Road noise rapid analysis method and system for vehicle body dynamic parameter adaptation

Abstract

The invention discloses a road noise rapid analysis method and system for vehicle body dynamic parameter adaptation, and relates to the field of vehicle body noise control, and the method comprises the steps: obtaining an input parameter of a road noise performance prediction model; building a road noise vehicle body system and a whole vehicle simulation model, and carrying out test verification on the simulation model; compiling codes to realize batch modification of simulation model parameters and simulation results; obtaining and generating road noise performance training sample data in batches; a multi-path multi-parameter multi-level data driving model is adopted to construct a road noise performance rapid analysis method oriented to vehicle body dynamic parameter adaptation, a vehicle body system is considered through the data prediction model on the basis of traditional suspension analysis, structure adjustment is carried out on a developed model architecture in combination with a current vehicle model, and the vehicle body dynamic parameter adaptation-oriented road noise performance rapid analysis method is established. New task requirements can be quickly met, the research efficiency is improved, and the research cost is remarkably reduced.

Description

[0001] Technical Field This application relates to the field of vehicle body noise control, and more particularly to a method, system, electronic device and storage medium for rapid road noise analysis adapted to vehicle body dynamic parameters.

[0002] Background Technology: Road noise not only significantly affects the subjective experience of drivers and passengers but also poses a threat to vehicle safety, directly impacting market competitiveness and sales. Therefore, in-depth research on road noise is of great significance. Traditional road noise analysis methods have limitations. Data-driven methods can learn from large-scale noise data, but they are constrained by sample size and require high-quality data. Furthermore, once established, predictive models often cannot quickly adapt to the rapid updates and replacements in the automotive industry. Therefore, rapid modeling and generalization of road noise data-driven models are crucial. Currently, many automakers have adopted the practice of sharing chassis and suspension structures to accelerate the iteration of new models. However, this approach still requires full-vehicle modeling and assembly for verification. Due to the difficulty of simulation modeling and the time-consuming simulation analysis, analyzing road noise in newly designed models remains challenging.

[0003] Among the publicly available information related to road noise, such as the Chinese invention patent, titled "An Active Road Noise Control Method," application number CN202110326109.4, this patent discloses an active road noise control method. This patent collects real-time sound signals from actual measurement points inside the vehicle, obtains a reference signal within the vehicle, and then feeds the reference signal and the virtual point sound signal into a time-frequency dual-thread adaptive filtering stage to obtain the excitation signal for the loudspeaker. This ensures the algorithm can be effectively implemented on a low-cost chip while solving the interference problem between the microphone and passengers in road noise control.

[0004] Chinese invention patent, invention title: Vehicle road noise improvement device, application number: CN202411547956.3

[0005] This patent discloses a vehicle road noise reduction device, which includes a vehicle body sheet metal, a vibration acceleration sensor, and a control assembly. The vehicle body sheet metal has an open cavity, the vibration acceleration sensor is mounted on the sheet metal, and the control assembly consists of a container and a pump. Through the coordinated operation of the vibration acceleration sensor and the pump, the injection or discharge volume of fluid is dynamically adjusted to effectively cope with different driving conditions and noise levels. All of the above inventions have employed experimental methods to investigate automotive road noise, but these methods suffer from high costs and low efficiency due to repeated testing.

[0006] Chinese invention patent, titled "Tire Modeling Method, Apparatus, and Equipment for Vehicle Road Noise Simulation," application number CN202411404837.2, discloses a tire simulation modeling method and optimization method based on road noise problems. This method first establishes a tire structure model and a tire cavity model, and then constructs a coupling relationship based on these models. The objective function is determined by comparing the correlation coefficient between the transfer function calculated by the finite element model and the transfer function obtained from bench testing. Finally, a tire model suitable for vehicle road noise simulation is obtained through objective optimization.

[0007] Chinese invention patent, titled "A Vehicle Road Noise Analysis Method Based on a Physical Tire Model," application number CN202110968521.6, discloses a random vibration road noise analysis method that combines road noise sensitive frequency band extraction based on a virtual sample database with noise normal distribution fitting. This method overcomes the problems of road noise simulation's inability to assess the deviation in noise calculation curves caused by tire parameter uncertainties, and the inability to guarantee the effectiveness of virtual road surface technology in predicting NVH performance in the early stages of vehicle development. While simulation model methods for analyzing road noise problems avoid the high costs and time consumption associated with repeated testing to some extent, the acquisition of key model parameters and the finite element modeling process remain overly complex.

[0008] Chinese invention patent, titled "Active Control Method and Device for Vehicle Road Noise Based on Elman Neural Network," application number CN202411349707.3, discloses a data-driven analysis method for vehicle road noise. This method first acquires vibration acceleration data and in-vehicle noise data during vehicle operation using sensors. Then, it trains an Elman neural network using this data. Next, the predicted noise signal output by the Elman neural network is used as a reference signal and input into a filter to generate a filtered signal. This filtered signal is played in the target noise reduction area, and residual noise is simultaneously collected as an error signal. By analyzing the error signal, the filter's weight parameters are adjusted, and the filter is updated to obtain an updated filtered signal when processing the reference signal. This process is repeated until the average noise reduction in the vehicle reaches a preset standard, thus obtaining the final filtered signal. Finally, the filtered signal is played to cancel out the in-vehicle noise. Building a road noise prediction model using this data-driven method can quickly and accurately obtain the in-vehicle noise level and precisely suppress it. Summary of the Invention

[0009] The purpose of this invention is to provide a method, system, electronic device, and storage medium for rapid road noise analysis adapted to vehicle dynamic parameters. It enables rapid location of the problem within a short time, narrowing down the possible causes, improving analysis efficiency, and providing graphical and textual displays of relevant data trends. It also provides intelligent judgments and conclusions, freeing up human resources and allowing even non-professionals to produce analysis results.

