CN114611211A - Vehicle suspension performance degradation parameter identification method - Google Patents

Vehicle suspension performance degradation parameter identification method Download PDF

Info

Publication number
CN114611211A
CN114611211A CN202210235674.4A CN202210235674A CN114611211A CN 114611211 A CN114611211 A CN 114611211A CN 202210235674 A CN202210235674 A CN 202210235674A CN 114611211 A CN114611211 A CN 114611211A
Authority
CN
China
Prior art keywords
vehicle
suspension
model
establishing
loop
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210235674.4A
Other languages
Chinese (zh)
Inventor
潘勇军
孙宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University
Original Assignee
Chongqing University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing University filed Critical Chongqing University
Priority to CN202210235674.4A priority Critical patent/CN114611211A/en
Publication of CN114611211A publication Critical patent/CN114611211A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Vehicle Body Suspensions (AREA)

Abstract

The invention discloses a vehicle suspension performance degradation parameter identification method, which comprises the following steps: 1) establishing a vehicle multi-body dynamic model; 2) establishing a deceleration strip road surface model, developing vehicle dynamics simulation, and acquiring vehicle state data; 3) establishing a deep learning DNN model based on the vehicle state and the corresponding suspension parameter data set; 4) identifying vehicle suspension parameters based on the DNN model and the actual vehicle state, and evaluating degradation of vehicle suspension performance; 5) analyzing the prediction precision of the DNN model based on different sample number sets and different hidden layer numbers; 6) the effect on the accuracy of the established DNN model is discussed as the input of different vehicle states. The method can provide model support and design reference for automatic driving automobile suspension parameter identification, suspension parameter optimization and the like; the research of the identification of the suspension parameters aiming at the performance degradation has very important engineering practical value.

