Simulation model accuracy improvement system, simulation model accuracy improvement method, and simulation model accuracy improvement program

Abstract

The present invention easily specifies a location and parameter causing deviation between an experimental result and a simulation result. The present invention is provided with: an experimental result data input unit for receiving input of collision experiment data of a real object; a simulation unit for simulating a collision experiment; a mode decomposition unit for performing mode decomposition on shape / behavior data and a simulation curve group created by the simulation unit; a coefficient calculation unit for calculating a coefficient for fitting the simulation curve group to the experimental result curve group; a reconstruction unit for reconstructing the simulation curve group and reconstructing virtual shape / behavior data; and a force output unit for calculating the force required for correcting the difference between the shape / behavior data and the virtual shape / behavior data.

Description

System for improving the accuracy of a simulation model, method for improving the accuracy of a simulation model, and program for improving the accuracy of a simulation model

[0001] The present invention relates to a system for improving the accuracy of a simulation model, a method for improving the accuracy of a simulation model, and a program for improving the accuracy of a simulation model.

[0002] Crash safety is extremely important in the automotive industry, and currently, crash safety is confirmed through actual vehicle testing. In actual vehicle testing, "safety" is defined as a crash at a specified speed and angle, with a specified occupant dummy on board, where the acceleration and force acting on the dummy are below a standard. However, in the real world, speeds, angles, and occupant body types vary widely, and there is a movement to change the definition of "safety" to reflect these real-world conditions.

[0003] However, conducting multiple experiments would significantly increase costs, so in order to achieve the definition of "safety" described above, it is necessary to utilize simulations at a highly accurate and predictable level in automobile development.

[0004] However, even though simulations are said to be highly accurate and predictable, there are still discrepancies with experimental results. Therefore, in order to define "safety" that corresponds to the real world as described above, it is necessary to show how reliable the simulation results are.

[0005] One scheme for comparing experimental and simulation results is ISO 18571 (formerly CORA). ISO 18571 evaluates the results by quantifying them based on whether the curve of the simulation results is within a corridor represented by two curves, whether the shapes of the curves of the experimental and simulation results are similar, whether there is a phase shift, whether the sizes are correct, etc. Furthermore, the injury level of the occupant dummy is evaluated using injury values ​​calculated from multiple curves.

[0006] TECHNICAL SPECIFICATION ISO / TS 18571,First edition 2014-08-01, Road vehicles - Objective rating metric for non-ambiguous signals

[0007] Furthermore, in the evaluation process at EuroNCAP, a European evaluation organization, the reliability of the simulation results will be evaluated based on ISO 18571 using a numerical value representing the degree of agreement between a large number of curves, and it is becoming extremely important for automobile manufacturers to obtain a numerical value above a certain level in this evaluation.

[0008] However, identifying the locations and parameters that cause the discrepancy between experimental and simulation results is an extremely difficult task, and the current situation is that the only way to find out is through repeated trial and error.

[0009] The present invention has been made to solve the above-mentioned problems, and aims to provide a system, a method, and a program for supporting improvement of accuracy of a simulation model, which can easily identify the locations and parameters that cause the discrepancy between experimental results and simulation results, and which can improve the accuracy of a simulation model while utilizing data from past experimental results and simulation results.

[0010] In order to solve the above-mentioned problems, a first aspect of the simulation model accuracy improvement support system of the present invention includes: an experiment result data input unit that inputs experiment result data from a collision experiment of a real object such as a vehicle; a simulation unit that uses CAE analysis software based on the finite element method to simulate a collision experiment for a virtual object corresponding to the real object, in relation to the collision experiment; a mode decomposition unit that performs mode decomposition on the shape / behavior data and simulation curves created by the simulation unit, i.e., the shape / behavior data representing the shape / behavior of the virtual object and the simulation curves representing changes in the predetermined locations of the virtual object, including acceleration outputs or displacement outputs or the like at the predetermined locations; a coefficient calculation unit that compares the simulation curves with the experiment result curves representing changes in the predetermined locations of the real object, including acceleration outputs or displacement outputs or the like at the predetermined locations, which are included in the experiment result data, and calculates coefficients by which the simulation curves fit to the experiment result curves; and a reconstruction unit that reconstructs the simulation curves based on the calculated coefficients, and reconstructs virtual shape / behavior data that is expected to have occurred in the collision experiment, based on the calculated coefficients. and a force output unit that compares the shape and behavior data created by the simulation unit with the shape and behavior data reconstructed by the reconstruction unit, and calculates or estimates the force required to correct the difference.

[0011] According to a first aspect of the simulation model accuracy improvement support system of the present invention, coefficients for fitting the group of simulation curves to the group of experimental result curves are calculated, the group of simulation curves are reconstructed based on the calculated coefficients, and virtual shape and behavior data that is expected to have occurred in the collision test is reconstructed based on the calculated coefficients, and the shape and behavior data are compared with the virtual shape and behavior data to calculate or estimate a force required to correct the difference. Therefore, it is possible to provide a simulation model accuracy improvement support system that can easily identify the locations and parameters that cause the discrepancy between the experimental results and the simulation results, and improve the accuracy of the simulation model.

[0012] In another aspect of the simulation model accuracy improvement support system of the present invention, a correlation calculation unit is provided that calculates the correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object, including plate thickness, etc. According to this other aspect, mode decomposition is performed on the shape / behavior data and the group of simulation curves, and the correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object is calculated. This makes it easy to understand the relationship between main behavior modes and input parameters that affect them, and makes it possible to evaluate and examine, for example, by quantifying the difference in results in intuitive expressions such as head behavior.

[0013] A second aspect of the simulation model accuracy improvement support system of the present invention includes: an experiment result data input unit that inputs experiment result data from a collision experiment of a real object such as a vehicle; a simulation unit that uses CAE analysis software based on the finite element method to simulate a collision experiment for a virtual object corresponding to the real object in relation to the collision experiment; a mode decomposition unit that performs mode decomposition on the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data that represents the behavior of the virtual object and the simulation curve group that represents changes in the predetermined locations of the virtual object, including acceleration output or displacement output, etc. at the predetermined locations; a difference pattern vector extraction unit that compares the experiment result curve group that represents changes in the predetermined locations of the real object, including acceleration output or displacement output, etc. at the predetermined locations, of the real object obtained by the collision experiment, with the simulation curve group, and extracts a difference pattern vector; a coefficient calculation unit that compares the experiment result curve group with the simulation curve group, and calculates coefficients for fitting the simulation curve group to the experiment result curve group; a reconstruction unit that reconstructs the group of simulation curves based on the calculated coefficients and also reconstructs virtual shape and behavior data that is expected to have occurred in the collision test based on the calculated coefficients; a force output unit that compares the shape and behavior data created by the simulation unit with the shape and behavior data reconstructed by the reconstruction unit and calculates or estimates the force required to correct the difference; a knowledge database unit that stores data including past experiment results, past simulation results, past proposal reports, and information regarding differences between past experiment results and simulations; and a similarity judgment unit that judges the similarity regarding differences between the current experiment results and simulations based on the experiment result data, the shape and behavior data created by the simulation unit, the shape and behavior data reconstructed by the reconstruction unit, or data from the knowledge database unit.and a suggestion unit that extracts countermeasures for cases where there is a high degree of similarity between the experimental results and the simulation based on the output of the similarity judgment unit, and that suggests how to change the parameters of the virtual object, including plate thickness, etc., to increase a predetermined evaluation score, based on the data in the knowledge database unit.

[0014] According to a second aspect of the simulation model accuracy improvement support system of the present invention, in addition to the first aspect described above, the system determines the degree of similarity between the current experiment results and the simulation based on the experimental result data, the shape and behavior data created by the simulation unit, the shape and behavior data reconstructed by the reconstruction unit, or data from the knowledge database unit. Then, in cases where the degree of similarity between the current experiment results and the simulation is high, countermeasures are extracted based on the data from the knowledge database unit, and suggestions are made on how to change the parameters to increase the specified evaluation score. Therefore, specific suggestions are made to bring the simulation results closer to the experimental results, making it easier for the user to solve problems, and a simulation model accuracy improvement support system that can improve the accuracy of the simulation model can be provided.

[0015] In another aspect of the simulation model accuracy improvement support system of the present invention, the system comprises a deviation visualization unit that, based on the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit, vectorizes information on a group of nodal coordinates in a mode space of a predetermined member in the shape and behavior data, and calculates a distance or direction in the mode space by decomposing the vector into modes, and visualizes the degree of deviation of the shape and behavior of the predetermined member from the direction of the difference pattern vector and whether the mode coordinate values ​​are close at a predetermined timing; and the suggestion unit, based on the output of at least one of the similarity determination unit and the deviation visualization unit, extracts countermeasures in cases where there is a high similarity between the difference between the current experiment result and the simulation, based on data in a knowledge database, and suggests how to change the parameters to increase the predetermined evaluation score. According to this aspect, in addition to the second aspect, the system vectorizes information on a group of nodal coordinates in modal space of a predetermined component in the shape and behavior data, decomposes the vectors into modes, calculates distances or directions in modal space, and visualizes the degree of deviation in the behavior of the predetermined component based on whether the modal coordinate values ​​are close at a predetermined timing. Furthermore, based on the output of at least one of the similarity determination unit and the deviation visualization unit, countermeasures for cases with high similarity between the experimental results and the simulation are extracted based on data in a knowledge database, and suggestions are made on how to change the parameters to improve the predetermined evaluation score. Therefore, specific suggestions are made to bring the simulation results closer to the experimental results, making it easier for users to solve problems, and a simulation model accuracy improvement support system that can improve the accuracy of the simulation model can be provided.