[0010] This invention provides the following solution:

[0011] According to one aspect of the present invention, a rapid road noise analysis method for vehicle body dynamic parameter adaptation is provided, comprising the following steps:

[0012] Step S1: Obtain the input parameters of the road noise performance prediction model;

[0013] Step S2: Build a road noise body system and a whole vehicle simulation model, and conduct experiments to verify the simulation model;

[0014] Step S3: Write code to implement batch modification of simulation model parameters and simulation results; batch acquire and generate road noise performance training sample data;

[0015] Step S4: Employ a multi-path, multi-parameter, and multi-level data-driven model to construct a rapid road noise performance analysis method adapted to vehicle dynamic parameters.

[0016] Furthermore, including:

[0017] Step S1 includes:

[0018] Step S11: Analyze the important influencing factors and paths of road noise based on the transmission path analysis method to obtain the variable factors of road noise vehicle body;

[0019] Among them, based on the front double wishbone and rear multi-link suspension, the variable factors of road noise and vehicle body were analyzed, and a multi-path, multi-parameter, multi-level road noise decomposition architecture was built based on the influencing factors and variable factors.

[0020] Step S12: Obtain the response curve and transfer function curve after modifying the variable factors using finite element simulation.

[0021] In step S12, simulation is performed based on a hierarchical decomposition architecture and relevant data is collected. The response curves include the active end acceleration response-frequency curve, the passive end acceleration response-frequency curve and the driver's right ear response-frequency curve. The transfer function curves include the active end IPI (origin dynamic stiffness)-frequency curve, the passive end IPI-frequency curve and the passive end to the driver's right ear NTF (noise transfer function)-frequency curve.

[0022] Step S13: Based on the changing trend of the finite element simulation results, the variable parameters of the vehicle body are finally determined;

[0023] In step S13, by changing the variable factors of the vehicle body and referring to the changing trend of the finite element simulation results in step S12, the elastic modulus of the sub-model at the connection between the vehicle body and the subframe is determined as a dynamic parameter variable.

[0024] Furthermore, including:

[0025] Step S2 includes:

[0026] S21. Based on the variable factors in S13, re-divide and establish the body system, and modify the elastic modulus of the body substructures involved in the changes.

[0027] In step S21, the model of the structural component connecting the vehicle body and the subframe is re-divided, the attributes are reassigned, and the materials are reassigned.

[0028] S22. Based on the substructure parameterized model established in step S21, integrate and assemble to form a road noise simulation model that adapts the vehicle body dynamic parameters and conduct experimental verification.

[0029] In step S22, the basic model is reassembled, and the simulation calculation conditions of the transfer function and the road noise response calculation conditions are set through the structure tree. The basic model includes: body model, acoustic cavity model, front and rear subframe models, tire model, and powertrain model.

[0030] The simulation calculation of the transfer function includes the calculation of the active end IPI, the passive end IPI, and the NTF response from the passive end to the driver's right ear. The road noise response calculation includes the calculation of the active end acceleration response, the passive end acceleration response, and the driver's right ear response.

[0031] Furthermore, including:

[0032] Step S3 includes:

[0033] Step S31: Based on the Latin hypercube method, a method for changing vehicle body dynamic parameters is used. At the same time, code is written to batch modify the vehicle body dynamic parameters and generate modified models in batches.

[0034] Step S32: Using the road noise model established in step S31, use a batch processing program to calculate the simulation model, submit the road noise model, and realize the rapid generation of model training sample data.

[0035] Furthermore, including:

[0036] Step S3 also includes:

[0037] In step S31, when designing the Latin hypercube, the factors of the Latin hypercube are set as the elastic modulus of the front subframe and the left front connection substructure of the vehicle body, the elastic modulus of the front subframe and the left rear connection substructure of the vehicle body, the elastic modulus of the front subframe and the right front connection substructure of the vehicle body, the elastic modulus of the front subframe and the right rear connection substructure of the vehicle body, the elastic modulus of the rear subframe and the left front connection substructure of the vehicle body, the elastic modulus of the rear subframe and the left rear connection substructure of the vehicle body, the elastic modulus of the rear subframe and the right front connection substructure of the vehicle body, and the elastic modulus of the rear subframe and the right rear connection substructure of the vehicle body. The horizontal level is set to 7 levels, including the original elastic modulus value and varying upward and downward by 20%, 30%, and 50% respectively from the original elastic modulus value. The target scheme is generated by sampling through the Latin hypercube.

[0038] In step S32, the vehicle body model is modified based on the vehicle body structure parameters of the Latin hypercube scheme, and the modified simulation model is used as input, with the road noise response and transfer function results mentioned in step S22 as output.

[0039] Furthermore, including:

[0040] Step S4 includes:

[0041] S41. A first-level chassis system performance prediction model is generated by training a deep neural network.

[0042] S42. A second-level vehicle body system performance prediction model is generated by training a deep neural network.

[0043] S43. A third-level road noise response prediction model is generated by training a deep neural network.

[0044] Further includes:

[0045] Step S4 also includes:

[0046] In step S41, the first-level deep neural network takes the wheel center force curve of the tire as input and the vibration acceleration response and IPI of the connection point of the chassis system, i.e. the active end, as the prediction output.

[0047] In step S42, the second-level deep neural network takes the output of step S41 plus the bushing dynamic stiffness parameter as input, and the vibration acceleration response and IPI of the connection point of the vehicle body system (i.e., the passive end) as the prediction output.

[0048] In step S43, the third-level deep neural network takes the output of step S42 plus the bushing dynamic stiffness parameter as input and the driver's right ear response as the road noise prediction output.

[0049] According to a second aspect of the present invention, a rapid road noise analysis system for vehicle body dynamic parameter adaptation is provided, comprising:

[0050] The module includes a parameter acquisition module, a model building and validation module, a batch processing module, and a multi-level prediction model module.

[0051] The parameter acquisition module is used to acquire the input parameters of the road noise performance prediction model;

[0052] The model building and verification module is used to build road noise body system and whole vehicle simulation models, and to test and verify the accuracy and reliability of the simulation models.

