PERFORMANCE PREDICTION APPARATUS, PERFORMANCE PREDICTION METHOD, AND PERFORMANCE PREDICTION PROGRAM

Abstract

To provide a performance prediction device, a performance prediction method, and a performance prediction program, capable of efficiently predicting a performance value of a product or part based on a three-dimensional shape of the product or part.SOLUTION: A performance prediction device includes: a storage unit that stores information on at least one of an analysis condition and an analysis result or a test condition and a test result regarding a predetermined object; an acquisition unit that acquires shape data of an object that is a prediction target for a performance value; a prediction unit that derives a prediction value for the performance value based on the information stored in the storage unit and the acquired shape data; and an output unit that outputs the derived prediction value.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a performance prediction device, a performance prediction method, and a performance prediction program. [Background technology]

[0002] With the development of machine learning technology (artificial intelligence technology), machine learning can now be used for various predictions. Non-Patent Document 1 discloses a design system for optimizing the aerodynamics and structure of an aircraft. In Non-Patent Document 1, an aerodynamic model and a structural model for optimizing the aerodynamics and structure are generated. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Hiroyuki Morino and 5 others, "Application of Multidisciplinary Optimization (MDO) to Environmentally Adaptive High Performance Small Aircraft Design", Mitsubishi Heavy Industries Techniques, Vol. 42, No. 5, pp. 216-219, 2005. Summary of the Invention [Problem to be solved by the invention]

[0004] When a machine learning method is used to train a model on test data (test results) obtained by conducting tests on an object (structure), there are limitations on the tests that can be performed (e.g., stiffness tests, stress tests, vibration tests) and the number of test data samples. In contrast, in simulations of objects using the Finite Element Method (FEM) and Computational Fluid Dynamics (CFD), analysis results for a larger number of samples than the test data can be obtained by analysis on a computer. However, the accuracy of the analysis results may be lower than the accuracy of the test results.

[0005] In addition, deriving performance values ​​that are determined according to the three-dimensional shape of a product (object) can take anywhere from several days to several weeks, depending on the performance of the computer. For this reason, improving the efficiency of pre-evaluating product performance values ​​is a major challenge in product development. The efficiency of simulations is directly linked to the efficiency of product development. However, it has been difficult to efficiently predict the performance values ​​of a product or part based on the three-dimensional shape of that product or part.

[0006] In view of the above circumstances, an object of the present invention is to provide a performance prediction device, a performance prediction method, and a performance prediction program that are capable of efficiently predicting the performance values ​​of a product or part based on the three-dimensional shape of the product or part. [Means for solving the problem]

[0007] One aspect of the present invention is a performance prediction device that includes a memory unit that stores at least one of information on analysis conditions and analysis results or test conditions and test results related to a predetermined object, an acquisition unit that acquires shape data of an object for which a performance value is to be predicted, a prediction unit that derives a predicted value of the performance value based on the information stored in the memory unit and the shape data acquired by the acquisition unit, and an output unit that outputs the derived predicted value.

[0008] The above-mentioned performance prediction device is capable of efficiently predicting the performance value of a product or part based on the three-dimensional shape of the product or part.

[0009] One aspect of the present invention is a performance prediction method including: an acquisition step of acquiring shape data of an object for which a performance value is to be predicted; a prediction step of deriving a predicted value of the performance value based on at least one of information on analysis conditions and analysis results or test conditions and test results related to a predetermined object and the acquired shape data; and an output step of outputting the derived predicted value.

[0010] One aspect of the present invention is a performance prediction program for causing a computer to execute an acquisition step of acquiring shape data of an object for which a performance value is to be predicted, a prediction step of deriving a predicted value of the performance value based on at least one of information on analysis conditions and analysis results or test conditions and test results related to a predetermined object and the acquired shape data, and an output step of outputting the derived predicted value. Effect of the Invention

[0011] According to the present invention, it is possible to efficiently predict the performance values ​​of a product or part based on the three-dimensional shape of the product or part. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram illustrating an example of the configuration of a performance prediction device in a first embodiment. [Diagram 2] FIG. 11 is a diagram showing an example of the reliability of a predicted value of a performance value in a case where second teacher data is not interpolated in the first embodiment. [Diagram 3] FIG. 11 is a diagram showing an example of the reliability of a predicted value of a performance value when second teacher data is interpolated in the first embodiment. [Figure 4] 4 is a flowchart showing an example of the operation of the learning device in the first embodiment. [Diagram 5] 4 is a flowchart showing an example of the operation of the prediction device in the first embodiment. [Figure 6] FIG. 11 is a diagram illustrating an example of the configuration of a learning device in a second embodiment. [Figure 7] 13 is a flowchart showing an example of the operation of the learning device in the second embodiment. [Figure 8] FIG. 13 is a diagram illustrating an example of the configuration of a prediction device in a third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail with reference to the drawings. (First embodiment) 1 is a diagram showing an example of the configuration of a performance prediction device 1. The performance prediction device 1 is a device that derives a predicted value of a performance value of an object based on three-dimensional shape data of the object to be predicted. The performance prediction device 1 includes a learning device 2 and a prediction device 3.

