Vehicle door stiffness prediction method and device, vehicle and storage medium

Abstract

The application relates to the technical field of vehicle engineering, in particular to a vehicle door stiffness prediction method and device, a vehicle and a storage medium, wherein the method comprises the following steps: acquiring vehicle door assembly data, establishing a vehicle door finite element model according to the vehicle door assembly data; performing sensitivity analysis on vehicle door parts to obtain a sensitivity analysis result, generating a plurality of vehicle door model samples according to the vehicle door finite element model, the sensitivity analysis result and a vehicle door deformation variable; calculating the stiffness analysis result of each vehicle door model sample according to a pre-set stiffness analysis working condition, generating a vehicle door stiffness data set according to the plurality of vehicle door model samples and the stiffness analysis result of each vehicle door model sample, training a prediction model by using the vehicle door stiffness data set, and predicting the vehicle door stiffness based on the trained prediction model. Therefore, the problems of lacking a standardized data processing method, difficulty in expanding the data set and low processing efficiency of complex models in the related technology of vehicle door stiffness research are solved.

Description

Technical Field

[0001] This application relates to the field of vehicle engineering technology, and in particular to a method, device, vehicle, and storage medium for predicting door stiffness. Background Technology

[0002] As the automotive industry moves towards lightweighting and intelligentization, door stiffness analysis and optimization have become crucial aspects of vehicle body design. Current methods for door stiffness research primarily rely on physical experiments and computer-aided engineering simulations. However, the introduction of machine learning methods in recent years has provided a new technological path for door stiffness prediction. Machine learning-based stiffness prediction often requires large datasets. Nevertheless, methods for processing datasets in door stiffness analysis suffer from several problems: a lack of standardized dataset processing methods, difficulty in scaling datasets, and low efficiency in processing complex models. Summary of the Invention

[0003] This application provides a method, device, vehicle, and storage medium for predicting the stiffness of a vehicle door, in order to solve the problems in the data processing methods of vehicle door stiffness research, such as the lack of standardized data processing methods, difficulty in expanding the data set, and low efficiency in processing complex models.

[0004] The first aspect of this application provides a method for predicting the stiffness of a vehicle door, comprising the following steps: acquiring vehicle door assembly data; establishing a finite element model of the vehicle door based on the vehicle door assembly data; performing sensitivity analysis on the vehicle door parts to obtain sensitivity analysis results; generating multiple vehicle door model samples based on the vehicle door finite element model, the sensitivity analysis results, and the vehicle door deformation variables; calculating the stiffness analysis results of each vehicle door model sample according to a pre-set stiffness analysis condition; generating a vehicle door stiffness dataset based on the multiple vehicle door model samples and the stiffness analysis results of each vehicle door model sample; training a prediction model using the vehicle door stiffness dataset; and predicting the vehicle door stiffness based on the trained prediction model.

[0005] According to one embodiment of this application, the door assembly data includes multiple components such as an outer door panel, an inner door panel, a window frame, a waistline reinforcement plate, a crash beam, a glass guide rail, a hinge reinforcement plate, a lock reinforcement plate, an upper hinge, and a lower hinge.

[0006] According to one embodiment of this application, establishing a finite element model of a car door based on door assembly data includes: establishing a three-dimensional model of the car door based on the door assembly data; and establishing a finite element model of the car door based on the three-dimensional model, material property definitions, and thickness information.

[0007] According to one embodiment of this application, sensitivity analysis is performed on a car door component to obtain sensitivity analysis results, including: acquiring thickness variation data of the car door component; analyzing the degree of influence of the thickness variation on the car door based on the thickness variation data; and determining the sensitivity analysis results based on the degree of influence value.

[0008] According to one embodiment of this application, multiple door model samples are generated based on the finite element model of the door, sensitivity analysis results, and door deformation variables, including: obtaining the numbering rules for coupling points, loading points, and constraint points of the door model samples; generating multiple target design parameters based on the sensitivity analysis results and door deformation variables; and generating multiple door model samples based on the multiple target design parameters and the numbering rules.

[0009] According to one embodiment of this application, before generating multiple door model samples based on the finite element model of the door, sensitivity analysis results, and door deformation variables, the method further includes: establishing a coordinate system for the loading point of each door model sample; placing the initial position coordinates of each loading point at the origin of the coordinate system; extracting the displacement value of the door model sample by reading the final position coordinates of the loading point; and determining the stiffness value of the door model sample based on the displacement value of the door model sample.