[0053] The batch processing module is used to write code to implement batch modification of simulation model parameters and batch acquisition of simulation results, and generate road noise performance training sample data;

[0054] The multi-level prediction model module adopts a multi-path, multi-parameter, multi-level data-driven model to build a rapid road noise performance analysis model adapted to vehicle dynamic parameters.

[0055] According to three aspects of the present invention, an electronic device is provided, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0056] The memory stores a computer program, which, when executed by the processor, causes the processor to perform steps of a rapid road noise analysis method adapted to vehicle dynamic parameters.

[0057] According to four aspects of the present invention, a computer-readable storage medium is provided that stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a method for rapid road noise analysis adapted to vehicle body dynamic parameters.

[0058] Compared with the prior art, the present invention has the following advantages:

[0059] This application constructs and validates a road noise fast prediction model adapted to vehicle dynamic parameters, thereby obtaining highly reliable performance prediction data samples.

[0060] This application uses a long short-term memory neural network (LSTM) to accurately predict road noise characteristics, thus eliminating the traditional, complex, and repetitive finite element simulation calculation process.

[0061] This application introduces a vehicle body system and employs a multi-level deep neural network to perform attentional learning on the transfer function data under dynamic changes in vehicle body parameters, thereby effectively predicting road noise performance.

[0062] This application incorporates the body system into the traditional suspension-based analysis using this data prediction model. By adjusting the structure of the existing model architecture to suit the current vehicle model, it can quickly adapt to new task requirements, improve research efficiency, and significantly reduce research costs. Attached Figure Description

[0063] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0064] Figure 1 This is a flowchart of a rapid road noise analysis method for vehicle body dynamic parameter adaptation provided by one or more embodiments of the present invention.

[0065] Figure 2 This is a structural diagram of a road noise rapid analysis system for vehicle body dynamic parameter adaptation provided by one or more embodiments of the present invention.

[0066] Figure 3 This is a flowchart of a road noise rapid analysis method for vehicle body dynamic parameter adaptation according to a specific embodiment of the present invention.

[0067] Figure 4 This is a specific embodiment of the multi-path, multi-parameter, multi-level road noise decomposition architecture of the present invention.

[0068] Figure 5 This is an example diagram illustrating the division and modification of the vehicle body substructure according to a specific embodiment of the present invention.

[0069] Figure 6 This is a diagram of a long short-term memory neural network structure according to a specific embodiment of the present invention.

[0070] Figure 7 This is a flowchart of a multi-objective equilibrium method for a road noise prediction model according to a specific embodiment of the present invention.

[0071] Figure 8 This is a flowchart illustrating the construction process of the prediction model at each level in a specific embodiment of the present invention.

[0072] Figure 9 This is a spectrum of the driver's right ear noise, which is the final target output of the road noise prediction model according to a specific embodiment of the present invention.

[0073] Figure 10 This is a block diagram of an electronic device for a rapid road noise analysis method adapted to vehicle dynamic parameters, provided by one or more embodiments of the present invention. Detailed Implementation

[0074] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] Figure 1 This is a flowchart of a rapid road noise analysis method for vehicle body dynamic parameter adaptation provided by one or more embodiments of the present invention.

[0076] like Figure 1 As shown, it includes the following steps:

[0077] Step S1: Obtain the input parameters of the road noise performance prediction model;

[0078] Step S2: Build a road noise body system and a whole vehicle simulation model, and conduct experiments to verify the simulation model;

[0079] Step S3: Write code to implement batch modification of simulation model parameters and simulation results; batch acquire and generate road noise performance training sample data;

[0080] Step S4: Employ a multi-path, multi-parameter, and multi-level data-driven model to construct a rapid road noise performance analysis method adapted to vehicle dynamic parameters.

[0081] Furthermore, including:

[0082] Step S1 includes:

[0083] Step S11: Analyze the important influencing factors and paths of road noise based on the transmission path analysis method to obtain the variable factors of road noise vehicle body;

[0084] Among them, based on the front double wishbone and rear multi-link suspension, the variable factors of road noise and vehicle body were analyzed, and a multi-path, multi-parameter, multi-level road noise decomposition architecture was built based on the influencing factors and variable factors.

[0085] Step S12: Obtain the response curve and transfer function curve after modifying the variable factors using finite element simulation.

[0086] In step S12, simulation is performed based on a hierarchical decomposition architecture and relevant data is collected. The response curves include the active end acceleration response-frequency curve, the passive end acceleration response-frequency curve and the driver's right ear response-frequency curve. The transfer function curves include the active end IPI (origin dynamic stiffness)-frequency curve, the passive end IPI-frequency curve and the passive end to the driver's right ear NTF (noise transfer function)-frequency curve.

[0087] Step S13: Based on the changing trend of the finite element simulation results, the variable parameters of the vehicle body are finally determined;

[0088] In step S13, by changing the variable factors of the vehicle body and referring to the changing trend of the finite element simulation results in step S12, the elastic modulus of the sub-model at the connection between the vehicle body and the subframe is determined as a dynamic parameter variable.

[0089] Furthermore, including:

[0090] Step S2 includes:

[0091] S21. Based on the variable factors in S13, re-divide and establish the body system, and modify the elastic modulus of the body substructures involved in the changes.

[0092] In step S21, the model of the structural component connecting the vehicle body and the subframe is re-divided, the attributes are reassigned, and the materials are reassigned.

[0093] S22. Based on the substructure parameterized model established in step S21, integrate and assemble to form a road noise simulation model that adapts the vehicle body dynamic parameters and conduct experimental verification.

[0094] In step S22, the basic model is reassembled, and the simulation calculation conditions of the transfer function and the road noise response calculation conditions are set through the structure tree. The basic model includes: body model, acoustic cavity model, front and rear subframe models, tire model, and powertrain model.

[0095] The simulation calculation of the transfer function includes the calculation of the active end IPI, the passive end IPI, and the NTF response from the passive end to the driver's right ear. The road noise response calculation includes the calculation of the active end acceleration response, the passive end acceleration response, and the driver's right ear response.

