A design method and system for a bracket structure of a new energy vehicle energy system

Abstract

The present invention relates to the field of new energy vehicle technology, and specifically to a design method and system for a bracket structure of a new energy vehicle energy system. The solution of the present application first creates a CAE finite element model based on the initial CAD geometric model of the bracket structure and performs a topological optimization analysis based on machine learning to determine the reinforcing rib structure of the CAD geometric model and form a new CAD geometric model, and then creates a refined CAE finite element model based on this and performs a size and shape optimization analysis based on machine learning to achieve the size and shape optimization of the bracket structure; finally, after reconstructing the CAE finite element model based on the optimized bracket structure, a safety working condition check is performed. When the check is successful, the size and shape optimization scheme is the optimal combination scheme. If it fails after the check, the size and shape optimization analysis is performed again until the requirements are met. By adopting the scheme of the application, the optimization results can be obtained more accurately and quickly, and the quality of the structure can be reduced quickly and accurately to the greatest extent.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy vehicles, and in particular to a design method and system for a bracket structure of an energy system of a new energy vehicle. Background Art

[0002] For new energy vehicles, the power system generally uses power batteries or hydrogen fuel cells. The energy system bracket is its key load-bearing component. The quality of its performance directly affects the safety of the energy system and indirectly affects driving safety and cruising range. The structural safety of the energy system bracket of heavy-duty new energy commercial vehicles must meet the structural strength requirements under three static conditions of sharp turns, emergency braking and bumps in the whole vehicle state, as well as the dynamic characteristics requirements of modal analysis.

[0003] The current structural design of the energy system bracket of heavy-duty commercial vehicles is mainly designed based on human experience. For example, the positions and sizes of various reinforcement ribs of the bracket structure are determined manually based on experience. The initial design structure is often difficult to meet the structural safety performance requirements through CAE analysis. In order to meet the requirements, many invalid reinforcement ribs are usually added or the thickness of the parts is increased, which ultimately leads to a large overall mass of the energy system bracket structure. If CAE optimization design is not performed, it will usually lead to a large redundancy in the structural design, a large design space for lightweight and weight reduction, or the strength does not meet the requirements, and there is no effective structural reinforcement plan; if CAE optimization design is performed, traditional optimization design methods are generally used, such as topology + size optimization, or only size optimization. This optimization result is greatly affected by the input of optimization parameters, and it takes a long time of repeated iterations to obtain the optimization result. However, this optimization design method has low accuracy, slow optimization speed and high optimization cost. Summary of the invention

[0004] The present invention provides a design method and system for a bracket structure of an energy system for a new energy vehicle, which effectively solves the problems of low accuracy, slow optimization speed and high optimization cost of the existing energy system bracket during optimization design.

[0005] According to the first aspect, an embodiment provides a method for designing a bracket structure of an energy system of a new energy vehicle, comprising:

[0006] Step 100: creating an initial CAE finite element model according to the initial CAD geometric model of the energy system bracket structure;

[0007] Step 200: performing a topology optimization analysis based on machine learning on the initial CAE finite element model to obtain a topology optimization analysis result;

[0008] Step 300: determining the structure of the reinforcing ribs of the initial CAD geometric model according to the topology optimization analysis result to form a new CAD geometric model;

[0009] Step 400: Creating a refined CAE finite element model according to the new CAD geometric model;

[0010] Step 500: Performing size and shape optimization analysis based on machine learning on the refined CAE finite element model to obtain size and shape optimization analysis results;

[0011] Step 600: optimizing the size and shape of the energy system bracket structure according to the size and shape optimization analysis results to obtain an optimized energy system bracket structure;

[0012] Step 700: Reconstructing a CAD geometric model according to the optimized energy system bracket structure;

[0013] Step 800: Creating a final CAE finite element model based on the reconstructed CAD geometric model;

[0014] Step 900: Perform safety condition verification on the final CAE finite element model. If the verification is successful, the optimized energy system bracket structure is the optimal structure and the optimization is completed. If the verification fails, return to step 500 until the verification is successful.

[0015] In an achievable implementation, the creating an initial CAE finite element model according to the initial CAD geometric model of the energy system bracket structure includes:

[0016] Filling the initial CAD geometric model with external structural materials to form a simplified geometric model based on the envelope of the outer contour of the energy system bracket structure;

[0017] The initial CAE finite element model is created according to the simplified geometric model.

[0018] In an achievable implementation, the performing a machine learning-based topology optimization analysis on the initial CAE finite element model includes:

[0019] The initial CAE finite element model is placed into a trained machine learning model for topology optimization analysis to obtain a prediction result;

[0020] Verifying the prediction result through finite element analysis or experiment to obtain a verification result;

[0021] The machine learning model is iteratively optimized according to the verification result until the optimal topology optimization analysis result is obtained.

[0022] In an achievable implementation, before placing the initial CAE finite element model into a trained machine learning model for topology optimization analysis, the process further includes:

[0023] Collecting a data set of a design case; the data set includes design parameters, load conditions, boundary conditions, and performance objectives;

[0024] Extracting target features from the data set as input to the machine learning model; the target features include geometric features, material properties, and load conditions;

[0025] Training the machine learning model according to the target features and their corresponding performance targets;

[0026] The expansion-splitting method is used to enrich the training data of the machine learning model to obtain a trained machine learning model.

[0027] In an achievable implementation, determining the structure of the reinforcing ribs of the initial CAD geometric model according to the topology optimization analysis result to form a new CAD geometric model includes:

[0028] Based on the topology optimization analysis results, the stiffeners of the initial CAD geometric model are designed to obtain a new stiffener structure; the stiffener structure includes size, thickness and arrangement position;

[0029] The initial CAD geometric model is reconstructed based on the new reinforcement rib structure to form the new CAD geometric model.