[0010] According to one embodiment of this application, the stiffness analysis results of each door model sample are calculated based on a pre-set stiffness analysis condition, including: determining the loading point location, loading size, constraint conditions, and output requirements based on the pre-set stiffness analysis condition; generating multiple header files based on the loading point location, loading size, constraint conditions, and output requirements; driving a finite element solver based on the multiple header files; performing finite element calculations on all door model samples through the finite element solver; and extracting the stiffness analysis results of each door model sample from the finite element calculation results.

[0011] A second aspect of this application provides a door stiffness prediction device, comprising: an acquisition module for acquiring door assembly data and establishing a door finite element model based on the door assembly data; a generation module for performing sensitivity analysis on door parts to obtain sensitivity analysis results and generating multiple door model samples based on the door finite element model, the sensitivity analysis results, and door deformation variables; and a prediction module for calculating the stiffness analysis results of each door model sample according to a pre-set stiffness analysis condition, generating a door stiffness dataset based on the multiple door model samples and the stiffness analysis results of each door model sample, training a prediction model using the door stiffness dataset, and predicting door stiffness based on the trained prediction model.

[0012] A third aspect of this application provides a vehicle, which includes a door assembly, the stiffness of which is predicted based on the door stiffness prediction method described above.

[0013] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the door stiffness prediction method as described in the above embodiments.

[0014] Therefore, this application has the following beneficial effects: This process involves acquiring door assembly data and establishing a finite element model of the door, achieving standardized digital modeling of the door structure and improving the standardization and uniformity of modeling. Sensitivity analysis is performed on door components to obtain the results. Based on the finite element model, sensitivity analysis results, and door deformation variables, multiple door model samples are generated. Sensitivity analysis accurately identifies components and parameters that significantly affect stiffness, and multiple sets of door model samples are automatically generated based on these parameters, making the construction of door model samples more scientific and easily expandable. Stiffness analysis results for each door model sample are calculated according to pre-set stiffness analysis conditions. A door stiffness dataset is generated based on multiple door model samples and the stiffness analysis results of each sample. This dataset is then used to train a prediction model, which predicts door stiffness. By standardizing operating conditions and automating batch calculations, the problem of uncontrollable data quality is solved. Ultimately, the prediction model achieves accurate prediction of door stiffness, significantly improving design efficiency and iteration speed.

[0015] This application constructs a high-quality, standardized vehicle door stiffness dataset based on standardized data preprocessing, sensitivity analysis of vehicle door components, component deformation and sample generation, and automated finite element calculation and extraction. This dataset can be used to match finite element models, point clouds, voxels, and other data to train prediction models. This solves the problems in vehicle door stiffness research related technologies, such as the lack of standardized data processing methods, difficulty in dataset expansion, and low efficiency in processing complex models.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a door stiffness prediction method according to an embodiment of this application; Figure 2 This is a schematic diagram showing the disassembly of a car door model according to an embodiment of this application; Figure 3 This is a schematic diagram illustrating the method of establishing a local coordinate system according to an embodiment of this application; Figure 4 This is a flowchart of a method for constructing a door stiffness dataset according to an embodiment of this application; Figure 5 This is an example diagram of a door stiffness prediction device according to an embodiment of this application. Detailed Implementation

[0018] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0019] The following description, with reference to the accompanying drawings, outlines a method, apparatus, vehicle, and storage medium for predicting door stiffness according to embodiments of this application. Addressing the issues mentioned in the background art regarding the lack of standardized data processing methods, difficulty in scaling datasets, and low efficiency in processing complex models in related technologies for door stiffness research, this application provides a method for predicting door stiffness. In this method, door assembly data is acquired, and a finite element model of the door is established based on the door assembly data. Sensitivity analysis is performed on the door components to obtain sensitivity analysis results. Multiple door model samples are generated based on the door finite element model, the sensitivity analysis results, and the door deformation variables. The stiffness analysis results of each door model sample are calculated according to pre-set stiffness analysis conditions. A door stiffness dataset is generated based on the multiple door model samples and the stiffness analysis results of each door model sample. A prediction model is trained using the door stiffness dataset, and the door stiffness is predicted based on the trained prediction model. This solves the problems of lacking standardized data processing methods, difficulty in scaling datasets, and low efficiency in processing complex models in related technologies for door stiffness research.

[0020] Specifically, Figure 1 A flowchart illustrating a door stiffness prediction method provided in an embodiment of this application.

