A spring tuning piece matching optimization method and system based on suspension natural frequency

Abstract

The present application relates to the technical field of vehicle design optimization, in particular to a spring adjustment piece matching optimization method and system based on suspension natural frequency, comprising extracting spring structure parameters and adjustment piece structure parameters associated with natural frequency response, and constructing a disturbance combination matrix; performing response analysis on the disturbance combination matrix based on a whole vehicle suspension system simulation model, obtaining first scale data and second scale data, and forming a multi-scale response data set; calculating the first sensitivity and the second sensitivity of the spring adjustment piece according to the multi-scale response data set, and constructing a multi-scale sensitivity index system; based on the multi-scale sensitivity index system, evaluating the independent contribution of each adjustment piece to the target natural frequency, identifying the key influence adjustment piece, and constructing a structure parameter-natural frequency response mapping model according to the independent contribution, to obtain the specific influence mode and influence degree of each adjustment piece; optimizing according to the multi-conflict design target, and outputting the spring and adjustment piece combination matching suggestion.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle design optimization, in particular to a spring tuning piece matching optimization method and system based on suspension natural frequency. BACKGROUND

[0002] As the core module of vehicle vertical dynamics performance, the frequency characteristics of the suspension system are directly related to the handling stability and ride comfort of the vehicle. The natural frequency response is the characteristic response frequency of the suspension system under typical excitation conditions, and is usually used as an important basis for dynamic performance evaluation and matching design. In the whole vehicle matching development process, the designer usually needs to adjust the parameter combination of the spring and its matching tuning piece in the suspension system according to the target natural frequency index, so as to realize the optimal matching of the whole vehicle dynamic performance.

[0003] However, the existing natural frequency matching process mostly depends on system-level simulation modeling and test verification, mainly focusing on the mapping relationship between the overall structure parameters and the system response, and lacks the modeling and response decomposition ability of the action path of the tuning piece level component. Under the condition of the joint action of multiple tuning pieces, the influence of each component parameter on the natural frequency has nonlinear superposition and coupling characteristics. The existing model cannot provide sensitivity differentiation and influence quantization for each component parameter, which makes it difficult to effectively track the source of the natural frequency deviation in engineering practice, and the optimization direction is unclear, the calculation efficiency is low, and the tuning granularity is coarse.

[0004] Therefore, an optimization method and system capable of identifying the sensitivity and influence contribution of the tuning piece level parameter to the suspension natural frequency are urgently needed, which can build a natural frequency response analysis model with controllable granularity, not only realize the traceable mapping from the system frequency characteristics to the component level action path, but also significantly improve the matching efficiency and tuning accuracy of the suspension system, so as to improve the explainability, controllability and automation level of the tuning process.

[0005] Therefore, a spring tuning piece matching optimization method and system based on suspension natural frequency are proposed. SUMMARY

[0006] The purpose of the present application is to provide a spring adjustment component matching optimization method and system based on suspension natural frequency, to realize traceable mapping from system frequency characteristics to component level action path by identifying the main effect of parameter changes of each adjustment component on the natural frequency, thereby improving the matching efficiency and adjustment accuracy of the suspension system. By extracting the spring structure parameters and adjustment component structure parameters associated with the natural frequency response, a disturbance combination matrix is constructed. Based on the whole vehicle suspension system simulation model, response analysis is performed on the disturbance combination matrix to obtain first scale data and second scale data, forming a multi-scale response data set. According to the multi-scale response data set, the first sensitivity and the second sensitivity of the spring adjustment component are calculated, and a multi-scale sensitivity index system is constructed. Based on the multi-scale sensitivity index system, the independent contribution of each adjustment component to the target natural frequency is evaluated, the key influence adjustment component is identified, and a structure parameter-natural frequency response mapping model is constructed according to the independent contribution, to obtain the specific influence mode and influence degree of each adjustment component. According to the multi-conflict design target, the structure parameter-natural frequency response mapping model and the independent contribution are optimized, and the spring and adjustment component combination matching suggestion is output.

[0007] To achieve the above purpose, the present application provides the following technical scheme:

[0008] A spring adjustment component matching optimization method based on suspension natural frequency, comprising:

[0009] extracting the spring structure parameters and adjustment component structure parameters associated with the natural frequency response, and constructing a disturbance combination matrix;

[0010] based on the whole vehicle suspension system simulation model, performing response analysis on each disturbance combination in the disturbance combination matrix to obtain first scale data and second scale data, forming a multi-scale response data set;

[0011] according to the multi-scale response data set, calculating the first sensitivity and the second sensitivity of the spring adjustment component, and constructing a multi-scale sensitivity index system;

[0012] based on the multi-scale sensitivity index system, evaluating the independent contribution of each adjustment component to the target natural frequency, identifying the key influence adjustment component, and constructing a structure parameter-natural frequency response mapping model according to the independent contribution, to obtain the specific influence mode and influence degree of each adjustment component;

[0013] according to the specific influence mode and influence degree of each adjustment component, optimizing the structure parameter-natural frequency response mapping model and the independent contribution based on the multi-conflict design target, and outputting the spring and adjustment component combination matching suggestion.

[0014] Preferably, the spring structure parameters include stiffness characteristic curve, effective number of turns, wire diameter, mean diameter and free length;

[0015] The structure parameters of the tuning parts include stiffness characteristic curves and intervention timing of auxiliary springs, material properties, height, acting area and shape characteristics of bumpers, diameter, arm length and connecting point hard point of stabilizer bars, and radial stiffness, axial stiffness and torsional stiffness of bushings;

[0016] The process of constructing the disturbance combination matrix is: determining the value range and level number of the spring structure parameters and the tuning part structure parameters, and generating an experiment scheme containing multiple parameter level combinations by using an orthogonal experiment design method to form the disturbance combination matrix.

[0017] Preferably, the whole vehicle suspension system simulation model includes: a suspension system simulation unit, a frequency response analysis unit and a multi-scale data acquisition unit.

[0018] The suspension system simulation unit simulates the whole vehicle suspension system based on the spring structure parameters and the tuning part structure parameters to obtain a three-dimensional entity model of the whole vehicle suspension system.

[0019] The frequency response analysis unit is integrated in the suspension system simulation unit and is used for frequency response analysis on each disturbance combination in the input disturbance combination matrix.

[0020] The multi-scale data acquisition unit is used to obtain first scale data and second scale data from the frequency response analysis results; wherein the first scale data is macro-scale data, including whole vehicle pitch modal frequency, whole vehicle roll modal frequency, whole vehicle vertical first order main frequency and whole vehicle vertical second order main frequency; and the second scale data is micro-scale data, including local response frequency of sub-assemblies.

[0021] Preferably, the process of constructing the multi-scale sensitivity index system is:

[0022] The sensitivity values of the spring structure parameters and the tuning part structure parameters at the first scale and the second scale are calculated respectively to obtain the first sensitivity and the second sensitivity.

[0023] The first sensitivity and the second sensitivity are normalized.

[0024] The normalized first sensitivity and the normalized second sensitivity are fused by using a nonlinear fusion algorithm to obtain a fused sensitivity value.

[0025] The fused sensitivity values of all structure parameters are collected to form the multi-scale sensitivity index system.

[0026] Preferably, the process of identifying the key tuning part is:

[0027] Quantify the influence relationship between each said tuning part structure parameter and the frequency response based on a multi-scale sensitivity index system;

[0028] Evaluate the independent contribution of each tuning part to the target frequency deviation using an attribution analysis method;

[0029] Set an independent contribution threshold, and mark the tuning part corresponding to the structure parameter higher than the independent contribution threshold as the key influence tuning part;

[0030] Based on the key influence tuning part, taking the structure parameters of the key influence tuning part as input and the frequency response as output, a structure parameter-frequency response mapping model is established using a graph neural network to output the specific influence mode and influence degree of each key influence tuning part.

[0031] Preferably, the structure parameter-frequency response mapping model includes a structure graph construction unit, a graph neural network modeling unit, a frequency response prediction unit, and an influence mode and degree analysis unit;

[0032] The structure graph construction unit constructs the spring and each tuning part into a structure graph, where each node represents a different tuning part, the features of each node represent the disturbance values of each tuning part in the disturbance combination matrix, and each edge represents the correlation between the tuning part structure parameters; The correlation between the tuning part structure parameters is obtained through correlation analysis;

[0033] The graph neural network modeling unit performs feature propagation and graph convolution calculation based on the structure graph to learn the influence relationship of each tuning part structure parameter on the suspension system frequency response;

[0034] The frequency response prediction unit decodes the high-dimensional embedding features output by the graph neural network modeling unit to obtain the predicted target frequency response value;

[0035] The influence mode and degree analysis unit performs a small disturbance on the parameter value of the key influence tuning part, and observes the change amount and direction of the predicted target frequency response value to obtain the specific influence mode and the influence degree.

[0036] Preferably, a spring tuning part matching optimization system based on suspension frequency deviation includes:

[0037] A parameter disturbance generation module extracts spring structure parameters and tuning part structure parameters associated with frequency response and constructs a disturbance combination matrix;

[0038] A multi-scale response analysis module performs response analysis on each disturbance combination in the disturbance combination matrix based on a whole vehicle suspension system simulation model, obtains first scale data and second scale data, and forms a multi-scale response data set;

[0039] A multi-scale sensitivity calculation module calculates a first sensitivity and a second sensitivity of the spring tuning part according to the multi-scale response data set, and constructs a multi-scale sensitivity index system;

[0040] A tuning part contribution evaluation module evaluates independent contribution degrees of each tuning part to the target frequency deviation based on the multi-scale sensitivity index system, identifies key influence tuning parts, and constructs a structure parameter-frequency deviation response mapping model according to the independent contribution degrees to obtain specific influence modes and influence degrees of each tuning part;

[0041] A spring tuning part matching optimization module optimizes the structure parameter-frequency deviation response mapping model and the independent contribution degrees based on multi-conflict design objectives according to the specific influence modes and influence degrees of each tuning part, and outputs spring and tuning part combination matching suggestions.

[0042] Compared with the prior art, the beneficial effects of the present application are:

[0043] 1、The present application adopts orthogonal experimental design method to construct a disturbance combination matrix, which systematically arranges the combination disturbance of springs and various tuning part parameters. Compared with global disturbance, this method significantly improves the structural resolution of the interaction path between tuning parts while ensuring disturbance coverage. By constructing a disturbance combination matrix through orthogonal experimental design method, the influence of each structure parameter on the frequency deviation response can be clearly separated under non-interfering design, realizing the explicitness of the independent action path of the tuning part. This strategy provides a more interpretable structural basis for sensitivity evaluation and subsequent contribution extraction, and also provides a solid support for the accurate modeling of the subsequent structure parameter-frequency deviation response mapping model.

[0044] 2、The present application introduces a double-scale response analysis mechanism, which obtains macro-scale and micro-scale data, constructs a multi-scale response data set, and uses normalization and nonlinear fusion algorithm to construct a sensitivity index system, realizing unified modeling of structure part-system level response. This double-scale response analysis mechanism significantly improves the expression ability of local tuning on the influence of main modal frequency deviation, effectively solves the problems of "only focusing on system response and ignoring local transmission path" and difficulty in capturing the microscopic driving path of local structure part to system frequency deviation, and can effectively reflect how local structural changes are transmitted to the main modal of the whole vehicle. The modeling error is significantly reduced, and the model expression and stability are significantly enhanced.

[0045] 3、The present application is based on a multi-scale sensitivity index system, the independent contribution of each tuning part to the frequency deviation is evaluated, and a structure parameter-frequency deviation response mapping model is constructed by using a graph neural network, the influence mode and degree of the key tuning parts are accurately quantified. At the same time, through the contribution degree sorting to guide the combination optimization process, the problem that the traditional "black box type" intelligent optimization path is difficult to land is effectively avoided, the controllability and explainability of the tuning scheme recommendation result are realized, and the matching efficiency and tuning accuracy of the suspension system in actual engineering application are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 A flowchart of a spring tuning part matching optimization method based on suspension frequency deviation provided by an embodiment of the present application is shown in the figure.

[0047] Figure 2 A structural diagram of a spring tuning part matching optimization system based on suspension frequency deviation provided by an embodiment of the present application is shown in the figure.

[0048] Figure 3 A working principle diagram of a whole vehicle suspension system simulation model provided by an embodiment of the present application is shown in the figure.

[0049] Figure 4 A structural diagram of a structure parameter-frequency deviation response mapping model provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0051] The present application provides a spring tuning part matching optimization method based on suspension frequency deviation, which is applied to a spring tuning part matching optimization system based on suspension frequency deviation, can identify the main effect of parameter change of each tuning part on frequency deviation, realize traceable mapping from system frequency characteristics to component level action path, and improve the matching efficiency and tuning accuracy of the suspension system. In order to illustrate that the method of the present application can identify the main effect of parameter change of each tuning part on frequency deviation, realize traceable mapping from system frequency characteristics to component level action path, the effectiveness of the present application will be illustrated from two embodiments below.

[0052] Embodiment one

[0053] In the embodiment of the present application, the method proposed by the present application is used to identify the main effect of parameter change of each tuning part on the frequency deviation, realize traceable mapping from the system frequency characteristic to the component level action path, and improve the matching efficiency and tuning accuracy of the suspension system. The present application is applied to the parameter tuning process of the front suspension system of a B-class car, Figure 1 The method is a specific flowchart, which includes: extracting the spring structure parameters and the tuning part structure parameters associated with the frequency deviation response, and constructing a disturbance combination matrix; performing response analysis on the disturbance combination matrix based on the whole vehicle suspension system simulation model, obtaining first scale data and second scale data, and forming a multi-scale response data set; calculating the first sensitivity and the second sensitivity of the spring tuning part according to the multi-scale response data set, and constructing a multi-scale sensitivity index system; based on the multi-scale sensitivity index system, evaluating the independent contribution of each tuning part to the target frequency deviation, identifying the key influence tuning part, and constructing a structure parameter-frequency deviation response mapping model according to the independent contribution, to obtain the specific influence mode and influence degree of each tuning part; and optimizing according to the multi-conflict design target, and outputting the spring and tuning part combination matching suggestion. Figure 2 The present application is a structure schematic diagram of a spring tuning part matching optimization system based on suspension frequency deviation. The following is described according to the contents of Figure 1 and Figure 2

[0054] extracting the spring structure parameters and the tuning part structure parameters associated with the frequency deviation response, and constructing a disturbance combination matrix;

[0055] The spring structure parameters include: stiffness characteristic curve, effective number of turns, wire diameter, mean diameter and free length;

[0056] The tuning part structure parameters include: stiffness characteristic curve of auxiliary spring and intervention timing, material properties, height, action area and shape characteristics of the buffer block, diameter, arm length and connecting point hard point of the stabilizer bar, and radial stiffness, axial stiffness and torsional stiffness of the bushing;

[0057] The process of constructing the disturbance combination matrix is:

[0058] determining the value range and the number of levels of the spring structure parameters and the tuning part structure parameters, generating a test scheme containing multiple parameter level combinations by using an orthogonal test design method, and forming the disturbance combination matrix.

[0059] Specifically, the effective number of turns of the main spring: the value range is 5-7 turns, and the number of levels is 3, which are 5, 6 and 7 respectively;

[0060] The diameter of the stabilizer bar: the value range is 20-24 mm, and the number of levels is 3, which are 20 mm, 22 mm and 24 mm respectively; ​

[0061] Buffer block height: the value range is 50-70mm, the horizontal number is 3, and the values are 50mm, 60mm and 70mm respectively;

[0062] Lower control arm rear bushing radial stiffness: the value range is 300-500N / mm, the horizontal number is 3, and the values are 300N / m, 400N / mm and 500N / mm respectively; the above data are selected as the four main disturbance parameters;

[0063] L9(3 4 ) orthogonal table is used to generate 9 groups of parameter disturbance combinations. For example, one group is: effective number of turns 6 turns, stabilizer bar diameter 22mm, buffer block height 50mm, and bushing stiffness 400N / mm. Each row represents a specific suspension parameter configuration.

[0064] Table 1 shows the influence of different disturbance design methods on the structural parameter path resolution capability. The key parameter independent influence identification accuracy is calculated by comparing the consistency of the simulation response results and the true parameter influence relationship under different disturbance design methods, and the proportion of correctly identified key structural parameters is calculated; the micro-scale problem positioning efficiency is obtained by recording the average calculation time required to complete one effective micro-scale frequency anomaly positioning for each disturbance design method.

[0065] Table 1 Influence of different disturbance design methods on structural parameter path resolution capability

[0066]

[0067] By using the orthogonal disturbance modeling mechanism, the traditional global disturbance method which may have parameter coupling effects is converted into an orthogonalized and high-structure-resolution tuning parameter disturbance strategy. This disturbance strategy can explicitly identify the independent influence path of each tuning component (such as spring, stabilizer bar, buffer block, and bushing), laying a solid foundation for subsequent contribution evaluation and optimization. For example, when evaluating the influence of buffer block height on the suspension end impact characteristics, the orthogonal design can effectively eliminate the confusion effect caused by changes in other parameters such as spring effective number of turns or stabilizer bar diameter, so that the net contribution of buffer block height as a single factor can be more accurately quantified.

[0068] Further, the response analysis is performed on each disturbance combination in the disturbance combination matrix based on the whole vehicle suspension system simulation model to obtain first scale data and second scale data, and a multi-scale response data set is formed;

[0069] The whole vehicle suspension system simulation model includes a suspension system simulation unit, a frequency response analysis unit and a multi-scale data acquisition unit; refer to Figure 3 ;

[0070] The suspension system simulation unit simulates the whole vehicle suspension system based on the spring structure parameters and the tuning piece structure parameters to obtain a three-dimensional entity model of the whole vehicle suspension system;

[0071] The frequency response analysis unit is integrated in the suspension system simulation unit and is configured to perform frequency response analysis on each group of disturbance combinations in the input disturbance combination matrix;

[0072] The multi-scale data acquisition unit is configured to acquire first scale data and second scale data from the frequency response analysis result; the first scale data is macro-scale data, including a whole vehicle pitch modal frequency, a whole vehicle roll modal frequency, a whole vehicle vertical first-order main frequency, and a whole vehicle vertical second-order main frequency; and the second scale data is micro-scale data, including a local response frequency of a subassembly.

[0073] Specifically, the suspension system simulation unit adopts a three-dimensional entity model of the whole vehicle suspension system containing detailed geometric and physical characteristics constructed by a multi-body dynamics modeling tool (e.g., Adams Car or Simcenter Amesim); on the basis of a standard whole vehicle body-suspension-wheel system structure, the structure parameters of springs and tuning pieces are defined and replaced; the suspension system simulation unit simulates the whole vehicle suspension system based on the input spring structure parameters and tuning piece structure parameters; each group of disturbance parameters input corresponds to a new simulation scenario.

[0074] The frequency response analysis unit arranges sensors on a whole vehicle pitch axis, a vertical axis, and a roll axis respectively to output body modal acceleration; displacement sensors and acceleration sensors are arranged on subassembly nodes to obtain local response curves; the subassembly nodes include auxiliary spring nodes, stabilizer bar end nodes, and bushing center nodes;

[0075] After each group of disturbance simulation, a complete frequency response curve is output, and corresponding frequency peak points are derived.

[0076] The multi-scale data acquisition unit performs frequency spectrum analysis on the whole vehicle pitch angle and the body vertical displacement, identifies main peak positions, and extracts local response frequencies of subassemblies to obtain first scale data and second scale data;

[0077] The first scale data is macro-scale data, including:

[0078] Pitch modal frequency: the maximum acceleration response frequency of the body around the horizontal axis;

[0079] Roll modal frequency: the maximum response frequency of the body around the vertical axis;

[0080] Vertical first-order main frequency: the first main frequency point of the body in the vertical direction;

[0081] Vertical second order natural frequency: the second significant response frequency point of the vehicle body in the vertical direction;

[0082] The second scale data is micro scale data, including local response frequencies of subassemblies.

[0083] By establishing a high-precision whole vehicle suspension system simulation model and performing multi-scale response analysis, the present application realizes accurate mapping from component parameters to system performance. Compared with the traditional single scale analysis method, multi-scale data acquisition can capture system level and component level response characteristics through efficient acquisition of whole vehicle and substructure frequency response under a large number of parameter combinations, enhancing the fineness of subsequent modeling; at the same time, the explainability of the source of frequency deviation change is improved, and the traceability of the relationship between structure parameters and target frequency is enhanced.

[0084] Further, according to the multi-scale response data set, the first sensitivity and the second sensitivity of the spring adjustment part are calculated, and a multi-scale sensitivity index system is constructed;

[0085] The process of constructing the multi-scale sensitivity index system is:

[0086] The sensitivity values of the spring structure parameters and the adjustment part structure parameters under the first scale and the second scale are calculated respectively, to obtain the first sensitivity and the second sensitivity;

[0087] The first sensitivity and the second sensitivity are normalized;

[0088] The first sensitivity and the second sensitivity after normalization are fused by using a nonlinear fusion algorithm to obtain a fusion sensitivity value;

[0089] The fusion sensitivity values of all structure parameters are collected to form the multi-scale sensitivity index system.

[0090] Specifically, the relationship between the structure parameters and the macro scale index is fitted using kernel regression, and the local disturbance slope is extracted as the sensitivity value under the first scale; the same regression analysis is performed on the micro scale index to obtain the sensitivity value of each parameter under the second scale.

[0091] Table 2 shows the influence comparison table of different sensitivity analysis methods on modeling accuracy. Among them, the relative error of the frequency deviation prediction error is obtained by adjusting a single structure parameter under different sensitivity analysis methods and predicting the frequency deviation change, and comparing with the actual simulation result; the local response consistency score is obtained by comparing the response trend of the prediction model under local parameter disturbance with the multi-scale simulation data.

[0092] Table 2 Influence comparison of different sensitivity analysis methods on modeling accuracy

[0093]

[0094] The multi-scale sensitivity fusion algorithm used in this embodiment can consistently model and quantitatively evaluate the local influence of the structural member and the system-level macro response, effectively solving the problem of "local influence cannot be effectively transmitted to the system main mode" or unclear transmission path. Through the normalization and nonlinear fusion strategy, the comprehensive consideration of the response influence of different physical meanings and different dimensions is realized, so that engineers can balance local detail optimization and overall performance improvement based on a unified framework, providing a scientific basis for subsequent key tuning part identification and optimization design, and significantly improving the efficiency of suspension tuning.

[0095] Further, based on the multi-scale sensitivity index system, the independent contribution of each tuning part to the target frequency deviation is evaluated, the key influence tuning part is identified, and a structure parameter-frequency response mapping model is constructed according to the independent contribution, to obtain the specific influence mode and influence degree of each tuning part;

[0096] The process of identifying the key influence tuning part is:

[0097] Based on the multi-scale sensitivity index system, the influence relationship between the structure parameter and the frequency response of each tuning part is quantified;

[0098] According to the influence relationship, the independent contribution of each tuning part to the target frequency deviation is evaluated using the attribution analysis method;

[0099] The independent contribution threshold is set, and the tuning part corresponding to the structure parameter higher than the independent contribution threshold is marked as the key influence tuning part;

[0100] Based on the key influence tuning part, the structure parameter-frequency response mapping model is established using the graph neural network with the structure parameter of the key influence tuning part as the input and the frequency response as the output, and the specific influence mode and influence degree of each key influence tuning part are output.

[0101] Specifically, in this embodiment, the independent contribution threshold is the top few parameters accounting for 80% of the total influence.

[0102] Through contribution evaluation and identification of key influence tuning parts, this embodiment can focus analysis resources and optimization efforts on a few parameters that have a decisive effect on performance, significantly improving the efficiency of tuning. Further, the graph neural network is used to construct the parameter-response mapping, which can deeply explore the nonlinear coupling effect between parameters; at the same time, the identification of key tuning parts provides clear tuning priority guidance for engineers, avoiding the waste of time and cost caused by blind adjustment of non-key parameters.

[0103] Further, the structure parameter-frequency response mapping model comprises a structure graph construction unit, a graph neural network modeling unit, a frequency response prediction unit, and an influence mode and degree analysis unit; refer to Figure 4 ;

[0104] The structure graph construction unit constructs the spring and each tuning part into a structure graph, wherein each node represents a different tuning part, the features of each node represent the disturbance values of each tuning part in the disturbance combination matrix, and each edge represents the correlation between the tuning part structure parameters; the correlation between the tuning part structure parameters is obtained through correlation analysis;

[0105] The graph neural network modeling unit learns the influence relationship of each tuning part structure parameter on the suspension system frequency response based on the structure graph through feature propagation and graph convolution calculation;

[0106] The frequency response prediction unit decodes the high-dimensional embedding features output by the graph neural network modeling unit to obtain the predicted target frequency response value;

[0107] The influence mode and degree analysis unit performs a small disturbance on the parameter value of the key influence tuning part and observes the change amount and change direction of the predicted target frequency response value to obtain the specific influence mode and the influence degree.

[0108] In this embodiment, the main spring and the identified key influence tuning part in the front suspension system of the B-class car are constructed into a structure graph.

[0109] In the structure parameter-frequency response mapping model of the B-class car, the key influence tuning part is the main spring, the stabilizer bar, and the lower control arm rear bushing; the input features are the normalized parameter disturbance values thereof; and the edges are defined according to their mechanical connection in the MacPherson suspension. A 2-layer GAT network is used, followed by a ReLU activation function in each layer. After graph embedding learning, an average pooling layer is used to obtain the overall representation of the graph, and then two fully connected layers are used to predict the pitch and roll modal frequencies, respectively. After training, for example, the stabilizer bar diameter and bushing stiffness are fixed, the effective number of turns of the main spring is changed from -1 (the lowest level) to +1 (the highest level), and the change curve of the pitch modal frequency output by the structure parameter-frequency response mapping model is observed to obtain the nonlinear influence mode of the effective number of turns on the pitch frequency and the specific influence degree in different intervals, i.e., the slope.

[0110] Table 3 shows the influence of different optimization mechanisms on the performance and explainability of the tuning scheme. The recommendation accuracy is obtained by comparing the tuning scheme output by each optimization method with the actual optimal frequency performance, and the matching accuracy is calculated; the parameter controllability score is calculated by expert scoring and standardized to the range of 0-1 according to the feasibility and engineering realization degree of the adjustment range of each structural parameter in the output scheme; the explainability score is obtained by averaging the scores of multiple experts according to whether the optimization mechanism can clearly indicate the influence path and direction of the parameters on the frequency response.

[0111] Table 3 Influence of different optimization mechanisms on the performance and explainability of the tuning scheme

[0112]

[0113] By constructing a structural parameter-frequency response mapping model based on a graph neural network, the present application realizes deep mining and accurate modeling of the complex and nonlinear coupling relationship between the parameters of each tuning part in a complex suspension system. This model not only accurately predicts the frequency response after parameter adjustment (with a prediction error of less than 5%), but also better captures the comprehensive influence of parameters on the final system response through multiple paths and multiple levels of transmission, and the influence mode and degree analysis unit can provide clear and quantitative parameter-response relationships, providing strong technical support for subsequent optimization and tuning decisions.

[0114] Further, based on the specific influence mode and influence degree of each tuning part, the structural parameter-frequency response mapping model and the independent contribution degree are optimized based on multiple conflict design objectives, and a spring and tuning part combination matching suggestion is output.

[0115] In this embodiment, the multiple conflict design objectives include a frequency target, a ride comfort target, a handling stability target, a durability target, and a weight consideration.

[0116] By systematically constructing parameter perturbations, performing multi-scale simulation response analysis, constructing a multi-scale sensitivity index system that integrates macro and micro effects, and combining an advanced structural parameter-frequency response mapping model to accurately evaluate the independent contribution degree and specific influence mode and degree of each tuning part, a clear and traceable mapping relationship from the overall vehicle target frequency characteristic to the specific component parameter action path is realized. This not only significantly improves the efficiency and tuning accuracy of suspension parameter matching design in dealing with multiple conflict performance objectives, but also outputs spring and tuning part optimization configuration suggestions with stronger engineering explainability and controllability, thereby effectively shortening the vehicle development cycle and improving the final overall vehicle dynamic performance.

[0117] Embodiment Two

[0118] In the embodiment one, the method of the present application successfully realizes the traceable mapping from the system frequency characteristics to the component level effect path by identifying the main effect of the parameter change of each adjusting part on the frequency deviation, thereby improving the matching efficiency and adjusting accuracy of the suspension system. In order to further verify the effectiveness of the present application, the parameter adjustment optimization of the front suspension system of another car is also carried out in the embodiment of the present application.

[0119] A spring adjusting part matching optimization system based on suspension frequency deviation, comprising:

[0120] A parameter disturbance generation module, which extracts the spring structure parameters and adjusting part structure parameters associated with the frequency response, and constructs a disturbance combination matrix.

[0121] The spring structure parameters include stiffness characteristic curve, effective number of turns, wire diameter, mean diameter and free length; and the adjusting part structure parameters include the stiffness characteristic curve and intervention timing of the auxiliary spring, the material property, height, action area and shape characteristic of the buffer block, the diameter, arm length and connecting point hard point of the stabilizer bar, and the radial stiffness, axial stiffness and torsional stiffness of the bushing.

[0122] The process of constructing the disturbance combination matrix is that the value range and level number of the spring structure parameters and the adjusting part structure parameters are determined, the orthogonal experimental design method is used to generate an experimental scheme containing multiple parameter level combinations, and the disturbance combination matrix is formed.

[0123] A multi-scale response analysis module, which performs response analysis on each group of disturbance combinations in the disturbance combination matrix based on a whole vehicle suspension system simulation model, acquires first scale data and second scale data, and forms a multi-scale response data set.

[0124] The whole vehicle suspension system simulation model includes a suspension system simulation unit, a frequency response analysis unit and a multi-scale data acquisition unit.

[0125] The suspension system simulation unit simulates the whole vehicle suspension system based on the spring structure parameters and the adjusting part structure parameters to obtain a three-dimensional entity model of the whole vehicle suspension system; the frequency response analysis unit is integrated in the suspension system simulation unit and is used to perform frequency response analysis on each group of disturbance combinations in the input disturbance combination matrix; the multi-scale data acquisition unit is used to acquire first scale data and second scale data from the frequency response analysis result; wherein the first scale data is macro-scale data, including the whole vehicle pitch modal frequency, the whole vehicle roll modal frequency, the whole vehicle vertical first order main frequency and the whole vehicle vertical second order main frequency; and the second scale data is micro-scale data, including the local response frequency of the subassembly.

[0126] a multi-scale sensitivity calculation module, configured to calculate first sensitivity and second sensitivity of the spring tuning part according to the multi-scale response data set, and construct a multi-scale sensitivity index system; the process of constructing the multi-scale sensitivity index system is:

[0127] respectively calculate the sensitivity values of the spring structure parameters and the tuning part structure parameters under the first scale and the second scale to obtain the first sensitivity and the second sensitivity; normalize the first sensitivity and the second sensitivity; fuse the normalized first sensitivity and the second sensitivity by using a nonlinear fusion algorithm to obtain a fused sensitivity value; and collect the fused sensitivity values of all structure parameters to form the multi-scale sensitivity index system.

[0128] a tuning part contribution evaluation module, configured to evaluate independent contribution degrees of each tuning part to the target frequency deviation based on the multi-scale sensitivity index system, identify key influence tuning parts, and construct a structure parameter-frequency response mapping model according to the independent contribution degrees to obtain specific influence modes and influence degrees of each tuning part; the process of identifying the key influence tuning parts is:

[0129] quantify the influence relationship between each tuning part structure parameter and the frequency response based on the multi-scale sensitivity index system; evaluate the independent contribution degrees of each tuning part to the target frequency deviation by using an attribution analysis method according to the influence relationship; set an independent contribution degree threshold, and mark the tuning part corresponding to the structure parameter higher than the independent contribution degree threshold as the key influence tuning part; based on the key influence tuning parts, use the structure parameters of the key influence tuning parts as inputs and the frequency response as outputs to establish a structure parameter-frequency response mapping model by using a graph neural network, and output the specific influence modes and influence degrees of each key influence tuning part.

[0130] The structure parameter-frequency response mapping model comprises a structure graph construction unit, a graph neural network modeling unit, a frequency response prediction unit, and an influence mode and degree analysis unit.

[0131] The structure graph construction unit constructs the spring and each tuning part into a structure graph, wherein each node represents a different tuning part, the features of each node represent the disturbance values of each tuning part in the disturbance combination matrix, and each edge represents the correlation between tuning part structure parameters; the correlation between tuning part structure parameters is obtained by correlation analysis.

[0132] The graph neural network modeling unit learns the influence relationship of each tuning part structure parameter on the frequency response of the suspension system based on the structure graph by feature propagation and graph convolution calculation;

[0133] The frequency offset response prediction unit decodes the high-dimensional embedding features output by the graph neural network modeling unit to obtain a predicted target frequency offset response value;

[0134] The influence mode and degree analysis unit performs a small perturbation on the parameter value of the key influence adjustment component and observes the change amount and change direction of the predicted target frequency offset response value to obtain the specific influence mode and the influence degree.

[0135] The spring adjustment component matching optimization module optimizes the structure parameter-frequency offset response mapping model and the independent contribution degree based on multiple conflict design objectives according to the specific influence mode and influence degree of each adjustment component, and outputs a spring and adjustment component combination matching suggestion.

[0136] Table 4 shows the technical innovation points and verification comparison table of the present scheme. Among them, the multi-scale sensitivity fusion technology combines the response characteristics of macro and micro scales; the graph neural network path modeling technology constructs a graph structure prediction model between the structure parameters and the frequency offset response, effectively capturing the complex relationship between the parameters; and the multi-objective combination optimization mechanism generates an adjustment combination suggestion with executability and engineering controllability.

[0137] Table 4 Technical innovation points and verification comparison table of the present scheme

[0138]

[0139] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A spring tuning piece matching optimization method based on suspension natural frequency, characterized by, The method comprises the following steps: extracting spring structure parameters and tuning piece structure parameters associated with the frequency response, and constructing a disturbance combination matrix; performing response analysis on each disturbance combination in the disturbance combination matrix based on a whole vehicle suspension system simulation model to obtain first scale data and second scale data, and forming a multi-scale response data set; the first scale data is macro-scale data, and the second scale data is micro-scale data; According to the multi-scale response data set, the first sensitivity and the second sensitivity of the spring tuning piece are calculated, and a multi-scale sensitivity index system is constructed; the specific process is: the sensitivity values of the spring structure parameters and the tuning piece structure parameters under the first scale and the second scale are calculated respectively, and the first sensitivity and the second sensitivity are obtained; normalizing the first sensitivity and the second sensitivity; using a nonlinear fusion algorithm to fuse the normalized first sensitivity and the second sensitivity to obtain a fusion sensitivity value; all structure parameter fusion sensitivity values are collected to form the multi-scale sensitivity index system; Based on the multi-scale sensitivity index system, the independent contribution of each tuning piece to the target frequency deviation is evaluated, the key influence tuning piece is identified, and a structure parameter-frequency response mapping model is constructed according to the independent contribution, and the specific influence mode and influence degree of each tuning piece are obtained; According to the specific influence mode and influence degree of each tuning piece, the structure parameter-frequency response mapping model and the independent contribution are optimized based on multi-conflict design goals, and the spring and tuning piece combination matching suggestion is output.

2. The method of claim 1, wherein, The spring structure parameters include: stiffness characteristic curve, effective number of turns, wire diameter, medium diameter and free length; the tuning piece structure parameters include: stiffness characteristic curve and intervention timing of auxiliary spring, material properties, height, action area and shape characteristics of buffer block, diameter, arm length and connection point hard point of stabilizer bar, and radial stiffness, axial stiffness and torsional stiffness of bushing; the process of constructing the disturbance combination matrix is: determining the value range and level number of the spring structure parameters and the tuning piece structure parameters, and generating a test scheme containing multiple parameter level combinations by using orthogonal test design method to form the disturbance combination matrix.

3. The method of claim 1, wherein, The whole vehicle suspension system simulation model comprises: a suspension system simulation unit, a frequency response analysis unit and a multi-scale data acquisition unit; the suspension system simulation unit simulates the whole vehicle suspension system based on the spring structure parameters and the tuning piece structure parameters to obtain a whole vehicle suspension system three-dimensional entity model; the frequency response analysis unit is integrated in the suspension system simulation unit and is used for frequency response analysis on each disturbance combination in the input disturbance combination matrix; the multi-scale data acquisition unit is used to obtain first scale data and second scale data from the frequency response analysis result; wherein, the first scale data includes whole vehicle pitch modal frequency, whole vehicle roll modal frequency, whole vehicle vertical first order main frequency and whole vehicle vertical second order main frequency; the second scale data includes sub-assembly local response frequency.

4. The method of claim 1, wherein, The process of identifying the key influence adjustment part is: based on a multi-scale sensitivity index system, the influence relationship between each adjustment part structure parameter and the frequency response is quantified; according to the influence relationship, the independent contribution of each adjustment part to the target frequency is evaluated by using the attribution analysis method; Set the independent contribution threshold, and mark the adjustment part corresponding to the structure parameter higher than the independent contribution threshold as the key influence adjustment part; based on the key influence adjustment part, the structure parameter of the key influence adjustment part is taken as the input and the frequency response is taken as the output, and a structure parameter-frequency response mapping model is established by using a graph neural network, and the specific influence mode and influence degree of each key influence adjustment part are output.

5. The method of claim 4, wherein, The structure parameter-frequency response mapping model includes: a structure graph construction unit, a graph neural network modeling unit, a frequency response prediction unit, and an influence mode and degree analysis unit; the structure graph construction unit constructs the spring and each adjustment part into a structure graph, wherein the nodes represent different adjustment parts, the characteristics of the nodes represent the disturbance values of each adjustment part in the disturbance combination matrix, and the edges represent the correlation between the adjustment part structure parameters; the correlation between the adjustment part structure parameters is obtained by correlation analysis; the graph neural network modeling unit performs feature propagation and graph convolution calculation based on the structure graph to learn the influence relationship of each adjustment part structure parameter on the suspension system frequency response; the frequency response prediction unit decodes the influence relationship to obtain the predicted target frequency response value; the influence mode and degree analysis unit performs a small disturbance on the parameter value of the key influence adjustment part, and observes the change amount and direction of the predicted target frequency response value to obtain the specific influence mode and the influence degree.

6. A suspension frequency-based spring alignment component matching optimization system, characterized by, A spring adjustment part matching optimization method based on suspension frequency is executed, as claimed in claim 1, comprising: A parameter disturbance generation module extracts the spring structure parameters and adjustment part structure parameters associated with the frequency response, and constructs a disturbance combination matrix; A multi-scale response analysis module performs response analysis on each group of disturbance combinations in the disturbance combination matrix based on the whole vehicle suspension system simulation model, obtains first scale data and second scale data, and forms a multi-scale response data set; A multi-scale sensitivity calculation module calculates the first sensitivity and the second sensitivity of the spring adjustment part according to the multi-scale response data set, and constructs a multi-scale sensitivity index system; An adjustment part contribution evaluation module evaluates the independent contribution of each adjustment part to the target frequency based on the multi-scale sensitivity index system, identifies the key influence adjustment part, and constructs a structure parameter-frequency response mapping model according to the independent contribution to obtain the specific influence mode and influence degree of each adjustment part; A spring adjustment part matching optimization module optimizes the structure parameter-frequency response mapping model and the independent contribution based on the multi-conflict design target according to the specific influence mode and influence degree of each adjustment part, and outputs the spring and adjustment part combination matching suggestion.

Images