[0014] The learning device 2 includes an acquisition unit 20, a storage unit 21, a control unit 22, a model generation unit 23, and an output unit 24. The model generation unit 23 includes a solid analysis unit 230, a fluid analysis unit 231, a test data derivation unit 232, and a reliability assignment unit 233. The prediction device 3 includes an acquisition unit 30, a prediction unit 31, a reliability assignment unit 32, and an output unit 33.

[0015] First, an overview of the learning device 2 will be described. In the learning stage, training data is input to the learning device 2. The training data includes three-dimensional shape data. This three-dimensional shape data is, for example, at least one of shape data (design dimension data) of a virtual object in CAD (Computer Aided Design) and measurement point cloud (point cloud) data representing the shape of a real object. The measurement point cloud is obtained, for example, by using a laser measuring instrument or the like.

[0016] The object is, for example, a product or a part. The product is not limited to a specific product, but is, for example, a reducer, a hydraulic valve, a compressor, or a door. The use of the product is not limited to a specific use, but is, for example, for a wind turbine, a ship, a construction machine, or a robot. The part is not limited to a specific part, but is, for example, a bolt, a spring, a housing, a bracket, a gear, or a frame. For example, the object is a device or part mounted on at least one of a wind turbine, a ship, a construction machine, and a robot.

[0017] The learning device 2 derives a predicted value of the performance value of the object based on the three-dimensional shape data included in the teacher data. The predicted value of the performance value is, for example, a predicted value of stiffness data, stress data, pressure loss data, flow data, and test data. The pressure loss data is data representing the pressure loss of the fluid around the object. The flow data is data representing the flow path and flow rate (flow velocity) of the fluid around the object. The test data is data representing the test result. The test result is not limited to a measurement value in a specific test, but is, for example, a displacement measurement value of the object against a force applied to the object, a stress measurement value obtained using a strain gauge under a predetermined deformation, a measurement value of the acceleration or displacement of the object against a time-series vibration input, a measurement value of the pressure difference between positions determined in a pipeline, or a measurement value of the force applied from the hydraulic oil to the object when the hydraulic oil flows.

[0018] The learning device 2 updates the parameters of the model under learning, for example, by using backpropagation, so as to reduce the difference between the correct label included in the training data and the predicted value of the performance value. When the model under learning becomes trained, the learning device 2 obtains a trained model.

[0019] The learning device 2 may derive the reliability of each predicted value of the performance value based on the accuracy of the analysis results or test results included in the teacher data. The reliability of the analysis results is determined based on the reliability of the test results under the same conditions. The accuracy of the analysis results or test results included in the teacher data may be determined in advance for each application of the product (e.g., wind turbine, ship, construction machine, or robot). For example, the accuracy of the analysis results or test results of a wind turbine, which is subject to many disturbances such as wind, may be lower than the accuracy of the analysis results or test results of construction machine.

[0020] Next, an overview of the prediction device 3 will be described. In the execution stage (prediction stage) after the learning stage, the prediction device 3 acquires the trained model from the learning device 2. The prediction device 3 acquires three-dimensional shape data of the object to be predicted. The prediction device 3 inputs the three-dimensional shape data of the object to be predicted into the trained model. As a result, the prediction device 3 acquires a predicted value of the performance value of the object as an output of the trained model.

[0021] The prediction device 3 may obtain accuracy data of the three-dimensional shape data of the object to be predicted. The prediction device 3 inputs the accuracy data of the three-dimensional shape data to the trained model. As a result, the prediction device 3 may derive the reliability of the predicted value of the performance value as an output of the trained model.

[0022] Next, the learning device 2 will be described in detail. In the learning stage, the learning device 2 generates a trained model using the teacher data. The trained model may be expressed using a mathematical formula including coefficients derived by statistical processing or the like, or may be expressed using a neural network. The trained model is generated for each product and each part. For example, a trained model is generated, such as a trained model for product "A".

