Research On Parameter Identification Method Of Hybrid Modeling Of Accelerometer In Frequency Domain

Accelerometer is a kind of sensor which is used to measure shock and vibration,its dynamic characteristics directly affect the accuracy of dynamic measurement,and the accelerometer dynamic model parameter identification is an important content of establishing dynamic model.In the linear range,the accelerometer is equivalent to a linear single mass oscillator with constant stiffness and damping,and its dynamic model is a two order linear system.The model parameters include damping ratio,natural frequency and acceleration to output parameters.When the nonlinear influence of noise and nonlinear can be ignored,the accelerometer dynamic model parameter identification methods are based on the two order linear model by many parameters identification methods,such as genetic algorithm,neural network algorithm,least squares algorithm,etc..The parameter identification method has certain accuracy.However,in the high precision application of accelerometer,noise and nonlinear of accelerometer is inevitable,when the noise and nonlinear effects cannot be ignored,the two order linear model based accelerometer dynamic model parameter identification method will produce large estimation errors.Therefore,it is of great theoretical significance and practical value to study how to reduce the influence of noise and nonlinearity on the identification of dynamic model parameters and improve the accuracy of dynamic model parameters identification.Based on two order linear dynamic model of accelerometer,a hybrid dynamic model of accelerometer with linear and nonlinear characteristics is constructed by introducing the nonlinear term.In view of the influence of spectrum leakage and measurement noise on the frequency response function of the accelerometer,the paper studied the method of frequency response function estimation,then a spectrum estimation method based on all phase fast Fourier transform(FFT)is proposed,which can effectively solve the influence of spectrum leakage and measurement noise on the estimation of the frequency characteristics of the accelerometer.Considering the bandwidth of the shock excitation signal,the method of the accelerometer hybrid dynamic model parameter identification based on weighted least squares and support vector machine is proposed.The accelerometer hybrid dynamic model parameters of the linear part is identified by WLS,the nonlinear characteristics is estimated by SVM,then the parameter identification of the hybrid dynamic model is realized.Based on virtual instrument technology,the dynamic model parameter identification system of accelerometer is developed,and the on-line identification of model parameters is realized.Experimental results show that the method based on all phase FFT spectrum can reduce spectrum leakage and measurement noise influence on frequency response function.Hybrid dynamic model parameter identification method based on weighted least squares and support vector machine can effectively reduce the influence of nonlinear characteristics and effectively improve the accelerometer hybrid dynamic model parameter identification accuracy.The accelerometer dynamic model parameter identification system based on the virtual instrument technology can realize the online parameters identification.