[0016] In another aspect of the simulation model accuracy improvement support system of the present invention, the similarity determination unit comprises: a deviation pattern similarity determination unit that determines the similarity of deviation patterns of curves based on the difference pattern vector; and a deformation mode similarity determination unit that, based on the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit, vectorizes information of a group of node coordinates in a mode space in the shape and behavior data, and calculates distances or directions in the mode space by decomposing the vectors into modes, and determines the similarity of deformation modes based on whether the mode coordinate values ​​are close at a predetermined timing; and the suggestion unit, based on output from at least one of the deviation pattern similarity determination unit, the deformation mode similarity determination unit, and the deviation visualization unit, extracts countermeasures in cases where there is a high similarity in the difference between the current experiment result and the simulation based on data in a knowledge database, and suggests how to change the parameters to increase the predetermined evaluation score. According to this aspect, in addition to the other aspects described above, a difference pattern vector between a group of experimental result curves and a group of simulation curves is extracted, and the similarity of the deviation patterns of the curves is determined based on the difference pattern vector. Furthermore, by vectorizing information on a group of nodal coordinates in a modal space in the shape and behavior data and decomposing the vector into modes, a distance or direction in the modal space is calculated, and the similarity of the deformation modes is determined at a predetermined timing by determining whether the modal coordinate values ​​are close to each other. Furthermore, based on the output of at least one of the deviation pattern similarity determination unit, the deformation mode similarity determination unit, and the deviation visualization unit, countermeasures for cases where there is a high similarity between the differences between the current experimental results and the simulation are extracted based on data in a knowledge database, and a proposal is made as to how to change the parameters to increase the predetermined evaluation score.Therefore, a user can learn specific measures to bring the simulation results closer to the experimental results based on the similarity of the deviation patterns of the curves, or based on the similarity of the deformation modes, or based on the output of the deviation visualization unit, or based on at least one of these, making it easier to solve problems and providing a simulation model accuracy improvement support system that can improve the accuracy of the simulation model.

[0017] In another aspect of the simulation model accuracy improvement support system of the present invention, a correlation calculation unit is provided that calculates the correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object, including plate thickness, etc. In this other aspect, mode decomposition is performed on the shape / behavior data and the group of simulation curves, and the correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object is calculated. This makes it easy to understand the relationship between main behavior modes and input parameters that affect them, and makes it possible to evaluate and examine, for example, by quantifying the difference in results in intuitive expressions such as head behavior.

[0018] In another aspect of the simulation model accuracy improvement support system of the present invention, the system further comprises an automatic execution unit that automatically executes the measures proposed by the proposal unit. In this other aspect, the user can more easily identify the locations and parameters that cause the discrepancy between the experimental results and the simulation results, and can improve the accuracy of the simulation model while using data from past experimental results and simulation results.

[0019] In order to solve the above-mentioned problems, a first aspect of the method for supporting improvement of simulation model accuracy of the present invention includes: a step of inputting, by an experiment result data input unit, experiment result data of a collision experiment of a real object such as a vehicle; a step of performing, by a simulation unit, a simulation of a collision experiment for the collision experiment on a virtual object corresponding to the real object using CAE analysis software based on the finite element method; a step of performing mode decomposition on the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data representing the shape / behavior of the virtual object and the simulation curve group representing changes in the predetermined locations of the virtual object, including acceleration output or displacement output, etc. at the predetermined locations; a step of comparing, by a coefficient calculation unit, the experiment result curve group representing changes in the predetermined locations of the real object, including acceleration output or displacement output, etc. at the predetermined locations, which are included in the experiment result data, with the simulation curve group, and calculating coefficients for fitting the simulation curve group to the experiment result curve group; The method comprises a step of reconstructing, by a reconstruction unit, the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape and behavior data that is expected to have occurred in the collision test based on the calculated coefficients; and a step of comparing, by a force output unit, the shape and behavior data created by the simulation unit with the shape and behavior data reconstructed by the reconstruction unit, and calculating or estimating the force required to correct the difference.

[0020] A first aspect of the method for supporting improvement of simulation model accuracy of the present invention includes calculating coefficients for fitting the group of simulation curves to the group of experimental result curves, reconstructing the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape and behavior data that is expected to have occurred in the collision test based on the calculated coefficients. Furthermore, the shape and behavior data is compared with the virtual shape and behavior data, and a force required to correct the difference is calculated or estimated. Therefore, it is possible to provide a method for supporting improvement of simulation model accuracy that can easily identify locations and parameters that cause discrepancies between experimental results and simulation results, thereby improving the accuracy of the simulation model.

[0021] Another aspect of the method for supporting improvement of simulation model accuracy of the present invention includes a step of calculating, by a correlation calculation unit, a correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object, including plate thickness, etc. According to this other aspect, mode decomposition is performed on the shape / behavior data and the group of simulation curves, and a correlation is calculated between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object. This makes it easy to understand the relationship between main behavior modes and input parameters that affect them, and makes it possible to evaluate and examine, for example, differences in results in intuitive expressions such as head behavior, by quantifying them.

[0022] In order to solve the above-mentioned problems, a second aspect of the method for supporting improvement of simulation model accuracy of the present invention includes the steps of: inputting, by an experiment result data input unit, experiment result data of a collision experiment of a real object such as a vehicle; performing, by a simulation unit, a simulation of a collision experiment for the collision experiment on a virtual object corresponding to the real object using CAE analysis software based on the finite element method; performing mode decomposition on the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data representing the behavior of the virtual object and the simulation curve group representing changes in the predetermined location of the virtual object, including acceleration output or displacement output, etc. at the predetermined location; and comparing, by a difference pattern vector extraction unit, the experiment result curve group representing changes in the predetermined location of the real object, including acceleration output or displacement output, etc. at the predetermined location, obtained by the collision experiment, with the simulation curve group, and extracting a difference pattern vector. a step of using a coefficient calculation unit to compare a group of experimental result curves, which are included in the experimental result data and which represent changes in the predetermined locations of the real object, including acceleration output or displacement output, etc., at the predetermined locations, with the group of simulation curves, and to calculate coefficients by which the group of simulation curves fits to the group of experimental result curves; a step of using a reconstruction unit to reconstruct the group of simulation curves based on the calculated coefficients, and to reconstruct virtual shape and behavior data that is expected to have occurred in the collision test based on the calculated coefficients; a step of using a force output unit to compare the shape and behavior data created by the simulation unit with the shape and behavior data reconstructed by the reconstruction unit, and to calculate or estimate a force required to correct the difference;The method comprises the steps of: determining the similarity regarding the difference between the current experiment result and the simulation by a similarity determination unit based on data from a knowledge database unit that stores the experiment result data, the shape / behavior data created by the simulation unit, the shape / behavior data reconstructed by the reconstruction unit, or past experiment results, past simulation results, past proposal reports, and data including information regarding the difference between the past experiment results and the simulation; and extracting countermeasures in cases where there is a high degree of similarity regarding the difference between the current experiment result and the simulation based on the data in the knowledge database and proposing how to change the parameters to increase the predetermined evaluation score based on the output of the similarity determination unit by a proposal unit.

[0023] A second aspect of the simulation model accuracy improvement support method of the present invention, in addition to the first aspect described above, determines the degree of similarity between the current experiment results and the simulation based on experimental result data, shape and behavior data created by the simulation unit, shape and behavior data reconstructed by the reconstruction unit, or data from the knowledge database unit. Then, in cases where the degree of similarity between the current experiment results and the simulation is high, measures are extracted based on the data from the knowledge database unit, and suggestions are made on how to change the parameters of the virtual object, including plate thickness, to increase a predetermined evaluation score. Therefore, specific suggestions are made to bring the simulation results closer to the experimental results, making it easier for the user to solve problems, and a simulation model accuracy improvement support method can be provided that can improve the accuracy of the simulation model.

[0024] Another aspect of the method for supporting improvement of simulation model accuracy of the present invention comprises a step of: using a deviation visualization unit, based on the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit, vectorizing information on a group of node coordinates in a mode space of a predetermined member in the shape and behavior data, and calculating a distance or direction in the mode space by decomposing the vector into modes, and visualizing whether or not the mode coordinate values ​​are close at a predetermined timing and the degree of deviation of the shape and behavior of the predetermined member from the direction of the difference pattern vector; and the suggesting step extracts, based on the output of at least one of the similarity determination unit and the deviation visualization unit, measures in cases where there is a high degree of similarity between the results of the current experiment and the simulation, based on data in the knowledge database, and suggests how to change the parameters to increase the predetermined evaluation score. According to this aspect, in addition to the second aspect, the system vectorizes information on a group of nodal coordinates in modal space of a predetermined component in the shape and behavior data, decomposes the vectors into modes, calculates distances or directions in modal space, and visualizes the degree of deviation in the behavior of the predetermined component based on whether the modal coordinate values ​​are close at a predetermined timing. Furthermore, based on the output of at least one of the similarity determination unit and the deviation visualization unit, countermeasures for cases with high similarity between the experimental results and the simulation are extracted based on the data in the knowledge database, and suggestions are made on how to change the parameters to improve the predetermined evaluation score. Therefore, specific suggestions are made to bring the simulation results closer to the experimental results, making it easier for the user to solve problems, and a simulation model accuracy improvement support system that can improve the accuracy of the simulation model can be provided.

[0025] In another aspect of the method for supporting improvement of simulation model accuracy of the present invention, the step of determining the similarity comprises: a step of determining the similarity of deviation patterns of curves based on the difference pattern vector by a deviation pattern similarity determination unit provided in the similarity determination unit; and a step of vectorizing information of a group of nodal coordinates in a mode space in the shape and behavior data based on the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit by a deformation mode similarity determination unit provided in the similarity determination unit, and decomposing the vector into modes to calculate distances in the mode space, and determining the similarity of deformation modes based on whether the mode coordinate values ​​are close at a predetermined timing; and the step of proposing extracts, based on output from at least one of the deviation pattern similarity determination unit, the deformation mode similarity determination unit, and the deviation visualization unit, countermeasures for cases where there is a high similarity between the results of the experiment and the simulation based on data in the knowledge database, and proposing how to change the parameters to increase the predetermined evaluation score. According to this aspect, in addition to the other aspects described above, a difference pattern vector between a group of experimental result curves and a group of simulation curves is extracted, and the similarity of the deviation patterns of the curves is determined based on the difference pattern vector. Furthermore, by vectorizing information on a group of nodal coordinates in a modal space in the shape and behavior data and decomposing the vector into modes, a distance or direction in the modal space is calculated, and the similarity of the deformation modes is determined at a predetermined timing by determining whether the modal coordinate values ​​are close to each other. Furthermore, based on the output of at least one of the deviation pattern similarity determination unit, the deformation mode similarity determination unit, and the deviation visualization unit, countermeasures for cases where there is a high similarity between the differences between the current experimental results and the simulation are extracted based on data in a knowledge database, and a proposal is made as to how to change the parameters to increase the predetermined evaluation score.Therefore, a user can learn specific measures to bring the simulation results closer to the experimental results based on the similarity of the deviation patterns of the curves, or based on the similarity of the deformation modes, or based on the output of the deviation visualization unit, or based on at least one of them, making it easier to solve problems and providing a method for supporting improvement of the accuracy of a simulation model that can improve the accuracy of the simulation model.