Description

Vehicle suspension performance degradation parameter identification method
Technical Field
The invention relates to a vehicle suspension performance degradation parameter identification method.
Background
With the rapid development of automobile industry in China, automobiles and traffic flow are greatly increased, and a suspension system is used as a key core component on the automobiles and plays a vital role in the aspect of riding comfort. Due to the harsh environment of a driving road, abrasion and aging of a suspension, the performance of the suspension can be degraded along with the change of a stiffness coefficient and a damping coefficient of the suspension, and the riding comfort of passengers is affected. In addition, if the degraded suspension is not maintained, the performance of other parts of the automobile is affected, and the fatigue life and riding comfort of the whole automobile are greatly affected.
The suspension is a component with elasticity and used for connecting a frame and an axle of an automobile, generally comprises components such as an elastic element, a guide mechanism, a shock absorber and the like, and is mainly used for relieving the impact transmitted to the automobile body from an uneven road surface so as to improve the riding comfort. The performance of the suspension is mainly determined by key parameters such as rigidity, damping and the like. The stiffness and damping of the suspension is difficult to assess during driving, which requires disassembly of the vehicle in order to perform cumbersome tests on the individual components. If the parameters of the suspension with degraded performance are identified by adopting a disassembly test method, the cost is too high.
Disclosure of Invention
The invention aims to provide a vehicle suspension performance degradation parameter identification method to solve the problems in the prior art.
The technical scheme adopted for achieving the aim of the invention is that the vehicle suspension performance degradation parameter identification method comprises the following steps:
1) establishing a vehicle multi-body dynamic model;
2) establishing a deceleration strip road surface model, developing vehicle dynamics simulation, and acquiring vehicle state data;
3) establishing a deep learning DNN model based on the vehicle state and the corresponding suspension parameter data set;
4) identifying vehicle suspension parameters based on the DNN model and the actual vehicle state, and evaluating degradation of vehicle suspension performance;
5) analyzing the prediction precision of the DNN model based on different sample number sets and different hidden layer numbers;
6) the effect on the accuracy of the established DNN model is discussed as the input of different vehicle states.
Further, the step 1) comprises the following sub-steps:
1-1) establishing a dynamic equation of an open-loop multi-body system;
1-2) establishing a semi-recursive kinetic equation of the closed-loop multi-body system;
1-3) combining an open-loop kinetic equation and a closed-loop semi-recursive kinetic equation to establish a vehicle semi-recursive multi-body kinetic model.
Further, the multi-body system is a vehicle.
Further, the dynamic equation expression of the open-loop multi-body system in the step 1-1) is as follows:
Figure BDA0003542050750000021
in the formula, matrix M∑Matrix Q∑Matrix P∑Respectively representing the accumulated total moment of inertia, external force and velocity-dependent inertial force of the multi-body system; rdRepresenting a first time velocity transformation matrix;
Figure BDA0003542050750000022
is a second derivative used to establish the relative coordinates of the open loop kinetic equation.
Further, the expression of the semi-recursive kinetic equation of the closed-loop multi-body system in the step 1-2) is as follows:
Figure BDA0003542050750000023
in the formula (I), the compound is shown in the specification,
Figure BDA0003542050750000024
is the average moment of inertia of the multi-body system; r iszIs a second velocity transformation matrix; t is a path matrix, RzRelative coordinates of closed loop system
Figure BDA0003542050750000025
Using a set of mutually independent relative coordinates
Figure BDA0003542050750000026
To describe;
Figure BDA0003542050750000027
are relative accelerations independent of each other.
Further, in the DNN model established in the step 3), vehicle states are used as input, suspension parameters are used as output, and the vehicle states comprise vehicle pitch angles, centroid vertical positions, vertical speeds and vertical accelerations.
The invention has the beneficial effects that:
1. according to the invention, a vehicle multi-body dynamic model is established based on relative coordinates, so that the dimension of a vehicle multi-body system dynamic equation is reduced, and more efficient real-time simulation and data acquisition of vehicle dynamics are realized;
