JP2023134063A - Performance prediction system and performance prediction method - Google Patents

Abstract

To provide a performance prediction system for stably and accurately predicting performance while significantly reducing time required for calculating an amount of flexure generated when loading panels for an automobile.SOLUTION: A performance prediction system 1 stores a dataset obtained by associating, for each load point, objective variables indicating an amount of flexure in the load points, with explanatory variables obtained by associating point cloud coordinate data of a point cloud around the load points arranged uniformly on a panel with information on a plate thickness and a material of points constituting the point cloud. A machine learning unit 23 generates, by machine learning, a performance prediction model for predicting the amount of flexure in performance prediction point using the dataset. The performance prediction unit receives input data generated by combining point cloud coordinate data of a point cloud around the performance prediction point with information on a plate thickness and material of points constituting the point cloud, and predicts the amount of flexure in the performance prediction point using the trained performance prediction model.SELECTED DRAWING: Figure 2

Description

The present invention relates to a performance prediction system and a performance prediction method for predicting the amount of deflection when a load is applied to various metal panels such as engine hoods and doors of automobiles.

For example, the engine hood of an automobile needs to be rigid enough to prevent denting or deformation even with slight impact, and is required to have high dent performance. is essential.

Patent Document 1 listed below discloses a dent stiffness prediction method for predicting dent stiffness from information on design drawings of various metal panels such as engine hoods and doors of automobiles. Specifically, when a load is applied to the measurement point, the area of the stress affected area defined around the measurement point is calculated as the deflection area, and the curvature of the measurement point of the plate member and the plate thickness of the measurement point of the plate member are calculated. A predetermined relational expression is determined by multiple regression analysis, etc., using the material of the plate member and the deflection area as factors representing the rigidity of the plate member. Then, by extracting the curvature of the measurement point of the plate member, the plate thickness of the plate member at the measurement point, the material of the plate member, and the deflection area from the design drawing, the displacement of the measurement point in the load direction is calculated based on a predetermined relational expression. (dent rigidity) should be appropriately calculated at the design stage.

In this way, the dent rigidity is determined in the first half of development, thereby preventing delays in development due to delays in evaluating dent rigidity.

JP2007-33067A (Patent No. 4568186)

In the dent stiffness prediction method described in Patent Document 1, in deriving the correlation equation (predetermined relational equation) and calculating the amount of deflection (dent stiffness), when a load is applied to the measurement point, the area around the measurement point is defined. There was a problem in that it was necessary to measure the curvature, plate thickness, material, and area of the stress-affected region each time, and this work took a lot of time. Further, there is a problem in that these measurement results vary depending on the operator, and this variation affects the accuracy of the predicted displacement (dent rigidity) of the measurement point.

The present invention was made in view of the above-mentioned problems, and it significantly shortens the time required to calculate the amount of deflection when a load is applied to various automotive panels, and also stabilizes the dent performance. The present invention also aims to provide a performance prediction system and a performance prediction method that can be predicted with high accuracy.

The performance prediction system according to the present invention predicts the performance of metal panels that constitute the body of an automobile. For example, data that associates the point cloud coordinate data of a point group around load points uniformly provided on the panel with information on the material and plate thickness of each point constituting the point group can be used as an explanatory variable, and each load A storage means is provided for storing a data set in which an explanatory variable at a point and an objective variable, which is the amount of deflection when a constant load is applied to the load point, are linked for each load point. Then, the model generation means uses the data set read from the storage means to machine-learn a performance prediction model that predicts the amount of deflection when a constant load is applied to a performance prediction point that is an arbitrary location on the panel. The performance prediction means receives input data in which the point cloud coordinate data of the point cloud around the performance prediction point is associated with information on the material and plate thickness of each point constituting the point cloud, and calculates the learned performance. It is characterized by using a prediction model to predict the amount of deflection when a constant load is applied to a performance prediction point.

Further, in the performance prediction system according to the present invention, the storage means includes 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and information on the material and thickness of the panel obtained by a known method. The amount of deflection at each load point is stored in advance. Then, the point cloud coordinate data generation means arranges the panel on a 3D coordinate space (X coordinate, Y coordinate, Z coordinate) based on the 3D design information of the panel, and places the load point or the performance prediction point as the center. Obtain multiple cross-sectional shapes of the panel, divide the obtained cross-sectional shapes into squares of a predetermined angle, generate a point group that is a collection of center points, and calculate the coordinates (X coordinates, Y coordinates, and Z coordinates) to generate the point group coordinate data.

