Hydraulic mount high-frequency parameter identification method based on genetic algorithm

Abstract

The invention relates to a hydraulic mount high-frequency parameter identification method based on a genetic algorithm. The method comprises the following steps that actually-measured dynamic characteristic curve data of a hydraulic mount are obtained; establishing a hydraulic mount lumped parameter high-frequency model; obtaining an initial value of a genetic algorithm through a traditional method; and taking the minimum relative error of the dynamic characteristic curve of the hydraulic mount and the lumped parameter model simulation curve as a fitness function, and obtaining optimal parameters through a genetic algorithm. The method has the following positive effects: 1) the simulation result has high precision, and the high-frequency dynamic characteristics of the hydraulic mount can be well simulated; (2) when the initial value of the genetic algorithm is obtained, the convergence speed of the genetic algorithm can be improved by adopting a traditional method, and the optimization accuracy is improved; and (3) the established parameter identification calculation method for the lumped parameter high-frequency model of the hydraulic mount based on the genetic algorithm has relatively good precision and relatively good efficiency, and can well improve the modeling speed of high-frequency dynamic characteristic analysis of the hydraulic mount.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of liquid resistance suspension research power assembly high-frequency damping and vibration isolation and improving the comfort of the automobile, and particularly relates to a hydraulic suspension high-frequency parameter identification method based on a genetic algorithm. BACKGROUND

[0002] The hydraulic suspension has the functions of damping, supporting, adjusting and protecting, plays a crucial role in the stability, safety and comfort of the vehicle, is widely used on traditional fuel vehicles, and the research on the dynamic performance thereof is relatively mature in the industry.

[0003] With the continuous progress of technology and the improvement of market demand, the hydraulic suspension is also widely used on new energy hybrid vehicles. Due to the presence of the motor, the external excitation frequency of the hydraulic suspension is higher than that of the traditional fuel vehicle, and therefore the dynamic performance of the hydraulic suspension under high-frequency excitation is also brought about. In actual engineering, the dynamic stiffness and the hysteresis angle of the hydraulic suspension are commonly used as the evaluation indexes of the dynamic performance thereof.

[0004] In the design and research of the high-frequency of the hydraulic suspension, the parameter identification of the high-frequency of the hydraulic suspension is crucial. An accurate parameter identification result can well predict the high-frequency dynamic performance of the hydraulic suspension, so as to optimize the structure and performance of the hydraulic suspension.

[0005] In order to obtain the accurate parameter values of the lumped parameters of the hydraulic suspension, the most common traditional identification methods at present include the experience formula method, the eigenvalue point method, the fixed point method, the finite element method and the test method.

[0006] The existing technical problems are as follows:

[0007] The experience formula method is easy to calculate, but the precision is not enough; the eigenvalue point method and the fixed point method require high experience and are complex in actual operation; the test method has high precision, but has a long identification period and high cost; and the identification precision of the finite element method depends on the precision of the finite element modeling. SUMMARY

[0008] In view of the problems in the prior art, the purpose of the present application is to provide a hydraulic suspension high-frequency parameter identification method based on a genetic algorithm, which has a short identification period, high identification precision, low experience requirement and does not need to depend on the precision of finite element modeling.

[0009] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0010] A hydraulic suspension high-frequency parameter identification method based on a genetic algorithm comprises the following steps,

[0011] S1: a high-frequency small-amplitude harmonic excitation is applied to the hydraulic suspension, the original curve of the dynamic stiffness and the hysteresis angle is measured, and the measured dynamic characteristic curve is obtained after the original curve is processed;

[0012] S2: using the measured dynamic characteristic curve obtained in S1, a lumped parameter high frequency model of the hydraulic mount is established, the model only retains eight parameters of rubber main spring stiffness K r , rubber main spring damping C r , equivalent pump pressure area A p , equivalent decoupling disc area A d , upper liquid chamber volume flexibility C1, lower liquid chamber volume flexibility C2, decoupling membrane passage liquid inertia I d and decoupling membrane passage liquid resistance coefficient R d ;