[0030] In an achievable implementation, the performing of size and shape optimization analysis based on machine learning on the refined CAE finite element model includes:

[0031] Determining the size design parameters and shape design parameters of the energy system bracket structure;

[0032] Determine the optimization target and constraint conditions to optimize the size design parameters and shape design parameters to obtain the optimization design results;

[0033] The optimized design results are used as input to the trained machine learning model, and a random forest algorithm is used to perform size and shape optimization analysis.

[0034] According to the second aspect, an embodiment provides a design system for a bracket structure of a new energy vehicle energy system, including:

[0035] An initial finite element model creation module, used to create an initial CAE finite element model based on an initial CAD geometric model of the energy system bracket structure;

[0036] A topology optimization analysis module is used to perform a topology optimization analysis based on machine learning on the initial CAE finite element model to obtain a topology optimization analysis result;

[0037] A CAD geometric model optimization module, used to determine the structure of the reinforcing ribs of the initial CAD geometric model according to the topology optimization analysis result to form a new CAD geometric model;

[0038] A refined finite element model creation module, used to create a refined CAE finite element model according to the new CAD geometric model;

[0039] A size and shape optimization analysis module is used to perform size and shape optimization analysis based on machine learning on the refined CAE finite element model to obtain size and shape optimization analysis results;

[0040] A structure optimization module, used to optimize the size and shape of the energy system bracket structure according to the size and shape optimization analysis results to obtain an optimized energy system bracket structure;

[0041] A CAD geometric model reconstruction module, used to reconstruct the CAD geometric model according to the optimized energy system bracket structure;

[0042] A final finite element model creation module, used to create a final CAE finite element model based on the reconstructed CAD geometric model;

[0043] The verification module is used to verify the safety conditions of the final CAE finite element model. If the verification is successful, the optimized energy system bracket structure is the optimal structure and the optimization is completed; if the verification fails, the step of size and shape optimization analysis is returned until the verification is successful.

[0044] In an achievable implementation, the initial finite element model creation module includes:

[0045] A filling unit, used for filling the initial CAD geometric model with external structural materials to form a simplified geometric model based on the envelope of the outer contour of the energy system bracket structure;

[0046] The initial finite element model creation unit is used to create the initial CAE finite element model according to the simplified geometric model.

[0047] In an achievable implementation, the topology optimization analysis module includes:

[0048] A result prediction unit is used to put the initial CAE finite element model into a trained machine learning model for topology optimization analysis to obtain a prediction result;

[0049] A verification unit, used to verify the prediction result through finite element analysis or experiment to obtain a verification result;

[0050] A topology optimization analysis unit is used to iteratively optimize the machine learning model according to the verification result until an optimal topology optimization analysis result is obtained.

[0051] According to a third aspect, an embodiment provides a computer program product, comprising a computer program and / or instructions, wherein the computer program and / or instructions implement the method as described above when executed by a processor.

[0052] According to a design method / system for a bracket structure of an energy system for a new energy vehicle in the above-mentioned embodiment, first, an initial CAE finite element model is created according to an initial CAD geometric model of the bracket structure, and a topological optimization analysis based on machine learning is performed on the finite element model. The structure of the reinforcing ribs of the CAD geometric model is determined according to the results of the topological optimization analysis, and a new CAD geometric model is formed. Then, a refined CAE finite element model is created according to the new CAD geometric model, and then a size and shape optimization analysis based on machine learning is performed on the refined CAE finite element model, and then the size and shape of the bracket structure are optimized according to the size and shape optimization analysis results; finally, the CAD geometric model is reconstructed according to the optimized bracket structure, and then a final CAE finite element model is created according to the geometric model, and the final CAE finite element model is checked for safety working conditions. When the checked CAE finite element model meets the safety performance requirements, the size and shape optimization scheme is the optimal combination scheme. If the checked CAE finite element model does not meet the safety performance requirements, the refined CAE finite element model is re-checked for size and shape optimization based on machine learning to determine a size and shape optimization scheme that meets the requirements. By adopting the applied design method / system, the optimization results can be obtained more accurately and quickly than traditional technologies, and the accuracy of obtaining the best result in one round of optimization can be improved. The optimization plan can also better meet the actual processing technology requirements, and can achieve the maximum degree of rapid and accurate reduction of the structure quality while meeting the safety performance requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of the design method of the bracket structure of the energy system of the new energy vehicle of this embodiment;

[0054] Figure 2 A flowchart for creating an initial CAE finite element model for this embodiment;

[0055] Figure 3 This is a flow chart of performing machine learning-based topology optimization analysis on the initial CAE finite element model in this embodiment;

[0056] Figure 4 A flow chart of forming a new CAD geometric model for this embodiment;

[0057] Figure 5 This is a flow chart of performing size and shape optimization analysis based on machine learning on a refined CAE finite element model in this embodiment;

[0058] Figure 6 (a), (b), and (c) are respectively the front view, side view, and top view of the CAD geometric model of the bracket structure of the original energy system of this embodiment;

[0059] Figure 7 (a) and (b) are CAD geometric models of the bracket structure of the original energy system of this embodiment at the first and second perspectives, respectively;

[0060] Figure 8 (a) is a simplified geometric model based on the outer contour envelope of the energy system bracket structure of this embodiment, and (b) is an internal structure diagram of the simplified geometric model;

[0061] Fig. 9 (a), (b), and (c) are respectively the front view, side view, and top view of the reconstructed CAD geometric model of the energy system bracket structure of this embodiment;

[0062] Fig.10 (a) and (b) are CAD geometric models of the reconstructed energy system bracket structure in the first and second perspectives, respectively;

[0063] Fig.11 It is a structural block diagram of the design system of the bracket structure of the new energy vehicle energy system of this embodiment.