2. the method combines vehicle multi-body dynamics simulation with deep learning to identify the parameters of the degraded suspension performance; various DNN prediction models are established based on different sample set numbers, different hidden layer numbers and different vehicle state input parameters, suspension parameters are predicted by utilizing the established DNN models based on actual vehicle states, and the prediction precision of each model is very high; the model prediction results of the DNN model established by the 3 hidden layers and the 2000 groups of sample sets are compared with multi-body model data, the average absolute percentage errors of the suspension stiffness and the damping are respectively less than 0.05 percent and 0.2 percent, and meanwhile, the consumed time is less. In addition, the vehicle pitch angle speed is added into the model input, so that the prediction precision of the model can be further improved;
3. the method combines a vehicle semi-recursion multi-body dynamic model and a deep learning method to identify the parameters of the suspension with degraded performance; the method can provide model support and design reference for automatic driving automobile suspension parameter identification, suspension parameter optimization and the like; the research of the identification of the suspension parameters aiming at the performance degradation has very important engineering practical value.
Drawings
FIG. 1 is a diagram of a dynamic simulated pavement model;
FIG. 2 is a cross-sectional view of a speed bump;
FIG. 3 is a schematic diagram of the forward/backward propagation of DNN;
FIG. 4 is a flow chart of DNN modeling for vehicle suspension parameter identification;
FIG. 5 is a graph of DNN and results of a multi-body model;
FIG. 6 is a box plot of absolute percent error for suspension parameters;
FIG. 7 is a graph comparing the results of different DNN models in terms of MAE;
FIG. 8 is a graph comparing the results of MAE at different DNN inputs.
Detailed Description
The present invention is further illustrated by the following examples, but it should not be construed that the scope of the above-described subject matter is limited to the following examples. Various substitutions and alterations can be made without departing from the technical idea of the invention and the scope of the invention is covered by the present invention according to the common technical knowledge and the conventional means in the field.
Example 1:
the embodiment discloses a vehicle suspension performance degradation parameter identification method, which comprises the following steps:
1) the method comprises the following steps of establishing a vehicle semi-recursion multi-body dynamic model:
1-1) describing the control equations of a multi-body system with a set of independent relative coordinates;
1-2) describing the characteristics of Cartesian coordinates and relative coordinates in the system based on a recursion formula, reducing the dimension of a vehicle multi-body system dynamic equation, and further improving the calculation efficiency of vehicle dynamic real-time simulation and control;
1-3) open-loop system multi-body dynamics modeling:
for a revolute or prismatic pair with one degree of freedom, the Cartesian coordinates (Z) of the upper part (i) of the mechanismi) The Cartesian coordinates (Z) of the lower part (i-1) can be usedi-1) And the relative coordinates between them
Figure BDA0003542050750000031
To show that other forms of kinematic pairs (such as universal pairs and ball pairs) can be decomposed into a one-degree-of-freedom revolute pair or moving pair and a plurality of virtual components without massCombining;
if the virtual reference point s on the component i extending to the global coordinate system origin is selected as the local reference coordinate system origin, the local coordinate systems of all the components in the multi-body system are kept consistent, which can avoid introducing a coordinate transformation matrix when the cartesian speeds of the connected components are recurred.
Figure BDA0003542050750000041
Figure BDA0003542050750000042
In the formula: biAs a vector related to velocity, diIs a vector related to acceleration.
For an open loop system or an open closed loop system, the cartesian coordinates (Z) of the system can be found using a recursive kinematic formula:
Figure BDA0003542050750000043
wherein: the matrix R is called the first velocity transformation matrix and can use the relative coordinates of the Cartesian velocity Z
Figure BDA0003542050750000044
To describe; the matrix T is a path matrix of the multi-body system, which reflects the connection relationship among the components of the multi-body system.
The first time velocity transformation matrix is introduced, and a Lagrange formula or a virtual work principle is utilized to deduce a recursive multi-body kinetic equation of an open-loop system:
Figure BDA0003542050750000045
wherein: matrix M∑、Q∑、P∑Respectively representing the total accumulated moment of inertia, external forces andthe inertial force that is dependent on the velocity,
Figure BDA0003542050750000046
is a second derivative used to establish the relative coordinates of the open loop kinetic equation.
