NEURAL NETWORK MODEL OF MAGNETORHEOLOGICAL DAMPER FOR VIBRATION ISOLATION PLATFORM OF WHOLE-SPACECRAFT

To improve the interface vibration environment of the satellite and the rocket, a six-pole vibration isolation platform is built, with the magnetorheological damper as the semi-active control device. The magnetorheological damper for the whole vibration isolation platform is tested to obtain the experimental data for the damping characteristics of the magnetorheological damper. The BP (back propagation) neural network with two hidden layers is established to model the damper for predicting the damping characteristics of the magnetorheological damper and for the design of the control system. A sequential algorithm is proposed to optimize the network structure, the weight and the threshold to ensure that the network has a better generalization ability and the stability of the network training. The simulation results show that compared with the parametric model, the proposed neural network model has less training error and higher generalization ability, and can well predict the damping characteristics of the damper.