[0021] like Figure 1 As shown, the method for predicting the stiffness of a car door includes the following steps: In step S101, the door assembly data is obtained, and a finite element model of the door is established based on the door assembly data.

[0022] Understandably, by acquiring door assembly data and establishing a finite element model of the door, the structural composition, geometry, material properties, connection relationships, and mechanical load-bearing characteristics of the door can be accurately reflected. This provides a high-precision digital foundation model for subsequent component sensitivity analysis, deformation data extraction, sample generation, and finite element simulation calculations. This step ensures the authenticity and standardization of the data from the source, avoiding adverse effects on subsequent dataset construction and AI (Artificial Intelligence) prediction model training due to model errors, thus laying the foundation for efficient and accurate prediction of door stiffness.

[0023] According to one embodiment of this application, the door assembly data includes multiple components such as an outer door panel, an inner door panel, a window frame, a waistline reinforcement plate, a crash beam, a glass guide rail, a hinge reinforcement plate, a lock reinforcement plate, an upper hinge, and a lower hinge.

[0024] Understandably, by acquiring complete door assembly data, including the outer door panel, inner door panel, window frame, waistline reinforcement plate, anti-collision beam, glass guide rail, hinge reinforcement plate, lock reinforcement plate, upper hinge, and lower hinge, it is possible to comprehensively and completely characterize the overall structural composition, component composition, assembly relationship, and mechanical force transmission path of the door. This ensures that the established finite element model of the door is structurally complete, with accurate parameters and reasonable boundary conditions, avoiding model distortion due to missing components or structural simplification. Data source-wise, the completeness and reliability of subsequent sensitivity analysis, model sample generation, stiffness calculation, and dataset construction are guaranteed, providing a solid data foundation for achieving high-precision and highly robust door stiffness prediction.

[0025] For example, this application establishes or collects finite element models based on the geometry of the vehicle's doors. The door assembly data includes the outer door panel, inner door panel, window frame, waistline reinforcement plate, anti-collision beam, glass guide rail, hinge reinforcement plate, lock reinforcement plate, and upper and lower hinges, among other structures. Figure 2 As shown, the car door assembly model is divided into four parts: upper hinge, lower hinge, lock, and car door body.

[0026] According to one embodiment of this application, establishing a finite element model of a car door based on door assembly data includes: establishing a three-dimensional model of the car door based on the door assembly data; and establishing a finite element model of the car door based on the three-dimensional model, material property definitions, and thickness information.

[0027] It is understandable that a car door finite element model refers to an assembly model constructed using car door assembly data, which can be used for mechanical simulation calculations. By establishing a 3D model of the car door based on the assembly data, and further establishing a finite element model of the car door by combining material property definitions and thickness information, the true physical characteristics and mechanical performance of the car door can be fully reproduced from multiple dimensions, making the established finite element model closer to the actual structure. This modeling method effectively improves the realism of the finite element model, ensuring the accuracy and reliability of subsequent part sensitivity analysis, model sample generation, stiffness calculation, and dataset construction from the source of modeling, providing a solid and stable model foundation for the subsequent construction of high-quality datasets and the training of high-precision AI prediction models.

[0028] For example, the finite element model uses shell elements to simulate sheet metal parts, solid elements to simulate castings, hexahedral solid elements to simulate weld joints, and RBE2 elements (Rigid Body Element 2) to simulate bolted connections. The model must include complete material property definitions and thickness information.

[0029] In step S102, sensitivity analysis is performed on the door parts to obtain the sensitivity analysis results. Based on the finite element model of the door, the sensitivity analysis results, and the deformation variables of the door, multiple door model samples are generated.

[0030] Understandably, by performing sensitivity analysis on car door components, the influence weight of each component on the door stiffness can be accurately identified, thus distinguishing between critical and non-critical components. Based on this, by combining the finite element model of the car door with the deformation variables of the car door, multiple car door model samples can be generated. This can expand the sample coverage and improve the sample diversity and representativeness while ensuring structural rationality, providing a sufficient and effective sample foundation for the subsequent construction of a high-quality, highly generalizable dataset.

[0031] According to one embodiment of this application, sensitivity analysis is performed on a car door component to obtain sensitivity analysis results, including: acquiring thickness variation data of the car door component; analyzing the degree of influence of the thickness variation on the car door based on the thickness variation data; and determining the sensitivity analysis results based on the degree of influence value.