As a result, the data set generating means includes the point cloud coordinate data generated by the point cloud coordinate data generating means, the 3D design information of the panel, and the material and plate thickness of the panel associated with the 3D design information. The data set for model generation can be generated based on the information and the amount of deflection at each load point obtained by the known method. Further, the input data generation means includes the point cloud coordinate data generated by the point cloud coordinate data generation means, 3D design information of the panel, and information on the material and thickness of the panel associated with the 3D design information. The input data for performance prediction can be generated based on the above.

Therefore, according to the performance prediction system of the present invention, by introducing a performance prediction model, the time required to calculate the amount of deflection when a load is applied to an arbitrary position (performance prediction point) of various automotive panels can be significantly reduced. Can be shortened. Furthermore, by using the performance prediction model, variations between workers are eliminated, so dent performance can be predicted stably and with high accuracy.

Further, in the performance prediction system according to the present invention, the result of predicting the amount of deflection when a constant load is applied to the performance prediction point is displayed, and the panel arranged on the 3D coordinate space is displayed in a 3D image. It is desirable to further include display means for displaying.

Further, the performance prediction method according to the present invention is a performance prediction method for predicting the performance of a metal panel constituting the body of an automobile. The explanatory variable is data that associates the point group coordinate data with information on the material and plate thickness of each point that makes up the point group, and the explanatory variable at each load point and the deflection when a constant load is applied to the load point. a data set storage step of storing in memory a data set in which the objective variable, which is the amount of A model generation step that uses machine learning to generate a performance prediction model that predicts the amount of deflection when a certain load is applied, and the point cloud coordinate data of the point cloud around the performance prediction point and the material of each point that makes up the point cloud. and a performance prediction step of accepting input data associated with board thickness information and predicting the amount of deflection when a certain load is applied to a performance prediction point using a learned performance prediction model, and the result of the prediction. and a display step of displaying on a display.

Further, in the performance prediction method according to the present invention, the memory stores 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and information on each panel obtained by a known method. The amount of deflection at the load point is stored in advance.

As pre-processing for generating the performance prediction model, a shape arrangement step of arranging the panel in a 3D coordinate space (X coordinate, Y coordinate, Z coordinate) based on 3D design information of the panel; a cross-sectional shape acquisition step of acquiring a plurality of cross-sectional shapes of the panel centered at a point; and a point cloud generation step of dividing the obtained cross-sectional shapes into squares of a predetermined angle to generate a point cloud that is a collection of center points. step, a point group coordinate data generation step of acquiring the coordinates (X coordinate, Y coordinate, Z coordinate) of each point forming the point group to generate point group coordinate data, and a point group coordinate data generation step of generating point group coordinate data. 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and the amount of deflection at each load point obtained by the known method. based on the data set generation step of generating the data set for model generation.

Furthermore, as pre-processing for the performance prediction step, the shape arrangement step, the cross-sectional shape acquisition step of acquiring a plurality of cross-sectional shapes of the panel centered at the performance prediction point, the point cloud generation step, and the point a group coordinate data generation step, the point group coordinate data generated in the point group coordinate data generation step, 3D design information of the panel, and information on the material and plate thickness of the panel associated with the 3D design information; and an input data generation step of generating the input data for performance prediction based on the above.

Therefore, according to the performance prediction method of the present invention, by introducing a performance prediction model, the time required to calculate the amount of deflection when a load is applied to an arbitrary position (performance prediction point) of various automotive panels can be significantly reduced. Can be shortened. Furthermore, by using the performance prediction model, variations between workers are eliminated, so dent performance can be predicted stably and with high accuracy.

Further, the performance prediction method according to the present invention preferably further includes a display step of displaying the panel arranged on the 3D coordinate space as a 3D image on a display.

According to the performance prediction system and performance prediction method according to the present invention, it is possible to significantly shorten the time required to calculate the amount of deflection when a load is applied to various automotive panels. Furthermore, dent performance can be predicted stably and with high accuracy.

FIG. 1 is a diagram showing an example of the hardware configuration of a computer that operates as a performance prediction system according to the present invention. FIG. 2 is a functional block diagram showing the functions of a control unit that performs learning model generation processing. FIG. 3 is a flowchart illustrating an example of learning model generation processing. FIG. 4 is a diagram showing an example of a 3D image of an engine hood. FIG. 5 is a diagram showing the acquisition position of the cross-sectional shape. FIG. 6 is a diagram showing an example of the cross-sectional shape of an engine hood and an image of the cross-sectional shape formed into a point group. FIG. 7 is a diagram showing an example of point group coordinate data. FIG. 8 is a diagram illustrating an example of a data set for generating a performance prediction model. FIG. 9 is a diagram showing the evaluation results of the model based on machine learning. FIG. 10 is a diagram showing prediction results by a trained performance prediction model using test data. FIG. 11 is a functional block diagram showing the functions of a control unit that performs performance prediction processing. FIG. 12 is a flowchart illustrating an example of performance prediction processing.