[0026] Another aspect of the method for supporting improvement of simulation model accuracy of the present invention includes a step of calculating, by a correlation calculation unit, a correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object, including plate thickness, etc. In this other aspect, mode decomposition is performed on the shape / behavior data and the group of simulation curves, and a correlation is calculated between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object. This makes it easy to understand the relationship between main behavior modes and input parameters that affect them, and makes it possible to evaluate and examine, for example, by quantifying the difference in results in intuitive expressions such as head behavior.

[0027] In another aspect of the method for supporting improvement of the accuracy of a simulation model of the present invention, the method further comprises the step of automatically executing the measures proposed by the suggestion unit by an automatic execution unit. In this other aspect, the user can more easily identify the locations and parameters that cause the discrepancy between the experimental results and the simulation results, and can improve the accuracy of the simulation model while using data from past experimental results and simulation results.

[0028] In order to solve the above-mentioned problems, a first aspect of the simulation model accuracy improvement support program of the present invention is a simulation model accuracy improvement support program, the program comprising the steps of: inputting experimental result data of a collision test of a real object such as a vehicle into a computer; simulating a collision test for the collision test for a virtual object corresponding to the real object using CAE analysis software based on the finite element method; performing mode decomposition of shape / behavior data and a group of simulation curves, i.e., the shape / behavior data representing the shape / behavior of the virtual object and a group of simulation curves representing changes in the predetermined locations of the virtual object, including acceleration outputs or displacement outputs, etc., at the predetermined locations; comparing the group of simulation curves with a group of experimental result curves representing changes in the predetermined locations of the real object, including acceleration outputs or displacement outputs, etc., which are included in the experimental result data and are obtained by the collision test, and calculating coefficients by which the group of simulation curves fits to the group of experimental result curves; reconstructing the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape / behavior data that is expected to have occurred in the collision test based on the calculated coefficients. A step of comparing the shape and behavior data created by the simulation with the reconstructed shape and behavior data and calculating a force required to correct the difference is executed.

[0029] In a first aspect of the simulation model accuracy improvement support program of the present invention, coefficients for fitting the group of simulation curves to the group of experimental result curves are calculated, the group of simulation curves are reconstructed based on the calculated coefficients, and virtual shape and behavior data that is expected to have occurred in the collision test is reconstructed based on the calculated coefficients, and the shape and behavior data is compared with the virtual shape and behavior data to calculate or estimate a force required to correct the difference. Therefore, it is possible to provide a simulation model accuracy improvement support program that can easily identify locations and parameters that cause discrepancies between experimental results and simulation results, and improve the accuracy of the simulation model.

[0030] In another aspect of the simulation model accuracy improvement support program of the present invention, the program causes the computer to execute a step of calculating a correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object, including plate thickness, etc. According to this other aspect, mode decomposition is performed on the shape / behavior data and the group of simulation curves, and a correlation is calculated between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object. This makes it easy to understand the relationship between main behavior modes and input parameters that affect them, and makes it possible to evaluate and examine, for example, differences in results in intuitive expressions such as head behavior, by quantifying them.

[0031] In order to solve the above-mentioned problems, a second aspect of the simulation model accuracy improvement support program of the present invention is a simulation model accuracy improvement support program, the program comprising the steps of: inputting experimental result data of a collision test of a real object such as a vehicle into a computer; simulating a collision test for the collision test with a virtual object corresponding to the real object using CAE analysis software based on the finite element method; performing mode decomposition of shape / behavior data and a group of simulation curves, i.e., the shape / behavior data representing the behavior of the virtual object and a group of simulation curves representing changes in the predetermined locations of the virtual object, including acceleration output or displacement output, etc. at the predetermined locations; comparing a group of experimental result curves representing changes in the predetermined locations of the real object, including acceleration output or displacement output, etc. at the predetermined locations, which are included in the experimental result data, with the group of simulation curves, and extracting a difference pattern vector; a step of comparing a group of experimental result curves, which are included in the experimental result data and which represent changes in the predetermined locations of the real object, including acceleration output or displacement output, etc., of the predetermined locations, with the group of simulation curves, and calculating coefficients by which the group of simulation curves fits to the group of experimental result curves; a step of reconstructing the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape and behavior data that is expected to have occurred in the collision test based on the calculated coefficients; a step of comparing the shape and behavior data created in the step of performing the simulation with the reconstructed shape and behavior data, and calculating or estimating a force required to correct the difference;The method executes the following steps: a step of determining the degree of similarity regarding the difference between the current experiment result and the simulation based on data from a knowledge database unit that stores the experiment result data, the shape and behavior data created by the step of performing the simulation unit, the shape and behavior data reconstructed by the step of reconstructing, or past experiment results, past simulation results, past proposal reports, and data including information regarding the difference between past experiment results and simulations; and a step of extracting countermeasures based on the data in the knowledge database in cases where there is a high degree of similarity regarding the difference between the current experiment result and the simulation based on the output of the step of determining the degree of similarity, and proposing how to change the parameters to increase the specified evaluation score.

[0032] A second aspect of the simulation model accuracy improvement support program of the present invention, in addition to the first aspect described above, determines the degree of similarity between the current experimental results and the simulation based on experimental result data, shape and behavior data created in the simulation step, shape and behavior data reconstructed in the reconstruction step, or data from a knowledge database. Then, in cases where the degree of similarity between the current experimental results and the simulation is high, measures are extracted based on the data from the knowledge database, and suggestions are made on how to change the parameters of the virtual object, including plate thickness, to increase a predetermined evaluation score. Therefore, specific suggestions are made to bring the simulation results closer to the experimental results, making it easier for the user to solve problems, and a simulation model accuracy improvement support program that can improve the accuracy of the simulation model can be provided.

[0033] In another aspect of the simulation model accuracy improvement support program of the present invention, the program causes the computer to execute the steps of: vectorizing information on a group of nodal coordinates in a mode space of a predetermined component in the shape and behavior data based on the shape and behavior data created in the simulating step and the shape and behavior data reconstructed in the reconstructing step; calculating a distance or direction in the mode space by decomposing the vector into modes; and visualizing the degree of deviation of the shape and behavior of the predetermined component from the direction of the difference pattern vector, based on whether the mode coordinate values ​​are close at a predetermined timing; and as the proposing step, extracting, based on the output of at least one of the steps of determining the similarity and visualizing the deviation, measures to be taken in cases where there is a high degree of similarity between the results of the experiment and the simulation, based on data in the knowledge database unit, and proposing how to change the parameters to increase the predetermined evaluation score. According to this aspect, in addition to the second aspect, the information on the nodal coordinates in modal space of a predetermined component in the shape and behavior data is vectorized, and the vectors are decomposed into modes to calculate distances or directions in modal space. The degree of deviation in the behavior of the predetermined component is visualized based on whether the modal coordinate values ​​are close at a predetermined timing. Furthermore, based on the output of at least one of the similarity determination unit and the deviation visualization unit, countermeasures for cases with high similarity between the experimental results and the simulation are extracted based on the data in the knowledge database, and suggestions are made on how to change the parameters to improve the predetermined evaluation score. Therefore, specific suggestions are made to bring the simulation results closer to the experimental results, making it easier for the user to solve problems, and a simulation model accuracy improvement support program that can improve the accuracy of the simulation model can be provided.

[0034] In another aspect of the simulation model accuracy improvement support program of the present invention, the program causes the computer to execute, as the similarity determining step, a step of determining a similarity of a deviation pattern of a curve based on the difference pattern vector, and a step of vectorizing information of a group of node coordinates in a modal space in the shape and behavior data based on the shape and behavior data created in the simulating step and the shape and behavior data reconstructed in the reconstructing step, calculating a distance or direction in the modal space by decomposing the vector into modes, and determining a similarity of a deformation mode at a predetermined timing based on whether the modal coordinate values ​​are close or not; and as the suggesting step, a step of extracting, based on data in the knowledge database, countermeasures in cases where there is a high similarity in the difference between the current experiment result and the simulation, based on output of at least one of the steps of determining the similarity of the deviation pattern, determining the similarity of the deformation mode, and visualizing the deviation, and proposing how to change the parameters to increase a predetermined evaluation score. According to this aspect, in addition to the other aspects described above, a difference pattern vector between a group of experimental result curves and a group of simulation curves is extracted, and the similarity of the deviation patterns of the curves is determined based on the difference pattern vector. Furthermore, by vectorizing information on a group of nodal coordinates in a modal space in the shape and behavior data and decomposing the vector into modes, a distance or direction in the modal space is calculated, and the similarity of the deformation modes is determined at a predetermined timing by determining whether the modal coordinate values ​​are close to each other. Furthermore, based on the output of at least one of the deviation pattern similarity determination unit, the deformation mode similarity determination unit, and the deviation visualization unit, countermeasures for cases where there is a high similarity between the differences between the current experimental results and the simulation are extracted based on data in a knowledge database, and a proposal is made as to how to change the parameters to increase the predetermined evaluation score.Therefore, a user can know specific measures for bringing the simulation results closer to the experimental results based on the similarity of the deviation patterns of the curves, or based on the similarity of the deformation modes, or based on either of them, making it easier to solve problems and providing a simulation model accuracy improvement support program that can improve the accuracy of the simulation model.

