training data 111a acquired from the training data acquisition unit 101a. The trained model 111b is a model generated by learning the relationship between the component parameters and manufacturing parameters of existing electrical steel sheets and the material properties of existing electrical steel sheets. The machine learning algorithm is not particularly limited. For example, the learning unit 102a may generate the trained model 111b using support vector regression, or it may generate the trained model 111b using a neural network.

[0036] The trained model 111b generated by the learning unit 102a is transmitted to the material property generation device 1b and stored in the storage unit 11b of the material property generation device 1b. Alternatively, the trained model 111b generated by the learning unit 102a may be stored in the storage unit 11a.

[0037] Next, the material property generation device 1b will be described. The material property generation device 1b includes a control unit 10b that controls all parts of the material property generation device 1b, and a storage unit 11b that stores various data used by the material property generation device 1b. The material property generation device 1b also includes a communication unit 12b for the material property generation device 1b to communicate with other devices, an input unit 13b that receives input of various data to the material property generation device 1b, and an output unit 14b for the material property generation device 1b to output various data. The control unit 10b also includes an acquisition unit 101b, an optimization calculation unit 110b, and an output control unit 105b. The optimization calculation unit 110b has the function of optimizing the parameters (component parameters and manufacturing parameters) that were given as an example above. This optimization calculation unit 110b includes a material property generation unit 102b, an equipment property calculation unit 103b, and an update unit 104b. As described above, the learned model 111b is stored in the storage unit 11b.

[0038] The material property generation device 1b may also be configured as a general-purpose computer, similar to the model generation device 1a. When the material property generation device 1b is configured as a general-purpose computer, the control unit 10b, storage unit 11b, communication unit 12b, input unit 13b, and output unit 14b may have the same configuration as the control unit 10a, storage unit 11a, communication unit 12a, input unit 13a, and output unit 14a of the model generation device 1a. That is, the storage unit 11b may also be provided on an external server instead of being provided in the material property generation device 1b. Embodiment 1 includes both cases where the storage unit 11b is physically connected to the material property generation device 1b and cases where the storage unit 11b is not physically connected to the material property generation device 1b.

[0039] The acquisition unit 101b acquires input (input data) to the trained model 111b, which is the generation model. The input data includes component parameters and manufacturing parameters. When processing starts, the acquisition unit 101b acquires initial parameters as initial values ​​of the input data. The initial parameters include component parameters and manufacturing parameters. The method by which the acquisition unit 101b acquires the initial parameters is not particularly limited. For example, the acquisition unit 101b may acquire the initial parameters from the storage unit 11b, or it may acquire data entered by the designer into the input unit 13b as initial parameters. Also, the initial parameters (component parameters and manufacturing parameters) may be parameters relating to existing electrical steel sheets. The acquisition unit 101b outputs the acquired initial parameters to the material property generation unit 102b included in the optimization calculation unit 110b. The initial parameters are the initial values ​​used in the optimization calculation. Since the optimization calculation is performed repeatedly, the input data used in the first optimization calculation among the input data input to the trained model 111b are the "initial parameters". Furthermore, after acquiring the initial parameters, the acquisition unit 101b acquires the updated input data if the input is updated by the update unit 104b, which will be described later.

[0040] The material property generation unit 102b generates material properties from initial parameters using a trained model 111b, which was created by learning the relationship between the component parameters and manufacturing parameters of existing electrical steel sheets and the material properties of existing electrical steel sheets. Specifically, the material property generation unit 102b generates material properties corresponding to the input data by inputting the input data acquired by the acquisition unit 101b to the trained model 111b. The material properties thus generated are the material properties output from the trained model 111b and are output to the equipment property calculation unit 103b as evaluation properties (material properties for evaluation).

[0041] The equipment characteristic calculation unit 103b calculates an evaluation index that can evaluate the characteristics of electrical equipment based on the evaluation characteristics output from the trained model 111b. The evaluation index includes at least one predetermined evaluation item of the electrical equipment. If the evaluation index includes multiple evaluation items, it becomes possible to comprehensively evaluate multiple evaluation items. For example, in the case of a motor, torque and efficiency are each evaluation items, and it becomes possible to evaluate the motor characteristics according to the specified content of the evaluation index. The method for calculating the evaluation items is not particularly limited, but for example, the finite element method (FEM) may be used. The data showing the evaluation items calculated by the equipment characteristic calculation unit 103b is output to the update unit 104b. However, this embodiment is not limited thereto, and the equipment characteristic calculation unit 103b may be configured to acquire evaluation indexes calculated based on evaluation characteristics by other devices, etc.

[0042] The update unit 104b updates the input to the trained model 111b until a predetermined termination condition is met. That is, the update unit 104b updates (changes) at least some of the component parameters and manufacturing parameters included in the input data according to a predetermined rule. When the input data is updated, the acquisition unit 101b described above acquires the input data updated by the update unit 104b. The target of the changes according to the predetermined rule may be only component parameters (component values) or only manufacturing parameters. Furthermore, the predetermined rule may be a change range, and may be defined for each type of component parameter or each type of manufacturing parameter.

[0043] Then, the material property generation unit 102b generates evaluation characteristics and the equipment property calculation unit 103b calculates evaluation indices for the updated input data, and the processing of the update unit 104b is repeatedly executed until the termination condition is met, thereby optimizing the input data. More specifically, such optimization processing may be performed using well-known optimization methods such as genetic algorithms, grid search, random search, and Bayesian optimization. This makes it possible to search for input data that yields the desired motor characteristics.

[0044] The calculation results from the material property generation unit 102b, the equipment property calculation unit 103b, and the update unit 104b, i.e., the calculation results from the optimization calculation unit 110b, are output to the output control unit 105b.

[0045] The output control unit 105b outputs the calculation results from the optimization calculation unit 110b to the output unit 14b. These calculation results may include information that identifies the content of one or more specified numbers of input data with superior motor characteristics calculated until the termination condition is met (such as the input data itself), or they may include other information such as motor characteristic values ​​corresponding to each input data (input data used in the repeatedly executed optimization calculation). The output method is arbitrary. For example, the output control unit 105b may output the calculation results in at least one of the following ways: display output, audio output, and print output. Note that the output destination device is not limited to the material property generation device 1b, but may be an external device.

[0046] Furthermore, the output control unit 105b may determine at least one evaluation characteristic from the evaluation characteristics generated by the material property generation unit 102b based on the evaluation index. Here, the "evaluation characteristic" may be the optimal value or the top few data points with high performance. The output control unit 105b may output output information that includes an input corresponding to the determined evaluation characteristic.

[0047] By reviewing the calculation results, designers can identify the optimized component and manufacturing parameters that ensure the motor possesses the desired characteristics. The material properties (magnetization properties and iron loss properties) of the electromagnetic steel sheet manufactured using these optimized component and manufacturing parameters may exhibit new properties that differ from those of existing electromagnetic steel sheets. A motor manufactured using such an electromagnetic steel sheet with new material properties is expected to exhibit improved performance compared to existing motors. Thus, according to Embodiment 1, it is possible to efficiently develop materials that realize new material properties that are highly feasible to manufacture and contribute to improving motor performance. Furthermore, since these new material properties are generated based on the material properties of existing materials, solutions that are not realistically feasible or impossible to manufacture (e.g., zero iron loss, infinite magnetic flux density, etc.) can be eliminated.

[0048] Although Figure 2 describes the model generation device 1a and the material property generation device 1b separately, the invention is not limited to this. A single information processing device that combines the functions of both the model generation device 1a and the material property generation device 1b is also included in the scope of this disclosure.

[0049] (Data flow during optimization) Figure 3 is a schematic diagram showing an example of the data flow when optimizing component parameters and manufacturing parameters.

[0050] As shown in Figure 3, the input data to the trained model 111b consists of the component parameters shown in block 100 (hereinafter referred to as component parameters 100) and the manufacturing parameters shown in block 101 (hereinafter referred to as manufacturing parameters 101). Note that, as described above, initial parameters are input to the trained model 111b during the first calculation.

[0051] The trained model 111b generates the material properties shown in block 102 (hereinafter referred to as material properties 102) from the component parameters 100 and the manufacturing parameters 101. Thus, the component parameters 100 and the manufacturing parameters 101 are explanatory variables in the trained model 111b, and the material properties 102 are the target variables in the trained model 111b. Note that the explanatory variables in the trained model 111b are not limited to the component parameters 100 and the manufacturing parameters 101, but any parameters that determine the material properties may be used in place of or in addition to these parameters.

[0052] The equipment characteristic calculation unit 103b calculates the equipment characteristics 103 (for example, torque and efficiency in the case of a motor) shown in block 103 using FEM or the like, based on the material characteristics 102. Then, component parameters and manufacturing parameters that can obtain the desired motor characteristics are searched for. At this time, the update unit 104b may search within the parameter range of the existing electrical steel sheet. For example, the update unit 104b may search within a range of specified values ​​for component parameters, or within the range of available values ​​for manufacturing parameters. Alternatively, if there is no numerical range for manufacturing parameters, the presence or absence of each manufacturing process described above may be treated as a parameter. This makes it possible to eliminate solutions that are realistically unlikely to be manufactured or are impossible to manufacture (for example, zero iron loss, infinite magnetic flux density, etc.).

[0053] (Process flow) Next, with reference to Figure 4, the processing flow performed by the material property generation device 1b will be explained. Figure 4 is a flowchart showing an example of the processing performed by the material property generation device 1b.

[0054] In step S101, the acquisition unit 101b acquires component parameters and manufacturing parameters as initial values ​​(initial parameters) for input to the trained model 111b.

[0055] In step S102, the material property generation unit 102b generates material properties (evaluation properties) corresponding to the input data by inputting the initial parameters obtained in the processing of step S101 into the trained model 111b.

[0056] In step S103, the equipment characteristics calculation unit 103b calculates evaluation indices such as torque and efficiency, which are capable of evaluating the characteristics of the motor, based on the evaluation characteristics generated in the processing of step S102. As mentioned above, the equipment characteristics calculation unit 103b may be configured to acquire evaluation indices calculated based on the evaluation characteristics by other devices or the like.