[0096] Furthermore, including:

[0097] Step S3 includes:

[0098] Step S31: Based on the Latin hypercube method, a method for changing vehicle body dynamic parameters is used. At the same time, code is written to batch modify the vehicle body dynamic parameters and generate modified models in batches.

[0099] Step S32: Using the road noise model established in step S31, use a batch processing program to calculate the simulation model, submit the road noise model, and realize the rapid generation of model training sample data.

[0100] Furthermore, including:

[0101] Step S3 also includes:

[0102] In step S31, when designing the Latin hypercube, the factors of the Latin hypercube are set as the elastic modulus of the front subframe and the left front connection substructure of the vehicle body, the elastic modulus of the front subframe and the left rear connection substructure of the vehicle body, the elastic modulus of the front subframe and the right front connection substructure of the vehicle body, the elastic modulus of the front subframe and the right rear connection substructure of the vehicle body, the elastic modulus of the rear subframe and the left front connection substructure of the vehicle body, the elastic modulus of the rear subframe and the left rear connection substructure of the vehicle body, the elastic modulus of the rear subframe and the right front connection substructure of the vehicle body, and the elastic modulus of the rear subframe and the right rear connection substructure of the vehicle body. The levels are set as seven levels, including the original elastic modulus value and varying upward and downward by 20%, 30%, and 50% respectively from the original elastic modulus value. The target scheme is generated by sampling through the Latin hypercube.

[0103] In step S32, the vehicle body model is modified based on the vehicle body structure parameters of the Latin hypercube scheme, and the modified simulation model is used as input, with the road noise response and transfer function results mentioned in step S22 as output.

[0104] Furthermore, including:

[0105] Step S4 includes:

[0106] S41. A first-level chassis system performance prediction model is generated by training a deep neural network.

[0107] S42. A second-level vehicle body system performance prediction model is generated by training a deep neural network.

[0108] S43. A third-level road noise response prediction model is generated by training a deep neural network.

[0109] Further includes:

[0110] Step S4 also includes:

[0111] In step S41, the first-level deep neural network takes the wheel center force curve of the tire as input and the vibration acceleration response and IPI of the connection point of the chassis system, i.e. the active end, as the prediction output.

[0112] In step S42, the second-level deep neural network takes the output of step S41 plus the bushing dynamic stiffness parameter as input, and the vibration acceleration response and IPI of the connection point of the vehicle body system (i.e., the passive end) as the prediction output.

[0113] In step S43, the third-level deep neural network takes the output of step S42 plus the bushing dynamic stiffness parameter as input and the driver's right ear response as the road noise prediction output.

[0114] In another embodiment, this embodiment provides a method for rapid road noise analysis adapted to vehicle dynamic parameters, including the following steps:

[0115] S1. Obtain the parameters required for the parametric model of active vibration damper performance prediction;

[0116] Building a road noise prediction model adapted to vehicle dynamic parameters requires obtaining necessary structural parameters. Since the whole vehicle model is complex and has many transmission paths, it is only necessary to select structural and performance parameters that are significantly related to the prediction target.

[0117] This step specifically includes the following sub-steps:

[0118] S11. Based on the transmission path analysis method, analyze the important influencing factors and paths of road noise to obtain the variable factors of road noise vehicle body;

[0119] This patent's road noise transmission path analysis is based on a vehicle with a front double wishbone and rear multi-link suspension. According to the road noise transmission path mechanism, the main influencing factors along the path are determined. Based on rules for top-down target decomposition and bottom-up target prediction, a multi-path, multi-parameter, multi-level road noise decomposition architecture is constructed, such as... Figure 4 As shown (taking the left side as an example).

[0120] S12. Obtain the response curve and transfer function curve after modifying the variable factors based on the finite element simulation method.

[0121] Since altering the elastic modulus of a certain substructure of the vehicle body is difficult to perform experimentally, simulation methods are used to calculate its response curve and transfer function curve. The response curve includes the active-end acceleration response-frequency curve, the passive-end acceleration response-frequency curve, and the driver's right ear response-frequency curve. The transfer function curve includes the active-end IPI (origin dynamic stiffness)-frequency curve, the passive-end IPI-frequency curve, and the passive-end to the driver's right ear NTF (noise transfer function)-frequency curve.

[0122] S13. Based on the changing trend of the finite element simulation results, the variable parameters of the vehicle body are finally determined.

[0123] By trying different body modification schemes and analyzing the changing trends of finite element simulation results, the elastic modulus of the sub-model at the connection between the body and the subframe was finally determined as a dynamic parameter variable.

[0124] S2. Build a road noise body system and whole vehicle simulation model and conduct experiments to verify the accuracy and reliability of the simulation model;

[0125] The vehicle body structure was redefined, and the substructures that are variables (the structural components that connect the subframe to the body) were newly created as components. Then, the complete vehicle model was reassembled. The dynamic parameters of the vehicle body sub-models could be modified by changing the simulation model, and the accuracy and reliability of this modification were verified through experiments.

[0126] This step specifically includes the following sub-steps:

[0127] S21. Based on the variable factors in S13, re-divide and establish the body system, and modify the elastic modulus of the structural components connecting the body and subframe that involve changes.

[0128] Create a new structural component connecting the body and subframe mentioned in S1, and reassign its properties and materials to facilitate later parameter modifications. A specific example is shown below. Figure 5 As shown.

[0129] S22. Based on the parameterized sub-model established in step S21, integrate and assemble to form a road noise simulation model that adapts the vehicle body dynamic parameters and conduct experimental verification.

[0130] The basic models, including the body model, acoustic cavity model, front and rear subframe models, tire model, and powertrain model, which define the structural components connecting the body and subframe, were reassembled. New hard points and connection points were created. Simulation conditions for transfer functions and road noise response were set using a structure tree. The transfer function simulation conditions included active-end IPI calculation, passive-end IPI calculation, and passive-end response to the driver's right ear NTF calculation. The road noise response calculation included active-end acceleration response calculation, passive-end acceleration response calculation, and the driver's right ear response calculation. The simulation results were compared and verified with experimental results.