[0032] Understandably, by acquiring thickness variation data of car door components, analyzing the impact of these variations on the door's stiffness, deformation, and other properties, and determining the sensitivity analysis results based on this impact value, it is possible to accurately quantify the contribution and influence weight of each component's thickness parameter to the overall mechanical performance of the car door. This allows for the effective differentiation between critical and secondary structural components. It provides a scientific basis for subsequent model sample generation and parameter adjustment, avoiding blindly adjusting parameters that could lead to invalid samples or computational redundancy, and further improving the training efficiency and prediction accuracy of subsequent AI prediction models.

[0033] For example, sensitivity analysis is performed on various components of a car door to identify key components that significantly impact stiffness performance. This sensitivity analysis employs either a single-factor perturbation method or the Morris screening method to analyze the impact of thickness variations in each component on the overall and local stiffness of the car door, providing a reference for subsequent parametric deformation and variable setting of components. The single-factor perturbation method, with its advantages of low computational cost and intuitive results, quickly identifies key variables, avoiding the computational overhead of indiscriminate optimization of all components. This provides a scientific basis for subsequent variable selection and leads to efficient utilization of computational resources.

[0034] Sensitivity analysis employs a single-factor perturbation method. By adjusting the thickness of various door components, the overall local stiffness of the door is calculated, and the change in stiffness compared to the original thickness state is compared. Components with a greater impact on stiffness due to thickness changes are considered to have high sensitivity under this condition. Analyzing the influence of each component on the overall and local stiffness of the door provides a reference for subsequent parameterized deformation and variable settings for components.

[0035] According to one embodiment of this application, multiple door model samples are generated based on the finite element model of the door, sensitivity analysis results, and door deformation variables, including: obtaining the numbering rules for coupling points, loading points, and constraint points of the door model samples; generating multiple target design parameters based on the sensitivity analysis results and door deformation variables; and generating multiple door model samples based on the multiple target design parameters and the numbering rules.

[0036] It is understandable that by obtaining the numbering rules of coupling points, loading points and constraint points of the door model samples, and generating multiple target design parameters based on the sensitivity analysis results and door deformation variables, and combining the numbering rules to generate multiple door model samples, it is possible to ensure that each model sample has a unified definition standard in terms of structural coupling and boundary constraints, and avoid the distortion of simulation results due to chaotic node definitions or inconsistent constraints.

[0037] For example, design variable space samples are generated based on parametric mesh deformation technology. These deformation variables include the shape, thickness, and stiffener positions of key components. Referring to the sensitivity analysis results mentioned above, an optimized Latin hypercube experimental design method is used to generate a sample matrix, ensuring effective coverage of the design space. Each sample point corresponds to a finite element model of a car door with specific design parameters. By deforming and generating samples of car door parts with different structural forms, more than 3000 car door model samples are created for training. This method allows for the rapid generation of a large number of design variants without reconstructing CAD (Computer-Aided Design) geometry, overcoming the drawbacks of high cost and long cycle time of traditional modeling methods. The use of the optimized Latin hypercube experimental design method to generate the sample matrix ensures the uniformity and coverage of samples in the design space, resulting in a significant improvement in data generation efficiency and sample quality.

[0038] According to one embodiment of this application, before generating multiple door model samples based on the finite element model of the door, sensitivity analysis results, and door deformation variables, the method further includes: establishing a coordinate system for the loading point of each door model sample; placing the initial position coordinates of each loading point at the origin of the coordinate system; extracting the displacement value of the door model sample by reading the final position coordinates of the loading point; and determining the stiffness value of the door model sample based on the displacement value of the door model sample.

[0039] Understandably, by establishing a coordinate system for the loading points of each door model sample before generating multiple door model samples, and placing the initial position coordinates of each loading point at the origin of the coordinate system, the position reference and coordinate definition of the loading points can be unified, ensuring that the reference remains consistent for different door model samples during loading, constraint, and displacement extraction. By reading the final position coordinates of the loading points to extract displacement values, and using this to determine the stiffness values ​​of the door model samples, high precision in the stiffness calculation process can be achieved. This effectively avoids calculation errors caused by inconsistent coordinate systems and references, improves the accuracy of stiffness results, and provides a reliable computational foundation and data guarantee for the subsequent generation of door stiffness datasets.