[0057] The process proceeds to step S104, where it is determined whether the motor evaluation index calculated in step S103 satisfies the termination condition. If it is determined that the motor evaluation index satisfies the termination condition (YES in step S104), the process proceeds to step S106. On the other hand, if it is determined that the motor evaluation index does not satisfy the termination condition (NO in step S104), the process proceeds to step S105. The termination condition can be set arbitrarily. For example, the termination condition may be set as the point at which the Pareto solution that balances torque and efficiency within the desired range is no longer updated, the completion of a predetermined number of updates, or the elapsed of a predetermined calculation time.

[0058] In step S105, the update unit 104b updates the component parameters and manufacturing parameters, and then the process returns to step S102. As a result, the acquisition unit 101b acquires the updated component parameters and manufacturing parameters as input data to the trained model 111b, and steps S102 to S105 are executed again. The method for updating the component parameters and manufacturing parameters and the termination conditions can be predetermined. For example, as described above, a genetic algorithm may be used. The processes in steps S102 to S105 are repeatedly executed until the termination condition is met, that is, until YES is determined in step S104.

[0059] In step S106, the output control unit 105b outputs the optimized calculation result obtained by repeating the processes in steps S102 to S105 to the output unit 14b.

[0060] As described above, the material property generation method according to Embodiment 1 includes an acquisition step (S101) of acquiring input to a generation model (a trained model 111b in this embodiment) generated based on information on multiple existing properties which are material properties of multiple existing materials used in electrical equipment; a generation step (S102) of generating evaluation properties which are evaluation material properties corresponding to the input acquired in the acquisition step using the generation model; and an evaluation step (S103) of acquiring an evaluation index that can evaluate the characteristics of electrical equipment, calculated based on the evaluation properties generated in the generation step.

[0061] Note that the processing flow shown in the flowchart in Figure 4 is just one example, and you may delete steps, add new steps, or change the processing order as long as it does not deviate from the main point.

[0062] (Regarding the analysis results) Next, an example of the analysis results obtained using the method described in Embodiment 1 will be explained with reference to Table 1 and Figure 5. Table 1 shows the analysis conditions for this example. As shown in Table 1, this example uses an IPMSM with 8 poles (IPMSM10 shown in Figure 1). Two evaluation indicators were selected: torque and efficiency, which are electrical characteristics of the motor.

[0063] [Table 1]

[0064] Figure 5 is a graph showing the relationship between torque and efficiency, comparing the analysis results using the method described in Embodiment 1 (Embodiment 1) with the analysis results for existing electrical steel sheets (Comparative Example 1). As is clear from Figure 5, Embodiment 1 provides highly feasible material properties based on the material properties of existing electrical steel sheets, and therefore, compared to the case of only existing electrical steel sheets shown in Comparative Example 1, Embodiment 1 provides a wider range of Pareto solutions. In other words, Embodiment 1 makes it possible to find material properties different from those of existing electrical steel sheets that can achieve both motor torque and efficiency. Furthermore, by generating material properties based on the material properties of existing electrical steel sheets to match motor design, it is possible to obtain a meaningful optimal solution that serves as a guideline for considering materials to be applied to motors and is also relatively easy to manufacture. Here, the motor designer can appropriately select and adopt the solution they consider most preferable from the obtained Pareto solutions. Alternatively, they can calculate the weighted sum of the two evaluation items, torque and efficiency, and adopt it as a new evaluation index, then select the solution for which this index has the optimal value. The weighting ratio of the multiple items can be appropriately selected by the designer.

[0065] However, this disclosure is not limited to the embodiments described above. The material property generation apparatus 1b (material property generation method shown in Figure 4) may not include an update unit 104b (update step S105).

[0066] (Effects and Benefits) As described above, the "generation model" in Embodiment 1 is a trained model 111b that has learned the relationship between parameters, which include at least one of component parameters indicating the components used in the existing material and manufacturing parameters indicating the manufacturing conditions of the existing material, and the existing properties.

[0067] According to the above configuration, by optimizing the component parameters and / or manufacturing parameters, a wider range of Pareto solutions can be obtained compared to existing materials (e.g., electrical steel sheets). This makes it possible to manufacture electrical equipment that meets a wide variety of needs.

[0068] Furthermore, with the above configuration, designers can grasp the optimized evaluation characteristics so that the electrical equipment has the desired characteristics. In addition, by including magnetization characteristics and iron loss characteristics in the material properties of existing materials, these can be optimized simultaneously, making it suitable for motors whose characteristics change according to magnetization characteristics and iron loss characteristics.

[0069] [Embodiment 2] Next, Embodiment 2 will be described.

[0070] (Example configuration of material property generation device 2) Figure 6 is a block diagram showing an example configuration of the material property generation apparatus 2 according to Embodiment 2.

[0071] The material property generation device 2 includes a control unit 20 that comprehensively controls all parts of the material property generation device 2, and a storage unit 21 that stores various data used by the material property generation device 2. The material property generation device 2 also includes a communication unit 22 for the material property generation device 2 to communicate with other devices, an input unit 23 that receives input of various data to the material property generation device 2, and an output unit 24 for the material property generation device 2 to output various data. The control unit 20 includes a selection unit 201, a setting unit 202, an acquisition unit 203, an optimization calculation unit 210, and an output control unit 207. The optimization calculation unit 210 includes an evaluation property set generation unit 204, a calculation unit 205, and an update unit 206.

[0072] The material property generation device 2 may also be configured as a general-purpose computer, similar to the material property generation device 1b described in Embodiment 1. When the material property generation device 2 is configured as a general-purpose computer, the control unit 20, storage unit 21, communication unit 22, input unit 23, and output unit 24 may have the same configuration as the control unit 10b, storage unit 11b, communication unit 12b, input unit 13b, and output unit 14b of the material property generation device 1b. That is, the storage unit 21 may also be provided on an external server instead of being provided on the material property generation device 2. Embodiment 2 also includes both cases in which the storage unit 21 is physically connected to the material property generation device 2 and cases in which the storage unit 21 is not physically connected to the material property generation device 2.

[0073] In this embodiment, the memory unit 21 described above stores a characteristic set generation model 213, which is a generation model. The characteristic set generation model 213 outputs a set of two or more characteristic item values ​​(a characteristic set for evaluation) as a material characteristic for evaluation corresponding to the input. The characteristic item values ​​are, for example, typical index values ​​used to evaluate material properties, and may be B50 for evaluating the magnetic flux density of electrical steel sheets, or W10 / 400 for evaluating iron loss. "B50" is the value of the magnetic flux density when the magnetic field strength is 5000 A / m. "W10 / 400" is the value of the iron loss when the frequency is 400 Hz and the magnetic flux density is 1.0 T. When the characteristic set generation model 213 receives an input, it outputs a set of characteristic item values, such as a set of B50 and W10 / 400 values, corresponding to the parameters specified in the input, as an evaluation characteristic.

[0074] In this embodiment shown in Figure 6, the memory unit 21 stores multiple types of material properties for each of the multiple electrical steel sheets (existing materials). The material property generation device 2 selects characteristic item values ​​for each of the multiple types of material properties for each electrical steel sheet, and generates a characteristic set generation model 213 that outputs an evaluation characteristic set from the input based on the set of existing characteristic item values ​​for each of the multiple electrical steel sheets obtained in this way (existing characteristic set). The reason for preparing multiple electrical steel sheets is as described above.

[0075] To elaborate on the embodiment shown in Figure 6, the memory unit 21 stores data indicating magnetization characteristics (hereinafter simply referred to as magnetization characteristics 211) and data indicating iron loss characteristics (hereinafter simply referred to as iron loss characteristics 212) as multiple types of material properties for each electrical steel sheet.

[0076] The selection unit 201 acquires the magnetization characteristics 211 and iron loss characteristics 212 of the existing electrical steel sheet from the storage unit 21. The acquisition method is not particularly limited, but for example, the method described in Embodiment 1 may be used.

[0077] The selection unit 201 selects multiple characteristic item values ​​for each material property of each electrical steel sheet. The characteristic item values ​​to be selected are specified in advance by the user or other party. The characteristic item values ​​are not particularly limited, but for example, "B50" may be selected for the magnetization property 211. For the iron loss property 212, "W10 / 400" may be selected. The selection unit 201 outputs an existing characteristic set (for example, a set of B50 and W10 / 400) which is a set of characteristic item values ​​selected for each electrical steel sheet to the setting unit 202. Note that since data exists for B50 and W10 / 400 for each steel type of electrical steel sheet, multiple existing characteristic sets corresponding to multiple steel types will be output to the setting unit 202.

[0078] The setting unit 202 generates a characteristic set generation model 213 using multiple existing characteristic sets selected by the selection unit 201, so as to uniquely output interpolation points for multiple existing characteristic sets according to the parameters specified by the input. In the characteristic set generation model 213, the input and output (interpolation points) correspond uniquely. The interpolation points are material properties within a range that does not deviate significantly from existing materials. Therefore, it is possible to optimize the evaluation index while limiting the input within a range that has a relatively high degree of real-world manufacturability. The general definition of interpolation is to obtain values ​​within an element by interpolating from known quantities at each point of the element, but the parameters are data that can specify the interpolation points.

[0079] Specifically, the characteristic set generation model 213 may be configured to divide the largest overall region enclosed by multiple existing characteristic sets into multiple non-overlapping subregions using three or more existing characteristic sets selected from among the multiple existing characteristic sets, and to identify interpolation points using identification numbers for the multiple subregions (e.g., the triangle numbers mentioned above) and weights that identify points inside the subregions. In this case, the input to the characteristic set generation model 213 includes the identification numbers for the subregions and the weights. This input is also acquired by the acquisition unit 203. Furthermore, the subregions may be formed by three existing characteristic sets that are adjacent to each other, and the subregions may not contain other existing characteristic sets. The existing characteristic sets may be characteristic item values ​​obtained from the magnetization characteristics and iron loss characteristics of existing materials, respectively.

[0080] Here, referring to Figures 7 and 8, we will explain an example of how to set parameters when dividing into multiple division regions using three existing characteristic sets.