[0064] Figure numerals: 10, initial finite element model creation module; 20, topology optimization analysis module; 30, CAD geometry model optimization module; 40, refined finite element model creation module; 50, size and shape optimization analysis module; 60, structure optimization module; 70, CAD geometry model reconstruction module; 80, final finite element model creation module; 90, verification module. DETAILED DESCRIPTION

[0065] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0066] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.

[0067] The serial numbers assigned to the components herein, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning. The "connection" and "coupling" mentioned in this application, unless otherwise specified, include direct and indirect connections (couplings).

[0068] The energy system bracket structure involved in this application is mainly installed on heavy-duty commercial new energy vehicles, and is the main bearing mechanism for energy system components such as power batteries and hydrogen fuel. Its installation position on the vehicle is generally between the front of the vehicle and the car body. Unlike passenger cars, the deadweight and load of heavy-duty commercial new energy vehicles are very large, resulting in a very large volume, weight and capacity of the energy system. It is externally located behind the front of the vehicle and is only covered by an external protective cover, which needs to meet relevant safety requirements. Lightweighting is the mainstream content in current automobile design. Under the premise that the energy material itself is difficult to change, it is particularly important to optimize the lightweight design of its bracket structure.

[0069] Among the many factors that enhance the automotive R&D capabilities, computer-aided engineering (CAE) is an indispensable numerical calculation tool in product structure analysis and optimization design. It mainly uses the finite element method for simulation analysis, and can analyze the structural mechanical properties of complex products through computer-aided solution, which plays a vital role in the optimization design of energy system brackets. Artificial Intelligence (AI) is an important driving force for the new round of scientific and technological revolution and industrial transformation. With the development of AI in recent years, this application applies AI technology to CAE optimization analysis, and uses AI optimization analysis to guide product structure design faster and more accurately.

[0070] refer to Figure 1 The present embodiment provides a method for designing a bracket structure of a new energy vehicle energy system, comprising the following steps:

[0071] Step 100: The initial finite element model creation module creates an initial CAE finite element model according to the initial CAD geometric model of the energy system bracket structure.

[0072] Specifically, Figure 6 and Figure 7 The figure shows the CAD geometric model of the bracket structure of the original energy system. A CAE finite element model is created for the CAD geometric model of the bracket structure of the original energy system. After that, an initial simulation analysis of the safety working conditions is performed on the CAE finite element model. The safety working conditions include three static working conditions of sharp turning, emergency braking, and bumping of the whole vehicle and one dynamic working condition of modal analysis. The evaluation criteria are that the maximum stress value of each static working condition does not exceed the yield strength of 355MPa of the material Q355 of the reinforcement ribs in the bracket structure and the yield strength of 500MPa of the material 610L of the main load-bearing beam, and the first-order modal frequency of the modal analysis is not less than 20Hz.

[0073] For example, according to the initial analysis results, the maximum structural stress of material Q355 in the sharp turning condition is 461MPa and the maximum structural stress of material 610L is 873MPa, the maximum structural stress of material Q355 in the emergency braking condition is 345MPa and the maximum structural stress of material 610L is 745MPa, the maximum structural stress of material Q355 in the bumping condition is 267MPa and the maximum structural stress of material 610L is 431MPa. Only the bumping condition meets the strength requirements, the first-order mode is 12Hz, and the modal requirement of not less than 20Hz is not met. Therefore, the initial structure of the bracket structure does not meet the safety performance requirements, and optimization analysis is required to obtain the best structural form solution that meets the safety performance and at the same time ensures that the total mass of the structure is minimized.

[0074] refer to Figure 2 , step 100 specifically includes the following steps:

[0075] Step 110: Filling the initial CAD geometric model with external structural materials through filling cells to form a simplified geometric model based on the envelope of the outer contour of the energy system bracket structure;

[0076] Step 120: The initial finite element model creation unit creates an initial CAE finite element model according to the simplified geometric model.

[0077] Specifically, the original CAD geometric model of the energy system bracket structure is filled with external structural materials, and a new geometric model is formed based on the outer contour of the original bracket structure, in which the internal cavity structure remains unchanged. From an external perspective, the entire bracket is basically simplified to a cube.

[0078] Since topology optimization is a mathematical method that optimizes material distribution in a given area based on given load conditions, constraints and performance indicators, it is also called structural layout optimization. With material distribution as the optimization object, through topology optimization, the best material distribution scheme can be found in the design space of uniformly distributed materials. The optimization analysis in this step mainly performs topology optimization on the reinforcement area around the outside of the bracket structure. The purpose is to find the best arrangement of the reinforcements around it while meeting the requirements of safety analysis, including size, position, quantity, etc. Therefore, before performing topology optimization, the geometric model of the bracket structure must be filled, the reinforcement structure around the outer surface of the original geometric model must be deleted, and the surrounding space must be filled with material. The filling thickness is the same as the thickness of the original reinforcement to form an outer contour envelope model of the bracket structure, such as Figure 8 Then, a corresponding CAE roughened finite element model is created for the bracket outer contour envelope CAD model for subsequent topology optimization analysis.

[0079] Step 200: The topology optimization analysis module performs a topology optimization analysis based on machine learning on the initial CAE finite element model to obtain a topology optimization analysis result.

[0080] Based on the above four safety analysis conditions (three static conditions of sharp turn, emergency braking and bump, and one dynamic condition of modal analysis), the machine learning-based topology optimization (MLTO) method is used to create optimization variables, optimization constraints and optimization targets for the initial CAE finite element model of the bracket structure and perform optimization analysis. The material of the bracket structure can be optimized through topology optimization analysis, and the density of units that have little impact on safety analysis can be reduced to 0, that is, the unit is deleted. The machine learning-based topology optimization (MLTO) method is used for multiple rounds of iterations to finally form the optimal material distribution. According to the optimal material distribution, it can be determined which positions have a higher density, that is, reinforcing ribs need to be arranged at these positions.