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a performance prediction system and a performance prediction method according to the present invention will be described in detail based on the drawings. Note that the present invention is not limited to this embodiment. Further, in the specification and drawings of the present application, elements that can be explained in the same manner may be designated by the same reference numerals, thereby omitting repeated explanation.

<System configuration>
FIG. 1 is a diagram showing an example of the hardware configuration of a computer that operates as a performance prediction system according to the present invention. The performance prediction system 1 of the present embodiment includes a process for predicting the amount of deflection (dent performance) when a load is applied to various metal panels for automobiles (hereinafter referred to as performance prediction process), and a process for predicting the amount of deflection. It operates as a host computer that performs processing to generate a learning model to be executed (hereinafter referred to as learning model generation processing).

In FIG. 1, a performance prediction system 1 includes a control unit 11 composed of a CPU (Central Processing Unit), an FPGA (Field Programmable Gate Array), etc., and various types of memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory). A storage section 12 including a memory, an input section 13 including a user interface such as a keyboard and a mouse, an interface (I/F) section 14 that performs input/output processing such as printing and scanning, and a display section 15 that is a display. It includes a communication section 16 that communicates with the outside via a predetermined network. Although FIG. 1 includes the input unit 13 including a user interface such as a keyboard and a mouse, the performance prediction system 1 of this embodiment is not limited to this, and the display unit 15 may have a touch panel function. By having the input section 13, a configuration may be provided in which the input section 13 is not provided, or a configuration in which the input section 13 is used together.

In FIG. 1, in order to realize performance prediction processing and learning model generation processing by the performance prediction system 1 of the present embodiment, the control unit 11 predicts, for example, the amount of deflection when a load is applied to various automobile panels. and a learning model generation program for generating a learning model for predicting the amount of deflection. The storage unit 12 stores programs (performance prediction program, learning model generation program) related to the performance prediction process and learning model generation process of this embodiment, and various information (3D shape data indicating the shapes of various panels, materials, plate thicknesses, and performance). , data sets for machine learning, etc.) and various data obtained in the process (point group coordinate data, etc.). The control unit 11 executes the performance prediction process and learning model generation process of this embodiment by reading various programs stored in the storage unit 12. Note that the storage unit 12 is not limited to an internal memory, and may be an external storage medium such as a DVD (Digital Versatile Disc) or an SD memory, or may include an internal memory and an external storage medium (such as a DVD or an SD memory). (memory, etc.). Furthermore, for convenience of explanation, the hardware configuration of the performance prediction system 1 of the present embodiment is a list of configurations related to the performance prediction process and the learning model generation process of the present embodiment. It does not represent all the functions of.

Furthermore, although the performance prediction system 1 of this embodiment assumes a general-purpose PC such as a desktop computer or a notebook computer, it is not limited to these, and may also be a mobile terminal such as a smartphone or a tablet terminal. .

<Learning model generation process>
Next, before explaining the performance prediction process in the performance prediction system 1 of this embodiment, the learning model generation process that is the premise will be explained in detail. FIG. 2 is a functional block diagram showing the functions of the control unit 11 that performs the learning model generation process, and FIG. 3 is a flowchart showing an example of the learning model generation process.

In FIG. 2, the control unit 11 includes a point cloud coordinate data generation unit 21, a data set generation unit 22, a machine learning unit 23 that generates a learning model using a machine learning algorithm, and a known CAD (computer-aided design) system 24 and a CAE (Computer Aided Engineering) system 25. In this embodiment, a case will be described in which a learning model is generated using 3D design information of an automobile engine hood as an example of various metal panels. That is, in this embodiment, as an example of the learning model, a performance prediction model is generated that predicts the amount of deflection when a predetermined load (for example, a constant load) is applied to an arbitrary position on the engine hood of an automobile. . This performance prediction model is generated using a known machine learning algorithm such as a neural network (CNN: Convolution Neural Network, RNN: Recurrent Neural Network, etc.).