[0081] Figure 7 is a plot of multiple existing characteristic sets CH1, CH2, CH3, ... selected by the selection unit 201 on a plane, where the existing characteristic sets CH1, CH2, CH3, ... are sets of characteristic item values ​​shown on each axis. In Figure 7, the horizontal axis represents B50 and the vertical axis represents W10 / 400. Figure 8 is a diagram illustrating the position of point 26a (interpolation point) within triangle 25a shown in Figure 7.

[0082] As shown in Figure 7, the setting unit 202 plots the existing characteristic sets CH1, CH2, CH3, ... obtained for each electrical steel sheet on a plane. Each plot corresponds to one electrical steel sheet. As mentioned above, since data exists for each type of electrical steel sheet, multiple points will be plotted. Next, the setting unit 202 divides the set of points into triangles. The division method is not particularly limited, but for example, Delaunay triangulation may be used. According to Delaunay triangulation, triangles can be generated such that no data (B50 and W10 / 400) exists inside the circumcircle of each triangle. In other words, each vertex of the triangle will represent the B50 and W10 / 400 of the three existing electrical steel sheets. The triangles divided in this way are assigned numbers. Figure 7 illustrates an example in which the divided triangles are assigned numbers such as 25a, 25b, 25c, ... Any point inside each triangle (triangle 25a, triangle 25b, triangle 25c, ...) can be represented by a weighted linear sum of the position vectors of each vertex. Therefore, by optimizing the selected triangles and weights, the optimal point can be found. Specifically, by changing the line segments r1 to r3 shown in Figure 8 so that their sum is 1, any point 26a inside triangle 25a can be represented. Regarding the magnetization characteristics, if BH1, BH2, and BH3 are assigned to the B50 values ​​at each vertex of triangle 25a, then the magnetization characteristics BH calc This can be calculated using the following equation 1 (weighted linear sum of the position vectors of each vertex). Note that the values ​​of each line segment r1 to r3 may be 0. That is, the values ​​of each line segment r1 to r3 are 0 ≤ r n The value ≤ 1 (n=1,2,3) is also acceptable.

[0083]

number

[0084] Furthermore, regarding the iron loss characteristics, if the W10 / 400 at each vertex of triangle 25a is BW1, BW2, and BW3, then the iron loss characteristics BW at any point 26a inside triangle 25a are...calc This can be calculated using the following equation 2 (weighted linear sum of the position vectors of each vertex).

[0085]

number

[0086] The setting unit 202 sets the triangle number and interpolation ratios r1 and r2 as parameters to be used in the optimization calculation (inputs to the characteristic set generation model 213). This method generates magnetization and iron loss characteristics different from those of existing electrical steel sheets, as a weighted average of existing electrical steel sheets. Therefore, solutions that are practically unmanufacturable or impossible to manufacture (e.g., zero iron loss, infinite magnetic flux density) can be excluded, and optimization can be performed by excluding unmanufacturable material characteristics. Figure 7 illustrates an example of dividing the point cloud set into triangles, but is not limited to this. For example, the minimum number of plots may be 2, meaning the point cloud set may be represented by straight lines. Furthermore, it may be divided into polygons other than triangles. In this case, it is preferable that no concave areas are formed in the polygons other than triangles. Also, Figure 7 illustrates an example of plotting in a two-dimensional plane, but is not limited to this; if other characteristic item values ​​are to be further considered, the plots may be in three-dimensional space. If the measurement points for magnetization and iron loss characteristics are not aligned, the data points may be aligned by interpolation. For example, if B50 is used and there are no data points with a magnetization force of 5000 A / m, B50 may be interpolated from measured values ​​under nearby conditions.

[0087] In this way, the setting unit 202 generates a characteristic set generation model 213 based on multiple existing characteristic sets, each consisting of two or more sets of characteristic item values ​​(for example, the set of B50 and W10 / 400 described above) for each of multiple existing materials used in electrical equipment (e.g., motors). The characteristic set generation model 213 may also be configured to output an evaluation characteristic set obtained by interpolation from at least two of the multiple existing characteristic sets specified by the input. Such an input may be an interpolation ratio that identifies the interpolation points of at least two existing characteristic sets. Such an input is acquired by the acquisition unit 203.

[0088] The evaluation characteristic set generation unit 204 uses the characteristic set generation model 213 to generate an evaluation characteristic set, which is an evaluation characteristic set corresponding to the input acquired by the acquisition unit 203. The evaluation characteristic set generated by the evaluation characteristic set generation unit 204 is output to the calculation unit 205.

[0089] The calculation unit 205 generates two or more material properties corresponding to two or more characteristic item values ​​based on the evaluation characteristic set generated by the evaluation characteristic set generation unit 204, and calculates an evaluation index that can evaluate the characteristics of electrical equipment using the generated material properties. An example of the "evaluation index" was explained in Embodiment 1, so the explanation is omitted here. The data showing the evaluation items calculated by the calculation unit 205 is output to the update unit 206. Note that the calculation unit 205 may be configured to acquire an evaluation index calculated using two or more material properties calculated using two or more characteristic item values, which are generated by another device or the like based on the evaluation characteristic set.

[0090] The update unit 206 updates the inputs to the characteristic set generation model 213 (for example, parameters such as triangle numbers and interpolation ratios r1, r2) until predetermined termination conditions are met. That is, the update unit 206 updates (changes) at least some of the parameters included in the input data according to predetermined rules.

[0091] Then, the evaluation characteristic set generation unit 204 generates evaluation characteristic sets and the calculation unit 205 calculates evaluation indices for the updated input data, and the processing of the update unit 206 is repeatedly executed until the termination condition is met, thereby optimizing the input data. More specifically, such optimization processing may be performed using well-known optimization methods such as genetic algorithms, grid search, random search, and Bayesian optimization. This makes it possible to search for parameters that will yield the desired motor characteristics.

[0092] The calculation results from the evaluation characteristic set generation unit 204, the calculation unit 205, and the update unit 206, i.e., the calculation results from the optimization calculation unit 210, are output to the output control unit 207.

[0093] The output control unit 207 outputs the calculation results from the optimization calculation unit 210 to the output unit 24. The output method is arbitrary. For example, the output control unit 207 may output the calculation results in at least one of the following ways: display output, audio output, and printed output. The output destination device is not limited to the material property generation device 2, but may be an external device. By checking the calculation results, the designer can understand the magnetization characteristics and iron loss characteristics of the existing electromagnetic steel sheet, which have been optimized so that the motor has the desired characteristics, and which are different from the magnetization characteristics and iron loss characteristics of the existing electromagnetic steel sheet.

[0094] Furthermore, the output control unit 207 may determine at least one evaluation characteristic set from the evaluation characteristic sets generated by the evaluation characteristic set generation unit 204 based on the evaluation index. Here, "evaluation characteristic set" may be data consisting of the top few sets with high performance. The output control unit 207 may output output information that includes the input corresponding to the determined evaluation characteristic set.

[0095] (Process flow) Next, with reference to Figure 9, the processing flow performed by the material property generation device 2 will be explained. Figure 9 is a flowchart showing an example of the processing performed by the material property generation device 2.

[0096] In step S201, the acquisition unit 203 acquires inputs for the characteristic set generation model 213 (for example, the triangle number and interpolation ratios r1 and r2 mentioned above).

[0097] In step S202, the evaluation characteristic set generation unit 204 uses the characteristic set generation model 213 to generate an evaluation characteristic set, which is an evaluation characteristic set corresponding to the input obtained in the processing of step S201.

[0098] In step S203, the calculation unit 205 generates two or more material properties corresponding to two or more characteristic item values ​​based on the evaluation characteristic set generated in the processing of step S202, and uses the generated material properties to calculate evaluation indices such as torque and efficiency that can evaluate the characteristics of the motor. As described above, the calculation unit 205 may be configured to obtain evaluation indices calculated using evaluation indices calculated using evaluation indices that are generated by another device or the like based on the evaluation characteristic set, which generates two or more material properties corresponding to two or more characteristic item values.

[0099] The process proceeds to step S204, where it is determined whether the motor evaluation index calculated in step S203 satisfies the termination condition. If it is determined that the motor evaluation index satisfies the termination condition (YES in step S204), the process proceeds to step S206. On the other hand, if it is determined that the motor evaluation index does not satisfy the termination condition (NO in step S204), the process proceeds to step S205. The termination condition can be set arbitrarily. For example, the termination condition may be set as the point at which the Pareto solution that balances torque and efficiency within the desired range is no longer updated, the completion of a predetermined number of updates, or the elapsed of a predetermined calculation time.

[0100] In step S205, the update unit 206 updates the parameters (e.g., triangle number, interpolation ratios r1, r2), and then the process returns to step S202. As a result, the acquisition unit 203 acquires the updated parameters as input data for the characteristic set generation model 213, and steps S202 to S205 are executed again. The parameter update method and termination conditions can be predetermined. For example, as described above, a genetic algorithm may be used. The processes in steps S202 to S205 are repeatedly executed until the termination condition is met, that is, until YES is determined in step S204.

[0101] In step S206, the output control unit 207 outputs the optimized calculation result obtained by repeating the processing in steps S202 to S205 to the output unit 24.

[0102] As described above, the material property generation method according to Embodiment 2 includes an acquisition step (S201) of acquiring input to a property set generation model 213 generated based on a plurality of existing property sets, each consisting of a set of two or more property item values ​​for a plurality of existing materials used in electrical equipment; a generation step (S202) of generating an evaluation property set, which is an evaluation property set corresponding to the input acquired in the acquisition step, using the property set generation model 213; and an evaluation step (S203) of acquiring an evaluation index that can evaluate the characteristics of electrical equipment, which is calculated using two or more material properties corresponding to two or more property item values ​​generated based on the evaluation property set.

[0103] (Regarding the analysis results) Next, an example of the analysis results obtained using the method described in Embodiment 2 will be explained with reference to Table 2 and Figure 10. Table 2 shows the analysis conditions for this example. As shown in Table 2, this example uses an IPMSM with 8 poles (IPMSM10 shown in Figure 1). Two evaluation indicators were selected: torque and efficiency, which are electrical characteristics of the motor.