[0081] Step 300: The CAD geometric model optimization module determines the structure of the reinforcement ribs of the initial CAD geometric model according to the topology optimization analysis results to form a new CAD geometric model.

[0082] Based on the results of topological optimization analysis, the dimensions of the new reinforcement rib structure are designed while meeting the requirements of the bracket structure processing technology. Then, the CAD geometric model of the new energy system bracket structure (i.e., the new CAD geometric model) is reconstructed, and the corresponding CAE refined finite element model is created before the safety working conditions are checked again.

[0083] refer to Figure 4 , step 300 specifically includes the following steps:

[0084] Step 310: Designing the stiffener of the initial CAD geometric model based on the topology optimization analysis results to obtain a new stiffener structure; the stiffener structure includes size, thickness and arrangement position;

[0085] Step 320: Reconstruct the initial CAD geometric model based on the new reinforcement rib structure to form a new CAD geometric model.

[0086] Specifically, first, based on the topology optimization results, the stiffeners of the initial CAD geometric model are designed. Under the condition of satisfying the processing technology of the bracket structure parts, the new stiffener structure is determined, including its size, thickness and position, etc. Then, the geometry is reconstructed based on the new stiffener layout to form a new CAD geometric model of the energy system bracket structure, such as Fig. 9 and Fig.10 shown.

[0087] Step 400: The refined finite element model creation module creates a refined CAE finite element model according to the new CAD geometric model.

[0088] Based on the reconstructed CAD geometric model of the energy system bracket structure, the corresponding CAE finite element model is created, and the safety condition analysis and verification are performed on it. The evaluation criteria of the simulation results remain unchanged. Specifically, the reconstructed bracket structure has a maximum structural stress of 365MPa for material Q355 and 571MPa for material 610L in the sharp turning condition, a maximum structural stress of 307MPa for material Q355 and 537MPa for material 610L in the emergency braking condition, and a maximum structural stress of 237MPa for material Q355 and 422MPa for material 610L in the bumping condition. Only the bumping condition still meets the strength requirements, and the first-order mode is 17.8Hz, which still does not meet the modal requirement of not less than 20Hz, but it has been significantly improved compared with the original structural results, and there is room for further optimization and strengthening, so further optimization design, namely size and shape optimization analysis, is carried out. It should be noted that even if the reconstructed result meets the stress requirements, subsequent size and shape optimization analysis can still be performed. Through size and shape optimization analysis, a lightweight structural size combination solution that meets the stress requirements can be obtained.

[0089] Step 500: Performing size and shape optimization analysis based on machine learning on the refined CAE finite element model through the size and shape optimization analysis module to obtain size and shape optimization analysis results.

[0090] Step 600: The structure optimization module optimizes the size and shape of the energy system bracket structure according to the size and shape optimization analysis results to obtain an optimized energy system bracket structure.

[0091] Specifically, the thickness and cross-sectional shape of the components of the bracket structure that need to be optimized are used as design parameters, the maximum structural stress of Q355 material is less than 355MPa as optimization constraint 1, the maximum structural stress of 610L material is less than 500MPa as optimization constraint 2, the first-order mode of the structure is greater than 20Hz as optimization constraint 3, and the minimum structural mass is used as the optimization target. The size and shape optimization analysis based on machine learning is adopted to obtain the optimal size combination scheme of the components of the bracket structure while meeting the safety performance requirements.

[0092] Step 700: The CAD geometric model reconstruction module reconstructs the CAD geometric model according to the optimized energy system bracket structure.

[0093] Step 800: The final finite element model creation module creates a final CAE finite element model based on the reconstructed CAD geometric model.

[0094] Step 900: The verification module verifies the safety conditions of the final CAE finite element model. If the verification is successful, the process proceeds to step 901: the optimized energy system bracket structure is the optimal structure and the optimization ends. If the verification fails, the process returns to step 500 until the verification is successful.

[0095] Based on the size and shape optimization analysis results of machine learning, the thickness size, shape width are regularized to reconstruct a new energy system bracket structure CAD structural geometry model while meeting the requirements of the bracket structure's parts processing technology and bracket structure assembly. Among them, after the sizes of all optimized design parts are updated, it is necessary to ensure that the installation space of the internal energy device is not affected and the external space does not interfere with other components such as the car's front and compartment.

[0096] Then, the corresponding CAE finite element model (and the final CAE finite element model) is generated based on the reconstructed geometric model, and the finite element model is subjected to safety condition analysis to check whether the reconstructed structure meets the safety condition requirements. If the structural results meet the performance requirements, the optimization is completed; if any safety condition after verification does not meet the safety performance requirements, that is, the maximum stress or the first-order mode does not meet the requirements, then return to step 500 and re-determine a new size combination scheme until the optimal size combination scheme that meets the requirements is found.

[0097] Specifically, the final result after verification is that the cross beam is 2.4mm thick and 25mm wide, the longitudinal beam is 2.4mm thick and 40mm wide, the vertical beam is 3.6mm thick and 70mm wide, the left and right oblique beams are 3.0mm thick and 25mm wide, the front and rear oblique beams are 3.0mm thick and 40mm wide, the bottom support beam is 5.5mm thick and 80mm wide, the maximum structural stress of the optimized sharp turning condition belonging to material Q355 is 254MPa, the maximum structural stress of the material 610L is 464MPa, and the emergency braking condition belongs to material The maximum structural stress of material Q355 is 309MPa, and the maximum structural stress of material 610L is 412MPa. The maximum structural stress of material Q355 under bumpy working condition is 247MPa, and the maximum structural stress of material 610L is 455MPa. The maximum stresses are all less than the corresponding material yield strength. The first-order mode is 21.5Hz, which is greater than 20Hz. Therefore, the safety working condition meets the analysis requirements. The size and shape optimization scheme is the final optimal combination scheme, and the energy system bracket optimization design is completed.