[0131] The formulas for calculating the origin acceleration admittance (IPI) are shown in equations (1)-(5). Since the origin dynamic stiffness and displacement admittance are reciprocals, the expression for the origin dynamic stiffness can be obtained as follows:

[0132] (1);

[0133] The origin acceleration admittance, which reflects the dynamic stiffness characteristics of the connection point, is collectively referred to as IPI, and can be obtained from the following expression:

[0134] (2);

[0135] (3);

[0136] (4);

[0137] Its amplitude is:

[0138] (5);

[0139] By combining the principles of finite element modal analysis, all the parameters in the acceleration admittance expression can be calculated, requiring only the excitation frequency of the system to be known. , excitation frequency Substituting into the expression, the value of the system's origin acceleration admittance IPI can be calculated. The NTF noise transfer function refers to the ratio of the Laplace change of the response (output) to the Laplace change of the excitation (input) in a linear system.

[0140] S3. Write code to implement batch modification of simulation model parameters and batch acquisition of simulation results, and generate road noise performance training sample data;

[0141] This step specifically includes the following sub-steps:

[0142] S31. Based on the Latin hypercube method and the changes in vehicle body dynamic parameters, write code to batch modify vehicle body dynamic parameters and generate modified models in batches.

[0143] The variable factors of the vehicle body are the elastic modulus of the front subframe and the left front connection substructure, the elastic modulus of the front subframe and the left rear connection substructure, the elastic modulus of the front subframe and the right front connection substructure, the elastic modulus of the front subframe and the right rear connection substructure, the elastic modulus of the rear subframe and the left front connection substructure, the elastic modulus of the rear subframe and the left rear connection substructure, the elastic modulus of the rear subframe and the right front connection substructure, and the elastic modulus of the rear subframe and the right rear connection substructure. Each factor has seven variable levels, specifically set to the original elastic modulus value and seven levels of upward and downward variation of 20%, 30%, and 50% respectively from the original elastic modulus value, ultimately generating 200 vehicle body modification schemes.

[0144] S32. Using the road noise model established in step S31, connect the simulation calculation interface using code, and achieve rapid generation of model training sample data by submitting road noise models in batches.

[0145] Automated modification code for vehicle body parameters was developed. Based on the Latin hypercube generation scheme, the vehicle body structural parameters are automatically modified on the vehicle body model. The modified simulation model is used as input, and the outputs are road noise response results and transfer function results. Finally, 200 sets of samples are generated for training the prediction model.

[0146] S4. A multi-path, multi-parameter, and multi-level data-driven model is adopted to construct a rapid road noise performance analysis method adapted to vehicle dynamic parameters.

[0147] This step specifically includes the following sub-steps:

[0148] S41. A first-level chassis system performance prediction model is generated by training a deep neural network.

[0149] The 200 sets of sample data obtained in step S32 are used as training samples. The wheel center force curve of the tire is used as input. The chassis system data obtained from the simulation, namely the vibration acceleration response and IPI of the connection point of the active end, are used as prediction outputs. The third-level chassis system prediction model is trained by a long short-term memory neural network.

[0150] Long Short-Term Memory (LSTM) neural networks consist of gating units. These gating units judge the input data, keeping data that conforms to the rules and forgetting data that does not. The internal structure is as follows: Figure 4 As shown. LSTM mainly consists of forget gates. Input gate Output gate and memory cells The components are calculated as shown in (6)-(9):

[0151] (6);

[0152] (7);

[0153] (8);

[0154] (9);

[0155] The state of the memory cells is updated using equation (10), and the output is: The calculation of the hidden state is shown in equation (11):

[0156] (10);

[0157] (11);

[0158] in, , and These are the weight matrices for the forget gate, input gate, and output gate, respectively. , and These are the bias terms for the forget gate, input gate, and output gate, respectively. , and These correspond to the current input data, current output data, and current cell state, respectively. and These represent the output data and cell state at the previous time step, respectively; ☉ represents the dot product. This represents the sigmoid activation function.

[0159] Because there are multiple input parameters and multiple target output prediction forms between layers, this invention combines a multi-target equalization method to optimize the output of the LSTM neural network, achieving independent optimization of the loss function for each target branch. Based on sharing some network layers, this enhances the model's generalization ability and prediction accuracy. The combination of LSTM and the multi-target equalization method is as follows: Figure 5 As shown, the general form of the multi-objective optimization problem is shown in equation (12).

[0160] (12);

[0161] in, Let x be the objective function and the sub-objective function. , ,…, ] is the decision vector. and These are the constraints for the inequalities and the constraints for the equality, respectively.

[0162] Dynamic weighted averaging methods typically focus only on learning speed, easily overlooking the differences in loss magnitude between different tasks. Therefore, a new method for determining the loss magnitude of the longitudinal and lateral control parameters is introduced based on the dynamic weighted averaging method, as shown in equation (13). By establishing a new method for weighting the loss function to balance the loss magnitude and learning speed, effective learning of the long short-term memory neural network is achieved, ultimately reaching the goal of multi-objective balanced prediction. The flowchart of multi-objective balanced prediction is shown below. Figure 7 As shown.

[0163] (13);

[0164] The process of building prediction models at each level is as follows: Figure 8 As shown.

[0165] S42. A second-level vehicle body system performance prediction model is generated by training a deep neural network.

[0166] The output of step S41 is combined with the bushing dynamic stiffness parameter as the input of the second level, and the vibration acceleration response and IPI of the connection point of the vehicle body system (i.e., the passive end) are used as the prediction output of the second level.

[0167] S43. A third-level road noise response prediction model is generated by training a deep neural network.