[0035] In another aspect of the simulation model accuracy improvement support program of the present invention, the program causes the computer to execute a step of calculating a correlation between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object, including plate thickness, etc. In this other aspect, mode decomposition is performed on the shape / behavior data and the group of simulation curves, and a correlation is calculated between the mode-decomposed group of simulation curves and shape / behavior data and parameters of the virtual object. This makes it easy to understand the relationship between main behavior modes and input parameters that affect them, and makes it possible to evaluate and examine, for example, differences in results in intuitive expressions such as head behavior, by quantifying them.

[0036] In another aspect of the simulation model accuracy improvement support program of the present invention, the program further causes the computer to automatically execute the measures proposed in the proposing step. In this other aspect, the user can more easily identify the locations or parameters that cause the discrepancy between the experimental results and the simulation results, and can improve the accuracy of the simulation model while using data from past experimental results and simulation results.

[0037] According to the present invention, it is possible to provide a system, a method, and a program for supporting improvement of accuracy of a simulation model, which can easily identify the locations and parameters that cause the discrepancy between experimental results and simulation results, and can improve the accuracy of a simulation model while utilizing data from past experimental results and simulation results.

[0038] 1 is a diagram illustrating a schematic configuration of a simulation model accuracy improvement support system according to a first embodiment of the present invention. FIG. 2 is a diagram illustrating functional blocks of the simulation model accuracy improvement support system according to the first embodiment. FIG. 3 is a diagram illustrating functional blocks of a database in the simulation model accuracy improvement support system according to the first embodiment. FIG. 4 is a block diagram illustrating a processing flow in the simulation model accuracy improvement support system according to the first embodiment. FIGS. 1A and 1B show the results of principal component analysis by DIFFCRASH on a node group of the head part at 100 ms. The behavior of each of two large clusters in first-order modal coordinates is shown. FIGS. 1A and 1B show correlation matrices between input parameters and modal contribution rates. The results of regression analysis of input parameters and modal coordinates are shown. FIGS. 2A and 2B are diagrams for explaining mode decomposition processing in the simulation model accuracy improvement support system according to the first embodiment. FIGS. 3A and 3B are diagrams for explaining processing by a reconstruction unit in the simulation model accuracy improvement support system according to the first embodiment. FIGS. 4A and 4B are diagrams for explaining processing by a deviation visualization unit in the simulation model accuracy improvement support system according to the first embodiment. FIG. 5 is a block diagram illustrating a processing flow for extracting countermeasures in cases where there is a high degree of similarity regarding the difference between the current experiment result and the simulation result in the simulation model accuracy improvement support system according to the first embodiment. Fig. 1 is a diagram for explaining the processing of a deviation pattern similarity determination unit in the simulation model accuracy improvement support system of the first embodiment. Fig. 2 is a diagram for explaining the processing of a deformation mode similarity determination unit in the simulation model accuracy improvement support system of the first embodiment. Fig. 3 is a diagram for explaining the processing of a deformation mode similarity determination unit in the simulation model accuracy improvement support system of the first embodiment. Fig. 4 is a block diagram showing the flow of processing to extract measures in a case where there is a high degree of similarity regarding the difference between the current experiment result and the simulation result in the simulation model accuracy improvement support system of a second embodiment of the present invention.

[0039] First Embodiment A simulation model accuracy improvement support system according to a first embodiment of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a diagram showing a schematic configuration of a simulation model accuracy improvement support system 100 according to this embodiment. As shown in FIG. 1, the simulation model accuracy improvement support system 100 of this embodiment includes a simulation model accuracy improvement support device 110 and a server 120 connected to the simulation model accuracy improvement support device 110 via a network 200 such as the Internet. As shown in FIG. 1, the simulation model accuracy improvement support device 110 includes a central processing unit 1, a display device 2, a storage device 3, an input device 4, an output device 5, and a communication device 6.

[0040] The central processing unit 1 is a device capable of executing a program, such as a personal computer, and includes a CPU, memory, etc. The display device 2 is a device capable of displaying characters and images, such as a liquid crystal display. The storage device 3 is a device capable of storing programs and data, such as a hard disk drive (HDD), and may also be an external server. The program of the present invention is stored in the storage device 3. The input device 4 is a device such as a keyboard that allows a user to input data or instructions. The output device 5 is a device such as a printer that can output characters and images. In the simulation model accuracy improvement support system 100 of this embodiment, the output device 5 may be omitted. The communication device 6 is an interface that enables communication with the server 120 and other devices via the network 200.

[0041] The server 120 is used as a storage destination for metadata (to be described later) and reports as simulation results. However, in the present invention, the storage destination for the metadata and reports does not necessarily have to be a server, but may be a local storage such as an HDD (Hard Disk Drive), a cloud server, or a network such as a blockchain.

[0042] 2 is a diagram showing functional blocks that function when the central processing unit 1 of the simulation model accuracy improvement support device 110 executes the program of the present invention. As shown in Fig. 2, the central processing unit 1 functions as a control unit 10. Furthermore, the control unit 10 functions as an experiment result data input unit 11, a simulation unit 12, a mode decomposition unit 13, a correlation calculation unit 14, a coefficient calculation unit 15, a difference pattern vector extraction unit 16, a reconstruction unit 17, a force output unit 19, a server registration unit 20, a deviation pattern similarity determination unit 21, a deformation mode similarity determination unit 22, a deviation visualization unit 23, and a proposal unit 24 in accordance with the program of the present invention.

[0043] In this embodiment, the program of the present invention can be executed in a simulation model accuracy improvement support device 110 .

[0044] The experiment result data input unit 11 inputs experiment result data from a collision experiment using a vehicle as an example of a real object into the simulation model accuracy improvement support device 110 .

[0045] The simulation unit 12 uses CAE analysis software based on the finite element method to simulate a collision test for a virtual object corresponding to the real object. In this embodiment, Ansys LS-DYNA (registered trademark) is used as an example of CAE analysis software. In the simulation, shape / behavior data representing the shape and behavior of the virtual object and a group of simulation curves representing changes in the predetermined locations of the virtual object, including acceleration output or displacement output, etc., at the predetermined locations, are created. The shape / behavior data can also be visualized, for example, as a 3D animation for user confirmation. The shape / behavior data can be visualized at any time by the user's selection.

[0046] The modal decomposition unit 13 performs modal decomposition on the shape / behavior data and the group of simulation curves created by the simulation unit 12 .

[0047] Mode decomposition is performed on the shape / behavior data and the group of simulation curves based on the following data: Shape / behavior data and the group of simulation curves created in a simulation under the same conditions as a crash test (a crash test conducted under certain speed and angle conditions is referred to as a "base test." Details of the "base test" will be described later). A simulation conducted under the same speed and angle conditions as the base test is referred to as a "base calculation." Details of the "base calculation" will be described later. Past shape / behavior data archive data and past simulation curve group archive data stored in a simulation database unit (described later). In addition to the shape / behavior data and the group of simulation curves created in one of the base calculations, another base experiment may be conducted under different speed and angle conditions from the one base experiment, and a separate base calculation corresponding to this separate base experiment may be performed. Furthermore, one or more variation cases may be simulated under different speed and angle conditions from the base test, and the mode decomposition may be based on the simulation of the one or more variation cases. Details of mode decomposition will be described later.

[0048] The correlation calculation unit 14 calculates the correlation between the group of mode-decomposed simulation curves and shape / behavior data and the parameters of the virtual object, including plate thickness, etc. The calculation of this correlation will be described in detail later.

[0049] The coefficient calculation unit 15 compares the group of simulation curves with a group of experiment result curves that are included in the experiment result data and that represent changes at predetermined locations of the vehicle, including acceleration outputs or displacement outputs at the predetermined locations obtained by a collision experiment, and then calculates coefficients that allow the group of simulation curves to fit to the group of experiment result curves.

[0050] The difference pattern vector extraction unit 16 compares a group of experimental result curves, which are included in the experimental result data and represent changes in the specified locations of the vehicle, including acceleration output or displacement output, etc., at the specified locations obtained by the collision experiment, with a group of simulation curves, and extracts a difference pattern vector.

[0051] The reconstruction unit 17 reconstructs a group of simulation curves based on the coefficients calculated as described above, and also reconstructs virtual shape and behavior data that is expected to have occurred in the collision test based on the coefficients calculated as described above.

[0052] The force output unit 19 compares the shape and behavior data with the virtual shape and behavior data, and calculates the force required to correct the difference.

[0053] The server registration unit 20 stores data such as data obtained by the simulation model accuracy improvement support device 110, input experimental results, and measures to correct discrepancies between experimental results and simulation results in a simulation database 121 and a knowledge database 122, which will be described later.

[0054] The deviation pattern similarity determining unit 21 determines the similarity of the deviation patterns of the curves based on the difference pattern vectors.

[0055] The deformation mode similarity determination unit 22 calculates the distance in mode space between simulation results by vectorizing information on nodal coordinate groups in the shape and behavior data in mode space in the shape and behavior data based on the shape and behavior data created by the simulation unit 12 and the shape and behavior data reconstructed by the reconstruction unit 17, and decomposing the vectors into modes. Then, the deformation mode similarity determination unit 22 determines the similarity of the deformation modes based on whether the mode coordinate values ​​are close at a predetermined timing.

[0056] Based on the shape and behavior data created by the simulation unit 12 and the shape and behavior data reconstructed by the reconstruction unit 17, the deviation visualization unit 23 vectorizes information on the nodal coordinates in the modal space of a specified component in the shape and behavior data, and decomposes the vectors into modes to calculate the distance in the modal space and quantify the deviation. The deviation visualization unit 23 may also use directional information in addition to distance to quantify the deviation. This is because the same distance will produce different results if the direction in the modal space is different. The deviation visualization unit 23 then visualizes the deviation in the behavior of the specified component using contours based on whether the modal coordinate values ​​are close at a specified timing. Note that visualization of the deviation is not limited to the method using contours. The deviation may also be expressed as a scalar value or a vector value. For example, a method of showing scalar values ​​with a color bar (e.g., a low scalar value is shown with a blue color bar, and a high scalar value is shown with a red color bar) may be adopted. Furthermore, it is also conceivable to present vector values ​​with arrows.