[0098] The design method of the bracket structure of the energy system of the new energy vehicle of this embodiment optimizes and analyzes the bracket structure of the energy system based on AI technology, specifically: firstly, the arrangement mode and position of the bracket outer surface reinforcement ribs are determined through the topological optimization (MLTO) analysis based on machine learning, and then the specific size and thickness of the reinforcement ribs are determined through the size and shape optimization analysis based on machine learning, so as to obtain the optimal bracket structure arrangement scheme while meeting the safety performance requirements of the whole vehicle, and at the same time achieve the lightweight of the bracket. Through CAE optimization analysis based on machine learning, the mass of the energy system bracket structure can be minimized while meeting the safety requirements, which can not only meet the safety requirements of the whole vehicle, but also minimize the manufacturing cost of the energy system bracket structure, and reduce the weight to improve the relevant performance of the vehicle.

[0099] refer to Figure 3 , a machine learning-based topology optimization analysis is performed on the initial CAE finite element model, which specifically includes the following steps:

[0100] Step 210: The result prediction unit puts the initial CAE finite element model into the trained machine learning model for topology optimization analysis to obtain a prediction result;

[0101] Step 220: The verification unit verifies the prediction result through finite element analysis or experiment to obtain a verification result;

[0102] Step 230: The topology optimization analysis unit iteratively optimizes the machine learning model according to the verification result until the optimal topology optimization analysis result is obtained.

[0103] Traditional topology optimization generally uses the SIMP variable density method, which only performs single optimization on the optimized structure according to given loads, constraints, design variables, design goals, etc. The machine learning-based topology optimization (MLTO) analysis of this embodiment is an advanced design technology that combines machine learning technology and topology optimization methods. It uses machine learning algorithms to predict and optimize material distribution in structural design to achieve optimal performance within a given design space. The core idea of ​​this method is to use a machine learning model to learn the relationship between design parameters, load conditions, boundary conditions, and performance goals, thereby guiding the topology optimization process.

[0104] Specifically, when the method of the present application is analyzed by topology optimization based on machine learning, first, the trained machine learning model is applied to the new optimization problem to directly predict the optimized topology of the basic unit with regular shape under arbitrary load and boundary conditions (that is, the initial CAE finite element model is placed in the trained machine learning model for topology optimization analysis). The sensitivity information is predicted by the machine learning model, thereby eliminating the need to solve the state equation in fine resolution discretization (i.e., design parameterized density variables) during the topology optimization process, significantly accelerating the optimization process. Then, the prediction results of the machine learning model are verified by finite element analysis or experiments. Finally, the machine learning model is adjusted according to the verification results, and iterative optimization is performed until the best design solution (i.e., the optimal topology optimization analysis results) is found. According to the optimal topology optimization analysis results, areas with high unit density can be obtained. According to the area and position of these areas, the force transmission path of the bracket structure can be known. At the same time, considering the processing requirements of the bracket structure, the layout position of the reinforcement ribs and the shape, size, thickness and number of the reinforcement ribs are preliminarily determined.

[0105] Furthermore, before step 210, the machine learning model needs to be trained, including:

[0106] Collect data sets for design cases; the data sets include design parameters, load conditions, boundary conditions, and performance objectives;

[0107] Extract target features from the dataset as input to the machine learning model; target features include geometric features, material properties, and load conditions;

[0108] Train the machine learning model based on the target features and their corresponding performance goals;

[0109] The expansion-split method is used to enrich the training data of the machine learning model to obtain a trained machine learning model.

[0110] Specifically, first, data sets of design cases are collected or generated. These data sets include design parameters, load conditions, boundary conditions, and performance targets (performance targets such as stress distribution, displacement, etc.). Taking this scheme as an example, the optimized design parameters of this scheme are the material density of the material filling area; the load conditions are the loads corresponding to three static working conditions (sharp turn, emergency braking, and bumping), among which the sharp turn working condition load is 3g gravity acceleration in the y direction and 5g gravity acceleration in the z direction, the emergency braking working condition load is 3g gravity acceleration in the x direction and 5g gravity acceleration in the z direction, and the bumping working condition load is 6g gravity acceleration in the z direction; the boundary conditions are to constrain the six degrees of freedom of the bolt holes at the connection position between the bracket and the frame longitudinal beam; the performance constraints are that the maximum stress of the structural area to which the Q355 material belongs is less than 355MPa, the maximum stress of the structural area to which the 610L material belongs is less than 500MPa, the first-order mode is greater than 20Hz, and the volume of the optimized area is reduced by at least 70%; the optimization goal is to minimize the comprehensive weighted strain energy of the four analysis conditions.

[0111] Secondly, feature extraction is performed: useful features (i.e., target features) are extracted from the data set. These target features may include geometric features, material properties, load conditions, etc. These target features will serve as input to the machine learning model.

[0112] Then, the extracted target features and the performance targets corresponding to the features are used to train machine learning models, including deep neural networks, support vector machines, random forests, etc. Machine learning models are trained to learn optimization strategies rather than just optimizing results. This may involve a long short-term memory neural network (SaConvLSTM) model that uses a self-attention mechanism and convolution operations.

[0113] Afterwards, the expansion-split method is used to enrich the training data of the model and enhance the generalization ability of the model. In addition, the splicing strategy is applied to process the design domain of arbitrary shape, split the design domain into multiple basic units, and then splice the final result according to the shape of the design domain.