Further, it is assumed that the engine hood of an automobile is designed by, for example, a CAD system 24 shown in FIG. 2, and 3D design information including 3D shape data is stored in the storage unit 12 in advance. Further, information such as the material and plate thickness of each component constituting the engine hood is also stored in advance in the storage unit 12 in a state associated with 3D design information (3D shape data).

In addition, in this embodiment, when a constant load is applied to predetermined positions (load points: for example, 143 locations) on the engine hood, the amount of deflection at each load point is obtained in advance, and the amount of deflection is calculated for each load point. It is stored in advance in the storage unit 12 in association with the load point. Specifically, for example, the CAE system 25 shown in FIG. 2 reads 3D design information of the engine hood from the storage unit 12, performs CAE analysis using the finite element method based on this 3D design information, and as the analysis result, , obtain the amount of deflection (performance information) at each load point. In CAE analysis using the finite element method, the 3D shape of the engine hood is divided into meshes, and a load is applied to the load points to analyze the deformation of each portion of the mesh.

Note that the 3D design by the CAD system 24 and the CAE analysis by the CAE system 25 are performed using known general methods, so detailed explanations will be omitted. Furthermore, in this embodiment, the amount of deflection at each load point is obtained in advance by performing CAE analysis using the finite element method based on the 3D design information, but the present invention is not limited to this. Any method may be used to obtain the amount of deflection at the load point. Further, in this embodiment, as an example, the number of load points to which loads are applied is 143, but the number is not limited to this, and depending on the accuracy of performance prediction required, for example, 143 or more points can be applied. It is desirable that In other words, the load points to which the load is applied need only be provided uniformly (uniformly distributed) over the entire panel. It can be set as appropriate depending on the range, required accuracy of performance prediction, etc.

Further, in FIG. 2, the CAD system 24 and the CAE system 25 are included in the control unit 11 as functional blocks of the control unit 11 of the performance prediction system 1, but the configuration is not limited to this, and for example, , the CAD system 24 and the CAE system 25 may be configured separately and connected to the performance prediction system 1 via a network.

In the present embodiment, on the premise that various types of information are stored in advance in the storage unit 12 as described above, the control unit 11 stores 3D design information of an engine hood of an automobile, for example, as shown in FIG. , a process of generating a performance prediction model (learning model generation process) is executed using information on the material and thickness of the engine hood and the amount of deflection at each load point acquired in advance.

In FIG. 3, first, when generation of a performance prediction model is instructed by the operation of the input unit 13 by the operator, the point cloud coordinate data generation unit 21 of the control unit 11 receives the 3D design information of the engine hood from the storage unit 12. The engine hood (shape) is read out and arranged on the 3D coordinate space (X coordinate, Y coordinate, Z coordinate) based on the 3D shape data obtained from the 3D design information (step S1). At this time, a 3D image of the engine hood is displayed on the display unit 15. FIG. 4 is a diagram showing an example of a 3D image of an engine hood.

Next, the point group coordinate data generation unit 21 acquires the cross-sectional shape of the engine hood around the load points (143 locations) from the 3D design information (step S2). Specifically, for example, a cross-sectional shape of the engine hood with a width of 150 mm in the front, rear, left and right directions centered around the load point is obtained. FIG. 5 is a diagram showing the acquisition position of the cross-sectional shape. In this embodiment, the cross-sectional shape of the engine hood has a width of 150 mm in the left-right direction centered at the load point (see the AA' straight line (parallel to the X-axis) centered at the load point C shown in FIG. 5). , the cross-sectional shape of the engine hood with a width of 150 mm in the longitudinal direction centered on the load point (see the BB' straight line (parallel to the Y axis) centered on the load point C shown in Fig. 5) is calculated for each load point. to get to.

6(a) and (b) are diagrams showing an example of the cross-sectional shape of the engine hood, in detail, (a) shows the AA' cross section of load point C, and (b) shows the load point A BB' cross section of C is shown. In this embodiment, as an example, the width of the cross-sectional shape to be obtained is 150 mm, but the width is not limited to this. It is only necessary to obtain the information, and it can be set as appropriate depending on, for example, the shape of the panel, the material and thickness, the stress range when a load is applied, the required accuracy of performance prediction, etc. Furthermore, in this embodiment, two orthogonal cross-sectional shapes are obtained, such as the AA' cross section and the BB' cross section shown in FIG. 5, but the present invention is not limited to this. For example, In order to improve performance prediction accuracy, cross-sectional shapes in other directions around the load point may be further acquired.