[0013] S3: based on the model established in S2, initial values of eight parameters of rubber main spring stiffness K r , rubber main spring damping C r , equivalent pump pressure area A p , equivalent decoupling disc area A d , upper liquid chamber volume flexibility C1, lower liquid chamber volume flexibility C2, decoupling membrane passage liquid inertia I d and decoupling membrane passage liquid resistance coefficient R d are obtained by traditional methods respectively;

[0014] S4: taking the relative error between the measured dynamic characteristic curve obtained in S1 and the simulation curve of the model in S2 as the fitness function, the initial values obtained in S3 are optimized by using genetic algorithm to obtain optimal parameters varying with frequency, and the stiffness curve and the hysteresis angle curve are simulated.

[0015] Further, in S1, multiple hydraulic mount parts of the same batch are selected for testing, and the input simple harmonic excitation of each part during testing is controlled to be the same, and after the obtained multiple sets of data are processed by outlier elimination and sliding average filtering, the data meeting the requirements in consistency are taken as the final measured dynamic characteristic curve data.

[0016] Further, the sliding average filtering window is at least 5.

[0017] Further, the rubber main spring stiffness K r is obtained from the static characteristic test of the hydraulic mount. First, the load-unload force and displacement curve of the hydraulic mount is obtained, the force-displacement curve in the preloading ± 150N range is taken, and the slope thereof is calculated as the rubber main spring stiffness K r ,

[0018]

[0019] In the formula, F is the force generated in the preloading ± 150N range; x is the displacement of the rubber main spring generated by the force generated in the preloading ± 150N range; η is the dynamic-static ratio of the rubber main spring, η = 1.1-1.4;

[0020] Rubber main spring damping element C r Obtained from the measured main spring.

[0021] Further, the equivalent pump area A p is expressed as

[0022]

[0023] where D1 and D2 represent the upper and lower rubber main spring frustum diameters, respectively.

[0024] Further, the equivalent decoupling disc area A d is expressed as

[0025]

[0026] where d d is the equivalent decoupling disc diameter.

[0027] Further, the upper liquid chamber volume compliance C1 is estimated using the low frequency fixed point frequency f p

[0028] where A i is the inertance passage cross-sectional area, p is the liquid density, l i is the flow passage length, and f p is the fixed point frequency.

[0029] The lower liquid chamber volume compliance C2 = 100*C1.

[0030] Further, the decoupling membrane passage liquid inertance I d and the decoupling membrane passage liquid resistance coefficient R d are estimated using the following formulas, respectively

[0031]

[0032] where l d is the decoupling membrane passage length, A d is the equivalent decoupling disc area, p is the liquid viscosity, and d f is the decoupling membrane passage diameter.

[0033] Further, the fitness function is

[0034]

[0035] where q is the parameter of the lumped parameter dynamic characteristic model, and q = (K r , C r , A p , A d , C1, C2, I d , R​d ), K d and φ are the experimental dynamic stiffness and damping lag angle, respectively, K d and φ^ are the dynamic stiffness and damping lag angle fitted by the lumped parameter model, respectively; ω1 and ω2 are weight coefficients, representing the importance of dynamic stiffness and damping lag angle in parameter identification, respectively.

[0036] Further, the genetic algorithm parameter identification process includes the following steps:

[0037] Initialize the population: based on the physical meaning and value range of the to-be-identified parameters in the hydraulic suspension lumped parameter high-frequency model, an initial population containing multiple parameter individuals is randomly generated, each individual corresponding to a combination of to-be-identified parameters;

[0038] Fitness value evaluation: the relative error between the measured dynamic characteristic curve of the hydraulic suspension and the simulation curve of the lumped parameter model is taken as the fitness function, and the fitness value of each parameter individual in the initial population is calculated, and the smaller the fitness value, the better the simulation effect of the model corresponding to the parameter individual;

[0039] Selection operation: according to the obtained fitness value, the tournament selection method is used to select the parameter individuals with better fitness value from the current population as the parents for subsequent genetic operation;

[0040] Cross operation: for the selected parent parameter individuals, arithmetic crossover operation is performed according to the preset crossover probability to generate child parameter individuals with the characteristics of the parent parameters;

[0041] Mutation operation: for the obtained child parameter individuals, self-adaptive Gaussian mutation is performed according to the preset mutation probability to introduce parameter mutation diversity to avoid the algorithm falling into local optimum;

[0042] Repeat iteration until the preset maximum iteration number is reached, and output the parameter individual with the smallest fitness value in the iteration process as the optimal identification parameter.

