Research on modeling of magneto-rheological damping dynamic inverse model based on support vector machine

Abstract

In order to improve the accuracy of magnetorheological dynamics modeling and increase its generalization ability, a model of magnetorheological damper inverse model based on support vector machine learning algorithm is proposed for different ambient temperature and vibration frequency. By setting up the test platform of the magnetic rheovariable damper, combined with the mechanical characteristic stoic data in the measured temperature-rise state, the nonlinear damping hysteresis characteristics of the magnetic rhemomorphic damper are analyzed. The BP neural network, RBF neural network and support vector machine technique are used to construct the inverse model of magnetorheological damper, and the error accuracy and fitting effect of the reverse model are compared. The results show when the condition of $0\sim 60^{\circ}C$, sinusoidal excitation frequency 0.5Hz, 1Hz, 2Hz, amplitude 10mm, current 0 $\sim$ 0.8A, the inverse root mean square accuracy of magneto-rheological damper under temperature-rise effect established by support vector machine compared with the inverse model of magneto-rheological damper established by BP neural network and RBF neural network, the error is reduced by 83.66% and 61.125%, respectively. The fitting effect is good and meets the needs of practical engineering applications.