[0168] The output from step S42, plus the bushing dynamic stiffness parameters of the second level, is used as the input to the first-level model. The driver's right ear response is used as the final target output of the road noise prediction model. Due to the large number of prediction results, only the driver's right ear noise response of the first level, the passive end acceleration response of the vehicle body and left front upper control arm of the second level, the passive end IPI of the vehicle body and left front upper control arm of the second level, the active end acceleration response of the vehicle body and left front upper control arm of the third level, and the active end IPI of the vehicle body and left front upper control arm of the third level are shown. Except for the acceleration response, all the above data include X, Y, and Z-axis accelerations. Specific results are as follows... Figure 9 As shown.

[0169] Specifically, by constructing and validating a fast road noise prediction model adapted to vehicle body dynamic parameters, highly reliable performance prediction data samples can be obtained. This method avoids cumbersome experimental conditions and high experimental costs. During model training, a Long Short-Term Memory (LSTM) neural network is used to accurately predict road noise characteristics, eliminating the traditional complex and repetitive finite element simulation calculation process. Simultaneously, by introducing the vehicle body system, a multi-level deep neural network is used to learn the transfer function data under dynamic changes in vehicle body parameters, thereby effectively predicting road noise performance. This data prediction model, based on traditional suspension-centric analysis, incorporates consideration of the vehicle body system. By structurally adjusting the existing model architecture to suit current vehicle models, it can quickly adapt to new task requirements, improve research efficiency, and significantly reduce research costs.

[0170] Figure 2 This is a structural diagram of a road noise rapid analysis system for vehicle body dynamic parameter adaptation provided by one or more embodiments of the present invention.

[0171] like Figure 2 As shown, it includes:

[0172] The module includes a parameter acquisition module, a model building and validation module, a batch processing module, and a multi-level prediction model module.

[0173] The parameter acquisition module is used to acquire the input parameters of the road noise performance prediction model;

[0174] The model building and verification module is used to build road noise body system and whole vehicle simulation models, and to test and verify the accuracy and reliability of the simulation models.

[0175] The batch processing module is used to write code to implement batch modification of simulation model parameters and batch acquisition of simulation results, and generate road noise performance training sample data;

[0176] The parameter acquisition module includes:

[0177] The path analysis unit is used to analyze the important influencing factors and paths of road noise based on the transmission path analysis method, and to obtain the variable factors of road noise vehicle body.

[0178] Finite element simulation unit, used to obtain response curves and transfer function curves after modifying variable factors according to finite element simulation method;

[0179] The parameter determination unit is used to determine the variable parameters of the vehicle body based on the changing trends of the finite element simulation results.

[0180] The path analysis unit, based on the front double wishbone and rear multi-link suspension, analyzes the variable factors of road noise and vehicle body, and builds a multi-path, multi-parameter, multi-level road noise decomposition architecture based on the influencing factors and variable factors.

[0181] The response curves obtained from the finite element simulation unit include the active end acceleration response-frequency curve, the passive end acceleration response-frequency curve, and the driver's right ear response-frequency curve; the transfer function curves include the active end IPI-frequency curve, the passive end IPI-frequency curve, and the passive end to the driver's right ear NTF-frequency curve.

[0182] The variable parameters of the vehicle body determined by the parameter determination unit are the elastic modulus of the sub-model at the connection between the vehicle body and the subframe.

[0183] The model building and validation module includes:

[0184] The body system construction unit is used to re-divide and build the body system according to the determined variable factors, and modify the elastic modulus of the body substructures involved in the changes.

[0185] The assembly verification unit is used to integrate and assemble a road noise simulation model that adapts the vehicle body dynamic parameters based on the established substructure parameterized model, and then conduct experimental verification.

[0186] The body system construction unit re-divides the model, reassigns attributes, and reassigns materials to the structural components connecting the body and subframe.

[0187] The assembly verification unit reassembles the vehicle body model, acoustic cavity model, front and rear subframe models, tire model, and powertrain model, and sets the transfer function simulation calculation conditions and road noise response calculation conditions through a structure tree. The transfer function simulation calculation conditions include active end IPI calculation, passive end IPI calculation, and passive end to driver's right ear response NTF calculation. The road noise response calculation includes active end acceleration response calculation, passive end acceleration response calculation, and driver's right ear response calculation.

[0188] The batch processing module includes:

[0189] The parameter modification unit is used to design a method for changing vehicle dynamic parameters based on the Latin hypercube method, and to write code to batch modify vehicle dynamic parameters and generate modified models in batches.

[0190] The sample generation unit is used to calculate the simulation model using the modified road noise model through a batch processing program, submit the road noise model, and realize the rapid generation of model training sample data.

[0191] The multi-level prediction model module adopts a multi-path, multi-parameter, multi-level data-driven model to build a rapid road noise performance analysis model adapted to vehicle dynamic parameters.

[0192] Specifically, for the first time, the vehicle body system is incorporated into the core road noise analysis framework: breaking through the limitations of the traditional suspension-based approach, a multi-level architecture including the vehicle body is built through path analysis units to more comprehensively reflect the road noise transmission mechanism.

[0193] Full-process automation and batch processing: From parameter modification (parameter modification unit) to sample generation (sample generation unit), automation is achieved through code and batch processing programs, greatly improving efficiency.

[0194] Data-driven approach combined with simulation: Samples are generated based on reliable simulation models, and multi-level neural network predictions replace complex finite element calculations, balancing accuracy and efficiency.

[0195] Dynamic parameter adaptation capability: By locking dynamic parameters such as the elastic modulus of the connection between the vehicle body and the subframe, and combining the adjustability of the model structure, rapid adaptation analysis of road noise for new vehicle models can be achieved.

[0196] The parameter acquisition module provides "precise parameter input" to the system and serves as the starting point for analysis.

[0197] This module solves the problems of "blind parameter selection and disconnect from the vehicle body system" in traditional road noise analysis by progressively operating through three levels of units (path analysis unit → finite element simulation unit → parameter determination unit), from "macro-factor identification" to "micro-parameter locking".