[0040] For example, a local coordinate system is established for the loading point of each car door model, such as... Figure 3 As shown, the origin of the local coordinate system is defined as the loading point, and the Z-axis is perpendicular to the tangent plane of the part at the loading point. Each loading point is placed in the local coordinate system to ensure its initial coordinates are (0, 0, 0). By establishing the local coordinate system and combining the relationship that stiffness equals force divided by displacement, the complex global displacement output is transformed into a simple and intuitive local coordinate displacement. Only the displacement of specific nodes in the local coordinate system needs to be analyzed, which standardizes and programmables the extraction of stiffness results, laying a solid foundation for automated extraction and directly improving extraction accuracy and efficiency.

[0041] In step S103, the stiffness analysis results of each door model sample are calculated according to the pre-set stiffness analysis conditions. A door stiffness dataset is generated based on multiple door model samples and the stiffness analysis results of each door model sample. A prediction model is trained using the door stiffness dataset, and the door stiffness is predicted based on the trained prediction model.

[0042] It is understandable that stiffness analysis cases refer to the predefined set of stiffness information for each component in finite element analysis, used to uniformly and systematically calculate structural stiffness and deformation. Calculating each door model sample according to the predefined stiffness analysis cases allows for the batch, efficient, and accurate acquisition of corresponding stiffness response results, thereby forming a standardized and high-quality door stiffness dataset. This fundamentally solves the problems of non-standardized and difficult-to-expand datasets in related technologies. The prediction model trained using this dataset can significantly improve the accuracy and efficiency of door stiffness prediction, enabling rapid prediction of door stiffness without repeatedly performing complex finite element calculations. This significantly reduces design costs, shortens the R&D cycle, and provides reliable data support and prediction methods for lightweight and forward design of doors.

[0043] According to one embodiment of this application, the stiffness analysis results of each door model sample are calculated based on a pre-set stiffness analysis condition, including: determining the loading point location, loading size, constraint conditions, and output requirements based on the pre-set stiffness analysis condition; generating multiple header files based on the loading point location, loading size, constraint conditions, and output requirements; driving a finite element solver based on the multiple header files; performing finite element calculations on all door model samples through the finite element solver; and extracting the stiffness analysis results of each door model sample from the finite element calculation results.

[0044] Understandably, by determining the loading point location, loading magnitude, constraint conditions, and output requirements based on pre-set stiffness analysis conditions, and generating multiple header files accordingly, and using these header files to drive the finite element solver to perform batch finite element calculations on all door model samples, the stiffness analysis process can be standardized. This ensures that different samples are calculated under the same conditions, constraints, and loading, significantly improving the reliability of the analysis results. Simultaneously, automatically driving the solver and extracting stiffness analysis results in batches through header files significantly reduces manual intervention, thereby improving computational efficiency and data extraction accuracy. This effectively solves the problem of low efficiency in processing complex models in traditional methods, providing efficient and reliable technical support for rapidly constructing large-scale door stiffness datasets.

[0045] For example, a unified set of door stiffness analysis cases is defined, and standardized header files are generated. These stiffness analysis cases include inner and outer waistline stiffness, front and rear window frame stiffness, door vertical stiffness, lock fixing point stiffness, upper and lower hinge fixing point stiffness, and local fixing point stiffness cases. Each header file determines its sub-files based on the case, explicitly defining the loading point location, loading magnitude, constraints, and output requirements for each case. The header files drive a finite element solver for batch calculations. The door finite element model and the corresponding header files are then packaged and submitted to the cloud computing platform via a client. The platform manages the task queue, schedules resources, and allocates nodes. It then calls the NASTRAN solver to perform batch linear static solutions, reading the header files and model data to construct stiffness matrices and load vectors, obtaining the displacement response in the global coordinate system, and outputting the .op2 result file. Standardized header files ensure that all samples are calculated under completely consistent analysis conditions, improving efficiency while eliminating data noise caused by differences in case settings, providing high-quality, consistent labeled data for machine learning.

[0046] Next, batch finite element calculations are performed and stiffness results are automatically extracted. The calculations use a batch processing approach to call the finite element solver. After the calculations are completed, a dedicated analysis program automatically identifies the load case name and its corresponding stiffness value from the result file and stores the results in a table. The result file format is .f06, .dat, or .odb. This fully automated calculation and extraction process completely liberates engineers from the tedious and repetitive manual work of submitting calculations, monitoring the process, searching for result files, and copying and pasting data. Its batch processing and high parallelism greatly improve the utilization of computing resources; the rule-based programmatic extraction eliminates errors and omissions that may be caused by manual operation. Together, this results in an exponential increase in data production speed and the elimination of human error, making the construction of large-scale databases feasible in terms of time and cost.