[0023] Hereinafter, training data in which the design shape data and analysis conditions of an object are used as input data and the analysis results based on the design shape data and analysis conditions of the object are used as correct labels will be referred to as "first training data." The first training data includes, as correct labels, at least one analysis result from, for example, the stiffness data of the object, the stress data of the object, the pressure loss data of the fluid around the object, and the flow data of the fluid.

[0024] Hereinafter, the teacher data, which has the measured shape data of the object and the test conditions as input data and the test results based on the measured shape data of the object and the test conditions as correct answer labels, is referred to as "second teacher data." The second teacher data includes, as correct answer labels, at least one test result from, for example, a displacement measurement value of the object, a stress measurement value of the object, a measurement value of the acceleration or displacement of the object in response to a time-series vibration input, a measurement value of a pressure difference between positions determined on the object, and a measurement value of a force applied to the object by hydraulic oil.

[0025] The acquisition unit 20 receives a first teacher data set (design shape data, analysis conditions, analysis results (correct answer labels), and accuracy data). In the first teacher data, for example, accuracy data is added to the analysis results in advance. Note that the accuracy of the analysis results using a virtual object may be lower than the accuracy of the test results using a real object.

[0026] The design shape data is three-dimensional shape data of a virtual object in CAD. The analysis conditions are conditions used in the analysis process (simulation), such as conditions related to continuous data such as stiffness, temperature, and humidity. The analysis results are results obtained by the analysis, and are the correct answer labels included in the first teacher data. The acquisition unit 20 outputs the acquired first teacher data to the solid analysis unit 230, the fluid analysis unit 231, and the memory unit 21.

[0027] The second teacher data set (measurement shape data, test conditions, test results (correct answer labels), and accuracy data) is input to the acquisition unit 20. In the second teacher data, for example, accuracy data is added to the test results in advance. Note that the accuracy of the test results using real objects may be higher than the accuracy of the analysis results using virtual objects.

[0028] The measured shape data is measured point group (point cloud) data that represents a three-dimensional shape of a real object. The test conditions are conditions used in the test (measurement), such as conditions related to continuous data such as stiffness, temperature, and humidity. The test results are results obtained by the test (measurement), and are the correct answer labels included in the second teacher data. The acquisition unit 20 outputs the acquired second teacher data to the test data derivation unit 232 and the memory unit 21.

[0029] The acquisition unit 20 may output accuracy data of the first teacher data (analysis results), accuracy data of the second teacher data (test results), the number of samples of the first teacher data, and the number of samples of the second teacher data to the reliability assignment unit 233.

[0030] The storage unit 21 outputs the model under training to the model generation unit 23 in response to an access from the model generation unit 23. The storage unit 21 stores a trained model generated from the model under training by the model generation unit 23. The storage unit 21 outputs the trained model to the prediction unit 31 and the reliability assignment unit 32 in response to an access from the prediction unit 31. The storage unit 21 stores the acquired first teacher data and second teacher data. That is, the storage unit 21 stores at least one of information on analysis conditions and analysis results or test conditions and test results regarding a predetermined object. The predetermined object may be, for example, an object on which no analysis or testing has been performed, or an object whose shape is similar to an object on which analysis or testing has been performed.

[0031] The control unit 22 controls the learning operation in the model generation unit 23. For example, the control unit 22 outputs an instruction to the model generation unit 23 to update parameters of the model under learning using, for example, backpropagation so as to reduce the difference between the predicted value output from the model under learning and the correct label of each teaching data. For example, the control unit 22 may output an instruction to the model generation unit 23 to update coefficients of a formula representing the model under learning so as to reduce the difference between the predicted value output from the model under learning and the correct label.

[0032] The control unit 22 may output an instruction to the model generation unit 23 to execute a transfer learning technique on the model under learning. For example, the model generation unit 23 may, in response to the instruction, use a transfer learning technique on the parameters of the neural network of the solid analysis unit 230 or the fluid analysis unit 231 to interpolate the parameters of the neural network of the test data derivation unit 232. For example, the model generation unit 23 may, in response to the instruction, mutually update the parameters of the stiffness model, the parameters of the stress model, and the parameters of the vibration model using correlation information between stiffness, stress, and vibration. For example, the model generation unit 23 may, in response to the instruction, correct the coefficients of the equations of the solid analysis unit 230 or the fluid analysis unit 231 using the coefficients of the equations of the test data derivation unit 232.

[0033] The model generation unit 23 generates a trained model using the first teacher data and the second teacher data. The output of this trained model is a predicted value of the performance value of the object. The input of this trained model is at least one of the design shape data and analysis conditions included in the first teacher data, and the measurement shape data and test conditions included in the second teacher data. A part of the first teacher data and the second teacher data may be missing.