[0114] Machine learning-based topology optimization (MLTO) relies on a large number of design cases and corresponding performance data. The relevant data used in this application are all from CAE analysis results, experimental data or historical design experience of a large number of similar structural products in the early stage. By training machine learning models (such as deep neural networks, support vector machines, random forests, etc.), MLTO can learn complex nonlinear relationships and predict the impact of different design parameters on performance. MLTO can significantly reduce the number of iterations and calculation time required for traditional topology optimization methods (such as SIMP method) because it can directly predict the optimal or near-optimal design instead of approximating it through step-by-step iteration. In addition, MLTO can also help designers quickly explore and evaluate different design schemes in a larger design space, improving the flexibility and innovation of the design. Moreover, MLTO can predict the performance of the structure under different loads and boundary conditions, providing a scientific basis for design decisions. The trained MLTO model has a certain generalization ability and can be applied to new and unseen design cases to provide guidance for design.

[0115] refer to Figure 5 , the size and shape optimization analysis of the refined CAE finite element model based on machine learning is carried out, which specifically includes the following steps:

[0116] Step 510: Determine the size design parameters and shape design parameters of the energy system bracket structure;

[0117] Step 520: Determine the optimization target and constraint conditions to optimize the size design parameters and shape design parameters to obtain the optimization design results;

[0118] Step 530: Use the optimized design results as input to the trained machine learning model and use the random forest algorithm to perform size and shape optimization analysis.

[0119] Specifically, first, a machine learning-based size and shape optimization analysis is performed on the refined CAE finite element model. According to the actual requirements of the processing technology and assembly of the parts of the bracket structure, the size design parameters of the bracket structure are determined to be the thickness of the six parts, namely the crossbeam, longitudinal beam, vertical beam, left and right diagonal beams, front and rear diagonal beams, and bottom support beam. The optimization design range of each size design parameter is 2mm to 6mm, that is, the minimum thickness is 2mm and the maximum is 6mm. The cross-section of each component is a square with the same length and width. The shape design parameters of the bracket structure are determined to be the cross-sectional width of the six parts, namely the cross-sectional width of the six parts, namely the longitudinal beam, vertical beam, left and right diagonal beams, front and rear diagonal beams, and bottom support beam. The optimization design range of each shape design parameter is 20mm to 80mm, that is, the minimum width is 20mm and the maximum is 80mm. A total of 12 design parameters are optimized for size and shape.

[0120] Secondly, the maximum structural stress of Q355 material is less than 355MPa as optimization constraint 1, the maximum structural stress of 610L material is less than 500MPa as optimization constraint 2, the first-order mode of the structure is greater than 20Hz as optimization constraint 3, and the minimum structural mass is taken as the optimization goal. According to this optimization goal and constraints, through optimization design, a result with minimum mass and stress that meets the safety analysis requirements can be finally obtained.

[0121] Then, after determining the optimization design inputs required for the first two steps, the random forest algorithm based on machine learning is used to perform size and shape optimization analysis based on safe working conditions. It should be noted that the random forest algorithm is an integrated learning method that improves the accuracy of classification or regression prediction by constructing multiple decision trees and combining their prediction results. Its core idea is to use randomness and diversity to enhance the generalization ability of the model. Before optimization, a DOE (Design of Experiments) analysis is first performed to provide a set of sample data for subsequent optimization. This time, Latin HyperCube sampling is used, which is a method of approximate random sampling from a multivariate parameter distribution. In a stratified sampling form, the value range of each component is divided into the same interval, and uniform sampling is performed in each interval to reduce the correlation between input variables, thereby improving the simulation accuracy. 500 samples are selected in this application, each sample includes 12 design parameters for optimization design, and there are a total of 500 size combinations within their respective design ranges, and the corresponding 500 sets of simulation analysis results are calculated. These 500 sets of data are used as inputs to the random forest algorithm for model training to learn the relationship between the input design variables and the optimization objectives. Then, the hyperparameters of the random forest model are determined, such as the number of trees, the maximum depth, the minimum sample ratio required for splitting, etc. Grid search, random search, or Bayesian optimization can be used to find the optimal hyperparameter combination in the defined hyperparameter space. Then, appropriate evaluation indicators such as accuracy and mean square error are used to evaluate the performance of the random forest model on the validation set, and then the results of the model prediction are analyzed to determine whether the model can accurately predict the size optimization target. Finally, after multiple rounds of iterations, the optimal design parameter combination that meets the optimization objectives and constraints can be obtained.

[0122] The optimal size combination scheme can be obtained through multiple iterations. In this experiment, the final optimization results are as follows: the cross beam is 2.1mm thick and 24mm wide, the longitudinal beam is 2.5mm thick and 36mm wide, the vertical beam is 3.7mm thick and 66mm wide, the left and right oblique beams are 3.1mm thick and 25mm wide, the front and rear oblique beams are 2.8mm thick and 41mm wide, and the bottom support beam is 5.5mm thick and 78mm wide. After optimization, the maximum structural stress of material Q355 in sharp turning conditions is 265MPa, and the maximum structural stress of material 610L is 471MPa, the maximum structural stress of material Q355 in emergency braking conditions is 322MPa, and the maximum structural stress of material 610L is 437MPa, the maximum structural stress of material Q355 in bumpy conditions is 287MPa, and the maximum structural stress of material 610L is 489MPa. The maximum stresses are all less than the corresponding material yield strength. The first-order mode is 21.1Hz, which is greater than 20Hz. Therefore, the safety conditions meet the analysis requirements. The total mass of the energy system bracket structure can be reduced by 8.3% based on the comparison before and after size and shape optimization, achieving the optimal quality. In addition, if the result of the last step of the optimization iteration does not meet the analysis requirements of the safety condition, you can check the previous step of the iteration until you find the optimal size combination solution that meets the analysis requirements.