[0047] Finally, the multi-source data are matched and associated to construct a complete sample dataset. Data matching includes associating finite element models, voxels, point clouds, or grayscale images with calculation results, pairing them with stiffness performance data, and mapping design parameters with performance responses. A unified indexing system is established to ensure that each data sample contains complete input features (design parameters) and output labels (stiffness values), forming a dataset that can be used to train various deep learning and traditional machine learning network models such as convolutional neural networks (CNN), fully connected networks (FCN), and Transformers, such as 2D CNN / ResNet / U-NET with 2D grayscale images or feature maps as input, and 3D CNN / PointNet with voxels or point clouds as input.

[0048] The following examples illustrate the door stiffness prediction method of this application. Figure 4 As shown.

[0049] In step S401, finite element models of doors for multiple car models are established and collected.

[0050] In step S402, the door assembly model is divided into four parts: upper hinge, lower hinge, lock, and door body. The model coupling points, loading points, and constraint points are distinguished according to the model area and working condition type. The node numbering rules are formulated, and the finite element model is modified according to the regulations.

[0051] In step S403, a local coordinate system is established for the loading point of each door model. The origin of the local coordinate system is defined as the loading point, and the Z-axis is perpendicular to the tangent plane of the part at the loading point. Then, an Assign operation is performed in Hypermesh (finite element pre- and post-processing software) to place each loading point into the local coordinate system, ensuring that the initial coordinates are (0, 0, 0). Using this method, the displacement can be extracted from the result file by directly reading the coordinates of the final position of the loading point, thus extracting the stiffness value.

[0052] In step S404, sensitivity analysis is performed on each component of the door to identify key components that have a significant impact on stiffness performance. The sensitivity analysis uses a single-factor perturbation method or the Morris screening method to analyze the degree of influence of thickness changes of each component on the overall and local stiffness of the door, providing a reference for subsequent Morph deformation (mesh deformation) and DOE (Design of Experiments) variable settings.

[0053] In step S405, design variable space samples are generated based on parametric mesh deformation technology. The deformation variables include the shape, thickness, and stiffener positions of key components. An optimized Latin hypercube experimental design method is used to generate a sample matrix to ensure effective coverage of the design space. Each sample point corresponds to a finite element model of a car door with specific design parameters. Through parametric deformation and sample generation of different structural forms, more than 3000 car door model samples are created for training.

[0054] In step S406, a unified set of door stiffness analysis conditions is defined and a standardized header file is generated. These conditions include inner and outer waistline stiffness, front and rear window frame stiffness, door vertical stiffness, lock fixing point stiffness, upper and lower hinge fixing point stiffness, and local fixing point stiffness conditions. Each header file determines its sub-files based on the conditions, explicitly defining the loading point locations, loading magnitudes, constraints, and output requirements for each condition. The header file drives the finite element solver for batch calculations, ensuring consistent analysis conditions across all models.

[0055] In step S407, batch finite element calculations are performed and stiffness results are automatically extracted. The calculations use a batch processing method to call the finite element solver. After the calculations are completed, a dedicated analytical program automatically identifies the working condition name and the Z-axis coordinate of the working condition loading point from the result file, outputs the corresponding stiffness values, and stores the results in a table.

[0056] In step S408, multi-source data are matched and correlated to construct a complete sample dataset. The data matching includes associating and pairing finite element models, point clouds, voxels, or grayscale images with calculation results to map design parameters to performance responses. An index file is created to ensure that each data sample contains complete input features and output labels, forming a dataset that can be used to train machine learning models. Input features refer to design parameters, and output labels refer to stiffness values.

[0057] According to the door stiffness prediction method proposed in this application, door assembly data is acquired, and a finite element model of the door is established based on the door assembly data. This achieves standardized digital modeling of the door structure, improving the standardization and uniformity of modeling. Sensitivity analysis is performed on the door parts to obtain sensitivity analysis results. Multiple door model samples are generated based on the door finite element model, sensitivity analysis results, and door deformation variables. Sensitivity analysis accurately identifies parts and parameters that significantly affect stiffness, and multiple sets of door model samples are automatically generated based on these parameters, making the construction of door model samples more scientific and easily expandable. The stiffness analysis results of each door model sample are calculated according to pre-set stiffness analysis conditions. A door stiffness dataset is generated based on multiple door model samples and the stiffness analysis results of each door model sample. The prediction model is trained using the door stiffness dataset, and the door stiffness is predicted based on the trained prediction model. By standardizing operating conditions and automating batch calculations, the problem of uncontrollable data quality is solved. Finally, the AI ​​prediction model achieves accurate prediction of door stiffness, significantly improving design efficiency and iteration speed.