Further, the point cloud coordinate data generation unit 21 divides the cross-sectional shape (150 mm width) obtained for each load point into sections at 5 mm intervals (5 mm x 5 mm squares), and divides the cross-sectional shape (5 mm width) for each load point into A collection of corner center points) is generated (step S3). The interval between each point constituting the point group was set to 5 mm from the viewpoint of shape expression accuracy (fillet R portion, etc.) and the balance of the number of features when the cross-sectional shape was formed into a point group. Then, the point cloud coordinate data generation unit 21 acquires the coordinates (X coordinate, Y coordinate, Z coordinate) of each point forming the point cloud based on the 3D shape data of the engine hood arranged on the 3D coordinate space. , generate point group coordinate data of the cross-sectional shape for each load point (step S4), and store it in the storage unit 12. 6(c) and (d) are, for example, schematic diagrams showing images of the state in which the cross-sectional shape obtained at the load point C in FIG. 5 is formed into a point group. (b) shows a point group image of the BB' cross section of the load point C. Moreover, FIG. 7 is a diagram showing an example of point group coordinate data. In this embodiment, as an example, the interval between each point in the point cloud is set to 5 mm, but it is not limited to this. It can be set as appropriate depending on the stress range of , the required accuracy of performance prediction, etc.

As described above, after the point group coordinate data generation unit 21 generates point group coordinate data for each load point, the data set generation unit 22 next extracts the point group coordinate data and 3D design information ( (including information on the material and plate thickness of each component that makes up the engine hood) and the amount of deflection (performance information) for each load point obtained and stored in advance through CAE analysis, and forms point cloud coordinate data. A data set for machine learning is generated by linking the coordinates of each point, the material and plate thickness of each point, and the amount of deflection for each load point (step S5). In other words, the data set used in the machine learning algorithm uses explanatory variables (input data), and the amount of deflection at that load point obtained in advance is used as the objective variable (teacher data), and the teacher data corresponding to each load point is individually linked to the input data prepared for each load point. This is a data set. Then, the data set generation unit 22 stores the generated data set in the storage unit 12.

FIG. 8 is a diagram showing an example of a data set for generating a performance prediction model (a data set for 120 locations among the data set for 143 load points). The machine learning unit 23 of the control unit 11 reads the data set for 120 load points shown in FIG. By performing this, for example, a performance prediction model that predicts the amount of deflection when a certain load is applied to any part of the engine hood of an automobile is generated (step S6). That is, in this embodiment, a machine learning algorithm is used to calculate the 3D shape data (for example, point group coordinate data for each load point), material and plate thickness of the engine hood of an automobile, and the information for each load point of the engine hood. By learning the correlation between the amount of deflection when a certain load is applied and the amount of deflection when a certain load is applied to any part of the engine hood of a car, a performance prediction model is generated that predicts the amount of deflection when a certain load is applied to any part of the engine hood of an automobile.

In addition, in this embodiment, as mentioned above, 143 data sets (load points: 143 locations) were prepared for machine learning. Among them, 120 data are used as a data set for generating a performance prediction model, that is, 80 data are used as learning data and 40 data are used as evaluation data. The remaining 23 items are then used as test data. Here, the training data is a dataset used for learning the performance prediction model, the evaluation data is a dataset used to verify the performance of the model, and the test data is a dataset that is obtained in advance. This data set is used to perform the final evaluation of the model by comparing the deflection amount. Further, the number of learning data is prepared such that a desired prediction accuracy (for example, the coefficient of determination R2 is in the range of 0.8 to 0.9) can be obtained without causing overfitting or insufficient learning. Further, in the performance evaluation of the performance prediction model in this embodiment, mean squared error (MSE) and mean absolute error (MAE), which are examples of evaluation indicators, are employed. In other words, it can be said that the smaller these values are, the less error there is in the model. Note that the evaluation index is not limited to this.

FIG. 9 is a diagram showing the evaluation results of the model by machine learning, and shows, for example, changes in the mean square error and the mean absolute error with respect to the degree of learning (Epochs). As shown in FIG. 9, in the performance prediction model of this embodiment, the mean square error and the mean absolute error become smaller as the learning progresses, so it can be confirmed that the learning progresses smoothly. In addition, the mean square error and the mean absolute error are sufficiently small after about 50 epochs, and there are few changes thereafter, so in this embodiment, the number of epochs: about 50 is suitable for learning the performance prediction model. I can say that there is.