[0034] The solid analysis unit 230 inputs the design shape data and analysis conditions included in the first teacher data to the learning model, thereby obtaining at least one of the stiffness data of the object and the stress data of the object as a predicted value. The solid analysis unit 230 updates the parameters of the learning model so as to reduce the difference between the predicted value output from the learning model and the correct label of the first teacher data. The solid analysis unit 230 may analyze the stiffness data of the object and the stress data of the object by using the finite element method on the design shape data and analysis conditions of the first teacher data.

[0035] The fluid analysis unit 231 inputs the design shape data and analysis conditions included in the first teacher data to the learning model to obtain at least one of the pressure loss data of the fluid around the object and the flow data of the fluid as a predicted value. The fluid analysis unit 231 updates the parameters of the learning model so as to reduce the difference between the predicted value output from the learning model and the correct label of the first teacher data. The fluid analysis unit 231 may analyze the pressure loss data of the fluid around the object and the flow data of the fluid by using a computational fluid dynamics method on the design shape data and analysis conditions included in the first teacher data.

[0036] The test data derivation unit 232 derives the test data of the object as a predicted value by inputting the measured shape data and the test conditions included in the second teacher data to the model under training. In addition, when an additional test is performed using, for example, an experimental design method, the test data derivation unit 232 may update the test data of the object by inputting the measured shape data and the test conditions included in the second teacher data of the additional test to the model under training.

[0037] The confidence assigning unit 233 assigns confidence to the predicted value of the performance value based on at least one of the accuracy data of the first teacher data and the accuracy data of the second teacher data. That is, the confidence assigning unit 233 performs weighting on the predicted value of the performance value. The confidence assigning unit 233 derives confidence based on the ratio between the number of samples of the first teacher data and the number of samples of the second teacher data. For example, when the number of samples of the first teacher data (analysis result) with an accuracy of "100%" is one and the number of samples of the second teacher data (test result) with an accuracy of "90%" is two, the confidence assigning unit 233 derives the confidence (accuracy) of the predicted value of the performance value as "93% (= (100 x 1 + 90 x 2) / (1 + 2)". Note that such a derivation method is one example.

[0038] The reliability assigning unit 233 may update the reliability of the predicted value of the performance value when at least one of the first teacher data and the second teacher data is added or changed to the storage unit 21. The reliability assigning unit 233 may also suggest to the user that the second teacher data for improving the reliability of the predicted value of the performance value be additionally input to the trained model. The content of the suggestion is displayed on the output unit 24, for example. By additionally inputting the second teacher data for improving the reliability to the trained model, the second teacher data input to the acquisition unit 20 is interpolated.

[0039] 2 is a diagram showing an example of the reliability of a predicted value of a performance value when the second teacher data is not interpolated (additionally input). The horizontal axis represents temperature as an example of an analysis condition or test condition. The vertical axis on the left represents the test result. The vertical axis on the right represents the analysis result (simulation result).

[0040] In Figure 2, since 50 degrees Celsius is close to both 40 degrees Celsius and 60 degrees Celsius, the accuracy of the analysis result at 50 degrees Celsius is high based on the test results at 40 degrees Celsius and the test results at 60 degrees Celsius. In contrast, since 80 degrees Celsius is far from both 40 degrees Celsius and 60 degrees Celsius, the accuracy of the analysis result at 80 degrees Celsius is lower than the accuracy of the analysis result at 50 degrees Celsius.

[0041] Therefore, the reliability assigning unit 233 suggests to the user that the second teacher data of the test result at, for example, 70 degrees Celsius (new test condition) close to 80 degrees Celsius be additionally input to the trained model or the trained model as second teacher data for improving the reliability of the predicted value of the performance value. For example, the reliability assigning unit 233 displays the suggestion on the output unit 24.

[0042] FIG. 3 is a diagram showing an example of the reliability of the predicted value of the performance value when the second teacher data is interpolated (additionally input). The horizontal axis represents temperature as an example of an analysis condition or a test condition. The vertical axis on the left represents the test result. The vertical axis on the right represents the analysis result. For example, using an experimental design technique, 70 degrees Celsius, which is close to 80 degrees Celsius, is selected as the temperature of the data for improving the reliability of the predicted value of the performance value. Then, an additional test is performed at the selected 70 degrees Celsius. Furthermore, the second teacher data including the test result at 70 degrees Celsius is additionally input to the trained model.