1-4) semi-recursive multi-body dynamics modeling of a closed-loop system:
in order to obtain the dynamic characteristics of the closed-loop system, on the basis of a recursion dynamic equation, a closed-loop constraint equation generated by the removed kinematic pair or the light connecting rod needs to be considered;
firstly, establishing a closed-loop constraint equation by using a natural coordinate system, and then mapping the closed-loop constraint equation into a relative coordinate system; the natural coordinate system adopts two Cartesian coordinate positions and directions to describe the position of the rigid body; the great advantage is that complex rotational modeling is avoided.
The Jacobian matrix of the closed-loop constraint equation can be calculated through a recursive kinematic formula, a Lagrange multiplier is introduced, and the multi-body kinetic equation of the closed-loop system can be expressed as follows:
Figure BDA0003542050750000047
in the formula: phiZλ is the lagrange operator for the jacobian matrix of the constraint equation.
The equation is a multi-body kinetic equation of a closed-loop system in the form of a differential algebraic equation system, in order to reduce the dimension of the kinetic equation, a group of mutually independent relative coordinates are extracted, and matrix blocking is carried out on a closed-loop constraint equation:
Figure BDA0003542050750000051
Figure BDA0003542050750000052
wherein: r iszIs a second velocity transformation matrix capable of transforming the relative coordinates of the closed-loop system
Figure BDA0003542050750000053
Using a set of mutually independent relative coordinates
Figure BDA0003542050750000054
To describe; the dimension of the system multi-body dynamic equation can be greatly reduced by combining the two velocity transformation matrixes.
The second time velocity transformation matrix is essentially a set of basis of constraint equation Jacobian null space, and the Lagrangian multiplier can be eliminated through the introduction of the second time velocity transformation matrix. And finally deducing an ordinary differential multi-body kinetic equation of the closed-loop multi-body system through a series of mathematical transformations:
Figure BDA0003542050750000055
in the formula:
Figure BDA0003542050750000056
is the average moment of inertia of the multi-body system.
1-5) based on the steps 1-1) to 1-4), and based on a two-step semi-recursive multi-body dynamic modeling method and program codes, a closed-loop multi-body system of the passenger vehicle with fourteen degrees of freedom is established.
2) Setting a deceleration strip road surface in the multi-body system dynamics model set up in the step 1), developing vehicle dynamics simulation, changing parameters of front and rear suspensions, and obtaining corresponding vehicle states. Fig. 1 shows the road surface condition of the dynamics simulation, a deceleration strip is arranged, and the acquired data are the vehicle state when the rear wheel of the vehicle just passes through the deceleration strip, including the vehicle pitch angle, the vertical position of the mass center, the vertical speed, the vertical acceleration and the like. Fig. 2 shows a model of the deceleration strip which is built up and has a height of 50mm and a radius of 205 mm. In addition, some parameters of the multi-body model of the vehicle are shown in table 1:
TABLE 1 vehicle model part parameters
Figure BDA0003542050750000057
3) Establishing a suspension parameter prediction DNN model, wherein in the process, according to different sample set numbers and different hidden layer numbers, multiple deep learning DNN models based on vehicle states and suspension parameter data sets are established, and the method specifically comprises the following steps:
3-1) DNN comprises three parts, namely an input layer, a hidden layer and an output layer. The neurons of the input layer are responsible for receiving and propagating data values forward into neurons of the middle layer of the neural network, i.e., the hidden layer. The weighted sum of the hidden layers is finally propagated forward to the output layer, and the output layer shows the output result of the neural network. Forward propagation and backward propagation are referred to herein as processes for training the DNN model. The DNN alternately performs forward propagation and backward propagation, and calculates gradient iterative model parameters from the backward propagation.
3-2) flow of forward and backward propagation of DNN is shown in FIG. 3.
3-3) in the forward propagation process, an input matrix, a weight matrix and a deviation matrix need to be defined, namely:
Z0＝(i1，i2，i3，···im) (8)
Wn＝(wn1，wn2，wn3，···wnm) (9)
Bn＝(bn1，bn2，bn3，···bnm) (10)
in the formula: m represents the number of sample sets; n represents the total number of hidden layers and output layers; z0Representing the input matrix, namely the vehicle state; wnAnd BnRespectively representing the weight matrix and the deviation matrix of the nth layer.
3-4) the forward propagation can be expressed as:
A0＝Z0 (11)
Zi＝Wi TAi-1+Bi，i＝1···n (12)
Ai＝fi(Zi)，i＝1···n (13)
in the formula: ziRepresents the input of the i-th layer, AiRepresents the output of the i-th layer, fiAnd represents the activation function of the ith layer. The Tansig function is chosen in the hidden layer and the Purelin function is used between the hidden layer and the output layer.
3-5) before back-propagation, an appropriate loss function needs to be defined to calculate the error between the DNN model result and the reference data. Mean Square Error (MSE) is chosen as a loss function:
Figure BDA0003542050750000061
in the formula: l represents the loss of the DNN model; y represents reference data; a. thenThe results of the DNN model are shown.
In the back propagation process, the weight matrix and the bias matrix of the ith layer are updated according to the following formula:
Figure BDA0003542050750000071
Figure BDA0003542050750000072
in the formula, α represents a learning rate. It is a very important hyper-parameter and a good learning rate can bring the DNN to the minimum of losses faster.
4) And taking the actual vehicle state (vehicle pitch angle, vertical centroid position, centroid speed and centroid acceleration) as the input of the established DNN model, predicting the suspension parameters corresponding to the actual vehicle state by utilizing the prediction function of the model, and evaluating the degradation of the vehicle suspension performance. FIG. 4 shows a DNN modeling flow for vehicle suspension parameter identification.
5) Analyzing the prediction precision of the DNN model based on different sample set numbers and different hidden layer numbers, which comprises the following steps:
5-1) establishing various DNN models by 2000 groups, 1500 groups and 1000 groups and 3 hidden layers, 4 hidden layers and 5 hidden layers respectively and training. And predicting vehicle suspension parameters by using the trained DNN model and taking the vehicle pitch angle, the vertical position of the mass center, the vertical speed and the vertical acceleration as input.
5-2) take 2000 sets of data, 3 hidden layer DNN model as an example. Fig. 5 depicts the relationship of DNN model inputs (vehicle state) and outputs (suspension parameters) and compares DNN predicted suspension parameters with reference data (multi-body model data). The abscissa and ordinate respectively represent the stiffness and damping of the rear suspension, and the vertical coordinate represents four vehicle states, wherein fig. 5a, 5b, 5c and 5d respectively correspond to four vehicle states of a pitch angle, a vertical position of a center of mass, a vertical velocity and a vertical acceleration.
5-3) the boxplot of FIG. 6 is used to describe the absolute percent error between the DNN results and the multiple-body model results, where FIGS. 6a, 6b, 6c, and 6d correspond to front suspension stiffness, front suspension damping, rear suspension stiffness, and rear suspension damping, respectively. Elements in the block diagram include a median, a mean, a lower quartile, and an upper quartile.
5-4) in order to express the accuracy of the DNN model in detail, the Mean Absolute Error (MAE), the Mean Absolute Percent Error (MAPE), the maximum absolute error (ME), the Root Mean Square Error (RMSE) and the coefficient of determination (R) are selected2) And the like as precision evaluation indexes:
Figure BDA0003542050750000073
Figure BDA0003542050750000081
Figure BDA0003542050750000082
Figure BDA0003542050750000083
Figure BDA0003542050750000084
wherein: y isiEach representing the true value of the ith sample,
Figure BDA0003542050750000085
each represents a predicted value of the ith sample,
Figure BDA0003542050750000086
representing predicted values
Figure BDA0003542050750000087
N each represents the number of samples. ME, MAE, MAPE, and RMSE are used to evaluate regression prediction models, whose values represent the associated errors, the smaller the error, the higher the accuracy of the model. R2Is used to measure the quality of the regression model, R2The larger the value, the better the performance of the model. Table 2, table 3, and table 4 describe evaluation indices of DNN model output using 2000, 1500, and 1000 data sets and 3, 4, and 5 hidden layers, respectively. In the table, DNN3_2000 represents a DNN model of a 3 hidden layer, 2000 group dataset.
Table 2.2000 DNN model accuracy of sets of data sets
Figure BDA0003542050750000088
Table 3.1500 DNN model accuracy of sets of data sets
Figure BDA0003542050750000091
Table 4.1000 DNN model accuracy of sets of data sets
Figure BDA0003542050750000092
As can be seen from the table, MAE, MAPE and RMSE are relatively small, R2Close to 1. The results show that these DNN models are able to accurately estimate suspension parameters from four states of the vehicle. Among these, the DNN model based on 2000 data sets is the most accurate.
5-5) FIG. 7 compares the results of the various DNN models, using MAE as an example. As can be seen from the figure, the DNN model with 3 hidden layers and 2000 groups of data is the optimal DNN model for suspension parameter identification, which not only has higher precision, but also has less time consumption for training the model due to less number of hidden layers.