Moreover, FIG. 10 is a diagram showing prediction results by a learned performance prediction model using 23 pieces of test data. Here, the predicted value (the amount of deflection) output from the performance prediction model and the teacher data (actual value: the amount of deflection obtained in advance) were plotted, and the results were observed. In FIG. 10, the range where the coefficient of determination R2 is equal to or greater than 0.8 is indicated by a dotted line with respect to the straight line of the measured values connecting the plot data of the circles. As a result, the approximate curve of the predicted values of the performance prediction model (deflection amounts at 23 locations indicated by x marks) satisfied the coefficient of determination R2=0.8 or more, and it can be said that good prediction results were obtained.

<Example of machine learning algorithm>
The machine learning algorithm used in this embodiment is, for example, a known neural network such as a convolution neural network (CNN) or a recurrent neural network (RNN). For example, the machine learning unit 23 of the control unit 11 calculates the mean square error between the prediction result (deflection amount) that is the output of the output layer of the neural network and the teaching data (deflection amount), and calculates the mean square error so that this error is minimized. Then, learning is repeated over a predefined number of epochs.

Note that in this embodiment, a neural network such as CNN or RNN is used as an example of a machine learning algorithm to generate the above-mentioned performance prediction model, but the machine learning algorithm for generating the performance prediction model is as follows. It is not limited to this. Other machine learning algorithms can also be used, for example random forests or boosted decision trees.

<Performance prediction processing>
Next, the performance prediction process in the performance prediction system 1 of this embodiment will be described in detail. FIG. 11 is a functional block diagram showing the functions of the control unit 11 that performs performance prediction processing, and FIG. 12 is a flowchart showing an example of the performance prediction processing.

In FIG. 11, the control unit 11 includes the above-described point group coordinate data generation unit 21, an input data generation unit 26, and a performance prediction unit 27 that operates as a performance prediction model. In this embodiment, similar to the above, the 3D design information of an automobile engine hood is used as an example of various metal panels, and the learning model (performance prediction model) generated above is used to determine the performance of the engine hood. A case will be described in which the amount of deflection at a prediction point (any location where the amount of deflection is desired to be known) is predicted. That is, in this embodiment, the learned performance prediction model is used to predict the amount of deflection when a constant load is applied to the performance prediction point of the engine hood of an automobile.

Specifically, first, when execution of the performance prediction process is instructed by the operator's operation of the input unit 13, the point cloud coordinate data generation unit 21 of the control unit 11 extracts 3D design information of the engine hood from the storage unit 12. is read out, and the engine hood (shape) is placed on the 3D coordinate space (X coordinate, Y coordinate, Z coordinate) based on the 3D shape data obtained from the 3D design information (step S11). At this time, a 3D image of the engine hood is displayed on the display unit 15 (see FIG. 4).

Next, when the operator operates the input unit 13 to set a performance prediction point that is an arbitrary position on the engine hood for which the amount of deflection is desired, the point cloud coordinate data generation unit 21 generates, from the 3D design information, The cross-sectional shape of the engine hood around the performance prediction point is acquired (step S12). Specifically, for example, a cross-sectional shape of the engine hood with a width of 150 mm in the front, rear, left and right directions centered around the performance prediction point is obtained. That is, in this embodiment, the cross-sectional shape of the engine hood with a width of 150 mm in the left-right direction centered on the performance prediction point, and the cross-sectional shape of the engine hood with a width of 150 mm in the longitudinal direction centered on the performance prediction point are obtained (Fig. 5 and Figure 6). Note that the width dimension of the cross-sectional shape obtained here is the same as the width dimension of the cross-sectional shape obtained at the time of generating the performance prediction model (see step S2 in FIG. 3).

Furthermore, the point cloud coordinate data generation unit 21 divides the cross-sectional shape (150 mm width) in the front-rear direction and the left-right direction centering on the performance prediction point, for example, at 5 mm intervals (5 mm x 5 mm squares), and A point group (a collection of center points of a 5 mm square) of the cross-sectional shape is generated (step S13). Then, the point cloud coordinate data generation unit 21 acquires the coordinates (X coordinate, Y coordinate, Z coordinate) of each point forming the point cloud based on the 3D shape data of the engine hood arranged on the 3D coordinate space. , generates point group coordinate data of the cross-sectional shape at the performance prediction point (step S14) (see FIGS. 6(c) and (d) and FIG. 7), and stores it in the storage unit 12. Note that the intervals between the points constituting the point cloud of the cross-sectional shape at the performance prediction point are the same as the intervals of the points constituting the point cloud of the cross-sectional shape acquired for each load point when generating the performance prediction model ( (See step S3 in FIG. 3).