[0057] The suggestion unit 24 extracts countermeasures based on data from the knowledge database 122 for cases where there is a high degree of similarity between the difference between the current experimental result and the simulation, based on the output of at least one of the deviation pattern similarity determination unit 21, the deformation mode similarity determination unit 22, and the deviation visualization unit 23. The suggestion unit 24 then proposes how to change the parameters to increase the predetermined evaluation score. The suggestion unit 24 may also classify the similarity of the deviation pattern determined by the deviation pattern similarity determination unit 21. For example, a deviation pattern is a numerical sequence created by comparing the experimental and simulation curves at a predetermined location. Possible classification methods include a single curve with a similar pattern, or a group of deviation patterns at multiple locations with a similar trend. Since similar patterns or trends may indicate deviations due to similar factors, it is possible to make similar suggestions for each such group. Furthermore, when making the proposal, the proposal unit 24 does not have to use all of the outputs of the deviation pattern similarity determination unit 21, the deformation mode similarity determination unit 22, and the deviation visualization unit 23. The proposal may be made using the output of at least one of the deviation pattern similarity determination unit 21, the deformation mode similarity determination unit 22, and the deviation visualization unit 23. Furthermore, when making the proposal, restrictions may arise on the materials that can be used, the shape layout, and the like. In such cases, the proposal unit 24 can be configured to make the proposal while taking such restrictions into consideration. Furthermore, the proposal unit 24 may make the proposal by taking into consideration the correlation between the simulation curve group and the shape / behavior data calculated by the correlation calculation unit 14 and the parameters of the virtual object, including the plate thickness, etc. This can improve the persuasiveness and accuracy of the proposal.

[0058] 3, the server 120 includes a simulation database 121 and a knowledge database 122. The simulation database 121 is a database that stores past simulation results obtained by the simulation unit 12.

[0059] The knowledge database 122 is a database that stores data including past experiment results, past simulation results, past proposal reports, and information regarding differences between past experiment results and simulations. The past data stored in the knowledge database 122 does not necessarily have to be created by the user himself, and may also include data on which a third party performed calculations in the past.

[0060] Next, an outline of the operation of this embodiment will be described with reference to Figures 4 to 12. As shown in Figure 4, in this embodiment, the program of the present invention includes a function for comparing experimental results with simulation results and a function for proposing measures to correct any discrepancy between the experimental results and the simulation results.

[0061] <Function for Comparing Experimental Results with Simulation Results> First, the function for comparing experimental results with simulation results in this embodiment will be described with reference to FIGS. 4 to 12. In this embodiment, a collision experiment is performed in which a vehicle is caused to collide at a predetermined speed and a predetermined angle, an occupant dummy is placed in the vehicle, and the acceleration and force acting on the dummy are measured. As an example, a collision experiment performed under certain speed and angle conditions is referred to as a base experiment. In this embodiment, the experiment result data input unit 11 inputs the experiment result data of the base experiment (FIG. 4: S1).

[0062] Next, the simulation unit 12 creates a virtual object corresponding to the vehicle used in the base experiment using finite element modeling (FEM) with Ansys LS-DYNA (registered trademark) as an example of CAE analysis software. The simulation unit 12 also performs a finite element simulation corresponding to the above-mentioned crash experiment. As an example, a simulation performed under the same speed and angle conditions as the base experiment is referred to as a base calculation. The simulation unit 12 performs the base calculation (FIG. 4: S2).

[0063] As a result of the above simulation, shape and behavior data representing the behavior of the virtual object and a group of simulation curves representing changes in predetermined locations of the virtual object, including acceleration output or displacement output, etc., at the predetermined locations, are generated as simulation results.

[0064] The mode decomposition unit 13 performs mode decomposition on the shape / behavior data and the group of simulation curves created by the simulation unit 12 as described above ( S3 in FIG. 4 ). The mode decomposition is performed based on the shape / behavior data and the group of simulation curves from the base calculation, as well as on past shape / behavior data and the group of past simulation curves from past simulations stored in the simulation database 121. Alternatively, as described above, the mode decomposition may be performed based on the shape / behavior data and the group of simulation curves created in a different base calculation by performing a different base experiment under different speed and angle conditions from a certain base experiment, and then performing a different base calculation corresponding to this different base experiment. Furthermore, the mode decomposition may be performed based on the simulation of one or more variation cases under different speed and angle conditions from the base experiment.

[0065] Next, the difference pattern vector extraction unit 16 compares a group of experimental result curves, which are included in the experimental result data of the base experiment and represent changes at predetermined locations of the vehicle, including acceleration output or displacement output at the predetermined locations, obtained through a crash experiment, with a group of simulation curves obtained through a simulation of the base calculation, and extracts a difference pattern vector ( FIG. 4 : S4). The reason for extracting the difference pattern vector is to lower the resolution to make it easier to find similar patterns, since the curves have very complex shapes. For example, it is conceivable to divide a phenomenon over a time period of 150 ms into 30 ms intervals, integrate the differences in the curves, and then divide by the unit time of 30 ms to obtain an average. As shown in FIG. 13 (described later), the intention is to lower the resolution and display the curves as a sequence, since the curves have very complex shapes. This is an example of a "difference pattern vector."

[0066] Next, the correlation calculation unit 14 calculates the correlation between the simulation curve group and shape / behavior data mode-decomposed by the mode decomposition unit 13 in step S4 and the parameters of the virtual object including plate thickness and the like (FIG. 4: S5).

[0067] In this embodiment, the correlation is calculated using, for example, DIFCRASH, a software developed by SIDACT GmbH of Germany. DIFCRASH is software that visualizes the degree of variation based on statistical methods from multiple calculation results obtained by a finite element method solver such as Ansys LS-DYNA (registered trademark) (DIFFCRASH reference (1): R. Brown, M. Bloomfield, C. Thole, L. Nikitina, “Analysis of the Scatter of a Deploying Airbag,” 12th International LS-DYNA Users Conference, 2012; DIFCRASH reference (2): D. Borsotto et al.: “Improving robustness of Chevrolet Silverado with exemplary design adaptations based on identified scatter sources,” 10th European LS-DYNA Conference, 2015).

[0068] By using DIFFCRASH, the following can be achieved: - Detecting major events and noteworthy points - Reducing the dimension of behavior when an event occurs using mode decomposition - Connecting input variables, intermediate variables (behavior modes), and output variables in chronological order to visualize their relationships

[0069] In the following, a specific example of out-of-position of a vehicle driver will be described. In this example, the friction between the airbag and the dummy and the position of the dummy are used as parameters.

[0070] (STEP 1) First, based on the shape and behavior data, the main events and the areas to be focused on are extracted. For example, the behavior of the dummy head is taken as the main event and the areas to be focused on.

[0071] (STEP 2) Next, a modal analysis of the dummy head behavior at 100 ms (final state), which is the main event, is performed using principal component analysis. This process reduces the dimensions of each calculation result to a two-dimensional or three-dimensional position vector. Figures 5(A) and 5(B) show the results of principal component analysis using DIFFCRASH on the node group of the head part at 100 ms.

[0072] The eigenvalue ranking in Figure 5(A) shows that in this example, the first mode is dominant compared to the second and subsequent modes. Each point in Figure 5(B) represents a single simulation result, and the coordinates of the first and second mode axes represent the contribution ratio, which indicates whether the behavior in each result can be expressed as a superposition of modes. Figure 5(B) shows that in this example, two large clusters are observed on the left and right of the first mode coordinate, indicating that bifurcation has occurred in the behavior.

[0073] The behavior of each cluster is shown in Figure 6. The cluster on the negative side of the primary mode axis is a mode in which the dummy is pushed away by the airbag (push-away mode), and the cluster on the positive side is a mode in which the dummy's head is pressed down by the airbag (push-down mode).

[0074] (STEP 3) Next, a regression analysis is performed on the input parameters and modes. Fluctuations in the parameters affect the behavior of the main parts for each event. This is expressed as a mode contribution rate in the graph, and by performing a regression analysis, it becomes possible to evaluate which parameters have the greatest impact on the behavior. First, the relationship between the input parameters and the mode contribution rate is analyzed.

[0075] 7A and 7B show the correlation matrices between the input parameters and the modal contributions. It can be seen from 7A and 7B that the position (DX_DMY) and friction (MU_PD) have a strong correlation only with the first mode.

[0076] In this example, the influence of input variations, such as the dummy position and the friction coefficient between the dummy and the airbag, on coordinates in mode space was determined by sensitivity analysis using regression analysis. The results of the regression analysis showed that when the dummy position was low or the friction coefficient was low, the mode on the left side of Figure 8 occurred stably. However, when the dummy moved even slightly above the zero point (base calculation) or the friction coefficient increased from the initial friction of 0.3, the mode on the right side also began to occur, and the results began to diverge. Mode bifurcation indicates increased uncertainty in the results that will occur during use. This means increased design costs to address this divergence, so the design must be revised to stabilize the behavior.

[0077] In this way, it is possible to visualize the relationship between the main behavior modes and the input parameters that affect them. This method makes it possible to quantify and evaluate the differences in results for intuitive expressions such as head behavior.

[0078] Next, the coefficient calculation unit 15 compares a group of experimental result curves, which are included in the base experiment and which represent changes in the predetermined locations of the vehicle, including acceleration output or displacement output, etc., obtained by the collision experiment, with a group of simulation curves of the base calculation. Then, the coefficient calculation unit 15 calculates coefficients that fit the group of simulation curves obtained by the simulation of the base calculation to the group of experimental result curves of the base experiment (FIG. 4: S6).

[0079] As shown in Fig. 9(A), a difference occurs between the simulation curve obtained by the simulation and the curve obtained from the experiment. Therefore, the mode decomposition unit 13 decomposes the simulation curve obtained by the simulation into modes as shown in Fig. 9(B). Then, the coefficient calculation unit 15 multiplies the decomposed modes by coefficients and adds the adjusted modes as shown in Fig. 9(C), thereby fitting the simulation curve obtained by the simulation to the curve obtained from the experiment.