[0058] This application constructs a high-quality, standardized vehicle door stiffness dataset based on standardized data preprocessing, sensitivity analysis of vehicle door components, component deformation and sample generation, and automated finite element calculation and extraction. This dataset can be used to match finite element models, point clouds, voxels, and other data to train prediction models. This solves the problems in vehicle door stiffness research related technologies, such as the lack of standardized data processing methods, difficulty in dataset expansion, and low efficiency in processing complex models.

[0059] Next, the door stiffness prediction device according to an embodiment of this application is described with reference to the accompanying drawings.

[0060] Figure 5 This is a block diagram of a door stiffness prediction device according to an embodiment of this application.

[0061] like Figure 5 As shown, the door stiffness prediction device 10 includes: an acquisition module 110, a generation module 120, and a prediction module 130.

[0062] The module 110 is used to acquire door assembly data and establish a finite element model of the door based on the door assembly data; the generation module 120 is used to perform sensitivity analysis on the door parts to obtain sensitivity analysis results, and generate multiple door model samples based on the door finite element model, sensitivity analysis results, and door deformation variables; the prediction module 130 is used to calculate the stiffness analysis results of each door model sample according to a pre-set stiffness analysis condition, generate a door stiffness dataset based on multiple door model samples and the stiffness analysis results of each door model sample, train a prediction model using the door stiffness dataset, and predict the door stiffness based on the trained prediction model.

[0063] According to one embodiment of this application, the acquisition module 110 is used to establish a three-dimensional model of the door based on the door assembly data; and to establish a finite element model of the door based on the three-dimensional model of the door, the material property definition and the thickness information.

[0064] According to one embodiment of this application, the generation module 120 is used to acquire thickness change data of the door parts; analyze the degree of influence of the thickness change on the door based on the thickness change data; and determine the sensitivity analysis result based on the degree of influence value.

[0065] According to one embodiment of this application, the generation module 120 is used to obtain the numbering rules of coupling points, loading points and constraint points of the door model sample; generate multiple target design parameters based on the sensitivity analysis results and door deformation variables; and generate multiple door model samples based on the multiple target design parameters and the numbering rules.

[0066] According to one embodiment of this application, the generation module 120 is used to establish a coordinate system for the loading point of each door model sample; place the initial position coordinates of each loading point into the origin of the coordinate system; extract the displacement value of the door model sample by reading the final position coordinates of the loading point, and determine the stiffness value of the door model sample based on the displacement value of the door model sample.

[0067] According to one embodiment of this application, the prediction module 130 is used to determine the loading point location, loading size, constraint conditions and output requirements according to the preset stiffness analysis conditions, generate multiple header files according to the loading point location, loading size, constraint conditions and output requirements, drive the finite element solver according to the multiple header files, perform finite element calculations on all door model samples through the finite element solver, and extract the stiffness analysis results of each door model sample from the finite element calculation results.

[0068] It should be noted that the foregoing explanation of the embodiment of the door stiffness prediction method also applies to the door stiffness prediction device of this embodiment, and will not be repeated here.

[0069] According to the door stiffness prediction device proposed in this application, the acquisition module is used to acquire door assembly data and establish a finite element model of the door based on the door assembly data, realizing standardized digital modeling of the door structure and improving the standardization and uniformity of modeling. The generation module is used to perform sensitivity analysis on the door parts to obtain sensitivity analysis results. Based on the door finite element model, sensitivity analysis results, and door deformation variables, multiple door model samples are generated. Through sensitivity analysis, parts and parameters that significantly affect stiffness are accurately identified, and multiple sets of door model samples are automatically generated based on the parameters, making the construction of door model samples more scientific and easily expandable. The prediction module is used to calculate the stiffness analysis results of each door model sample according to the pre-set stiffness analysis conditions. Based on multiple door model samples and the stiffness analysis results of each door model sample, a door stiffness dataset is generated. The prediction model is trained using the door stiffness dataset, and the door stiffness is predicted based on the trained prediction model. By standardizing the conditions and automating batch calculations, the problem of uncontrollable data quality is solved. Finally, the AI ​​prediction model achieves accurate prediction of door stiffness, significantly improving design efficiency and iteration speed.