[0043] In Figure 3, because 80 degrees Celsius is a temperature close to 70 degrees Celsius (new test condition), the accuracy of the analysis result at 80 degrees Celsius is improved compared to the accuracy of the analysis result at 80 degrees Celsius in Figure 2. By executing model learning using the first training data including the analysis result with improved accuracy in this way and the second training data, the reliability of the predicted value of the performance value output from the trained model is improved.

[0044] Based on the results of Bayesian optimization, hyperparameters (e.g., learning rate, batch size, number of learning iterations) of a trained model or a model in the midst of training (neural network) may be determined. This is expected to improve the performance value of the object (product).

[0045] The output unit 24 shown in Fig. 1 is a display unit such as a liquid crystal display. The output unit 24 displays the predicted values ​​(stiffness data, stress data, pressure loss data, flow data, test data) and the reliability output from the model generation unit 23. The output unit 24 may display the predicted values ​​output from the model generation unit 23 in the form of a graph. The output unit 24 may be an audio output unit equipped with a speaker. The output unit 24 may output the predicted values ​​and the reliability output from the model generation unit 23 by voice.

[0046] Next, the prediction device 3 will be described in detail. In the execution stage (prediction stage) after the learning stage, the prediction device 3 shown in Fig. 1 inputs 3D shape data of the object to be predicted into the trained model. As a result, the prediction device 3 obtains a predicted value of the performance value of the object as an output of the trained model.

[0047] The acquisition unit 30 acquires three-dimensional shape data of an object for which a performance value is to be predicted (hereinafter referred to as "target shape data"). The acquisition unit 30 outputs the target shape data to the prediction unit 31. The target shape data may be provided with accuracy data.

[0048] The prediction unit 31 acquires from the storage unit 21 a trained model generated based on at least one of information on analysis conditions and analysis results or test conditions and test results for a predetermined object. The prediction unit 31 inputs three-dimensional shape data (shape data acquired by the acquisition unit 30) of an object for which a performance value is to be predicted, to the trained model. As a result, the prediction unit 31 derives an output of the trained model as a predicted value of the performance value of the prediction target, based on the information stored in the storage unit 21 and the three-dimensional shape data acquired by the acquisition unit 20. The output of the trained model is, for example, stiffness data of the object, stress data of the object, pressure loss data of the fluid around the object, flow data of the fluid around the object, and test data of the object.

[0049] The confidence assigning unit 32 acquires the trained model from the storage unit 21. For example, the confidence assigning unit 32 inputs accuracy data assigned to the three-dimensional shape data of the object to be predicted into the trained model. As a result, the confidence assigning unit 32 derives the output of the trained model as the confidence of the performance value of the prediction target.

[0050] The output unit 33 is a display unit such as a liquid crystal display. The output unit 33 displays the predicted values ​​(stiffness data, stress data, pressure loss data, flow data, test data) output from the prediction unit 31 as prediction results. The output unit 33 may display the reliability output from the prediction unit 31 as prediction results. The output unit 33 may display the predicted value and reliability output from the prediction unit 31 in a graph. The output unit 33 may be an audio output unit equipped with a speaker. The output unit 33 may output the predicted value and reliability output from the prediction unit 31 by audio.

[0051] Next, an example of the operation of the performance prediction device 1 will be described. 4 is a flowchart showing an example of the operation of the learning device 2. The acquiring unit 20 outputs at least one of the first teacher data and the second teacher data to the model generating unit 23 (step S101).

[0052] The model generation unit 23 inputs the three-dimensional shape data and conditions included in each teacher data output from the acquisition unit 20 to the solid analysis unit 230, the fluid analysis unit 231, and the test data derivation unit 232 as a model under learning. The model under learning may be a machine learning model or a model represented by a formula derived using a statistical method. The model generation unit 23 may input the accuracy data included in the first teacher data, the accuracy data included in the second teacher data, and the sample number data of each teacher data to the reliability assignment unit 233 as a model under learning (step S102).

[0053] The model generation unit 23 updates the parameters of the learning model so as to reduce the difference between the predicted values ​​(e.g., stiffness data, stress data, pressure loss data, flow data, test data) output from the learning model and the correct label of each teaching data output from the acquisition unit 20. The model generation unit 23 may update the coefficients of the formula representing the learning model so as to reduce the difference between the predicted value of the performance value and the correct label. A transfer learning technique may be performed on the learning model. The learning model with updated parameters in this way becomes a trained model. The model generation unit 23 may derive the reliability of the predicted value based on the accuracy data (step S103).

[0054] The output unit 24 displays the predicted value of the performance value of the product or part and the reliability of the predicted value of the performance value (step S104). The model generation unit 23 records the trained model in the storage unit 21 (step S105).