[0123] refer to Fig.11 The present embodiment provides a design system for a bracket structure of an energy system of a new energy vehicle, comprising an initial finite element model creation module 10, a topology optimization analysis module 20, a CAD geometric model optimization module 30, a refined finite element model creation module 40, a size and shape optimization analysis module 50, a structural optimization module 60, a CAD geometric model reconstruction module 70, a final finite element model creation module 80 and a verification module 90.

[0124] The initial finite element model creation module 10 is used to create an initial CAE finite element model according to the initial CAD geometric model of the energy system bracket structure; the topology optimization analysis module 20 is used to perform a topology optimization analysis based on machine learning on the initial CAE finite element model to obtain a topology optimization analysis result; The CAD geometric model optimization module 30 is used to determine the structure of the reinforcement ribs of the initial CAD geometric model according to the topological optimization analysis results to form a new CAD geometric model; the refined finite element model creation module 40 is used to create a refined CAE finite element model according to the new CAD geometric model; the size and shape optimization analysis module 50 is used to perform size and shape optimization analysis based on machine learning on the refined CAE finite element model to obtain the size and shape optimization analysis results; the structure optimization module 60 is used to optimize the size and shape of the energy system bracket structure according to the size and shape optimization analysis results to obtain the optimized energy system bracket structure; the CAD geometric model reconstruction module 70 is used to reconstruct the CAD geometric model according to the optimized energy system bracket structure; the final finite element model creation module 80 is used to create the final CAE finite element model according to the reconstructed CAD geometric model; the verification module 90 is used to perform safety working condition verification on the final CAE finite element model. If the verification is successful, the optimized energy system bracket structure is the optimal structure and the optimization is completed; if the verification fails, it returns to step 500 of the size and shape optimization analysis until the verification is successful.

[0125] The design system of the energy system bracket structure of the new energy vehicle of this embodiment optimizes and analyzes the energy system bracket structure based on AI technology, specifically: firstly, the arrangement mode and position of the bracket outer surface reinforcement ribs are determined through topological optimization (MLTO) analysis based on machine learning, and then the specific size and thickness of the reinforcement ribs are determined through size and shape optimization analysis based on machine learning, so as to obtain the optimal bracket structure arrangement scheme while meeting the safety performance requirements of the whole vehicle, and at the same time achieve lightweight brackets. CAE optimization analysis based on machine learning can minimize the mass of the energy system bracket structure while meeting the safety requirements, which can not only meet the safety requirements of the whole vehicle, but also minimize the manufacturing cost of the energy system bracket structure, and reduce the weight to improve the relevant performance of the vehicle.

[0126] In view of the fact that the roles and functions of the initial finite element model creation module 10, the topology optimization analysis module 20, the CAD geometric model optimization module 30, the refined finite element model creation module 40, the size and shape optimization analysis module 50, the structural optimization module 60, the CAD geometric model reconstruction module 70, the final finite element model creation module 80 and the verification module 90 have been described in detail in the above embodiment, this embodiment will not go into details here.

[0127] Furthermore, the initial finite element model creation module 10 includes a filling unit and an initial finite element model creation unit. The filling unit is used to fill the initial CAD geometric model with external structural materials to form a simplified geometric model based on the outer contour envelope of the energy system bracket structure; the initial finite element model creation unit is used to create an initial CAE finite element model based on the simplified geometric model.

[0128] Specifically, the external structural material is filled into the original CAD geometric model of the energy system bracket structure through the filling unit, and a new geometric model is formed based on the outer contour of the original structure of the bracket, in which the internal cavity structure remains unchanged. From an external perspective, the entire bracket is basically simplified to a cube.

[0129] Since topology optimization is a mathematical method that optimizes material distribution in a given area based on given load conditions, constraints and performance indicators, it is also called structural layout optimization. With material distribution as the optimization object, through topology optimization, the best material distribution scheme can be found in the design space of uniformly distributed materials. The optimization analysis in this step mainly performs topology optimization on the reinforcement area around the outside of the bracket structure. The purpose is to find the best arrangement of the reinforcements around it while meeting the requirements of safety analysis, including size, position, quantity, etc. Therefore, before performing topology optimization, the geometric model of the bracket structure must be filled, the reinforcement structure around the outer surface of the original geometric model must be deleted, and the surrounding space must be filled with material. The filling thickness is the same as the thickness of the original reinforcement to form an outer contour envelope model of the bracket structure, such as Figure 8 Then, the initial finite element model creation unit is used to create a corresponding CAE roughened finite element model for the bracket outer contour envelope CAD model for subsequent topology optimization analysis.

[0130] Furthermore, the topology optimization analysis module 20 includes a result prediction unit, a verification unit and a topology optimization analysis unit. The result prediction unit is used to put the initial CAE finite element model into the trained machine learning model for topology optimization analysis to obtain a prediction result; the verification unit is used to verify the prediction result through finite element analysis or experiment to obtain a verification result; the topology optimization analysis unit is used to iteratively optimize the machine learning model according to the verification result until the optimal topology optimization analysis result is obtained.

[0131] Specifically, when applying the machine learning-based topology optimization for analysis, the present embodiment first applies the trained machine learning model to the new optimization problem, and directly predicts the optimized topology of the basic unit with regular shape under arbitrary load and boundary conditions through the result prediction unit (that is, the initial CAE finite element model is placed in the trained machine learning model for topology optimization analysis). The machine learning model predicts the sensitivity information, thereby eliminating the need to solve the state equation in the fine resolution discretization (i.e., the density variable of the design parameterization) during the topology optimization process, and significantly accelerates the optimization process. Then, the verification unit verifies the prediction results of the machine learning model through finite element analysis or experiments. Finally, the topology optimization analysis unit adjusts the machine learning model according to the verification results and performs iterative optimization until the best design solution (i.e., the optimal topology optimization analysis result) is found. According to the optimal topology optimization analysis result, the area with high unit density can be obtained. According to the area and position of these areas, the force transmission path of the bracket structure can be known. At the same time, considering the processing requirements of the bracket structure, the arrangement position of the stiffener and the shape, size, thickness and quantity of the stiffener are preliminarily determined.