[0080] Next, the reconstruction unit 17 reconstructs a group of simulation curves based on the coefficients calculated as described above, and also reconstructs hypothetical shape and behavior data that are expected to have occurred in the base experiment based on the coefficients calculated as described above (FIG. 4: S6).

[0081] The reconstruction is performed by multiplying the decomposed mode vectors by coefficients obtained by optimization calculations and then adding the results together. One possible example is the reconstruction technology (Reference: European Patent No. 2924585) by SIDAC GmbH of Germany, shown in Figure 10. This makes it possible to reconstruct the vehicle behavior that cannot be measured in experiments as shape and behavior data.

[0082] Next, the force output unit 19 compares the shape and behavior data of the base calculation with the virtual shape and behavior data corresponding to the base experiment reconstructed as described above, and calculates the force required to correct the difference (FIG. 4: S7).The force output unit 19 also compares the shape and behavior data obtained from the calculation with the virtual shape and behavior data corresponding to the experiment reconstructed as described above, and calculates the force required to correct the difference (FIG. 4: S7).

[0083] This force output processing is performed by calculating the force required to bridge the difference between the virtual shape and behavior data corresponding to the reconstructed base experiment and the shape and behavior data of the base calculation, and visualizing it as a contour (Figure 11). One possible application is the Grey-Box-Processing technology of the Fraunhofer EMI Institute in Germany (Reference 1: Grey-Box-Processing: a novel validation method for use in vehicle safety applications, Thomas Soot1, Michael Dlugosch, Jens Fritsch, Noriyo Ichinose, Stefan Hiermaier, Fabian Duddeck, Published: 18 April 2022, volume 39, pages 2677-2698 (2023) ; Reference 2: JP 2023-524694 A). Gray-Box-Processing technology uses data assimilation to calculate the force required to bridge the difference between the behavior of an experiment and a simulation, and visualizes the force as a contour. In this embodiment, this technology is applied to calculate the force required to bridge the difference between the behavior of the virtual shape and behavior data corresponding to the base experiment reconstructed as described above and the shape and behavior data of the base calculation, and visualize the force as a contour. By performing this processing, it is possible to provide the user with hints on measures required to bring the simulation results closer to the experimental results in areas where they deviate.

[0084] <Function for Proposing Countermeasures to Correct Deviations Between Experimental and Simulation Results> Next, the function for proposing countermeasures to correct deviations between experimental and simulation results in this embodiment will be described with reference to Fig. 4 and Figs. 12 to 15. In this embodiment, as shown in Fig. 4, by referring to past data stored in the knowledge database 122 (Fig. 4: S9), cases with high similarity in terms of differences between the current experimental results and the simulation are extracted, and countermeasures for those cases are further extracted (Fig. 4: S10).

[0085] Details of step S10 in Fig. 4 are shown in Fig. 12. First, the deviation pattern similarity determination unit 21 determines the similarity of the deviation patterns of the curves based on the difference pattern vector extracted in step S5 in Fig. 4 (Fig. 12: S21).

[0086] In this way, to determine the similarity of the deviation patterns of the curves, as shown in Fig. 13, the deviation patterns of the experiment and simulation are quantified, and the deviation pattern in the current calculation results (Fig. 12: S20) is compared to determine whether it is similar to the deviation pattern stored in the experiment result database 123 of the knowledge database 122. Specifically, the deviation pattern is extracted as a difference pattern vector, and the similarity is evaluated by a comparison method such as an inner product with a difference pattern vector extracted from the same part. This is based on the idea that if the deviations are similar in the same output channel, the factors of the deviations are likely to be similar as well.

[0087] Next, the deformation mode similarity determination unit 22 vectorizes the information of the nodal coordinate group in the modal space in the shape and behavior data from the simulation based on the current calculation result (S20 in FIG. 12) and data from the experiment result database 123 in the knowledge database 122. Then, as shown in FIG. 14, the deformation mode similarity determination unit 22 calculates the distance in the modal space by decomposing the vector into modes. In this way, the deformation mode similarity determination unit 22 determines the similarity of the deformation modes based on whether the modal coordinate values ​​are similar at a predetermined timing (S21 in FIG. 12). This is based on the idea that if the modal coordinate values ​​are similar at the same timing, the behaviors will also be similar.

[0088] Next, the deviation visualization unit 23 vectorizes the information on the nodal coordinates in the modal space of a specific component in the simulated shape and behavior data based on the current calculation results ( S20 in FIG. 12 ) and data from the experimental result database 123 in the knowledge database 122. The deviation visualization unit 23 then calculates the distance or direction in the modal space by decomposing the vectors into modes, as shown in FIG. 15 . In other words, the concept of vectorizing the information on the nodal coordinates in the shape calculated by the finite element method and decomposing the vectors into modes can be applied not only to the entire shape, i.e., the entire product (e.g., when considering the entire automobile body as one unit), but also to parts (e.g., narrowing down to a specific component or assembly (e.g., a bumper)). Even if the entire vehicle body appears to be moving in the same way, certain parts may be moving differently. In such cases, local differences may be overlooked by simply looking at the entire body. However, evaluating each part individually can prevent such oversights. In this way, the deviation visualization unit 23 visualizes the deviation of the behavior of the specified component as a contour based on whether the modal coordinate values ​​are close at a specified timing ( S21 in FIG. 12 ). Increasing the resolution in this way enables similarity searches, such as searches for similar behaviors of certain parts. The visualization of deviation is not limited to the contour method described above. Deviations may also be expressed as scalar values ​​or vector values. For example, scalar values ​​may be displayed using a color bar (e.g., a blue color bar indicates a low scalar value, and a red color bar indicates a high scalar value). Furthermore, vector values ​​may also be displayed using arrows.

[0089] When the search for the curve deviation pattern, the similarity of the deformation mode, and the portion with large deviation (S21 in FIG. 12) is completed in this manner, a comparison is made with past problem cases in the troubleshooting database 124 of the knowledge database 122 based on the curve deviation pattern, the similarity of the deformation mode, and the portion with large deviation, and an evaluation is performed (S22 in FIG. 12). Specifically, information close to the result obtained in step S21 is searched for among the information on past curve deviation patterns, the similarity of the deformation mode, and the portion with large deviation in the troubleshooting database 124.

[0090] Then, based on the output of at least one of the deviation pattern similarity determination unit 21, the deformation mode similarity determination unit 22, and the deviation visualization unit 23, the proposal unit 24 extracts and lists countermeasures for cases with high similarity between the current experimental results and the simulation based on the data in the troubleshooting database 124 of the knowledge database 122 ( FIG. 12 : S23). For example, the troubleshooting database may store reports of past countermeasures taken in association with the difference pattern vectors at that time, and may combine a pattern recognition AI trained to determine whether the associated pattern is similar to the current pattern, with a summarization AI and proposal AI that summarizes the associated report and presents it to the user when it is determined to be similar. The proposal unit 24 also outputs documents of successful countermeasures and documents of unsuccessful countermeasures in order of effectiveness based on past reports and comments stored in the troubleshooting database 124 ( FIG. 12 : S24).

[0091] Then, based on the correlation of the parameters calculated in step S6 of FIG. 4, and the documents of the cases where the countermeasures were successful and the documents of the cases where the countermeasures were unsuccessful, the suggestion unit 24 proposes a list of candidates for how to change the parameters to increase the specified evaluation score (ISO 18571) (FIGS. 4 and 12: S8).

[0092] As described above, according to this embodiment, a candidate list of parameter changes that will increase a predetermined evaluation score (ISO 18571) is presented to the user. Therefore, the user can easily identify the locations and parameters that are causing the discrepancy between the experimental results and the simulation results, and can improve the accuracy of the simulation model while using data from past experimental and simulation results.

[0093] The user then makes changes based on the suggestions above, and sends back results and reports indicating whether the discrepancy between the experimental results and the simulation results was reduced or not to the simulation database 121 and knowledge database 122 (FIG. 4: S11, S12). By doing this, more data is accumulated, and the accuracy of the simulation model can be improved.

[0094] Second Embodiment Next, a second embodiment of the present invention will be described with reference to FIG.

[0095] In the first embodiment, the proposing unit 24 proposes a candidate list of how to change parameters to increase the score in ISO 18571, and the user makes decisions and implements the proposals.

[0096] However, in this embodiment, an automatic execution unit is provided, and as shown in Fig. 16, the proposed measures are automatically executed by the automatic execution unit (Fig. 16: S14). The automatic execution unit then presents the obtained results to the user (Fig. 16: S15). In this case, the user is responsible for determining the results.

[0097] As described above, according to this embodiment, not only is a candidate list of parameter changes that will increase a predetermined evaluation score (CORA) proposed, but the automatic execution unit also executes the countermeasures. Therefore, the user can more easily identify the parts and parameters that are causing the discrepancy between the experimental results and the simulation results, and can improve the accuracy of the simulation model while using data from past experimental and simulation results.

[0098] (Modifications) The above-described embodiment is merely illustrative, and various modifications are possible without departing from the scope of the present invention. In the above-described embodiment, an example using one base experiment and one base calculation was described, but the present invention is not limited to such an example. For example, a second base experiment with different speed and angle conditions from the first base experiment and a second base calculation corresponding to this second base experiment may be added (FIG. 4: S1, S2). Such information may be obtained from a simulation unit or, if necessary, from past data in the simulation database. Furthermore, simulation and mode decomposition may be performed by adding multiple variation cases with different speed and angle conditions (FIG. 4: S3).

[0099] In the above-described embodiment, the correlation calculation unit 14 has been described as an essential component, but the present invention is not limited to this configuration. The simulation model accuracy improvement support system 100 described above may be configured without the correlation calculation unit 14. However, including the correlation calculation unit 14 can improve the persuasiveness and accuracy of the above-described proposals.

[0100] In the above-described embodiment, the mode decomposition unit 13, correlation calculation unit 14, coefficient calculation unit 15, and reconstruction unit 17 are described as applying technologies from SIDACT, but the present invention is not limited to this. Technologies other than those from SIDACT may be used as long as the functions of the above-described units are realized.