[0043] Overall, the present application has the following advantages:

[0044] 1) According to the method for establishing the hydraulic suspension lumped parameter high-frequency model, the simulation result has high precision and can well simulate the high-frequency dynamic characteristics of the hydraulic suspension.

[0045] 2) When obtaining the initial value of the genetic algorithm, the traditional method is used, which can improve the convergence speed of the genetic algorithm and improve the accuracy of optimization.

[0046] 3) The established parameter identification method of the hydraulic suspension lumped parameter high-frequency model based on genetic algorithm has low experience requirement, does not need to depend on the precision of finite element modeling, has good precision, has good efficiency, can improve the modeling speed of the high-frequency dynamic characteristic analysis of the hydraulic suspension, reduces the test cost of parameter identification, and provides theoretical guidance and technical support for high-frequency dynamic characteristics. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is a lumped parameter high-frequency model of an inertia passage-decoupling membrane type hydraulic suspension of the present application;

[0048] Figure 2 is a suspension static characteristic loading-unloading force and displacement curve;

[0049] Figure 3 is an equivalent structure diagram of a suspension upper liquid chamber;

[0050] Figure 4 is a principle diagram of a suspension parameter identification method of the present application;

[0051] Figure 5 is a flow chart of a genetic algorithm parameter identification of the present application;

[0052] Fig. 6(a) is a comparison diagram of measured stiffness and genetic algorithm fitted stiffness;

[0053] Fig. 6(b) is a comparison diagram of measured hysteresis angle and genetic algorithm fitted hysteresis angle;

[0054] Fig. 6(c) is a diagram of optimal parameters identified by the genetic algorithm changing with frequency;

[0055] Fig. 7(a) is a comparison diagram of measured stiffness and genetic algorithm fitted stiffness;

[0056] Fig. 7(b) is a comparison diagram of measured hysteresis angle and genetic algorithm fitted hysteresis angle;

[0057] Fig. 7(c) is a diagram of optimal parameters identified by the genetic algorithm changing with frequency;

[0058] Figure 8 is a comparison diagram of stiffness simulation and measurement taking the average value of optimal parameters;

[0059] Figure 9 is a comparison diagram of hysteresis angle simulation and measurement taking the average value of optimal parameters. DETAILED DESCRIPTION

[0060] The present application establishes a high-frequency lumped parameter model of a hydraulic suspension, and proposes a method for identifying the high-frequency hydraulic suspension lumped parameters by using a genetic algorithm, thereby saving the time of traditional parameter identification.

[0061] The present application will be further described in detail below.

[0062] Taking an inertia channel-decoupling membrane type hydraulic suspension as an example, the high-frequency parameter identification method of the hydraulic suspension based on the genetic algorithm proposed by the present application comprises the following steps:

[0063] (1) Obtain the measured dynamic characteristic curve data of the hydraulic suspension: In actual engineering, the dynamic characteristic data of the hydraulic suspension is obtained by actual measurement on an MTS device. The hydraulic suspension is clamped on the MTS through a suitable tool, one end of which is connected to an actuator head, and the other end is fixed to a fixed table. The actuator head first applies a preload force equivalent to a power assembly, and then the actuator head applies a simple harmonic excitation x(t) = X0 sin(2πf0t) to the hydraulic suspension, and the force sensor is used to measure the reaction force F T (t) of the fixed end of the hydraulic suspension, wherein x(t) is the engine displacement excitation, and F T (t) is the force transmitted to the vehicle frame or body. Finally, the dynamic characteristic curve data of the dynamic stiffness K d and the lag angle φ of the test hydraulic suspension are obtained by the transfer function method.