[0104] [Table 2]

[0105] Figure 10 is a graph showing the relationship between torque and efficiency, comparing the analysis results using the method described in Embodiment 2 (Embodiment 2) with the analysis results for existing electrical steel sheets (Comparative Example 2). As is clear from Figure 10, Embodiment 2 provides highly feasible material properties based on the material properties of existing electrical steel sheets, and therefore, compared to the case of only existing electrical steel sheets shown in Comparative Example 2, Embodiment 2 provides a wider range of Pareto solutions. In other words, Embodiment 2 makes it possible to find material properties different from those of existing electrical steel sheets that can achieve both motor torque and efficiency. Furthermore, by generating material properties based on the material properties of existing electrical steel sheets to match motor design, it is possible to obtain a meaningful optimal solution that serves as a guideline for considering materials to be applied to motors and is also relatively easy to manufacture. Here, the motor designer can appropriately select and adopt the solution they consider most preferable from the obtained Pareto solutions. Alternatively, they can calculate the weighted sum of the two evaluation items, torque and efficiency, and adopt it as a new evaluation index, then select the solution for which this index has the optimal value. The weighting ratio of the multiple items can be appropriately selected by the designer.

[0106] However, this disclosure is not limited to the embodiments described above. The material property generation apparatus 2 (material property generation method shown in Figure 7) may not include an update unit 206 (update step S205).

[0107] (Effects and Benefits) As described above, the "generation model" in Embodiment 2 is a characteristic set generation model 213 generated based on a plurality of existing characteristic sets, each consisting of a set of two or more characteristic item values ​​for each of a plurality of existing materials used in electrical equipment.

[0108] According to the above configuration, for example, by optimizing the interpolation ratios r1 and r2, a wider range of Pareto solutions can be obtained compared to existing materials (e.g., electrical steel sheets), making it possible to manufacture electrical equipment that meets a wide variety of needs.

[0109] [Embodiment 3] Next, Embodiment 3 will be described.

[0110] (Example configuration of model generation device 3a and material property generation device 3b) Figure 11 is a block diagram showing an example configuration of the model generation apparatus 3a and the material property generation apparatus 3b according to Embodiment 3.

[0111] As shown in Figure 11, the model generation device 3a comprises a control unit 30a that centrally controls all parts of the model generation device 3a, and a storage unit 31a that stores various data used by the model generation device 3a. The model generation device 3a also comprises a communication unit 32a for the model generation device 3a to communicate with other devices, an input unit 33a that receives various data inputs to the model generation device 3a, and an output unit 34a for the model generation device 3a to output various data. The control unit 30a also includes a learning data acquisition unit 301a and a learning unit 302a. The storage unit 31a stores learning data 311a.

[0112] The model generation device 3a may also be configured as a general-purpose computer, similar to the model generation device 1a described in Embodiment 1. When the model generation device 3a is configured as a general-purpose computer, the control unit 30a, storage unit 31a, communication unit 32a, input unit 33a, and output unit 34a may have the same configuration as the control unit 10a, storage unit 11a, communication unit 12a, input unit 13a, and output unit 14a of the model generation device 1a described in Embodiment 1. That is, the storage unit 31a may also be provided on an external server instead of being provided on the model generation device 3a. Embodiment 3 also includes both cases in which the storage unit 31a is physically connected to the model generation device 3a and cases in which the storage unit 31a is not physically connected to the model generation device 3a.

[0113] The training data 311a includes data showing the material properties of existing electrical steel sheets. In the example shown in Figure 11, such material properties include magnetization properties and iron loss properties. Therefore, the training data 311a includes data showing the magnetization properties and iron loss properties of existing electrical steel sheets (hereinafter sometimes simply referred to as "existing data").

[0114] The learning data acquisition unit 301a acquires the learning data 311a from the storage unit 31a. The learning data acquisition unit 301a outputs the acquired learning data 311a to the learning unit 302a.

[0115] The learning unit 302a generates a compression model for dimensionality reduction of existing data and a restoration model for restoring the compressed information obtained by the compression model through learning. The learning method is not particularly limited, but for example, principal component analysis (PCA) may be used. By using principal component analysis, it is possible to generate a compression model for dimensionality reduction and a restoration model for restoring the compressed information obtained by this compression model through machine learning without supervised data.

[0116] Here, the existing data is a multidimensional quantity that includes multiple data (multiple discrete data) that form (constitute) material properties such as measurement results at multiple points. If such a multidimensional quantity is handled as is, there are concerns about increased computation time and trapping into local minima. Therefore, in Embodiment 3, the existing data is reduced in dimensionality using a compression model, and the resulting low-dimensional vector (compressed information) is optimized. In other words, optimization is performed by updating the low-dimensional vector, restoring the updated low-dimensional vector using a restoration model, and calculating evaluation indices using these restored material properties as evaluation characteristics. In this embodiment, the magnetization properties and iron loss properties are optimized simultaneously. With this configuration, by limiting the search range to the region where low-dimensional vectors of existing electrical steel sheets exist, it is possible to exclude solutions that are realistically unlikely to be manufactured or are impossible to manufacture (e.g., zero iron loss, infinite magnetic flux density, etc.), and to optimize while excluding material properties that are impossible to manufacture.

[0117] Furthermore, dimensionality reduction and restoration methods are not limited to principal component analysis; for example, an autoencoder (AE) may be used. When using an autoencoder, the compression model corresponds to the encoder portion of the autoencoder, and the restoration model corresponds to the decoder portion of the autoencoder. Other dimensionality reduction methods capable of both compression and restoration may also be used.

[0118] The compressed model 311b and the restored model 312b generated by the learning unit 302a are transmitted to the material property generation device 3b and stored in the storage unit 31b of the material property generation device 3b. Alternatively, the compressed model 311b and the restored model 312b generated by the learning unit 302a may be stored in the storage unit 31a.

[0119] Next, the material property generation device 3b will be described. The material property generation device 3b includes a control unit 30b that controls all parts of the material property generation device 3b, and a storage unit 31b that stores various data used by the material property generation device 3b. The material property generation device 3b also includes a communication unit 32b for the material property generation device 3b to communicate with other devices, an input unit 33b that receives input of various data to the material property generation device 3b, and an output unit 34b for the material property generation device 3b to output various data. The control unit 30b also includes an acquisition unit 301b, an optimization calculation unit 310b, and an output control unit 305b. The optimization calculation unit 310b includes a material property generation unit 302b, an equipment property calculation unit 303b, and an update unit 304b. As described above, the storage unit 31b stores the compressed model 311b and the restored model 312b.

[0120] The material property generation device 3b may also be configured as a general-purpose computer, similar to the material property generation device 1b described in Embodiment 1. When the material property generation device 3b is configured as a general-purpose computer, the control unit 30b, storage unit 31b, communication unit 32b, input unit 33b, and output unit 34b may have the same configuration as the control unit 10b, storage unit 11b, communication unit 12b, input unit 13b, and output unit 14b of the material property generation device 1b. That is, the storage unit 31b may also be provided on an external server instead of being provided on the material property generation device 3b. Embodiment 3 also includes both cases in which the storage unit 31b is physically connected to the material property generation device 3b and cases in which the storage unit 31b is not physically connected to the material property generation device 3b.

[0121] The acquisition unit 301b acquires the input to the restoration model 312b, i.e., the compressed information compressed by the compression model 311b. The acquisition unit 301b outputs the acquired input data (compressed information) to the material property generation unit 302b included in the optimization calculation unit 310b. The input to the compression model 311b may be existing properties. However, the input to the compression model 311b does not need to be existing properties, and any data is acceptable, as long as the data format is correct. As an example of the data format, if multiple data are input, the format may be arranged in a single line, or it may be arranged in a predetermined order (for example, property A followed by property B).

[0122] The material property generation unit 302b reconstructs compression information using the reconstruction model 312b, that is, it generates evaluation characteristics, which are material properties for evaluation that correspond to the compression information. As a result, for example, magnetization characteristics and iron loss characteristics different from those of existing electrical steel sheets are generated. The evaluation characteristics generated by the material property generation unit 302b are output to the equipment property calculation unit 303b.

[0123] The equipment characteristics calculation unit 303b calculates evaluation indices that can evaluate the characteristics of electrical equipment based on the material characteristics for evaluation. An example of an "evaluation index" was explained in Embodiment 1, so the explanation is omitted here. The data showing the evaluation items calculated by the equipment characteristics calculation unit 303b is output to the update unit 304b. The equipment characteristics calculation unit 303b may also be configured to acquire evaluation indices calculated based on evaluation characteristics by other devices, etc.

[0124] The update unit 304b updates the input to the reconstruction model 312b until a predetermined termination condition is met. That is, the update unit 304b updates (changes) at least some of the parameters related to the low-dimensional vectors included in the input data according to a predetermined rule. Once the input data is updated, the acquisition unit 301b described above acquires the input data updated by the update unit 304b.

[0125] Then, the material property generation unit 302b generates evaluation characteristics and the equipment property calculation unit 303b calculates evaluation indices for the updated input data, and the processing of the update unit 304b is repeatedly executed until the termination condition is met, thereby optimizing the input data. More specifically, such optimization processing may be performed using well-known optimization methods such as genetic algorithms, grid search, random search, and Bayesian optimization. This makes it possible to search for input data that yields the desired motor characteristics.

[0126] The calculation results from the material property generation unit 302b, the equipment property calculation unit 303b, and the update unit 304b, i.e., the calculation results from the optimization calculation unit 310b, are output to the output control unit 305b.

[0127] The output control unit 305b outputs the calculation results from the optimization calculation unit 310b to the output unit 34b. The output method can be arbitrary. For example, the output control unit 305b may output the calculation results in at least one of the following ways: display output, audio output, and printed output. The output destination device is not limited to the material property generation device 3b, but may be an external device. By checking the calculation results, the designer can understand the magnetization characteristics and iron loss characteristics of the existing electromagnetic steel sheet, which have been optimized so that the motor has the desired characteristics, and which are different from the magnetization characteristics and iron loss characteristics of the existing electromagnetic steel sheet.