[0198] Path Analysis Unit: Serving as a "navigator" for parameter acquisition, based on the front double wishbone and rear multi-link suspension structure, it identifies key influencing factors of road noise (such as the connection structure between the vehicle body and the suspension) and their transmission paths through transmission path analysis. It also establishes a decomposition architecture of "multi-path (transfer function, vibration acceleration, etc.) - multi-parameter (bulb dynamic stiffness, etc.) - multi-level (from steering knuckle to driver's right ear)", which for the first time incorporates the vehicle body system into the core influencing factors of road noise, breaking through the limitations of traditional methods that only focus on the suspension.

[0199] Finite element simulation unit: As a "data generator" for parameter acquisition, based on the variable factors locked by the path analysis unit, it obtains response curves (acceleration-frequency at the active end / passive end, response-frequency at the driver's right ear) and transfer function curves (IPI, NTF, etc.) through simulation, providing a quantitative basis for the correlation analysis of parameters and road noise performance, avoiding the high cost and low efficiency of traditional test data acquisition.

[0200] Parameter determination unit: As the "decision-maker" for parameter acquisition, based on the changing trend of the finite element simulation results, it ultimately locks the "elastic modulus of the sub-model at the connection between the vehicle body and the subframe" as the core dynamic parameter, ensuring the relevance of parameters in subsequent analyses and avoiding interference from redundant parameters.

[0201] Model building and verification module: Provides a "reliable simulation foundation" for the system and is the core carrier of analysis.

[0202] This module receives the output (determined variable parameters of the vehicle body) from the parameter acquisition module and ensures the accuracy of the simulation model through "structural reconstruction-assembly verification," providing a reliable "digital twin" foundation for subsequent batch processing and prediction.

[0203] Body system building unit: As a "structural reconfigurator" for model building, it re-divides and reassigns attributes and materials to the structural components connecting the body and subframe based on the elastic modulus parameters locked by the parameter determination unit, so that the body sub-model has "parameter modifiability", which solves the problem that the body structure is fixed in traditional simulation models and is difficult to adapt to dynamic changes in parameters.

[0204] Assembly Verification Unit: Acting as the "quality inspector" of the model, this unit integrates the vehicle body model, acoustic cavity model, subframe model, etc., into a complete vehicle road noise simulation model. It then verifies the model's accuracy by setting transfer function calculation conditions (IPI, NTF) and road noise response calculation conditions (acceleration, right ear noise) and comparing the results with experimental data. This step ensures that subsequent batch processing and prediction are based on a "reliable model," avoiding analytical errors caused by distortion of simulation results.

[0205] Batch processing module: Provides "efficient sample support" for the system and is the "data engine" for prediction.

[0206] This module solves the pain point of "slow sample generation and small number" in traditional road noise analysis by batch operation of two-level units (parameter modification unit → sample generation unit), and provides sufficient and diverse training data for multi-level prediction models.

[0207] Parameter Modification Unit: Serving as a "scheme designer" for the sample, this unit uses the Latin hypercube method to set the elastic modulus of eight body connection substructures (the front / rear subframes and the left front / left rear / right front / right rear connections of the body) as variables. Each variable contains seven levels (original value and ±20% / 30% / 50%), generating 200 sets of parameter schemes. The unit also enables automated batch modification of parameters through code, overcoming the inefficiency of manual parameter modification.

[0208] Sample generation unit: Acting as a "batch producer" of samples, it submits 200 sets of models generated by the parameter modification unit for simulation calculation through a batch processing program, batch-acquires response and transfer function data, and finally generates 200 sets of training samples. This step achieves full automation of "parameter-simulation-data", significantly improving sample generation efficiency.

[0209] Multi-level prediction model module: Provides the system with "rapid prediction capabilities" and serves as the "output terminal" for analysis.

[0210] Based on samples generated by the batch processing module, this module adopts a "multi-path, multi-parameter, multi-level data-driven model" to achieve rapid prediction of road noise performance, solving the problems of complex calculations and difficulty in adapting to new vehicle models in traditional finite element simulation.

[0211] The model architecture adopts a three-level progressive prediction: the first level (chassis system) uses the tire wheel center force as input to predict the active end acceleration and IPI; the second level (body system) combines the output of the first level with the bushing dynamic stiffness to predict the passive end acceleration and IPI; the third level (road noise response) finally outputs the noise in the driver's right ear.

[0212] This architecture is the first to link the body system and suspension system for analysis. It uses a deep neural network (combined with LSTM time-series feature extraction) to focus on the transfer function features of dynamic changes in body parameters, enabling the model to quickly adapt to new models through structural adjustments, thereby improving generalization ability and prediction efficiency.

[0213] It is worth noting that although this system only discloses the parameter acquisition module, model building and verification module, batch processing module, and multi-level prediction model module, it does not mean that this device is limited to the above-mentioned basic functional modules. On the contrary, what this invention intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can add one or more functional modules in combination with existing technology to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. It should not be assumed that the scope of protection of the claims of this invention is limited to the above-disclosed basic functional modules just because this embodiment only discloses a few basic functional modules.

[0214] Figure 10This is a block diagram of an electronic device for a rapid road noise analysis method adapted to vehicle dynamic parameters, provided by one or more embodiments of the present invention.

[0215] like Figure 10 As shown, this application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0216] The memory stores a computer program, which, when executed by the processor, causes the processor to perform steps of a road noise rapid analysis method adapted to vehicle dynamic parameters.

[0217] This application also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform steps of a road noise rapid analysis method adapted to vehicle dynamic parameters.

[0218] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0219] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0220] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid road noise analysis method for vehicle body dynamic parameter adaptation, characterized in that, Includes the following steps: Step S1: Obtain the input parameters of the road noise performance prediction model; Step S2: Build a road noise body system and a whole vehicle simulation model, and conduct experiments to verify the simulation model; Step S3: Write code to implement batch modification of simulation model parameters and simulation results; batch acquire and generate road noise performance training sample data; Step S4: Employ a multi-path, multi-parameter, and multi-level data-driven model to construct a rapid road noise performance analysis method adapted to vehicle dynamic parameters.