[0055] 5 is a flowchart showing an example of the operation of the prediction device 3. The prediction unit 31 acquires a learned model from the storage unit 21 (step S201). The acquisition unit 30 acquires target shape data (three-dimensional shape data of a prediction target) from a predetermined external device (not shown) (step S202). The acquisition unit 30 outputs the target shape data to the prediction unit 31 (step S203).

[0056] The prediction unit 31 inputs the target shape data to the trained model (step S204). The prediction unit 31 derives the output of the trained model as a predicted value of the performance value of the prediction target. The reliability assignment unit 32 derives the reliability of the predicted value of the performance value based on the accuracy data assigned to the target shape data (step S205). The output unit 33 displays the predicted value of the performance value of the prediction target and the reliability of the predicted value of the performance value (step S206).

[0057] As described above, in the performance prediction device 1, the model generation unit 23 generates a trained model based on the first teacher data and the second teacher data. The first teacher data includes design shape data, analysis conditions, and analysis results (correct answer labels). The second teacher data includes measurement shape data, test conditions, and test results (correct answer labels). The first teacher data may include accuracy data of the analysis results. The second teacher data may include accuracy data of the test results. The accuracy data of the analysis results may be based on the accuracy data of the test results (e.g., 100%).

[0058] The input of the trained model is at least one of first teacher data (design shape data, etc.) and second teacher data (measurement shape data, etc.). A part of the first teacher data and the second teacher data may be missing. The trained model outputs a predicted value of the performance value of the object. The prediction unit 31 inputs three-dimensional shape data of the prediction target to the trained model. As a result, the prediction unit 31 derives the output of the trained model as a predicted value of the performance value of the prediction target. The trained model may output the reliability of the predicted value. The reliability assignment unit 32 inputs three-dimensional shape data of the prediction target to the trained model. As a result, the reliability assignment unit 32 derives the output of the trained model as the reliability of the predicted value. The output unit 33 outputs the derived predicted value and reliability.

[0059] This makes it possible to efficiently predict the performance values ​​of a product (equipment, part) based on the product's three-dimensional shape.

[0060] The learning device 2 uses the analysis results (simulation results) and the test data (test results) as teacher data for a model such as machine learning. The learning device 2 learns in advance the correlation between the three-dimensional shape data and the design performance, and generates a trained model. This enables the prediction device 3 to use the trained model to quickly predict the design performance of a product based on the target shape data of the product.

[0061] The accuracy of the test result is often higher than that of the analysis result. Also, the number of samples of the measurement points of the test result is often smaller than the number of samples of the analysis result. Therefore, by having the analysis result and the test result learn simultaneously in the model under learning, the correlation information between the analysis result and the test result is reflected in the trained model. This is expected to improve the reliability of the predicted value compared to the case where one of the analysis result and the test result is learned in the model under learning.

[0062] Product development becomes more efficient, enabling products to be introduced to customers more quickly. It becomes possible to try out various design proposals, making it possible to develop attractive products. Rapid verification enables design changes to be made early, facilitating coordination between suppliers and manufacturers. It becomes possible to efficiently predict the performance values ​​of wind turbines, which are subject to many external disturbances such as wind, and of ships, which are subject to many external disturbances such as ocean currents.

[0063] In order to predict performance values ​​(rigidity, stress, pressure loss), the prediction algorithm is trained based on previously conducted analysis information (analysis conditions such as fixed conditions and load conditions, analysis results) and product information corresponding to that analysis information (product type, part configuration, 3D shape, etc.). By inputting the 3D shape, product type, and part configuration of the object whose performance is to be predicted, it is possible to predict the performance values ​​of the product or part in a shorter time than if a new analysis were to be performed.

[0064] Second embodiment The second embodiment differs from the first embodiment in that the learning device executes learning without using teacher data (unsupervised learning). The second embodiment will be described focusing on the differences from the first embodiment.

[0065] 6 is a diagram showing a configuration example of the learning device 2a. The learning device 2a derives a predicted value of a performance value of an object based on three-dimensional shape data included in the input first data and second data. The first data includes design shape data, analysis conditions, and analysis results. The second data includes measured shape data, test conditions, and test results.

[0066] The learning device 2a includes an acquisition unit 20, a storage unit 21, a control unit 22, a model generation unit 23a, and an output unit 24. The model generation unit 23a includes a solid analysis unit 230, a fluid analysis unit 231, a test data derivation unit 232, and a reliability assignment unit 233.