[0132] The present embodiment provides a computer program product, including a computer program and / or instructions, which implement the above method when executed by a processor. Since the design method of the bracket structure of the energy system of the new energy vehicle has been described in detail in the above embodiment, this embodiment will not be described in detail here.

[0133] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0134] The above specific examples are used to illustrate the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the art, according to the concept of the present invention, some simple deductions, modifications or substitutions can be made.

Claims

1. A design method for a bracket structure of a new energy vehicle energy system, characterized in that: include: Step 100: creating an initial CAE finite element model according to the initial CAD geometric model of the energy system bracket structure; Step 200: Collecting a data set of design cases; The data set includes design parameters, load conditions, boundary conditions, and performance targets; target features are extracted from the data set as inputs to the machine learning model; the target features include geometric features, material properties, and load conditions; Training the machine learning model according to the target features and their corresponding performance targets; Enriching the training data of the machine learning model by using an expansion-splitting method to obtain a trained machine learning model; The initial CAE finite element model is placed in a trained machine learning model for topology optimization analysis to obtain a prediction result; the prediction result is verified by finite element analysis or experiment to obtain a verification result; the machine learning model is iteratively optimized according to the verification result until an optimal topology optimization analysis result is obtained; Step 300: determining the structure of the reinforcing ribs of the initial CAD geometric model according to the topology optimization analysis result to form a new CAD geometric model; Step 400: Creating a refined CAE finite element model according to the new CAD geometric model; Step 500: Determine the size design parameters and shape design parameters of the energy system bracket structure; Determine the optimization target and constraint conditions to optimize the size design parameters and shape design parameters to obtain the optimization design results; The optimization design result is used as the input of the trained machine learning model, and a random forest algorithm is used to perform size and shape optimization analysis to obtain size and shape optimization analysis results; Step 600: optimizing the size and shape of the energy system bracket structure according to the size and shape optimization analysis results to obtain an optimized energy system bracket structure; Step 700: Reconstructing a CAD geometric model according to the optimized energy system bracket structure; Step 800: Creating a final CAE finite element model based on the reconstructed CAD geometric model; Step 900: performing safety condition verification on the final CAE finite element model. If the verification is successful, the optimized energy system bracket structure is the optimal structure, and the optimization ends. If the verification fails, the process returns to step 500 until the verification succeeds.

2. The design method according to claim 1, characterized in that: The step of creating an initial CAE finite element model based on the initial CAD geometric model of the energy system bracket structure comprises: Filling the initial CAD geometric model with external structural materials to form a simplified geometric model based on the envelope of the outer contour of the energy system bracket structure; The initial CAE finite element model is created according to the simplified geometric model.

3. The design method according to claim 1, characterized in that: Determining the structure of the reinforcing ribs of the initial CAD geometric model according to the topology optimization analysis result to form a new CAD geometric model includes: Based on the topology optimization analysis results, the stiffeners of the initial CAD geometric model are designed to obtain a new stiffener structure; the stiffener structure includes size, thickness and arrangement position; The initial CAD geometric model is reconstructed based on the new reinforcement rib structure to form the new CAD geometric model.

4. A design system for a bracket structure of a new energy vehicle energy system, used to implement the method as described in any one of claims 1 to 3, characterized in that: The design system includes: An initial finite element model creation module, used to create an initial CAE finite element model based on an initial CAD geometric model of the energy system bracket structure; A topology optimization analysis module is used to perform a topology optimization analysis based on machine learning on the initial CAE finite element model to obtain a topology optimization analysis result; A CAD geometric model optimization module, used to determine the structure of the reinforcing ribs of the initial CAD geometric model according to the topology optimization analysis result to form a new CAD geometric model; A refined finite element model creation module, used to create a refined CAE finite element model according to the new CAD geometric model; A size and shape optimization analysis module is used to perform size and shape optimization analysis based on machine learning on the refined CAE finite element model to obtain size and shape optimization analysis results; A structure optimization module, used to optimize the size and shape of the energy system bracket structure according to the size and shape optimization analysis results to obtain an optimized energy system bracket structure; A CAD geometric model reconstruction module, used to reconstruct the CAD geometric model according to the optimized energy system bracket structure; A final finite element model creation module, used to create a final CAE finite element model based on the reconstructed CAD geometric model; The verification module is used to verify the safety conditions of the final CAE finite element model. If the verification is successful, the optimized energy system bracket structure is the optimal structure and the optimization is completed; if the verification fails, the step of size and shape optimization analysis is returned until the verification is successful.

5. The design system according to claim 4, characterized in that The initial finite element model creation module includes: A filling unit, used for filling the initial CAD geometric model with external structural materials to form a simplified geometric model based on the envelope of the outer contour of the energy system bracket structure; The initial finite element model creation unit is used to create the initial CAE finite element model according to the simplified geometric model.

6. The design system according to claim 4, characterized in that The topology optimization analysis module includes: A result prediction unit is used to put the initial CAE finite element model into a trained machine learning model for topology optimization analysis to obtain a prediction result; A verification unit, used to verify the prediction result through finite element analysis or experiment to obtain a verification result; A topology optimization analysis unit is used to iteratively optimize the machine learning model according to the verification result until an optimal topology optimization analysis result is obtained.

7. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or the instructions are executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images