[0070] This application constructs a high-quality, standardized vehicle door stiffness dataset based on standardized data preprocessing, sensitivity analysis of vehicle door components, component deformation and sample generation, and automated finite element calculation and extraction. This dataset can be used to match finite element models, point clouds, voxels, and other data to train prediction models. This solves the problems in vehicle door stiffness research related technologies, such as the lack of standardized data processing methods, difficulty in dataset expansion, and low efficiency in processing complex models.

[0071] This application also provides a vehicle, which includes a door assembly, the stiffness of which is predicted based on the above-described door stiffness prediction method.

[0072] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described door stiffness prediction method.

[0073] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0074] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0075] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0076] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0077] Those skilled in the art will understand that all or part of the steps of the methods implementing the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0078] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for predicting the stiffness of a vehicle door, characterized in that, Includes the following steps: Obtain the door assembly data, and establish a finite element model of the door based on the door assembly data; Sensitivity analysis is performed on the door components to obtain sensitivity analysis results. Based on the finite element model of the door, the sensitivity analysis results, and the deformation variables of the door, multiple door model samples are generated. The stiffness analysis results of each door model sample are calculated based on the pre-set stiffness analysis conditions. A door stiffness dataset is generated based on multiple door model samples and the stiffness analysis results of each door model sample. A prediction model is trained using the door stiffness dataset, and the door stiffness is predicted based on the trained prediction model.

2. The door stiffness prediction method according to claim 1, characterized in that, The door assembly data includes multiple components such as the outer door panel, inner door panel, window frame, waistline reinforcement plate, anti-collision beam, glass guide rail, hinge reinforcement plate, lock reinforcement plate, upper hinge, and lower hinge.

3. The door stiffness prediction method according to claim 1, characterized in that, The step of establishing a finite element model of the door based on the door assembly data includes: A three-dimensional model of the door is created based on the door assembly data. A finite element model of the car door is established based on the 3D model of the car door, the definition of material properties, and the thickness information.

4. The door stiffness prediction method according to claim 1, characterized in that, The sensitivity analysis results obtained by performing sensitivity analysis on the door components include: Obtain the thickness variation data of the door component; Based on the thickness variation data, analyze the degree of impact of thickness variation on the car door; The sensitivity analysis results are determined based on the influence level value.

5. The method for predicting door stiffness according to claim 1, characterized in that, The process of generating multiple door model samples based on the finite element model of the door, the sensitivity analysis results, and the door deformation variables includes: Obtain the numbering rules for the coupling points, loading points, and constraint points of the door model sample; Based on the sensitivity analysis results and the door deformation variables, multiple target design parameters are generated, and multiple door model samples are generated based on the multiple target design parameters and the numbering rules.

6. The method for predicting door stiffness according to claim 5, characterized in that, Before generating multiple door model samples based on the aforementioned finite element model of the door, the aforementioned sensitivity analysis results, and the door deformation variables, the process further includes: Establish the coordinate system for the loading point of each car door model sample; Place the initial position coordinates of each loading point at the origin of the coordinate system; The displacement value of the door model sample is extracted by reading the final position coordinates of the loading point, and the stiffness value of the door model sample is determined based on the displacement value of the door model sample.

7. The method for predicting door stiffness according to claim 1, characterized in that, The calculation of stiffness analysis results for each door model sample based on pre-set stiffness analysis conditions includes: The loading point location, loading size, constraint conditions, and output requirements are determined based on the pre-set stiffness analysis conditions, and multiple header files are generated based on the loading point location, loading size, constraint conditions, and output requirements. The finite element solver is driven by multiple header files. Finite element calculations are performed on all door model samples by the finite element solver, and the stiffness analysis results of each door model sample are extracted from the finite element calculation results.

8. A door stiffness prediction device, characterized in that, include: The acquisition module is used to acquire door assembly data and establish a finite element model of the door based on the door assembly data; The generation module is used to perform sensitivity analysis on the door parts to obtain sensitivity analysis results, and generate multiple door model samples based on the door finite element model, the sensitivity analysis results, and the door deformation variables. The prediction module is used to calculate the stiffness analysis results of each door model sample according to the pre-set stiffness analysis conditions, generate a door stiffness dataset based on multiple door model samples and the stiffness analysis results of each door model sample, train a prediction model using the door stiffness dataset, and predict the door stiffness based on the trained prediction model.

9. A vehicle, characterized in that, The vehicle includes a door assembly, the stiffness of which is predicted based on the door stiffness prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed, they implement the door stiffness prediction method according to any one of claims 1-7.

Images