[0067] The acquisition unit 20 outputs the acquired first data to the solid analysis unit 230, the fluid analysis unit 231, and the storage unit 21. The acquisition unit 20 outputs the acquired second data to the test data derivation unit 232 and the storage unit 21. The storage unit 21 stores the acquired first data and second data. That is, the storage unit 21 stores at least one of information on analysis conditions and analysis results or test conditions and test results related to a predetermined object.

[0068] The model generation unit 23 generates a trained model by unsupervised learning using the first data and the second data. The input of this trained model is at least one of the design shape data and analysis conditions included in the first data, and the measurement shape data and test conditions included in the second data. A part of the first data and the second data may be missing. The output of this trained model is a predicted value of the performance value of the object.

[0069] The model generation unit 23 performs clustering between the first data and the second data as unsupervised learning using a trained model. The model generation unit 23 performs clustering between the first data and the second data using, for example, the "k-means method." The value of the number of clusters "k" may be predetermined by a user, or may be estimated by the model generation unit 23 by deriving a posterior distribution of the number of clusters "k" using a Dirichlet mixture process.

[0070] This allows the prediction unit 31 to identify which cluster the target shape data input to the trained model belongs to in the prediction stage. Based on which cluster the target shape data belongs to, the trained model can output a predicted value of the performance value of the object in the prediction stage.

[0071] The model generation unit 23 may perform principal component analysis (dimensionality reduction) on the first data and the second data as unsupervised learning using the trained model. This allows the prediction unit 31 to extract the main factors that affect the predicted value of the performance value of the product or part from the target shape data in the prediction stage. That is, in the prediction stage, the trained model can output the predicted value of the performance value of the object based on the main factors in the target shape data input to the trained model.

[0072] 7 is a flowchart showing an example of the operation of the learning device 2a. The acquisition unit 20 outputs at least one of the first data and the second data to the model generation unit 23 (step S301).

[0073] The model generation unit 23 inputs the three-dimensional shape data and conditions included in each data output from the acquisition unit 20 to the solid analysis unit 230, the fluid analysis unit 231, and the test data derivation unit 232 as a model under learning. The model under learning may be a machine learning model, or a model (model formula) represented by a formula derived using a statistical method. The model generation unit 23 may input the accuracy data included in the first data, the accuracy data included in the second data, and the sample number data of each data to the reliability assignment unit 233 as a model under learning (step S302).

[0074] The model generation unit 23 may, for example, perform clustering on the first data and the second data. The model generation unit 23 may also, for example, perform principal component analysis on the first data and the second data. In this way, the model under training with updated parameters becomes a trained model. The model generation unit 23 may derive the reliability of the predicted value based on the accuracy data (step S303).

[0075] The output unit 24 displays the predicted value of the performance value of the product or part and the reliability of the predicted value of the performance value (step S304). The model generation unit 23 records the trained model in the storage unit 21 (step S305).

[0076] As described above, the model generation unit 23 generates a trained model by unsupervised learning using the first data and the second data. This makes it possible to efficiently predict the performance value of a product (equipment, part) based on the three-dimensional shape of the product.

[0077] Third embodiment The third embodiment differs from the first and second embodiments in that a prediction device derives a predicted value by using an arbitrary trained model prepared externally. The third embodiment will be described focusing on the differences from the first and second embodiments.

[0078] 8 is a diagram showing a configuration example of a prediction device 3b. The prediction device 3b includes an acquisition unit 30, a prediction unit 31b, a reliability assignment unit 32, and an output unit 33. The prediction unit 31b acquires an arbitrary trained model prepared externally. This arbitrary trained model is a model that receives three-dimensional shape data of an object to be predicted as an input and outputs a predicted value of the performance value of the object to be predicted. The trained model may be expressed, for example, by a model formula, or may be expressed, for example, by using a neural network.

[0079] The prediction unit 31b inputs the three-dimensional shape data of the object to be predicted to an arbitrary trained model prepared externally. As a result, the prediction unit 31 derives the output of the trained model as a predicted value of the performance value of the prediction target. Here, the technical feature is that the three-dimensional shape data of the object to be predicted is input to the trained model, and the trained model outputs a predicted value of the performance value of the prediction target, and the parameter values ​​of the trained model are not limited to specific values.

[0080] As described above, the prediction unit 31b inputs the three-dimensional shape data of the object to be predicted into a trained model that receives the three-dimensional shape data of the object to be predicted and outputs the predicted value of the performance value of the prediction target. This makes it possible to efficiently predict the performance value of a product (equipment, part) based on the three-dimensional shape of the product.