As described above, after the point cloud coordinate data generation unit 21 generates the point cloud coordinate data at the performance prediction point, the input data generation unit 26 generates the point cloud coordinate data and 3D data at the performance prediction point from the storage unit 12. Read design information (including information on the material and thickness of each part that makes up the engine hood), link the coordinates of each point forming the point cloud coordinate data, and the material and thickness of each point, Input data to be input to the performance prediction model is generated (step S15). The input data generation unit 26 then inputs the generated input data to the performance prediction unit 27 that operates as a performance prediction model.

The performance prediction unit 27 receives the input data generated by the input data generation unit 26, and uses the performance prediction model to predict the amount of deflection at the performance prediction point (any location where you want to know the amount of deflection) in the engine hood. Step S16). That is, in this embodiment, the learned performance prediction model is used to predict the amount of deflection (the amount of deflection at the performance prediction point) when a constant load is applied to the performance prediction point of the engine hood of an automobile. The amount of deflection at the performance prediction point is displayed on the display unit 15 as a performance prediction value.

<Effect>
As described above, the performance prediction system 1 of the present embodiment predicts the amount of deflection under load, which is one of the performances of metal panels (for example, the engine hood of a car) that constitute the body of a car. It is. Specifically, the point cloud coordinate data of the point cloud around the load points uniformly provided on the engine hood (see Figure 7) was associated with information on the material and plate thickness of each point making up the point cloud. Data is used as an explanatory variable, and a data set ( The machine learning unit 23 uses the data set read from the storage unit 12 to predict the amount of deflection when a constant load is applied to the performance prediction point of the engine hood. Generate a performance prediction model using machine learning. Then, the performance prediction unit 27 generates input data in which point group coordinate data of a point group around the performance prediction point, which is an arbitrary location on the engine hood, is associated with information on the material and plate thickness of each point constituting the point group. is accepted, and the learned performance prediction model is used to predict the amount of deflection when a constant load is applied to the performance prediction point.

Furthermore, the storage unit 12 stores 3D design information of the engine hood, information on the material and plate thickness of the engine hood associated with the 3D design information, and the amount of deflection at each load point obtained in advance by a known method. It was decided to memorize this in advance. Then, the point cloud coordinate data generation unit 21 places the engine hood on the 3D coordinate space (X coordinate, Y coordinate, Z coordinate) based on the 3D design information of the engine hood, and calculates the load point or performance prediction point. Obtain multiple cross-sectional shapes of the panel as a center, divide the obtained cross-sectional shapes into squares of a predetermined angle to generate a point group that is a collection of center points, and calculate the coordinates of each point forming the point group. (X coordinate, Y coordinate, Z coordinate) to generate point group coordinate data.

Thereby, the data set generation unit 22 generates the point cloud coordinate data generated by the point cloud coordinate data generation unit 21, the 3D design information of the engine hood, and the material and plate thickness of the engine hood associated with the 3D design information. A data set for generating a learning model can be generated based on the information and the amount of deflection at each load point acquired in advance. The input data generation unit 26 also generates the point cloud coordinate data generated by the point cloud coordinate data generation unit 21, the 3D design information of the engine hood, and the material and plate thickness of the engine hood associated with the 3D design information. Based on the information, input data for performance prediction can be generated.

As described above, according to the performance prediction system 1 of the present embodiment, by introducing a performance prediction model, the time required to calculate the amount of deflection when a load is applied to an arbitrary position (performance prediction point) of various automotive panels can be reduced. can be significantly shortened. Further, by using the performance prediction model, variations between workers are eliminated, so dent performance can be predicted stably and with high accuracy.

Note that in this embodiment, an automobile engine hood is used as an example of various metal panels, but the invention is not limited to this, and the learning model generation process and performance prediction process of this embodiment are as follows. For example, in addition to the engine hood, the present invention is applicable to all metal panels constituting the car body, such as front fenders, roof panels, and doors.

In addition, in this embodiment, a performance prediction model was created that predicts the amount of deflection when a certain load is applied to an arbitrary position on a metal panel, but for example, load information (load It is also possible to generate a performance prediction model that predicts the amount of deflection corresponding to the load size and load direction.