[0101] In the above-described embodiment, the force output unit 19 applies technology from Fraunhofer to compare the shape and behavior data of the base calculation with the virtual shape and behavior data corresponding to the reconstructed base experiment, and calculates the force required to correct the difference. However, the present invention is not limited to this configuration. For example, after sufficient data has been accumulated, the required force can be estimated by searching the knowledge database 122 using deviation patterns in modal space as a key. This is because different deviation patterns may lead to similar patterns of insufficient or excessive force, and this information can be mapped to the target shape. Another example of force output is a simple method of calculating the required force using Newtonian mechanics (F = m * a) based on the difference in the center of gravity positions and weights of parts in each calculation.

[0102] In the above-described embodiment, an example of the similarity determination unit has been described in which the deviation pattern similarity determination unit 21 and the deformation mode similarity determination unit 22 are provided. However, the present invention is not limited to this configuration. The similarity determination unit may be configured to include either the deviation pattern similarity determination unit 21 or the deformation mode similarity determination unit 22. Furthermore, the thresholds and weightings used when determining the similarity of past cases using deviation patterns of shapes and curves may be changed by a system administrator or user as needed. Furthermore, the determination of similarity may be based on strict logic or may be based on heuristics such as deep learning.

[0103] In the above-described embodiment, the case where both the deviation visualization unit 23 and the proposal unit 24 are provided has been described. However, the present invention is not limited to this aspect. When at least the proposal unit 24 is provided as described above, the user can improve the accuracy of the simulation model based on the proposal by the proposal unit 24. However, when the deviation visualization unit 23 is provided, after viewing the proposal by the proposal unit 24, the user can also check the information from the deviation visualization unit 23 to verify its authenticity. Alternatively, some users may be able to solve the problem using only the information visualized by the deviation visualization unit 23. Furthermore, during each of the above-described processes, the user may be able to visually check the results of each process upon request. Furthermore, the user may be able to interrupt each of the above-described processes and perform recalculation, change the data used, or the like at their discretion.

[0104] In the above-described embodiment, a case where finite element simulation is used to evaluate the crashworthiness of a vehicle has been described as an example, but the present invention is not limited to such a case. For example, the present invention can be applied to any evaluation that is performed using simulation, such as material property evaluation, structural evaluation, seismic resistance evaluation, and electromagnetic property evaluation.

[0105] Furthermore, in the above-described embodiment, a case has been described in which vehicles are used as real objects and virtual objects, but the present invention is not limited to this, and it is also possible to use airplanes, ships, or trains as real objects and virtual objects.

[0106] The simulation model accuracy improvement support program according to the above aspect can be provided in a form stored on a computer-readable recording medium and installed on a computer. The recording medium is, for example, a non-transitory recording medium, and is exemplified by an optical recording medium such as a CD-ROM, but can also include any known type of recording medium, such as a semiconductor recording medium or a magnetic recording medium. It is also possible to provide the above-mentioned program in the form of distribution via a communication network and install it on a computer.

[0107] The above describes the simulation model accuracy improvement support system, simulation model accuracy improvement support method, and simulation model accuracy improvement support system program according to embodiments of the present invention. However, the present invention is not limited to this, and various modifications are possible within the scope of the gist of the present invention.

[0108] DESCRIPTION OF SYMBOLS 10 Control unit 11 Experiment result data input unit 12 Simulation unit 13 Mode decomposition unit 14 Correlation calculation unit 15 Coefficient calculation unit 16 Difference pattern vector extraction unit 17 Reconstruction unit 19 Force output unit 20 Server registration unit 21 Deviation pattern similarity determination unit 22 Deformation mode similarity determination unit 23 Deviation visualization unit 24 Proposal unit 100 Simulation model accuracy improvement support system 110 Simulation model accuracy improvement support device 121 Simulation database 122 Knowledge database 123 Experiment result database 124 Troubleshooting database

Claims

1. An experiment result data input unit for inputting experiment result data in a collision experiment of a real object such as a vehicle; a simulation unit for simulating a collision experiment for the collision experiment for a virtual object corresponding to the real object by using CAE analysis software based on the finite element method; a mode decomposition unit for mode decomposing the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data representing the shape / behavior of the virtual object and the simulation curve group representing changes in the specified locations of the virtual object including acceleration output or displacement output or the like; a coefficient calculation unit for comparing the experiment result curve group representing changes in the specified locations of the real object including acceleration output or displacement output or the like obtained by the collision experiment, which is included in the experiment result data, with the simulation curve group, and calculating coefficients for fitting the simulation curve group to the experiment result curve group; a reconstruction unit for reconstructing the simulation curve group based on the calculated coefficients, and reconstructing virtual shape / behavior data expected to have occurred in the collision experiment based on the calculated coefficients; a force output unit that compares the shape / behavior data created by the simulation unit with the shape / behavior data reconstructed by the reconstruction unit, and calculates or estimates a force required to correct a difference.

2. The system for supporting improvement of accuracy of a simulation model according to claim 1, further comprising a correlation calculation unit for calculating the correlation between the group of mode-decomposed simulation curves and shape / behavior data and parameters of the virtual object including plate thickness, etc.

3. An experiment result data input unit for inputting experiment result data in a collision experiment of a real object such as a vehicle; a simulation unit for simulating a collision experiment for the collision experiment for a virtual object corresponding to the real object by using CAE analysis software based on the finite element method; a mode decomposition unit for mode decomposing the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data representing the shape / behavior of the virtual object and the simulation curve group representing changes in the specified location including the acceleration output or displacement output of the specified location of the virtual object; a difference pattern vector extraction unit for comparing the experiment result curve group representing changes in the specified location including the acceleration output or displacement output of the specified location of the real object obtained by the collision experiment, which is included in the experiment result data, with the simulation curve group and extracting a difference pattern vector; a coefficient calculation unit for comparing the experiment result curve group with the simulation curve group and calculating coefficients for fitting the simulation curve group to the experiment result curve group; a reconstruction unit which reconstructs the group of simulation curves based on the calculated coefficients, and reconstructs virtual shape and behavior data that is expected to have occurred in the collision experiment based on the calculated coefficients; a force output unit which compares the shape and behavior data created by the simulation unit with the shape and behavior data reconstructed by the reconstruction unit, and calculates or estimates the force required to correct the difference; a knowledge database unit which stores data including past experiment results, past simulation results, past proposal reports, and information regarding differences between past experiment results and simulations; and a similarity judgment unit which judges the similarity regarding the differences between the current experiment results and a simulation based on the experiment result data, the shape and behavior data created by the simulation unit, the shape and behavior data reconstructed by the reconstruction unit, or data from the knowledge database unit.a proposal unit that extracts, based on an output of the similarity determination unit, countermeasures in cases where there is a high degree of similarity between the experimental results and the simulation, based on data in the knowledge database, and proposes how to change parameters of the virtual object, including plate thickness, etc., in order to increase a predetermined evaluation score.

4. A deviation visualization unit which, based on the shape and behavior data created by said simulation unit and the shape and behavior data reconstructed by said reconstruction unit, vectorizes information on a group of nodal coordinates in a mode space of a specified component in said shape and behavior data, calculates a distance or direction in the mode space by decomposing the vector into modes, and visualizes the degree of deviation of the shape and behavior of the specified component from the direction of the difference pattern vector and whether or not the mode coordinate values ​​are close at a specified timing; and said suggestion unit, based on the output of at least one of said similarity judgment unit and said deviation visualization unit, extracts countermeasures in cases where there is a high degree of similarity between the difference between the experimental results and the simulation, based on data in the knowledge database, and proposes how to change the parameters to increase a specified evaluation score.

5. The system for supporting improvement of accuracy of a simulation model as set forth in claim 4, wherein the similarity determination unit comprises: a deviation pattern similarity determination unit which determines a similarity of deviation patterns of curves based on the difference pattern vector; and a deformation mode similarity determination unit which, based on the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit, vectorizes information of a group of nodal coordinates in a mode space in the shape and behavior data, calculates a distance or direction in the mode space by decomposing the vector into modes, and determines a similarity of deformation modes depending on whether or not the mode coordinate values ​​are close at a predetermined timing; and the suggestion unit extracts measures in cases where there is a high similarity regarding the difference between the current experimental result and the simulation based on output of at least one of the deviation pattern similarity determination unit, the deformation mode similarity determination unit, and the deviation visualization unit, based on data in a knowledge database, and proposes how to change the parameters to increase a predetermined evaluation score.

6. A system for supporting improvement of accuracy of a simulation model according to any one of claims 3 to 5, further comprising a correlation calculation unit that calculates the correlation between the group of mode decomposed simulation curves and shape / behavior data and parameters of the virtual object including plate thickness, etc.

7. The simulation model accuracy improvement support system according to any one of claims 3 to 5, further comprising an automatic execution unit that automatically executes the measures proposed by the proposal unit.

8. The system for supporting improvement of accuracy of a simulation model according to claim 6, further comprising an automatic execution unit that automatically executes the measures proposed by said proposal unit.

9. A step of inputting experimental result data of a collision experiment of a real object such as a vehicle by an experimental result data input unit; a step of simulating a collision experiment for the collision experiment by a simulation unit using CAE analysis software based on the finite element method for a virtual object corresponding to the real object; a step of mode decomposing the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data representing the shape / behavior of the virtual object and the simulation curve group representing changes in the predetermined location of the virtual object including the acceleration output or displacement output or the like of the predetermined location, by mode decomposition; a step of comparing the experimental result curve group representing changes in the predetermined location of the real object including the acceleration output or displacement output or the like obtained by the collision experiment and the simulation curve group by a coefficient calculation unit, which is included in the experimental result data, with the simulation curve group, and calculating a coefficient for fitting the simulation curve group to the experimental result curve group; a step of reconstructing the simulation curve group based on the calculated coefficient, and reconstructing virtual shape / behavior data expected to have occurred in the collision experiment based on the calculated coefficient, by a reconstruction unit. a step of comparing, by a force output unit, shape / behavior data created by the simulation unit with shape / behavior data reconstructed by the reconstruction unit, and calculating or estimating a force required to correct a difference.