[0128] Furthermore, the output control unit 305b may determine at least one evaluation characteristic from the evaluation characteristics generated by the material property generation unit 302b based on the evaluation index. The "evaluation characteristic set" here may be data consisting of the top few sets with high performance. The output control unit 305b may output output information that includes an input corresponding to the determined evaluation characteristic.

[0129] Although Figure 11 describes the model generation device 3a and the material property generation device 3b separately, the invention is not limited to this. A single information processing device that combines the functions of both the model generation device 3a and the material property generation device 3b is also included in the scope of this disclosure.

[0130] (Data flow during optimization) Figure 12 is a schematic diagram showing an example of the data flow when optimizing parameters for low-dimensional vectors.

[0131] As shown in Figure 12, the input data to the compression model 311b is existing data (hereinafter referred to as existing data 300), indicated by block 300. The compression model 311b generates a low-dimensional vector 310 from the existing data 300 by compressing its dimensions. As described above, the parameters (compression information) related to the low-dimensional vector 310 are the parameters that are the target of the optimization calculation. This compression information is acquired by the acquisition unit 301b.

[0132] The reconstructed model 312b reconstructs the low-dimensional vector 310 and generates material properties corresponding to the low-dimensional vector 310, shown in block 320 (hereinafter referred to as material properties 320). In the example shown in Figure 12, the material properties 320 include magnetization properties and iron loss properties, as described above.

[0133] The equipment characteristics calculation unit 303b calculates the equipment characteristics 330 (for example, torque and efficiency in the case of a motor) shown in block 330 using FEM or the like, based on the material characteristics 320. Then, parameters that yield the desired motor characteristics are searched for. Since these parameters are searched from within the parameter range of existing electrical steel sheets, solutions that are realistically unlikely to be manufactured or are impossible to manufacture (for example, zero iron loss, infinite magnetic flux density, etc.) can be excluded, and optimization can be performed by excluding material characteristics that are impossible to manufacture.

[0134] (Process flow) Next, with reference to Figure 13, the processing flow performed by the material property generation device 3b will be explained. Figure 13 is a flowchart showing an example of the processing performed by the material property generation device 3b.

[0135] In step S301, the acquisition unit 301b acquires the input to the restoration model 312b, that is, the compressed information compressed by the compression model 311b.

[0136] In step S302, the material property generation unit 302b uses the restoration model 312b to generate evaluation properties, which are material properties for evaluation that correspond to the compression information obtained in the processing of step S301.

[0137] In step S303, the equipment characteristics calculation unit 303b calculates evaluation indices such as torque and efficiency, which are capable of evaluating the characteristics of the motor, based on the evaluation characteristics generated in the processing of step S302. As described above, the equipment characteristics calculation unit 303b may be configured to acquire evaluation indices calculated based on the evaluation characteristics by other devices or the like.

[0138] The process proceeds to step S304, where it is determined whether the motor evaluation index calculated in step S303 satisfies the termination condition. If it is determined that the motor evaluation index satisfies the termination condition (YES in step S304), the process proceeds to step S306. On the other hand, if it is determined that the motor evaluation index does not satisfy the termination condition (NO in step S304), the process proceeds to step S305. The termination condition can be set arbitrarily. For example, the termination condition may be set as the point at which the Pareto solution that balances torque and efficiency within the desired range is no longer updated, the completion of a predetermined number of updates, or the elapsed of a predetermined calculation time.

[0139] In step S305, the update unit 304b updates the parameters (compression information) related to the low-dimensional vector, and then the process returns to step S302. As a result, the acquisition unit 301b acquires the updated parameters as input compression information for the reconstructed model 312b, and steps S302 to S305 are executed again. The parameter update method and termination conditions can be predetermined. For example, as described above, a genetic algorithm may be used. In addition, when updating the parameters related to the low-dimensional vector, the parameters may be restricted to the range and vicinity of the low-dimensional vector obtained by compressing the existing material. For example, if the parameters deviate from the range of the low-dimensional vector obtained by compressing the existing material, a penalty may be imposed on the evaluation index. The process in steps S302 to S305 is repeatedly executed until the termination condition is met, that is, until YES is determined in step S304.

[0140] In step S306, the output control unit 305b outputs the optimized calculation result obtained by repeating the processing in steps S302 to S305 to the output unit 34b.

[0141] As described above, the material property generation method according to Embodiment 3 includes an acquisition step (S301) in which input compressed information is acquired as input to a restoration model 312b that can reconstruct compressed information obtained by compressing a plurality of data that form the material properties, and the restoration model 312b is generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment; a generation step (S302) in which evaluation properties are material properties for evaluation corresponding to the input compressed information acquired in the acquisition step are generated using the restoration model 312b; and an evaluation step (S303) in which an evaluation index is obtained that can evaluate the characteristics of electrical equipment and is calculated based on the evaluation properties generated in the generation step.

[0142] Note that the processing flow shown in the flowchart in Figure 13 is just one example, and you may delete steps, add new steps, or change the processing order as long as it does not deviate from the main point.

[0143] (Regarding the analysis results) Next, an example of the analysis results obtained using the method described in Embodiment 3 will be explained with reference to Table 3 and Figure 14. Table 3 shows the analysis conditions for this example. As shown in Table 3, this example uses an IPMSM with 8 poles (IPMSM10 shown in Figure 1). Two evaluation indicators were selected: torque and efficiency, which are electrical characteristics of the motor.

[0144] [Table 3]

[0145] Figure 14 is a graph showing the relationship between torque and efficiency, comparing the analysis results using the method described in Embodiment 3 (Embodiment 3) with the analysis results for existing electrical steel sheets (Comparative Example 3). As is clear from Figure 14, Embodiment 3 provides highly feasible material properties based on the material properties of existing electrical steel sheets, and therefore, compared to the case of only existing electrical steel sheets shown in Comparative Example 3, Embodiment 3 provides a wider range of Pareto solutions. In other words, Embodiment 3 makes it possible to find material properties different from those of existing electrical steel sheets that can achieve both motor torque and efficiency. Furthermore, by generating material properties based on the material properties of existing electrical steel sheets to match motor design, it is possible to obtain a meaningful optimal solution that serves as a guideline for considering materials to be applied to motors and is also relatively easy to manufacture. Here, the motor designer can appropriately select and adopt the solution they consider most preferable from the obtained Pareto solutions. Alternatively, they can calculate the weighted sum of the two evaluation items, torque and efficiency, and adopt it as a new evaluation index, then select the solution for which this index has the optimal value. The weighting ratio of the multiple items can be appropriately selected by the designer.

[0146] However, this disclosure is not limited to the embodiments described above. The material property generation apparatus 3b (material property generation method shown in Figure 13) may not include an update unit 304b (update step S305).

[0147] (Effects and Benefits) As described above, the "generation model" in Embodiment 3 is a model generated based on multiple existing properties, which are material properties of multiple existing materials used in electrical equipment, and is a restored model 312b that can restore compressed information obtained by compressing multiple data that form the material properties.

[0148] According to the above configuration, by optimizing the compression information, a wider range of Pareto solutions can be obtained compared to existing materials (e.g., electrical steel sheets), making it possible to manufacture electrical equipment that meets a wide variety of needs.

[0149] [Embodiment 4] Next, Embodiment 4 will be described.

[0150] (Example configuration of material property generation device 4) Figure 15 is a block diagram showing an example configuration of the material property generation apparatus 4 according to Embodiment 4.

[0151] The material property generation device 4 includes a control unit 40 that comprehensively controls each part of the material property generation device 4, and a storage unit 41 that stores various data used by the material property generation device 4. The material property generation device 4 also includes a communication unit 42 for the material property generation device 4 to communicate with other devices, an input unit 43 that receives input of various data to the material property generation device 4, and an output unit 44 for the material property generation device 4 to output various data. The control unit 40 includes an approximation unit 401, a setting unit 402, an acquisition unit 403, an optimization calculation unit 410, and an output control unit 407. The optimization calculation unit 410 includes a material property generation unit 404, an equipment property calculation unit 405, and an update unit 406. The storage unit 41 stores data showing the magnetization characteristics of multiple existing electrical steel sheets (hereinafter simply referred to as magnetization characteristics 411) and data showing the iron loss characteristics (hereinafter simply referred to as iron loss characteristics 412). The reason for preparing multiple electrical steel sheets is as described above.

[0152] Furthermore, the memory unit 41 stores a function model 413. The function model 413 is one of the models included in the "generating model," and is a predetermined function with multiple coefficients that approximates the material properties of existing materials. Details of the function model 413 will be described later.

[0153] The material property generation device 4 may also be configured as a general-purpose computer, similar to the material property generation device 1b described in Embodiment 1. When the material property generation device 4 is configured as a general-purpose computer, the control unit 40, storage unit 41, communication unit 42, input unit 43, and output unit 44 may have the same configuration as the control unit 10b, storage unit 11b, communication unit 12b, input unit 13b, and output unit 14b of the material property generation device 1b. That is, the storage unit 41 may also be provided on an external server instead of being provided on the material property generation device 4. Embodiment 4 also includes both cases in which the storage unit 41 is physically connected to the material property generation device 4 and cases in which the storage unit 41 is not physically connected to the material property generation device 4.

[0154] The approximation unit 401 acquires the existing characteristics of the electrical steel sheet from the memory unit 41. In the example shown in Figure 15, the approximation unit 401 acquires magnetization characteristics 411 and iron loss characteristics 412 as existing characteristics. The acquisition method is not particularly limited, but for example, the method described in Embodiment 1 may be used. Hereinafter, the existing characteristics to be acquired will be described as magnetization characteristics 411 and iron loss characteristics 412.

[0155] The approximation unit 401 approximates the magnetization characteristics 411 and iron loss characteristics 412 of the acquired existing electrical steel sheet with a predetermined function. The reason for this approximation is explained below. As mentioned above, the material properties that contribute to motor characteristics consist of magnetization characteristics and iron loss characteristics, and these characteristics are multidimensional quantities that include measurement results at multiple points. If such multidimensional quantities are handled directly, there are concerns about increased calculation time and trapping into local optima. Therefore, in Embodiment 4, in order to reduce the number of parameters subject to optimization calculation, the material properties (magnetization characteristics and iron loss characteristics) are approximated with a predetermined function. For example, the number of parameters for a linear function: y = ax + b is two, coefficient a and coefficient b. With such a configuration, the number of parameters can be suppressed by approximating with a function, thus shortening the time required for optimization calculation. In addition to reducing the number of parameters, by limiting the range of parameter search to a range that can be expressed as a weighted linear sum of parameters, solutions that are realistically unlikely to be manufactured or are impossible to manufacture (e.g., zero iron loss, infinite magnetic flux density, etc.) are excluded, and the system is optimized while excluding material properties that are impossible to manufacture.

[0156] The function used for approximation is preferably one that can accurately represent the material properties, but it is not particularly limited. A good condition for accuracy is, for example, the coefficient of determination R between the material properties represented by the approximation function and the original material properties. 2 A function that satisfies the condition of 0.95 or higher, preferably 0.98 or higher, and more preferably 0.99 or higher, is desirable. As approximation functions, linear functions, power series functions, exponential functions, logarithmic functions, power functions, etc., can be used. These functions may be used individually or in combination depending on the interval of the material properties. Specific examples of approximation functions will be described later.

[0157] The setting unit 402 sets the coefficients of the approximation function as parameters to be optimized. In this way, the approximation unit 401 and the setting unit 402 generate a function model 413 based on multiple existing properties, which are material properties of multiple existing materials used in electrical equipment (e.g., motors). The functions included in the function model 413 may be functions specified in advance by the user. If no function is specified by the user, the approximation unit 401 may have a function to search for the optimal approximation function from among the functions included in the function model 413. If no function is specified by the user, such a search function has the advantage of increasing the likelihood of obtaining a function with higher accuracy (coefficient of determination close to 1), but it also has the disadvantage of increasing the computation time required to search for a function. When generating the function model 413 using analog data, discrete sampled values ​​can be created, and the function approximation can be done using these discrete sampled values.

[0158] The acquisition unit 403 acquires input to the function model 413, that is, multiple data that form the material properties of multiple existing materials.

[0159] The material property generation unit 404 uses a function model 413 to generate evaluation characteristics, which are material properties for evaluation corresponding to the input acquired by the acquisition unit 403. As an example, the material property generation unit 404 uses the function model 413 to generate magnetization characteristics 411 and iron loss characteristics 412 that are different from those of an existing electrical steel sheet, based on the parameters (coefficient part of the function) set by the setting unit 402. The data showing magnetization characteristics 411 and iron loss characteristics 412 that are different from those of an existing electrical steel sheet, generated by the material property generation unit 404, and the data showing iron loss characteristics are output to the equipment property calculation unit 405.

[0160] The equipment characteristics calculation unit 405 calculates evaluation indices that can evaluate the characteristics of electrical equipment based on the evaluation characteristics generated by the material characteristics generation unit 404. An example of an "evaluation indices" was explained in Embodiment 1, so the explanation is omitted here. The data showing the evaluation items calculated by the equipment characteristics calculation unit 405 is output to the update unit 406. The equipment characteristics calculation unit 405 may also be configured to acquire evaluation indices calculated based on evaluation characteristics by other devices, etc.

[0161] The update unit 406 updates the input to the function model 413 until a predetermined termination condition is met. That is, the update unit 406 updates (changes) at least some of the parameters of the approximation function included in the input data according to a predetermined rule. When the input data is updated, the acquisition unit 403 described above acquires the input data updated by the update unit 406.

[0162] Then, the material property generation unit 404 generates evaluation characteristics and the equipment property calculation unit 405 calculates evaluation indices for the updated input data, and the processing of the update unit 406 is repeatedly executed until the termination condition is met, thereby optimizing the input data. More specifically, such optimization processing may be performed using well-known optimization methods such as genetic algorithms, grid search, random search, and Bayesian optimization. This makes it possible to search for input data that yields the desired motor characteristics.

[0163] The calculation results from the material property generation unit 404, the equipment property calculation unit 405, and the update unit 406, that is, the calculation results from the optimization calculation unit 410, are output to the output control unit 407.

[0164] The output control unit 407 outputs the calculation results from the optimization calculation unit 410 to the output unit 44. The output method can be arbitrary. For example, the output control unit 407 may output the calculation results in at least one of the following ways: display output, audio output, and printed output. The output destination device is not limited to the material property generation device 4, but may be an external device. By checking the calculation results, the designer can understand the magnetization characteristics and iron loss characteristics of the existing electromagnetic steel sheet, which have been optimized so that the motor has the desired characteristics, and which are different from the magnetization characteristics and iron loss characteristics of the existing electromagnetic steel sheet.

[0165] Furthermore, the output control unit 407 may determine at least one evaluation characteristic from the evaluation characteristics generated by the material property generation unit 404 based on the evaluation index. Here, the "evaluation characteristic" may be the optimal value or the top few data points with high performance. The output control unit 407 may output output information that includes an input corresponding to the determined evaluation characteristic.

[0166] (An example of an approximation function that approximates magnetization properties) Next, with reference to Figure 16, an example of an approximation function for approximating magnetization characteristics will be described. Figure 16 is a graph showing an example of magnetization characteristics, where the vertical axis represents magnetic flux density and the horizontal axis represents magnetization force (magnetic field strength). In the magnetization characteristics shown in Figure 16, for example, magnetic flux density measured within a predetermined range with a predetermined step size may be used. The predetermined range of magnetic flux density may be, for example, 0 or more and 2T or less. The predetermined step size of magnetic flux density may be, for example, 0.1T. In this case, the magnetization characteristics exhibit discrete characteristics that do not take continuous values, but are not limited to this. For example, the magnetization characteristics may be characteristics that are shown as continuous values ​​that approximate the measurement results. Note that the level of magnetic flux density to be measured may be coarser than described above.

[0167] A linear function can be used as an example of the approximation function for magnetization characteristics. As shown in Figure 16, the magnetization characteristics can, for example, divide the range of magnetic flux density into three regions: a low magnetic field region R1, a medium magnetic field region R2, and a high magnetic field region R3, and the characteristics differ in each of the plurality of regions. Therefore, when a linear function is used as the approximation function for the magnetization characteristics, approximation by a polygonal line is possible according to each region (the range of magnetic flux density), which is preferable because the approximation can be performed with higher accuracy than in the case of a single region. The low magnetic field region R1, medium magnetic field region R2, and high magnetic field region R3 are not particularly limited regions, and may be appropriately selected by a designer, and the number of regions does not have to be three. For example, as shown in Figure 16, when the low magnetic field region R1 is 0 (T) or more and less than A (T), the medium magnetic field region R2 is A (T) or more and less than B (T), and the high magnetic field region R3 is B (T) or more, the designer may appropriately select a numerical value that satisfies A<B. When the magnetization characteristics are divided into the low magnetic field region R1, medium magnetic field region R2, and high magnetic field region R3, the number of straight lines used for approximation is three, but the number of straight lines is not limited to three. That is, the magnetization characteristics may be divided into four or more regions to increase the number of straight lines. However, increasing the number of straight lines increases the number of parameters, which improves the approximation accuracy but causes an increase in calculation time. In the case of approximation with three straight lines, since each straight line can be expressed by two parameters, the approximation can be performed with a total of six parameters. This is significantly less than, for example, 20 data points measured at 0.1T intervals of magnetic flux density up to 2T. Although the linear function does not strictly correspond to the physical phenomena in the magnetization of the material, it is known that magnetization is caused by domain wall displacement in the low magnetic field region R1 and the medium magnetic field region R2, and by magnetization rotation in the high magnetic field region R3. If necessary on this point, please refer to "Study on Modeling of Magnetization Curves in Non-Oriented Electrical Steel Sheets", IEEJ Transactions on Fundamentals and Materials, Vol.131, No.6, pp.466-471 (2011). For this reason, the linear function may be corrected by associating the range for dividing the magnetic flux density region and the linear function to be approximated with the physical phenomena in the magnetization of the material.

[0168] As another example, a power series function may be used as an example of the approximation function for magnetization characteristics (for example, H=a1×B+a2×B 2 +a3×B3 +···). Since the magnetization characteristics are point-symmetric with respect to the positive and negative magnetic fields, an odd function form is desirable (for example, H=a1×B+a2×B 3 +a3×B 5 +···). The upper limit of the order is not particularly limited, but increasing the order increases the number of parameters, which becomes a factor of increasing calculation time. In addition, if the order is increased excessively, the number of harmonic components increases, and the approximation accuracy may conversely decrease. Considering these factors, it is preferable that the maximum order is about 9 to 13. In this way, the coefficient of determination R between the material characteristics expressed by the approximation function and the original material characteristics can be set 2 to 0.99 or higher. For example, when approximated by an odd function with a maximum order of 13, the number of parameters is 7, which is significantly smaller than, for example, 20 pieces of data measured up to 2T in increments of magnetic flux density of 0.1T.

[0169] Further, as the magnetization characteristics, an approximation function may be obtained based on H-μ r characteristics instead of the BH characteristics. Here, μ r is the relative permeability. Also for the H-μ r characteristics, an approximation function with a large coefficient of determination may be appropriately selected.

[0170] (Example of approximation function for approximating iron loss characteristics) Next, an example of an approximation function for approximating iron loss characteristics will be described with reference to FIG. 17. FIG. 17 is a graph showing an example of iron loss characteristics, where the vertical axis represents the iron loss value and the horizontal axis represents the magnetic flux density. In the iron loss characteristics shown in FIG. 17, magnetic flux densities measured in a predetermined range at a predetermined step size at each of 50 Hz, 100 Hz, and 400 Hz may be used. The predetermined range and the predetermined step size are not particularly limited, and may be, for example, the same as the values described with reference to FIG. 16.

[0171] As an example of the approximation function for the iron loss characteristics, a power series function can be used. When using a power series function, as an example, the iron loss characteristics may be assumed to be proportional to the square of the magnetic flux density, and a proportionality coefficient may be derived for each frequency.