2. The method for rapid road noise analysis based on vehicle dynamic parameters according to claim 1, characterized in that, Step S1 includes: Step S11: Analyze the important influencing factors and paths of road noise based on the transmission path analysis method to obtain the variable factors of road noise vehicle body; Among them, based on the front double wishbone and rear multi-link suspension, the variable factors of road noise and vehicle body were analyzed, and a multi-path, multi-parameter, multi-level road noise decomposition architecture was built based on the influencing factors and variable factors. Step S12: Obtain the response curve and transfer function curve after modifying the variable factors using finite element simulation. In step S12, simulation is performed based on a hierarchical decomposition architecture and relevant data is collected. The response curves include the active end acceleration response-frequency curve, the passive end acceleration response-frequency curve and the driver's right ear response-frequency curve. The transfer function curves include the active end IPI-frequency curve, the passive end IPI-frequency curve and the passive end to the driver's right ear NTF-frequency curve. Step S13: Determine the variable parameters of the vehicle body based on the changing trend of the finite element simulation results; In step S13, by changing the variable factors of the vehicle body and referring to the changing trend of the finite element simulation results in step S12, the elastic modulus of the sub-model at the connection between the vehicle body and the subframe is determined as a dynamic parameter variable.

3. The method for rapid road noise analysis based on vehicle dynamic parameters according to claim 1, characterized in that, Step S2 includes: S21. Based on the variable factors in S13, re-divide and establish the body system, and modify the elastic modulus of the body substructures involved in the changes. In step S21, the model of the structural component connecting the vehicle body and the subframe is re-divided, the attributes are reassigned, and the materials are reassigned. S22. Based on the substructure parameterized model established in step S21, integrate and assemble to form a road noise simulation model that adapts the vehicle body dynamic parameters and conduct experimental verification. In step S22, the basic model is reassembled, and the simulation calculation conditions of the transfer function and the road noise response calculation conditions are set through the structure tree. The basic model includes: body model, acoustic cavity model, front and rear subframe models, tire model, and powertrain model. The simulation calculation of the transfer function includes the calculation of the active end IPI, the passive end IPI, and the NTF response from the passive end to the driver's right ear. The road noise response calculation includes the calculation of the active end acceleration response, the passive end acceleration response, and the driver's right ear response.

4. The method for rapid road noise analysis based on vehicle dynamic parameters according to claim 1, characterized in that, Step S3 includes: Step S31: Based on the Latin hypercube method, a method for changing vehicle body dynamic parameters is used. At the same time, code is written to batch modify the vehicle body dynamic parameters and generate modified models in batches. Step S32: Using the road noise model established in step S31, use a batch processing program to calculate the simulation model, submit the road noise model, and realize the rapid generation of model training sample data.

5. The method for rapid road noise analysis based on vehicle dynamic parameters according to claim 4, characterized in that, Step S3 further includes: In step S31, when designing the Latin hypercube, the factors of the Latin hypercube are set as the elastic modulus of the front subframe and the left front connection substructure of the vehicle body, the elastic modulus of the front subframe and the left rear connection substructure of the vehicle body, the elastic modulus of the front subframe and the right front connection substructure of the vehicle body, the elastic modulus of the front subframe and the right rear connection substructure of the vehicle body, the elastic modulus of the rear subframe and the left front connection substructure of the vehicle body, the elastic modulus of the rear subframe and the left rear connection substructure of the vehicle body, the elastic modulus of the rear subframe and the right front connection substructure of the vehicle body, and the elastic modulus of the rear subframe and the right rear connection substructure of the vehicle body. The horizontal level is set to 7 levels, including the original elastic modulus value and varying upward and downward by 20%, 30%, and 50% respectively from the original elastic modulus value. The target scheme is generated by sampling through the Latin hypercube. In step S32, the vehicle body model is modified based on the vehicle body structure parameters of the Latin hypercube scheme, and the modified simulation model is used as input, with the road noise response and transfer function results mentioned in step S22 as output.

6. The method for rapid road noise analysis based on vehicle dynamic parameters according to claim 1, characterized in that, Step S4 includes: S41. A first-level chassis system performance prediction model is generated by training a deep neural network. S42. A second-level vehicle body system performance prediction model is generated by training a deep neural network. S43. A third-level road noise response prediction model is generated by training a deep neural network.

7. The method for rapid road noise analysis based on vehicle dynamic parameters according to claim 6, characterized in that, Step S4 further includes: In step S41, the first-level deep neural network takes the wheel center force curve of the tire as input and the vibration acceleration response and IPI of the connection point of the chassis system, i.e. the active end, as the prediction output. In step S42, the second-level deep neural network takes the output of step S41 plus the bushing dynamic stiffness parameter as input, and the vibration acceleration response and IPI of the connection point of the vehicle body system (i.e., the passive end) as the prediction output. In step S43, the third-level deep neural network takes the output of step S42 plus the bushing dynamic stiffness parameter as input and the driver's right ear response as the road noise prediction output.

8. A rapid road noise analysis system for adapting to vehicle dynamic parameters, characterized in that, include: The module includes a parameter acquisition module, a model building and validation module, a batch processing module, and a multi-level prediction model module. The parameter acquisition module is used to acquire the input parameters of the road noise performance prediction model; The model building and verification module is used to build road noise body system and whole vehicle simulation models, and to test and verify the accuracy and reliability of the simulation models. The batch processing module is used to write code to implement batch modification of simulation model parameters and batch acquisition of simulation results, and generate road noise performance training sample data; The multi-level prediction model module adopts a multi-path, multi-parameter, multi-level data-driven model to build a rapid road noise performance analysis model adapted to vehicle dynamic parameters.

9. An electronic device, characterized in that, include: The processor, communication interface, memory, and communication bus are connected, with the processor, communication interface, and memory communicating with each other via the communication bus. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the road noise rapid analysis method for vehicle body dynamic parameter adaptation as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a road noise rapid analysis method for vehicle body dynamic parameter adaptation as described in any one of claims 1-7.

Images