6) The vehicle state of the vehicle pitch angle and the vehicle speed is also used as one input of the DNN model, a plurality of DNN suspension parameter identification models with 5 inputs are established, and the established DNN model is subjected to further precision analysis, specifically:
6-1) compare it with the 4-input DNN model, taking the 3 hidden layers, 2000 sets of data as an example, and the comparison results are shown in Table 5. DNN4 → 4 denotes a DNN model with 4 inputs (vehicle state) and 4 outputs (suspension parameters). DNN5 → 4 represents a DNN model with 5 inputs and 4 outputs. As can be seen from the table, the accuracy of the 5-input DNN model is much higher.
TABLE 5.4 INPUT AND 5 INPUT DNN MODEL ACCURACY ANALYSIS
Figure BDA0003542050750000101
6-2) FIG. 8 shows the comparison results of the 4-input and 5-input DNN models with MAE as the comparison term. It is clear from the figure that adding vehicle pitch angle rate to the DNN model input will significantly improve the accuracy of the prediction of suspension parameters. However, it is very difficult to actually measure vehicle pitch angle rate, and therefore, the 4-input DNN model may be an optimal suspension parameter identification model in view of cost.
In summary, the present embodiment considers the problem of identifying the degradation parameter of the suspension performance from multiple aspects. The results show that the accuracy of several DNN models is very good, but 3 hidden layers, 2000 sets of data are the best models in view of time consuming and better prediction accuracy. In addition, continuing to increase the number of sample data sets may further improve accuracy. The embodiment also finds that the accuracy of the identification of the suspension parameters can be further improved by adding the vehicle pitch angle speed to the training of the model.
Example 2:
the embodiment discloses a vehicle suspension performance degradation parameter identification method, which comprises the following steps:
1) establishing a vehicle multi-body dynamic model;
2) establishing a deceleration strip road surface model, developing vehicle dynamics simulation, and acquiring vehicle state data;
3) establishing a deep learning DNN model based on the vehicle state and the corresponding suspension parameter data set;
4) identifying vehicle suspension parameters based on the DNN model and the actual vehicle state, and evaluating degradation of vehicle suspension performance;
5) analyzing the prediction precision of the DNN model based on different sample number sets and different hidden layer numbers;
6) the effect on the accuracy of the established DNN model is discussed as the input of different vehicle states.
Example 3:
the main steps of this embodiment are the same as those of embodiment 2, and further, step 1) includes the following sub-steps:
1-1) establishing a dynamic equation of an open-loop multi-body system;
1-2) establishing a semi-recursive kinetic equation of the closed-loop multi-body system;
1-3) combining an open-loop kinetic equation and a closed-loop semi-recursive kinetic equation to establish a vehicle semi-recursive multi-body kinetic model.
Example 4:
the main steps of this embodiment are the same as those of embodiment 3, and further, the multi-body system is a vehicle.
Example 5:
the main steps of this embodiment are the same as those of embodiment 3, and further, the equation of dynamics of the open-loop multi-body system in step 1-1) is expressed as:
Figure BDA0003542050750000111
in the formula, matrix M∑Matrix Q∑Matrix P∑Respectively representing the accumulated total moment of inertia, external force and velocity-dependent inertial force of the multi-body system; rdRepresenting a first time velocity transformation matrix;
Figure BDA0003542050750000121
is a second derivative used to establish the relative coordinates of the open loop kinetic equation.
Example 6:
the main steps of this embodiment are the same as those of embodiment 3, and further, the expression of the semi-recursive kinetic equation of the closed-loop multi-body system in step 1-2) is as follows:
Figure BDA0003542050750000122
in the formula (I), the compound is shown in the specification,
Figure BDA0003542050750000123
is the average moment of inertia of the multi-body system; rzIs a second velocity transformation matrix; t is a path matrix, RzRelative coordinates of closed loop system
Figure BDA0003542050750000124
Using a set of mutually independent relative coordinates
Figure BDA0003542050750000125
To describe;
Figure BDA0003542050750000126
are relative accelerations independent of each other.
Example 7:
in the main steps of this embodiment, as in embodiment 2, further, in the DNN model established in step 3), the vehicle state is used as input, the suspension parameters are used as output, and the vehicle state includes vehicle pitch angle, vertical position of the center of mass, vertical velocity and vertical acceleration.