1 Performance prediction system 11 Control unit 12 Storage unit 13 Input unit 14 Interface (I/F) unit 15 Display unit 16 Communication unit 21 Point cloud coordinate data generation unit 22 Data set generation unit 23 Machine learning unit 24 CAD system 25 CAE system 26 Input data generation section 27 Performance prediction section

Claims (5)

In a performance prediction system that predicts the performance of metal panels that make up the car body,
The explanatory variable is data that associates the point group coordinate data of a point group around the load points uniformly provided on the panel with information on the material and plate thickness of each point constituting the point group, and the a storage means for storing a data set in which an explanatory variable and an objective variable, which is the amount of deflection when a constant load is applied to the load point, are linked for each load point;
a model generation means for generating, by machine learning, a performance prediction model that predicts the amount of deflection when a constant load is applied to a performance prediction point, which is an arbitrary location on the panel, using the data set read from the storage means; ,
It accepts input data that associates the point cloud coordinate data of the point cloud around the performance prediction point with information on the material and plate thickness of each point constituting the point cloud, and uses the trained performance prediction model to calculate the performance prediction point. a performance prediction means for predicting the amount of deflection when a constant load is applied to the
Equipped with
A performance prediction system characterized by: The storage means stores in advance 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and the amount of deflection at each load point obtained by a known method. This year,
moreover,
Based on the 3D design information of the panel, arrange the panel on a 3D coordinate space (X coordinate, Y coordinate, Z coordinate), and obtain a plurality of cross-sectional shapes of the panel centered on the load point or performance prediction point. Then, by dividing the obtained cross-sectional shape into squares with a predetermined angle, a point group is generated that is a collection of center points, and the coordinates (X coordinate, Y coordinate, Z coordinate) of each point forming the point group are obtained. point group coordinate data generation means for generating the point group coordinate data;
Point cloud coordinate data generated by the point cloud coordinate data generation means, 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and acquired by the known method. a data set generating means for generating the data set for model generation based on the amount of deflection at each load point;
Performance prediction is performed based on the point cloud coordinate data generated by the point cloud coordinate data generation means, 3D design information of the panel, and information on the material and thickness of the panel associated with the 3D design information. input data generation means for generating the input data for;
Equipped with
The performance prediction system according to claim 1, characterized in that: A performance prediction method for predicting the performance of a metal panel constituting an automobile body, the method comprising:
The explanatory variable is data that associates the point group coordinate data of a point group around the load points uniformly provided on the panel with information on the material and plate thickness of each point constituting the point group, and the a data set storage step of storing in memory a data set in which an explanatory variable and an objective variable, which is the amount of deflection when a constant load is applied to the load point, are linked for each load point;
a model generation step of generating, by machine learning, a performance prediction model that predicts the amount of deflection when a constant load is applied to a performance prediction point that is an arbitrary location on the panel, using the data set read from the memory;
Accepts input data that associates the point cloud coordinate data of the point cloud around the performance prediction point with information on the material and plate thickness of each point constituting the point cloud, and uses the trained performance prediction model to calculate the performance prediction point. a performance prediction step of predicting the amount of deflection when a constant load is applied to the
a display step of displaying the prediction result on a display;
including,
A performance prediction method characterized by: The memory stores in advance 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and the amount of deflection at each load point obtained by a known method. year,
Furthermore, as preprocessing for generating the performance prediction model,
a shape arrangement step of arranging the panel on a 3D coordinate space (X coordinate, Y coordinate, Z coordinate) based on 3D design information of the panel;
a cross-sectional shape acquisition step of acquiring a plurality of cross-sectional shapes of the panel centered on a load point;
a point cloud generation step of generating a point cloud that is a collection of center points by dividing the obtained cross-sectional shape into squares of a predetermined angle;
a point group coordinate data generation step of obtaining the coordinates (X coordinate, Y coordinate, Z coordinate) of each point forming the point group to generate point group coordinate data;
The point cloud coordinate data generated in the point cloud coordinate data generation step, 3D design information of the panel, information on the material and thickness of the panel associated with the 3D design information, and acquired by the known method. a data set generation step of generating the data set for model generation based on the amount of deflection at each load point;
including,
The performance prediction method according to claim 3, characterized in that: Furthermore, as a preprocessing of the performance prediction step,
the shape arrangement step;
the cross-sectional shape acquisition step of acquiring a plurality of cross-sectional shapes of the panel centered on the performance prediction point;
The point cloud generation step;
the point group coordinate data generation step;
Performance prediction is performed based on the point cloud coordinate data generated in the point cloud coordinate data generation step, 3D design information of the panel, and information on the material and thickness of the panel associated with the 3D design information. an input data generation step of generating the input data for;
including,
5. The performance prediction method according to claim 4.

Images