10. A method for supporting improvement of accuracy of a simulation model as set forth in claim 9, further comprising a step of calculating, by a correlation calculation unit, a correlation between the group of mode decomposed simulation curves and shape / behavior data and parameters of the virtual object including plate thickness, etc.

11. A step of inputting experimental result data of a collision experiment of a real object such as a vehicle by an experimental result data input unit; a step of simulating a collision experiment for the collision experiment by a simulation unit using CAE analysis software based on the finite element method for a virtual object corresponding to the real object; a step of mode decomposing the shape / behavior data and simulation curve group created by the simulation unit, i.e., the shape / behavior data expressing the behavior of the virtual object and the simulation curve group expressing changes in the specified location of the virtual object including the acceleration output or displacement output, etc. of the specified location; a step of comparing the experimental result curve group expressing changes in the specified location of the real object including the acceleration output or displacement output, etc. of the specified location obtained by the collision experiment, which is included in the experimental result data, with the simulation curve group and extracting a difference pattern vector by a difference pattern vector extraction unit; a step of a coefficient calculation unit comparing a group of experimental result curves, which are included in the experimental result data and which represent changes in a predetermined location of the real object, including an acceleration output or a displacement output, etc., of the predetermined location obtained by the collision experiment, with the group of simulation curves, and calculating coefficients for fitting the group of simulation curves to the group of experimental result curves; a step of a reconstruction unit reconstructing the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape / behavior data that is expected to have occurred in the collision test based on the calculated coefficients; a step of a force output unit comparing the shape / behavior data created by the simulation unit with the shape / behavior data reconstructed by the reconstruction unit, and calculating or estimating a force required to correct the difference;a step of determining a similarity regarding a difference between the current experimental result and the simulation by a similarity determination unit based on data from a knowledge database unit that stores the experimental result data, shape / behavior data created by the simulation unit, shape / behavior data reconstructed by the reconstruction unit, or past experimental results, past simulation results, past proposal reports, and data including information regarding a difference between the past experimental results and a simulation, and a step of extracting a measure in a case where the similarity regarding the difference between the current experimental result and the simulation is high based on the data from the knowledge database unit and proposing how to change parameters of the virtual object, including plate thickness, etc., in order to increase a predetermined evaluation score, based on an output of the similarity determination unit by a proposal unit.

12. A method for supporting improvement of accuracy of a simulation model as described in claim 11, comprising a step of: using a deviation visualization unit, based on the shape and behavior data created by said simulation unit and the shape and behavior data reconstructed by said reconstruction unit, vectorizing information of a group of nodal coordinates in a mode space of a specified component in the shape and behavior data, calculating a distance or direction in the mode space by decomposing the vector into modes, and visualizing the degree of deviation of the shape and behavior of the specified component from the direction of the difference pattern vector and whether the mode coordinate values ​​are close at a specified timing; wherein the suggesting step extracts measures in cases where there is a high degree of similarity between the difference between the experimental results and the simulation based on data in the knowledge database, based on the output of at least one of said similarity judgment unit and said deviation visualization unit, and suggests how to change the parameters to increase a specified evaluation score.

13. The method for supporting improvement of accuracy of a simulation model according to claim 12, wherein the step of judging the similarity comprises: a step of judging the similarity of deviation patterns of curves based on the difference pattern vector by a deviation pattern similarity judgment unit provided in the similarity judgment unit; and a step of vectorizing information of a group of nodal coordinates in a mode space in the shape and behavior data in the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit by a deformation mode similarity judgment unit provided in the similarity judgment unit, based on the shape and behavior data created by the simulation unit and the shape and behavior data reconstructed by the reconstruction unit, calculating distances in the mode space by decomposing the vectors into modes, and judging the similarity of deformation modes based on whether or not the mode coordinate values ​​are close at a predetermined timing; and the step of proposing comprises: extracting measures in cases where there is a high similarity regarding the difference between the current experimental result and the simulation based on data in the knowledge database, based on output from at least one of the deviation pattern similarity judgment unit, the deformation mode similarity judgment unit, and the deviation visualization unit, and proposing how to change the parameters to increase the predetermined evaluation score.

14. A method for supporting improvement of accuracy of a simulation model according to any one of claims 11 to 13, comprising a step of calculating, by a correlation calculation unit, a correlation between the group of mode decomposed simulation curves and shape / behavior data and parameters of the virtual object including plate thickness, etc.

15. A method for supporting improvement of accuracy of a simulation model according to any one of claims 11 to 13, further comprising a step of automatically executing the measures proposed by the suggestion unit using an automatic execution unit.

16. The method for supporting improvement of accuracy of a simulation model according to claim 14, further comprising the step of automatically executing the measures proposed by the suggestion unit using an automatic execution unit.

17. A program for supporting improvement of accuracy of a simulation model, comprising the steps of: inputting experimental result data in a collision experiment of a real object such as a vehicle; simulating a collision experiment for the collision experiment with a virtual object corresponding to the real object by CAE analysis software based on the finite element method; performing mode decomposition of shape / behavior data and a group of simulation curves, i.e., shape / behavior data representing the shape / behavior of the virtual object and a group of simulation curves representing changes in the predetermined locations of the virtual object, including acceleration output or displacement output, etc., of the predetermined locations; comparing a group of experimental result curves representing changes in the predetermined locations of the real object, including acceleration output or displacement output, etc., obtained by the collision experiment, contained in the experimental result data, with the group of simulation curves, and calculating coefficients for fitting the group of simulation curves to the group of experimental result curves; reconstructing the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape / behavior data expected to have occurred in the collision experiment based on the calculated coefficients. comparing the shape / behavior data created by the simulation with the reconstructed shape / behavior data, and calculating or estimating a force required to correct the difference.

18. The accuracy improvement support program for a simulation model according to claim 17, wherein the program causes the computer to execute a step of calculating correlations between the group of mode-decomposed simulation curves and shape / behavior data and parameters of the virtual object including plate thickness, etc.

19. A program for supporting improvement of accuracy of a simulation model, comprising the steps of: inputting experimental result data in a collision experiment of a real object such as a vehicle; simulating a collision experiment for the collision experiment with a virtual object corresponding to the real object by CAE analysis software based on the finite element method; performing mode decomposition of shape / behavior data and a group of simulation curves, i.e., shape / behavior data expressing the shape / behavior of the virtual object and a group of simulation curves expressing changes in the predetermined locations of the virtual object, including acceleration output or displacement output, etc., of the predetermined locations; comparing a group of experimental result curves expressing changes in the predetermined locations of the real object, including acceleration output or displacement output, etc., obtained by the collision experiment, contained in the experimental result data, with the group of simulation curves, and extracting a difference pattern vector; comparing a group of experimental result curves expressing changes in the predetermined locations of the real object, including acceleration output or displacement output, etc., obtained by the collision experiment, contained in the experimental result data, with the group of simulation curves, and calculating a coefficient for fitting the group of simulation curves to the group of experimental result curves; a step of reconstructing the group of simulation curves based on the calculated coefficients, and reconstructing virtual shape and behavior data that is expected to have occurred in the collision experiment based on the calculated coefficients; a step of comparing the shape and behavior data created in the simulation step with the reconstructed shape and behavior data, and calculating or estimating the force required to correct the difference; and a step of determining the similarity regarding the difference between the current experiment result and the simulation based on data from a knowledge database unit that stores the experiment result data, the shape and behavior data created in the simulation step, the shape and behavior data reconstructed in the reconstruction step, or data including past experiment results, past simulation results, past proposal reports, and information regarding the difference between past experiment results and simulations.and a step of extracting, based on the output of the step of determining the degree of similarity, measures to be taken in cases where there is a high degree of similarity between the experimental results and the simulation, based on the data in the knowledge database, and proposing how to change the parameters of the virtual object, including plate thickness, etc., in order to increase a predetermined evaluation score.

20. The program for supporting improvement of accuracy of a simulation model according to claim 19, wherein the program causes the computer to execute the steps of: vectorizing information on a group of nodal coordinates in a mode space of a specified component in the shape and behavior data based on the shape and behavior data created in the step of performing the simulation and the shape and behavior data reconstructed in the step of reconstructing, calculating a distance or direction in the mode space by decomposing the vector into modes, and visualizing the degree of deviation of the shape and behavior of the specified component from the direction of the difference pattern vector by determining whether or not the mode coordinate values ​​are close at a specified timing; and as the step of proposing, extracting measures in cases where there is a high degree of similarity between the difference between the current experimental result and the simulation based on data in the knowledge database unit, and proposing how to change the parameters to increase a specified evaluation score, based on at least one output of the step of determining the similarity and the step of visualizing the deviation.

21. The program causes the computer to execute, as the step of determining the degree of similarity, a step of determining a degree of similarity in deviation patterns of curves based on the difference pattern vector; and a step of vectorizing information of a group of nodal coordinates in a mode space in the shape and behavior data based on the shape and behavior data created in the step of performing the simulation and the shape and behavior data reconstructed in the step of reconstructing, calculating a distance or direction in the mode space by decomposing the vector into modes, and determining a degree of similarity in deformation modes based on whether or not the mode coordinate values ​​are close at a predetermined timing; and, as the step of proposing, a step of extracting, based on data in the knowledge database, countermeasures in cases where there is a high degree of similarity regarding the differences between the results of the experiment and the simulation, based on output of at least one of the steps of determining the degree of similarity of the deviation patterns, determining the degree of similarity of the deformation modes, and visualizing the degree of deviation, and proposing how to change the parameters to increase a predetermined evaluation score, as the step of proposing.

22. A program for supporting improvement of accuracy of a simulation model according to any one of claims 19 to 21, wherein the program causes the computer to execute a step of calculating a correlation between the group of mode decomposed simulation curves and shape / behavior data and parameters of the virtual object including plate thickness, etc.

23. A program for supporting improvement of accuracy of a simulation model as claimed in any one of claims 19 to 21, wherein the program further causes a computer to execute a step of automatically executing the measures proposed in the proposing step.

24. The program for supporting improvement of accuracy of a simulation model according to claim 22, further causing a computer to execute a step of automatically executing the measures proposed in the proposing step.

Images