[0064] The collected data is filtered by removing data with large errors and eliminating outliers, and then a sliding average filter is used for filtering processing, and in the embodiment, the filter window is set to 5, thereby increasing the data reliability.

[0065] (2) Establish a high-frequency model of the hydraulic suspension lumped parameters: The high-frequency model of the hydraulic suspension lumped parameters is shown in Figure 1 , which comprises a rubber main spring stiffness K r , a rubber main spring damping C r , an equivalent pump area A p , an equivalent decoupling disc area A d , a liquid volume compliance C1 of the upper liquid chamber, a liquid volume compliance C2 of the lower liquid chamber, an inertia channel liquid inertia I i , an inertia channel liquid resistance coefficient R i , a decoupling membrane channel liquid inertia I d , and a decoupling membrane channel liquid resistance coefficient R dComposed of parameters such as frequency and amplitude, the inertial channel parameter Q is not considered due to the high frequency and small amplitude. i Without considering the impact of the decoupling disk on F d Q i =0,F d =0.

[0066] (3) Obtaining initial values ​​for the genetic algorithm using traditional methods: According to the high-frequency stiffness formula of the lumped parameters of the hydraulic suspension, the parameters affecting stiffness and hysteresis angle include K. r C r A p A d C1, C2, I d and R d The initial value for each parameter is obtained using traditional methods.

[0067] Rubber spring stiffness K r The static characteristics of the liquid suspension are obtained by first obtaining the loading-unloading force and displacement curves of the liquid resistance suspension static characteristics test. The force-displacement curves within the range of ±150N of preload are taken, and their slopes are calculated as the stiffness of the suspension.

[0068] Right now

[0069]

[0070] In the formula, F is the force generated within the preload range of ±150N; x is the displacement of the rubber main spring generated by the force generated within the preload range of ±150N; η is the dynamic-to-static ratio of the rubber main spring, which is related to the structure and material of the main spring, and is usually η=1.1-1.4.

[0071] Rubber main spring damping C r Obtained from actual measurement of the main spring.

[0072] The equivalent area of ​​the rubber main spring is defined as the volume of liquid in the upper chamber that the main spring displaces per unit displacement. In this embodiment, the equivalent area of ​​the main spring is estimated based on a simplified schematic diagram of the upper chamber volume change.

[0073] Equivalent pump pressure area A p It can be represented as

[0074]

[0075] Where D1 and D2 represent the diameters of the upper and lower rubber main spring cones, respectively.

[0076] Equivalent decoupling disk area A d It can be represented as

[0077]

[0078] d d The equivalent decoupling disk diameter.

[0079] The upper liquid chamber volume compliance C1 refers to the ratio of the upper liquid chamber volume change amount of the hydraulic mount to the upper liquid chamber liquid pressure change amount. In the present application, a low frequency fixed point is used for estimation.

[0080]

[0081] In the formula, A i is the inertia passage cross-sectional area, ρ is the liquid density, l i is the flow passage length, f p is the fixed point frequency. The lower liquid chamber volume compliance is generally two orders of magnitude larger than the upper liquid chamber, and C2 = 100 * C1.

[0082] The liquid sensing and liquid resistance of the decoupling membrane passage are respectively estimated by the following formulas

[0083]

[0084] In the formula, l d is the decoupling membrane passage length, A d is the equivalent decoupling disc area, μ is the liquid viscosity, and d f is the decoupling membrane passage diameter.

[0085] (4) The relative error between the hydraulic mount dynamic characteristic curve and the lumped parameter model simulation curve is minimized as the fitness function, and the optimal parameters are obtained by genetic algorithm: the dynamic characteristics of the hydraulic mount include dynamic stiffness and hysteresis angle. First, the minimum data error of the dynamic stiffness curve is taken as the fitness function, and the optimal parameters are obtained by genetic algorithm, and the stiffness curve and hysteresis angle curve are simulated. Then, the minimum data error of the hysteresis angle curve is taken as the fitness function, and the optimal parameters are obtained by genetic algorithm, and the stiffness curve and hysteresis angle curve are simulated.

[0086] The essence of system parameter identification is to adjust the undetermined coefficients in the mathematical model to minimize the deviation between the theoretical calculation results and the experimental measurement values. For the hydraulic vibration isolation element, this process needs to determine the key physical quantities contained in its lumped parameter model, so as to ensure that the dynamic response characteristics obtained by simulation analysis and the bench test data have good consistency in the specified frequency range.

[0087] As an intelligent optimization method based on probability search, genetic algorithm has excellent global exploration ability due to its parallel processing mechanism. When solving complex optimization problems, the algorithm can effectively jump out of the local optimal trap and obtain an approximate global optimal solution. In particular, when dealing with the parameter identification of nonlinear systems, genetic algorithm shows unique advantages. For the hydro-elastic suspension system with significant nonlinear characteristics, the use of genetic algorithm for equivalent parameter identification can not only ensure the global convergence of parameter identification, but also significantly improve the calculation efficiency of the optimization process and avoid the problem of local extremum easily trapped by traditional optimization methods.

[0088] The basic principle of genetic algorithm applied to hydro-elastic suspension parameter identification is shown in Figure 4 The error value between the dynamic characteristic curve obtained by experiment and the lumped parameter dynamic characteristic curve is taken as the objective function of genetic algorithm, and the parameter identification problem of hydro-elastic suspension is converted into finding a set of optimal parameters in the feasible region to minimize the error between the experimental data and the lumped parameter dynamic characteristic data.

[0089] The fitness function selected by the present application is

[0090]

[0091] In the formula, θ is the parameter of the lumped parameter dynamic characteristic model, and the present application takes θ=(K r , C r , A p , A d , C1, C2, I d , R d ), K d and φ are the dynamic stiffness and damping lag angle of the experiment, K^ d and φ^ are the dynamic stiffness and damping lag angle fitted by the lumped parameter model; ω1 and ω2 are weight coefficients, representing the importance of dynamic stiffness and damping lag angle in parameter identification, and the present application first takes ω1=1 and ω2=0 to minimize the measured stiffness error as the fitness function, and then takes ω1=0 and ω2=1 to minimize the measured lag angle error as the fitness function; the optimal parameters are observed to change with frequency by genetic algorithm. Δθ is the relative error sum of squares between the experimental dynamic characteristic curve and the lumped parameter dynamic characteristic curve, the range of the first six parameters is taken between 20% of the upper and lower limits of the initial value, and the range of the last two parameters is taken between 50% of the upper and lower limits of the initial value, Δθ0 is the initial value of the parameters, and the result identified by the traditional method is taken as the initial value.

[0092] Figure 5 The genetic algorithm parameter identification process is shown in

[0093] Initialization is the process of generating an initial population containing a number of potential individuals. The present invention uses boundary constraints to generate the initial population randomly according to the initial parameter random generation method. Each individual is represented as a parameter vector by real coding. The set of parameter vectors is called a population, and the present invention takes the population as 40.

[0094] Fitness evaluation is the process of quantifying the quality of individuals. It provides the basis for selection operation and distinguishes individuals with high potential. Individuals with high fitness are more likely to be inherited to the next generation, and vice versa.

[0095] Selection, also known as replication, is the process of selecting high-quality individuals from the population to generate a new population. The present invention uses the tournament mechanism for selection, preserving individuals with high fitness and passing on good genes.

[0096] Crossover, also known as recombination, is the process of generating new individuals through gene recombination. The present invention uses arithmetic crossover for parameter mixing and controls the recombination frequency through probability.

[0097] Mutation is the process of maintaining population diversity through random disturbance. The present invention uses adaptive Gaussian mutation to dynamically adjust the mutation strength.

[0098] In the calculation of the present invention, the maximum iteration number is taken as 80, the crossover probability is 0.9, and the mutation probability is 0.1.

[0099] Taking the minimum stiffness error as the fitness function, the results are shown in Figures 6(a), 6(b), and 6(c).

[0100] From Figure 6(c), the curve of each parameter with frequency can be understood, and the present invention selects the average value of each parameter as the fitting parameter as shown in Table 1.

[0101] Table 1 Average value of each optimal parameter with minimum stiffness error

[0102] K r (N / m) C r (N.s / m)]]> A p (m 2 )]]> A d (m 2 )]]> 217981.4 24.8 0.002454 0.000709 [C1(m 5 / N)]]> [C2(m 5 / N)]]> I d (kg / m 4 )]]> [R d (N.s / m 5 )]]> 1.99e-11 2.03e-9 34320.3 14439382

[0103] Taking the minimum lag angle error as the fitness function, the results are shown in Figures 7(a), 7(b), and 7(c).

[0104] From Figure 7(c), the curve of each parameter with frequency can be understood, and the present invention selects the average value of each parameter as the fitting parameter as shown in Table 2.

[0105] Table 2 Average value of each optimal parameter with minimum lag angle error

[0106] K r (N / m) C r (N.s / m)]]> A p (m 2 )]]> A d (m 2 )]]> 223955.5 24.3 0.002435 0.000728 [C1(m 5 / N)]]> [C2(m 5 / N)]]> I d (kg / m 4 )]]> [R d (N.s / m 5 )]]> 2e-11 2.03e-9 35385.1 12300873

[0107] As Figure 8 and Figure 9The simulation and measured curve comparison chart of the optimal parameter average value is shown. It can be seen that the simulation results of the two methods using the average value can better simulate the dynamic characteristic trend, providing a method for further studying the parameter and frequency relationship.

[0108] As a kind of hydraulic suspension high-frequency parameter identification method based on genetic algorithm, aiming at the pain points of low precision, poor efficiency and high test cost of traditional hydraulic suspension high-frequency parameter identification, through the technical path of measured data optimization-precise model construction-initial value scientific acquisition-intelligent algorithm optimization, multi-dimensional technical breakthrough is realized:

[0109] First, the high-frequency dynamic characteristic simulation precision is significantly improved, meeting the engineering analysis requirements

[0110] Through the construction of a targeted hydraulic suspension lumped parameter high-frequency model, the high-frequency small amplitude excitation scenario is fully adapted. Considering that in this scenario, the inertia channel is long and narrow, the liquid cannot flow, and the decoupling disc cannot respond and there is no impact, the inertia channel parameters are excluded, and only 8 core parameters such as the main spring stiffness K of the rubber r , the liquid resistance R of the decoupling membrane channel d , etc. are retained, so that the model is both simplified in calculation and conforms to the real physical process.

[0111] At the same time, the relative error between the measured dynamic characteristic curve (dynamic stiffness, hysteresis angle) and the model simulation curve is minimized as the target, and the optimal parameters (or their average value) obtained through genetic algorithm iterative optimization are input into the model. The simulation curve has high fitting degree with the measured curve, and can accurately reproduce the dynamic response of the hydraulic suspension under high-frequency excitation, providing reliable model support for high-frequency dynamic characteristic analysis of hydraulic suspension, solving the problems of low precision of traditional empirical formula method and dependence on modeling accuracy of finite element method.

[0112] Second, the convergence speed and optimization accuracy of genetic algorithm are improved, and the efficiency and reliability are considered

[0113] Traditional intelligent algorithms (such as simple genetic algorithm) are prone to slow convergence and even fall into local optimum in parameter identification due to random initial values and wide search range. This application innovatively adopts a combination strategy of traditional method to obtain initial value and genetic algorithm optimization:

[0114] For the 8 core parameters in the model, the initial values are accurately obtained through engineering traditional method, and the initial search starting point of genetic algorithm is taken as the initial value, which greatly reduces the parameter search range and avoids algorithm iteration in invalid interval. At the same time, genetic algorithm is designed through the process of tournament selection, arithmetic crossover and adaptive Gaussian mutation, which not only ensures the population diversity to jump out of local optimum, but also speeds up the convergence through reasonable parameter setting. Finally, the global approximate optimal solution is obtained in a short calculation period, solving the efficiency pain points of long cycle of traditional test method and complex operation of characteristic value point method.

[0115] Three, reduce parameter identification test cost, promote the landing of hydraulic suspension high frequency design and development

[0116] Traditional parameter identification relies on a large number of physical tests (such as multiple bench tests to obtain dynamic characteristic data under different working conditions), which not only consumes a large number of same batch parts, but also needs to invest equipment and time cost. The present application reduces the cost through two paths:

[0117] 1. Data reliability optimization reduces repeated tests: In the actual measurement stage, the same batch of parts are controlled under the same simple harmonic excitation, combined with outlier rejection and 5 window moving average filtering, to ensure that the data of a single test is reliable, and there is no need to repeatedly test the data;

[0118] 2. Simulation replaces part of the physical test: Through an accurate parameter identification model, high frequency dynamic characteristic simulation analysis can be completed based on a small amount of measured data, without the need to verify the effect of different parameter combinations through a large number of physical tests, significantly reducing part consumption and equipment occupation time, and reducing research and development costs.

[0119] 3. In addition, the parameter output by the present application varies with the frequency rule (such as the value trend of K r , R d at different frequencies), which can also provide clear theoretical guidance for hydraulic suspension high frequency structure optimization (such as adjusting the main spring cone diameter and decoupling membrane channel length), avoiding design blindness and accelerating the research and development landing of hydraulic suspensions in new energy hybrid vehicle high frequency demand scenarios.

[0120] Four, adapt to high frequency excitation new scenarios, expand the application range of hydraulic suspension

[0121] With the popularity of new energy hybrid vehicles, the high frequency excitation brought by the motor (compared to traditional gasoline vehicles) puts higher requirements on the dynamic performance of the hydraulic suspension. The present application specially constructs a model and optimizes the identification method for high frequency small amplitude scenarios, solves the problem of model inapplicability and inaccurate parameters in high frequency scenarios of traditional identification technology, and enables the hydraulic suspension to adapt to the high frequency damping requirements of new energy vehicles, providing key technical support for the application expansion of hydraulic suspensions from traditional gasoline vehicles to new energy vehicles.

[0122] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited by the above embodiments, and any changes, modifications, substitutions, combinations and simplifications made without departing from the spirit and principles of the present application shall be equivalent replacement methods, and all shall be included in the protection scope of the present application.

Claims

1. A hydraulic mount high-frequency parameter identification method based on a genetic algorithm, characterized by: The method comprises the following steps of S1: applying a high-frequency small-amplitude harmonic excitation to the hydraulic mount, measuring an original curve of dynamic stiffness and hysteresis angle, and obtaining a measured dynamic characteristic curve after processing the original curve; S2: using the measured dynamic characteristic curve obtained by S1, a lumped parameter high frequency model of the hydraulic mount is established, the model only retains the rubber main spring stiffness K r , the rubber main spring damping C r , the equivalent pump pressure area A p , the equivalent decoupling disc area A d , the upper liquid chamber volume flexibility C1, the lower liquid chamber volume flexibility C2, the decoupling membrane passage liquid inertia I d and the decoupling membrane passage liquid resistance coefficient R d eight parameters; S3: Based on the model established in S2, the stiffness K of the rubber main spring is obtained by a traditional method r , the damping C of the rubber main spring r , the equivalent pump area A p , the equivalent decoupling disc area A d , the upper liquid chamber volume flexibility C1, the lower liquid chamber volume flexibility C2, the decoupling membrane passage liquid inertia I d and the decoupling membrane passage liquid resistance coefficient R d The initial values of the eight parameters; S4: taking the minimum relative error between the measured dynamic characteristic curve obtained in S1 and the model simulation curve in S2 as a fitness function, optimizing the initial values obtained in S3 by using a genetic algorithm, obtaining optimal parameters varying with frequency, and simulating to obtain the stiffness curve and the hysteresis angle curve.

2. The method of claim 1, wherein: In S1, multiple hydraulic mount parts of the same batch are selected for testing, and the harmonic excitation input during testing of each part is controlled to be the same. After the obtained multiple sets of data are subjected to outlier elimination and sliding average filtering, the data meeting the requirements in consistency are taken as the final measured dynamic characteristic curve data.

3. The method of claim 2, wherein: The sliding average filtering window is at least 5.

4. The method of claim 1, wherein: Rubber main spring stiffness K r From the liquid suspension static characteristic test, first, the load-unload force and displacement curve of the liquid resistance suspension static characteristic test is obtained, the force-displacement curve in the preloading ± 150 N range is taken, and the slope is calculated as the rubber main spring stiffness K r , In the formula, F is the force generated in the preloading ±150N range; x is the displacement of the rubber main spring generated by the force generated in the preloading ±150N range; η is the dynamic-static ratio of the rubber main spring, η=1.1-1.4; Rubber main spring damping element C r Obtained from the measured main spring.

5. The method of claim 1, wherein: Equivalent pump area A p is represented by In the formula, D1 and D2 respectively represent the upper rubber main spring taper diameter and the lower rubber main spring taper diameter.

6. The method of claim 1, wherein: Equivalent decoupling disc area A d is represented as where d d is the equivalent decoupling disc diameter.

7. The method of claim 1, wherein: The volume compliance C1 of the liquid uptake chamber is determined at the low frequency fixed point frequency f p Estimation, where A i is the cross-sectional area of the inertial passage, p is the liquid density, l i is the length of the flow passage, f p is the fixed point frequency; The lower liquid chamber volume flexibility C2=100*C1.

8. The method of claim 1, wherein: Decoupling membrane channel liquid inertia I d and decoupling membrane channel liquid resistance coefficient R d are estimated respectively by the following equations where l d Aeffis the effective decoupling disc area, μ is the liquid viscosity, d d is the decoupling membrane channel length, A f effis the equivalent decoupling disc area, μ is the liquid viscosity, d 9. The method of claim 1, wherein: The fitness function is where θ is the parameter of the lumped parameter dynamic characteristic model, and θ = (K r , C r , A p , A d , C1, C2, I d , R d ) is the parameter vector of the dynamic characteristic model, K d and φ are the dynamic stiffness and damping lag angle of the test, respectively, K^ d and φ^ are the dynamic stiffness and damping lag angle of the lumped parameter model fitting, respectively, and ω1 and ω2 are the weight coefficients, representing the importance of the dynamic stiffness and damping lag angle in parameter identification, respectively.

10. The method of claim 1, wherein: The genetic algorithm parameter identification process comprises the following steps: Initial population: based on the physical meaning and value range of the to-be-identified parameters in the hydraulic mount lumped parameter high-frequency model, an initial population containing multiple parameter individuals is randomly generated, and each individual corresponds to a combination of to-be-identified parameters; Fitness value evaluation: taking the relative error between the measured dynamic characteristic curve of the hydraulic mount and the simulation curve of the lumped parameter model as the fitness function, the fitness value of each parameter individual in the initial population is calculated, and the smaller the fitness value, the better the simulation effect of the model corresponding to the parameter individual; Selection operation: according to the obtained fitness value, the tournament selection method is used to select parameter individuals with better fitness values from the current population as parents for subsequent genetic operations; Cross operation: the selected parent parameter individuals are subjected to arithmetic crossover operation according to a preset crossover probability to generate child parameter individuals with the characteristics of the parent parameter individuals; Mutation operation: the obtained child parameter individuals are subjected to adaptive Gaussian mutation according to a preset mutation probability to introduce parameter mutation diversity to avoid the algorithm falling into local optimum; Repeat iteration until a preset maximum iteration number is reached, and output the parameter individual with the smallest fitness value in the iteration process as the optimal identification parameter.

Images