Claims (6)

1. A vehicle suspension performance degradation parameter identification method is characterized by comprising the following steps:
1) establishing a vehicle multi-body dynamic model;
2) establishing a deceleration strip road surface model, developing vehicle dynamics simulation, and acquiring vehicle state data;
3) establishing a deep learning DNN model based on the vehicle state and the corresponding suspension parameter data set;
4) identifying vehicle suspension parameters based on the DNN model and the actual vehicle state, and evaluating degradation of vehicle suspension performance;
5) analyzing the prediction precision of the DNN model based on different sample number sets and different hidden layer numbers;
6) the effect on the accuracy of the established DNN model is discussed as the input of different vehicle states.
2. The vehicle suspension performance degradation parameter identification method according to claim 1, wherein: the step 1) comprises the following sub-steps:
1-1) establishing a dynamic equation of an open-loop multi-body system;
1-2) establishing a semi-recursive kinetic equation of the closed-loop multi-body system;
1-3) combining an open-loop kinetic equation and a closed-loop semi-recursive kinetic equation to establish a vehicle semi-recursive multi-body kinetic model.
3. The vehicle suspension performance degradation parameter identification method according to claim 2, characterized in that: the multi-body system is a vehicle.
4. The vehicle suspension performance degradation parameter identification method according to claim 1, wherein: the dynamic equation expression of the open-loop multi-body system in the step 1-1) is as follows:
Figure FDA0003542050740000011
in the formula, matrix M∑Matrix Q∑Matrix P∑Respectively representing the accumulated total moment of inertia, external force and velocity-dependent inertial force of the multi-body system; rdRepresenting a first time velocity transformation matrix;
Figure FDA0003542050740000013
is a second derivative used to establish the relative coordinates of the open loop kinetic equation.
5. The vehicle suspension performance degradation parameter identification method according to claim 2, wherein: the expression of the semi-recursive kinetic equation of the closed loop multi-body system in the step 1-2) is as follows:
Figure FDA0003542050740000012
in the formula (I), the compound is shown in the specification,
Figure FDA0003542050740000021
is the average moment of inertia of the multi-body system; rzIs a second velocity transformation matrix; t is a path matrix, RzRelative coordinates of closed loop system
Figure FDA0003542050740000022
Using a set of mutually independent relative coordinates
Figure FDA0003542050740000023
To describe;
Figure FDA0003542050740000024
are relative accelerations independent of each other.
6. The vehicle suspension performance degradation parameter identification method according to claim 1, wherein: and 3) in the DNN model established in the step 3), the vehicle state is used as input, the suspension parameters are used as output, and the vehicle state comprises the vehicle pitch angle, the mass center vertical position, the vertical speed and the vertical acceleration.
CN202210235674.4A 2022-03-11 2022-03-11 Vehicle suspension performance degradation parameter identification method Pending CN114611211A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210235674.4A CN114611211A (en) 2022-03-11 2022-03-11 Vehicle suspension performance degradation parameter identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210235674.4A CN114611211A (en) 2022-03-11 2022-03-11 Vehicle suspension performance degradation parameter identification method