[0172] Since the iron loss characteristic is symmetric with respect to the positive and negative values ​​of the magnetic flux density, an even function form is desirable (for example, W = a¹ × B). 2 +a² × B 4 +a3×B 6 +···). There is no particular upper limit to the order, but increasing the order increases the number of parameters, which increases the computation time. Also, if the order is increased too much, the harmonic components increase, which may actually decrease the approximation accuracy. Considering these factors, it is preferable to set the maximum order to around 2 to 6. In this way, the coefficient of determination R between the material properties expressed by the approximation function and the original material properties can be improved. 2 This can be increased to 0.99 or higher. For example, if we approximate with an even function of the highest order being 6th, the number of parameters is 3, which is significantly less than, for example, 20 data points measured in 0.1T increments up to 2T.

[0173] (Process flow) Next, with reference to Figure 18, the processing flow performed by the material property generation device 4 will be explained. Figure 18 is a flowchart showing an example of the processing performed by the material property generation device 4.

[0174] In step S401, the acquisition unit 403 acquires an input to a function model 413 generated based on multiple existing properties, which are material properties of multiple existing materials used in electrical equipment (e.g., motors), and the input includes multiple data that form the material properties.

[0175] In step S402, the material property generation unit 404 uses the function model 413 to generate evaluation properties, which are evaluation material properties corresponding to the input obtained in the processing of step S401.

[0176] In step S403, the equipment characteristics calculation unit 405 calculates an evaluation index that can evaluate the characteristics of the electrical equipment based on the evaluation characteristics generated in the processing of step S402. As described above, the equipment characteristics calculation unit 405 may be configured to acquire an evaluation index calculated based on the evaluation characteristics by another device or the like.

[0177] The process proceeds to step S404, where it is determined whether the motor evaluation index calculated in step S403 satisfies the termination condition. If it is determined that the motor evaluation index satisfies the termination condition (YES in step S404), the process proceeds to step S406. On the other hand, if it is determined that the motor evaluation index does not satisfy the termination condition (NO in step S404), the process proceeds to step S405. The termination condition can be set arbitrarily. For example, the termination condition may be set as the point at which the Pareto solution that balances torque and efficiency within the desired range is no longer updated, the completion of a predetermined number of updates, or the elapsed of a predetermined calculation time.

[0178] In step S405, the update unit 406 updates the parameters (coefficients of the approximation function), and then the process returns to step S402. As a result, the acquisition unit 403 acquires the updated parameters as input data to the function model 413, and steps S402 to S405 are executed again. The method for updating the parameters and the termination conditions can be predetermined. For example, as described above, a genetic algorithm may be used. The processes in steps S402 to S405 are repeatedly executed until the termination condition is met, that is, until YES is determined in step S404.

[0179] In step S406, the output control unit 407 outputs the optimized calculation result obtained by repeating the processing in steps S402 to S405 to the output unit 44.

[0180] As described above, the material property generation method according to Embodiment 4 includes: an acquisition step (S401) in which an input is obtained for a function model 413 generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment, and which includes a plurality of data that form the material properties; a generation step (S402) in which evaluation properties which are material properties for evaluation corresponding to the input obtained in the acquisition step are generated using the function model 413; an evaluation step (S403) in which an evaluation index capable of evaluating the characteristics of electrical equipment is obtained calculated based on the evaluation properties generated in the generation step; and a determination step (S406) in which at least one evaluation property is determined from the evaluation properties generated in the generation step based on the evaluation index.

[0181] Note that the processing flow shown in the flowchart in Figure 18 is just one example, and steps may be deleted, new steps added, or the processing order rearranged as long as it does not deviate from the main point.

[0182] (Regarding the analysis results) Next, an example of the analysis results obtained using the method described in Embodiment 4 will be explained with reference to Table 4 and Figure 19. Table 4 shows the analysis conditions for this example. As shown in Table 4, this example uses an IPMSM with 8 poles (IPMSM10 shown in Figure 1). Two evaluation indicators were selected: torque and efficiency, which are electrical characteristics of the motor.

[0183] [Table 4]

[0184] Figure 19 is a graph showing the relationship between torque and efficiency, comparing the analysis results using the method described in Embodiment 4 (Embodiment 4) with the analysis results for existing electrical steel sheets (Comparative Example 4). As is clear from Figure 19, Embodiment 4 provides highly feasible material properties based on the material properties of existing electrical steel sheets, and therefore, compared to the case of only existing electrical steel sheets shown in Comparative Example 4, Embodiment 4 provides a wider range of Pareto solutions. In other words, Embodiment 4 makes it possible to find material properties different from those of existing electrical steel sheets that can achieve both motor torque and efficiency. Furthermore, by generating material properties based on the material properties of existing electrical steel sheets to match motor design, it is possible to obtain a meaningful optimal solution that serves as a guideline for considering materials to be applied to motors and is also relatively easy to manufacture. Here, the motor designer can appropriately select and adopt the solution they consider most preferable from the obtained Pareto solutions. Alternatively, they can calculate the weighted sum of the two evaluation items, torque and efficiency, and adopt it as a new evaluation index, then select the solution for which this index has the optimal value. The weighting ratio of the multiple items can be appropriately selected by the designer.

[0185] However, this disclosure is not limited to the embodiments described above. The material property generation apparatus 4 (material property generation method shown in Figure 18) does not necessarily have to include an update unit 406 (update step S405).

[0186] (Effects and Benefits) As described above, the "generation model" in Embodiment 4 is a function model 413 having multiple coefficients that approximates the material properties formed by multiple data points.

[0187] According to the above configuration, by approximating material properties with a predetermined function and optimizing the coefficients of that function, the number of data points can be reduced, and the computation time required for optimization can be shortened. Alternatively, even with the same computation time, optimization can be performed using a larger number of existing materials as source data. As a result of such optimization, a wider range of Pareto solutions can be obtained compared to existing materials (e.g., electrical steel sheets), making it possible to manufacture electrical equipment that meets a wide variety of needs.

[0188] Furthermore, material properties may also be properties that show the relationships between multiple variables. In addition, the function model 413 may include multiple functions that approximate the relationships in each of the multiple regions when the range of one variable in the material properties is divided into multiple regions.

[0189] As shown in the above configuration, approximation in each domain is expected to improve approximation accuracy.

[0190] Furthermore, the material properties may include magnetization properties that show the relationship between magnetic flux density and magnetic field strength. The multiple functions may be three different functions that approximate the magnetization properties in each of the three regions when the magnetization properties are divided into a low-field region R1, a medium-field region R2, and a high-field region R3 according to the magnitude of the magnetic flux density.

[0191] With the above configuration, as explained in Figure 16, it is possible to accurately approximate the low-field region R1, the medium-field region R2, and the high-field region R3, each with different characteristics.

[0192] Furthermore, the coefficient of determination R between the material properties expressed as a function and the original material properties is also important. 2 It is preferable that the value is 0.95 or higher. More preferably, it is 0.98 or higher, and even more preferably, 0.99 or higher.

[0193] 〔summary〕 According to the embodiments 1 to 4 described above, the following effects can be obtained.

[0194] Generating new material properties different from existing properties using a generative model (e.g., trained model 111b, property set generation model 213, reconstruction model 312b, function model 413) generated based on information about multiple existing properties leads to the generation of new material properties that are highly feasible and superior in their properties. Electrical equipment manufactured from materials with such new material properties is expected to have improved performance compared to existing electrical equipment. Thus, with the above configuration, it is possible to efficiently develop materials that realize new material properties that contribute to improving the performance of electrical equipment.

[0195] Furthermore, according to Embodiments 1 to 4, it is possible to obtain optimal material properties from a set that includes not only the material properties of existing materials, but also hypothetical material properties within a range that is difficult for even the engineer to easily estimate, but which the engineer can estimate to be manufacturable based on the results. Although verification and experiments are necessary to determine whether the estimated hypothetical material can actually be manufactured, it is possible to significantly increase the possibility of achieving even better properties than those obtained by applying existing materials to electrical equipment.

[0196] As mentioned above, the electrical equipment may be a motor. Furthermore, the existing materials used in the electrical equipment may be core materials used in motor cores (e.g., electrical steel sheets).

[0197] According to the above configuration, the methods described in Embodiments 1 to 4 can be suitably applied to motors.

[0198] Furthermore, the material properties of existing materials may include magnetization properties that show the relationship between magnetic flux density and magnetic field strength, and iron loss properties that show the relationship between magnetic flux density and iron loss value.

[0199] According to the above configuration, both magnetization characteristics and iron loss characteristics can be optimized simultaneously. As a result, the methods described in Embodiments 1 to 4 can be suitably applied to motors whose characteristics change according to the magnetization characteristics and iron loss characteristics.

[0200] Furthermore, the hypothetical material properties within the range that the engineer in question is presumed to be able to manufacture may differ from the magnetization properties and iron loss properties of existing materials. The magnetization properties of existing materials and magnetization properties different from those of existing materials each have different magnetization curves. Also, the iron loss properties of existing materials and iron loss properties different from those of existing materials each have different iron loss curves.

[0201] [Examples of implementation using software] The functions of the model generation device 1a and material property generation device 1b according to Embodiment 1, the material property generation device 2 according to Embodiment 2, the model generation device 3a and material property generation device 3b according to Embodiment 3, and the material property generation device 4 according to Embodiment 4 (hereinafter collectively referred to as "information processing device, etc.") are programs (material property generation programs) that cause a computer to function as an information processing device, etc., and can be realized by programs that cause a computer to function as each control block of the information processing device, etc.

[0202] In this case, the information processing device, etc., includes a computer having at least one device (e.g., a processor) and at least one storage device (e.g., memory) as hardware for executing a program. The computer executes the program, thereby realizing each of the functions described in each embodiment.

[0203] The program may be recorded on one or more computer-readable recording media, rather than temporarily. These recording media may or may not be present in the information processing device. In the latter case, the program may be supplied to the information processing device via any wired or wireless transmission medium.

[0204] Furthermore, some or all of the functions of each control block can be implemented by logic circuits. For example, an integrated circuit in which logic circuits functioning as each control block are formed is also included in the scope of this disclosure. In addition, it is also possible to implement the functions of each control block using, for example, a quantum computer.

[0205] Furthermore, each process described in each embodiment may be executed by AI (Artificial Intelligence). In this case, the AI ​​may operate on an information processing device or other device (for example, an edge computer or a cloud server).

[0206] This disclosure is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of this disclosure.

[0207] For example, in embodiments 1 to 4 described above, the case in which torque and efficiency are evaluated as electrical characteristics of a motor (electrical equipment) by electromagnetic field analysis simulation was explained. However, the characteristics of electrical equipment may also be evaluated by comprehensively combining the results of multiple analyses, such as a combination of electromagnetic field analysis and at least one of structural analysis or heat transfer analysis. In this case, the characteristics of electrical equipment are evaluated using evaluation indices that include the necessary evaluation characteristics obtained from each of the multiple analyses. Specifically, by combining electromagnetic field analysis and structural analysis, evaluation indices may be selected that aim to improve electrical characteristics such as torque and efficiency, as well as mechanical characteristics such as vibration reduction and remaining within the range of elastic deformation (not undergoing plastic deformation). Alternatively, by combining electromagnetic field analysis and heat transfer analysis, temperature conditions may be added as evaluation indices.

[0208] For example, the component parameters and manufacturing parameters for realizing the new magnetization and iron loss characteristics described in Embodiments 2 to 4 above can be determined by performing an inverse analysis of the method described in Embodiment 1. As a specific example, an inverse model can be constructed by swapping the inputs (each parameter) and outputs (material properties) of the model described in Embodiment 1, and the component parameters and manufacturing parameters for realizing the new magnetization and iron loss characteristics can be determined. The material properties (magnetization and iron loss characteristics) of the electromagnetic steel sheet manufactured using the component parameters and manufacturing parameters determined in this way may exhibit new characteristics that differ from the material properties of existing electromagnetic steel sheets. A motor manufactured using an electromagnetic steel sheet with such new material properties is expected to have improved characteristics compared to existing motors. Thus, by combining Embodiment 1 with any of Embodiments 2 to 4, it is possible to efficiently develop materials that realize new material properties that contribute to improving the characteristics of motors. Furthermore, since such new material properties are generated based on the material properties of existing materials, solutions that are realistically unfeasible or impossible to manufacture (e.g., zero iron loss, infinite magnetic flux density, etc.) can be excluded, and it is possible to optimize by excluding material properties that cannot be manufactured.

[0209] Furthermore, the material property generation device 1b of Embodiment 1, the material property generation device 2 of Embodiment 2, the material property generation device 3b of Embodiment 3, and the material property generation device 4 of Embodiment 4 can each be used independently, that is, in a standalone manner without communicating with other devices.

[0210] [Additional Notes] Furthermore, this disclosure can also be expressed as follows:

[0211] A material property generation apparatus according to Embodiment 1 of the present disclosure includes: an acquisition unit that acquires an input to a function model generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment, the input including a plurality of data that form the material properties; a generation unit that uses the function model to generate an evaluation property which is an evaluation material property corresponding to the input acquired by the acquisition unit; an evaluation unit that acquires an evaluation index capable of evaluating the characteristics of the electrical equipment, calculated based on the evaluation property generated by the generation unit; and an output control unit that determines at least one of the evaluation properties from the evaluation property generated by the generation unit based on the evaluation index.

[0212] The material property generation apparatus according to Embodiment 2 of the present disclosure, in Embodiment 1 described above, further comprises an update unit that updates at least a portion of the plurality of coefficients until a predetermined termination condition is met, wherein the function model is a predetermined function having a plurality of coefficients that approximates the material property formed by the plurality of data.

[0213] In the material property generation apparatus according to embodiment 3 of the present disclosure, in embodiment 1 or 2 described above, the material property is a property that shows the relationship between a plurality of variables, and the function model may include a plurality of functions that approximate the relationship in each of the regions when the range of one of the variables in the material property is divided into a plurality of regions.

[0214] The material property generation apparatus according to Embodiment 4 of the present disclosure, in Embodiment 3 above, includes a magnetization characteristic that shows the relationship between magnetic flux density and magnetic field strength, and the plurality of functions may be three different functions that approximate the magnetization characteristic in each of the low magnetic field region, medium magnetic field region and high magnetic field region when the magnetization characteristic is divided into a low magnetic field region, a medium magnetic field region and a high magnetic field region according to the magnitude of the magnetic flux density.

[0215] In the material property generation apparatus according to aspect 5 of the present disclosure, in aspect 2 described above, the coefficient of determination R2 between the material property expressed by the function and the original material property may satisfy the condition that is 0.95 or higher, preferably 0.98 or higher, and more preferably 0.99 or higher.

[0216] In the material property generation apparatus according to embodiment 6 of this disclosure, in any of embodiments 1 to 5 described above, the properties of the electrical equipment may be the electrical properties of the electrical equipment.

[0217] The material property generation apparatus according to Embodiment 7 of the present disclosure, in Embodiment 6 above, wherein the electrical equipment is a motor, the existing material used in the electrical equipment is a core material used in the core of the motor, and the electrical properties may include at least one of the torque or efficiency of the motor.

[0218] The material property generation program according to aspect 8 of this disclosure is a material property generation program for causing a computer to function as the material property generation device in aspect 1 described above, and the computer may function as the acquisition unit, the generation unit, the evaluation unit, and the output control unit.

[0219] A material property generation method according to aspect 9 of the present disclosure is a material property generation method performed by a material property generation apparatus, comprising: an acquisition step of acquiring an input to a function model generated based on a plurality of existing properties which are material properties of a plurality of existing materials used in electrical equipment, the input including a plurality of data that form the material properties; a generation step of generating an evaluation property which is an evaluation material property corresponding to the input acquired in the acquisition step using the function model; an evaluation step of acquiring an evaluation index which is calculated based on the evaluation property generated in the generation step and which is capable of evaluating the characteristics of the electrical equipment; and a determination step of determining at least one evaluation property from the evaluation property generated in the generation step based on the evaluation index. [Explanation of symbols]

[0220] 1a, 3a Model generation device 1b, 2, 3b, 4 Material property generator 10a, 10b, 20, 30a, 30b, 40 Control Unit 11a, 11b, 21, 31a, 31b, 41 Storage section 12a, 12b, 22, 32a, 32b, 42 Communication Department 13a, 13b, 23, 33a, 33b, 43 Input section 14a, 14b, 24, 34a, 34b, 44 Output section 101a, 301a Training data acquisition unit 101b, 203, 301b, 403 Acquisition Department 102a, 302a Learning Department 102b, 302b, 404 Material property generation section 103b, 303b, 405 Device characteristics calculation section 104b, 206, 304b, 406 update section 105b, 207, 305b, 407 Output control unit 110b, 210, 310b, 410 Optimization Calculation Unit 111a, 311a Training data 111b Pre-trained model 201 Selection Section 202, 402 Settings section 204 Evaluation Characteristics Set Generation Unit 205 Calculation Section 213 Characteristic Set Generation Model 311b compression model 312b Reconstruction Model 401 Approximation part 413 Function Models 10 IPMSM 11 rotors 111 Rotor Core 12 staters 121 Stator Core 121a Coreback 121b Teeth 131 Permanent Magnets

Claims

1. An acquisition unit that acquires an input to a function model generated based on multiple existing material properties of multiple existing materials used in electrical equipment, the input including multiple data that form the material properties, A generation unit generates evaluation characteristics, which are material properties for evaluation corresponding to the input acquired by the acquisition unit, using the function model described above. An evaluation unit that acquires an evaluation index capable of evaluating the characteristics of the electrical equipment, calculated based on the evaluation characteristics generated by the generation unit, A material property generation apparatus comprising: an output control unit that determines at least one of the evaluation characteristics from the evaluation characteristics generated by the generation unit based on the evaluation index; and

2. The function model is a predetermined function having a plurality of coefficients that approximates the material properties formed by the plurality of data, The material property generation apparatus according to claim 1, further comprising an update unit that updates at least a portion of the plurality of coefficients until a predetermined termination condition is met.

3. The aforementioned material properties are characteristics that show the relationship between multiple variables, The material property generation apparatus according to claim 1 or 2, wherein the function model includes a plurality of functions that approximate the relationship in each of the regions when the range of one of the variables in the material property is divided into a plurality of regions.

4. The aforementioned material properties include magnetization properties that show the relationship between magnetic flux density and magnetic field strength. The material property generation apparatus according to claim 3, wherein the plurality of functions are three different functions that approximate the magnetization characteristics in each of the low magnetic field region, medium magnetic field region, and high magnetic field region when the magnetization characteristics are divided into a low magnetic field region, a medium magnetic field region, and a high magnetic field region according to the magnitude of the magnetic flux density.

5. The material property generation apparatus according to claim 2, wherein the coefficient of determination R2 between the material property expressed by the function and the original material property is 0.95 or greater, preferably 0.98 or greater, and more preferably 0.99 or greater.

6. The material property generation apparatus according to claim 1, wherein the characteristics of the electrical equipment are the electrical characteristics of the electrical equipment.

7. The aforementioned electrical device is a motor, The existing material used in the aforementioned electrical equipment is a core material used in the core of the motor, The material property generating apparatus according to claim 6, wherein the electrical properties include at least one of the torque or efficiency of the motor.

8. A material property generation program for causing a computer to function as a material property generation apparatus according to claim 1, wherein the computer functions as the acquisition unit, the generation unit, the evaluation unit, and the output control unit.

9. A method for generating material properties performed by a material property generation apparatus, An acquisition step to obtain an input to a function model generated based on multiple existing properties which are material properties of multiple existing materials used in electrical equipment, the input including multiple data that form the material properties, A generation step that generates evaluation characteristics, which are material properties for evaluation corresponding to the input obtained in the acquisition step, using the function model described above, An evaluation step to obtain an evaluation index that can evaluate the characteristics of the electrical equipment, calculated based on the evaluation characteristics generated in the generation step, A method for generating material properties, comprising: a determination step of determining at least one of the evaluation properties generated in the generation step based on the evaluation index.

Images