[0081] Some or all of the functional units (computers) of the learning device and prediction device are realized by a processor such as a CPU (Central Processing Unit) executing a program stored in a storage unit. The storage unit is preferably a non-volatile recording medium (non-temporary recording medium) such as a flash memory or a HDD (Hard Disk Drive). The storage unit may include a volatile recording medium such as a RAM (Random Access Memory). Some or all of the functional units of the learning device and prediction device may be realized using hardware such as an LSI (Large Scale Integrated circuits) or an ASIC (Application Specific Integrated Circuit).

[0082] Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and designs that do not deviate from the gist of the present invention are also included. [Explanation of symbols]

[0083] 1...performance prediction device, 2, 2a...learning device, 3, 3b...prediction device, 20...acquisition unit, 21...storage unit, 22...control unit, 23...model generation unit, 24...output unit, 30...acquisition unit, 31...prediction unit, 32...confidence assignment unit, 33...output unit, 230...solid analysis unit, 231...fluid analysis unit, 232...test data derivation unit, 233...confidence assignment unit

Claims

1. A storage unit that stores at least one of information on analysis conditions and analysis results or information on test conditions and test results for a predetermined object; An acquisition unit that acquires shape data of an object for which a performance value is to be predicted; a prediction unit that derives a predicted value of the performance value based on the information stored in the storage unit and the shape data acquired by the acquisition unit; a reliability assigning unit that assigns reliability to the predicted value derived by the prediction unit; an output unit that outputs the predicted value to which the reliability is assigned; A performance prediction device comprising:

2. A model generation unit that generates a trained model based on the information stored in the storage unit, The prediction unit derives a predicted value of a performance value of the object based on the shape data acquired by the acquisition unit and the trained model. The performance prediction device according to claim 1 .

3. The reliability assigning unit derives the reliability based on at least one of accuracy of first teacher data representing the analysis conditions and the analysis results stored in the storage unit and accuracy of second teacher data representing the test conditions and the test results. The performance prediction device according to claim 1 .

4. The reliability assigning unit derives the reliability based on a ratio between a number of samples of accuracy of the first teacher data stored in the storage unit and a number of samples of accuracy of the second teacher data. The performance prediction device according to claim 3 .

5. the reliability assigning unit updates the reliability assigned to the predicted value when at least one of the first teacher data and the second teacher data is added to the storage unit. The performance prediction device according to claim 4.

6. the reliability assigning unit suggests to a user that the test results under the new test conditions be added to the storage unit as the second teacher data; The performance prediction device according to any one of claims 3 to 5.

7. the first teacher data includes, as a correct answer label, at least one of the analysis results of stiffness data of the object, stress data of the object, pressure loss data of a fluid around the object, and flow data of the fluid; The performance prediction device according to any one of claims 3 to 6.

8. The second teacher data includes at least one of the test results as a correct answer label, the test results being a displacement measurement value of the object, a stress measurement value of the object, a measurement value of the acceleration or displacement of the object in response to a time-series vibration input, a measurement value of a pressure difference between positions determined on the object, and a measurement value of a force applied to the object by hydraulic oil. The performance prediction device according to any one of claims 3 to 7.

9. the object is a device or component, the device is at least one of a reducer, a hydraulic valve, a compressor, and a door; The part is at least one of a bolt, a spring, a housing, a bracket, a gear, and a frame. The performance prediction device according to any one of claims 1 to 8.

10. The object is the equipment or the part mounted on at least one of a wind turbine, a ship, a construction machine, and a robot. The performance prediction device according to claim 9.

11. A performance prediction method executed by a performance prediction device including a storage unit that stores at least one of information on analysis conditions and analysis results or test conditions and test results for a predetermined object, an acquisition unit, a prediction unit, a reliability assignment unit, and an output unit, An acquisition step in which the acquisition unit acquires shape data of an object for which a performance value is to be predicted; a prediction step in which the prediction unit derives a predicted value of the performance value based on at least one of information of the analysis conditions and the analysis results or the test conditions and the test results related to the predetermined object and the acquired shape data; a reliability assigning step in which the reliability assigning unit assigns reliability to the predicted value derived by the prediction unit; an output step in which the output unit outputs the predicted value to which the reliability is assigned; A performance prediction method comprising:

12. On the computer, an acquisition step for acquiring shape data of an object for which a performance value is to be predicted; a prediction step of deriving a predicted value of the performance value based on at least one of information on analysis conditions and analysis results or test conditions and test results for a predetermined object and the acquired shape data; a step of assigning a reliability to the derived predicted value; an output step of outputting the predicted value to which the reliability is assigned; A performance prediction program for executing the above.

Images