Publications (1)

Publication Number Publication Date
CN114611211A true CN114611211A (en) 2022-06-10

Family

ID=81862987

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210235674.4A Pending CN114611211A (en) 2022-03-11 2022-03-11 Vehicle suspension performance degradation parameter identification method

Country Status (1)

Country Link
CN (1) CN114611211A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020185827A1 (en) * 2001-04-04 2002-12-12 Stmicroelectronics S.R.I. Method and apparatus for controlling a vehicle suspension system based on sky hook approach
CN109927501A (en) * 2019-03-12 2019-06-25 辽宁科技大学 A kind of intelligent control method of Vehicle Semi-active Suspension System
US20200384980A1 (en) * 2019-06-04 2020-12-10 Pony.ai, Inc. Systems and methods for implementing a preemptive suspension control for an autonomous vehicle to improve ride quality
CN113985728A (en) * 2021-09-08 2022-01-28 重庆大学 Vertical vibration suppression method for multi-body system of vehicle continuously bulging or bulging on road

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020185827A1 (en) * 2001-04-04 2002-12-12 Stmicroelectronics S.R.I. Method and apparatus for controlling a vehicle suspension system based on sky hook approach
CN109927501A (en) * 2019-03-12 2019-06-25 辽宁科技大学 A kind of intelligent control method of Vehicle Semi-active Suspension System
US20200384980A1 (en) * 2019-06-04 2020-12-10 Pony.ai, Inc. Systems and methods for implementing a preemptive suspension control for an autonomous vehicle to improve ride quality
CN113985728A (en) * 2021-09-08 2022-01-28 重庆大学 Vertical vibration suppression method for multi-body system of vehicle continuously bulging or bulging on road

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
NING HONG ET AL: "High-Speed Rail Suspension System Health Monitoring Using Multi-Location Vibration Data", IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEM, pages 2943 - 2955 *
YONGJU PAN: "maneuver-based deep learning parameter identification of vehicle suspension subjected to performance degradation", VEHICLE SYSTEM SYNAMICS, pages 1260 - 1276 *
YONGJUN PAN: "Machine learning approaches to estimate suspension parameters for performance degradation assessment using accurate dynamic simulations", RELIABILITY ENGINEERING AND SYSTEM SAFETY, pages 1 - 11 *
李海侠: "汽车前悬架隔振性能的实验分析与神经网络辨识", 桂林工学院学报, vol. 27, no. 01, pages 133 - 135 *
范政武 等: "基于人工鱼群算法的车辆平顺性优化分析", 农业工程学报, vol. 32, no. 06, pages 107 - 113 *

Similar Documents

Publication Publication Date Title
CN108020427B (en) Pure electric vehicle gear shifting quality evaluation method based on GA-BP neural network
CN109190171B (en) Vehicle motion model optimization method based on deep learning
CN112287551B (en) Driving performance system level index decomposition method based on whole vehicle conceptual model
CN113570057B (en) Vehicle wheel center vertical displacement measuring method and device based on model training
CN111444623B (en) Collaborative optimization method and system for damping nonlinear commercial vehicle suspension dynamics
CN110414052A (en) A kind of vehicle structure fatigue life calculation method composed based on digital sample car and virtual road
CN114379583A (en) Automatic driving vehicle trajectory tracking system and method based on neural network dynamics model
CN114722731B (en) Method and system for predicting drag viscous drag resistance of towed body
US20210124806A1 (en) Vehicle suspension
CN116560223A (en) Intelligent vehicle dynamics model, ILQR control algorithm and track tracking controller based on physical information neural network
CN113987678B (en) Road noise hybrid prediction and optimization method based on virtual wheel center force
Widner et al. Framework for vehicle dynamics model validation
CN114611211A (en) Vehicle suspension performance degradation parameter identification method
CN116415415A (en) Multi-suspension dynamic parameter distribution optimization design method for crawler-type vehicle
CN116859731A (en) Method for enhancing punctuality of high-speed rail automatic driving control system based on reinforcement learning
CN116415494A (en) Road noise optimization method, system and storage medium
CN116538286A (en) Commercial vehicle gear shifting system and method considering NVH characteristics
CN114278727B (en) Vehicle automatic gear shifting control method, system, medium and equipment under limit working condition
CN115659492A (en) Complex equipment nonlinear vibration model parameter optimization method
CN108427855B (en) Method for designing performance parameters of wheel-holding type automobile carrier
CN115130218A (en) Self-learning prediction method for road noise of automotive suspension structure
Shang et al. Design on a composite mobile system for exploration robot
Das Vehicle Dynamics Modelling: Lateral And Longitudinal
CN117930666B (en) Motor train unit control method, device and medium based on rapid power approach law
Seeger Neural Network-Based Wheel Suspension Modeling

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination