Research Of Random Dynamic Load Identification Based On Road Noise Of Whole Vehicle

At present,in the performance development process of vehicle noise,vibration and harshness(NVH)at home and abroad,NVH performance diagnostic development of the later prototype vehicle is still an important part,and the development cycle is long.In the integrated analysis and development of vehicle NVH,the performance of vehicle body is analyzed separately,and the influence of random excitation of actual chassis or road surface is seldom considered.The performance objectives of vehicle body and subsystem are determined based on the performance and empirical values of the previous generation.Therefore,it is of great practical significance to obtain road load quickly and accurately through random dynamic load identification,provide reliable source input for vehicle road noise analysis,reduce vehicle development cost and shorten development cycle,and realize accurate prediction and evaluation of vehicle NVH performance development in the early stage.However,due to the statistical characteristics of random dynamic loads,traditional load identification methods cannot be directly applied to the identification of random excitations,and it is necessary to take into account the uncertainty of the structural model and the illposedness in the inverse process.In order to solve these problems,this paper investigates the random load identification for uncertain structures and its application in vehicle road random excitation identification as well as road noise analysis and optimization.The main research work and innovations are as follows:(1)A novel improved regularization method is proposed for the random dynamic loads identification of deterministic structures,which solves the ill-posed problems in the inverse process and greatly reduces the identification error.Firstly,a forward problem model for identifying partially coherent stationary random excitations is established by using the inverse pseudo excitation method.The complex problem of identifying random loads is transformed into the problems of identifying multi-order pseudo excitations.By deducing the formula of relative error of random load identification,the influence of condition number of frequency response function(FRF)matrix and measurement noise on the relative error of random load identification is studied.According to the characteristics of pseudo excitation and the effect of regularization parameter on residual norm and solution norm,a novel modified regularization parameter is formulated to solve the ill-conditioned problem in the inverse process of random loads identification.Compared with the Moore-Penrose pseudo method and traditional Tikhonov regularization method,the proposed improved regularization method improves the accuracy and stability of random load identification,effectively reduces the huge identification error in the low frequency range,prevent the overly regularization in the medium-high frequency range,and can always match the actual excitations from low to high frequency domains.(2)A composite random dynamic loads identification method for uncertain structures is developed based on matrix perturbation method,and the problem of identifying random loads for uncertain structures containing random parameter with small coefficient of variation is studied.The random dynamic load is expressed as a function of frequency and random parameters in the frequency domain,and the forward problem model for composite random excitation identification is constructed.Based on Taylor series first-order expansion and matrix perturbation,the problem of random excitation identification for uncertain structures is transformed into two kinds of deterministic inverse problems,namely the pseudo excitation identification on the mean value of structural random parameters and the sensitivity identification of pseudo excitation with respect to each random parameter,and then the sensitivity of random excitation to each random parameter is deduced by taking partial derivative.Combining with the proposed novel improved regularization method to avoid the uncertainty in the inverse process,the statistical characteristics of identifying random excitation are obtained.This method can effectively and reliably identify the boundary of composite random loads when the variation of random parameters is small.(3)A random load identification method for uncertain structural-acoustic coupling system based on sound pressure is proposed.The problem of random system parameters with large coefficient of variation and multimodal distribution are solved based on evidence theory.Considering the weak structural-acoustic coupling,the forward problem model between random excitation and acoustic pressure response is established firstly based on the theory of structural-acoustic coupling system.According to the inverse pseudo excitation method,the non-contact measured acoustic pressure response is used to replace the structural vibration response to identify the random dynamic load.Aiming at the random parameters with large coefficient of variation and multimodal distribution in the coupled system,a method of composite random load identification based on evidence theory and interval analysis is studied.This method is independent of the variation level and distribution types of random parameters,and can effectively evaluate the identification results of random excitation.In addition,the non-contact measurement based on sound pressure is more accurate and convenient in comparison with the contact measurement.(4)Based on the proposed random excitation identification method,a platform for the spindle load identification and vehicle road noise forward development is established,which combines simulation with test.This platform breaks through the predicament of unpredictable road noise due to the load deficiency in the early stage of vehicle development.Through developing the software of random load identification,the spindle loads of a SUV are identified.Meanwhile,the TPA(Transfer path analysis)method based on pseudo excitation is proposed to overcome the problem of phase information missing in random excitation auto-spectrum.By imposing the identified random excitation on the wheel center of the high-precision vehicle simulation model,the right ear noise of the driver under the working condition of 6 gear 60 km/h is analyzed.Through the structural optimization of the liftgate door,the improvement and upgrading of the low-frequency road noise problem in the vehicle are realized,and is verified by the test.The forward development platform solves the difficulties of random excitation acquisition in the early stage of vehicle development and large identification error in low frequency range,realizes road noise prediction and the integrated design of diagnosis as well as optimization,greatly improves the prediction accuracy and shortens the overall design cycle.