Adaptive Inverse Control On Random Vibration Of A Vibration-Centrifuge System

Experiment on vibration-centrifuge compound environment, which is proposed to satisfy the requirement of structural or model's reliability, is a important measure to find out the failure mechanism of a structure and the validity of it's model. The environment can be constructed by fixing a vibrator on a centrifuge, but noise caused by high linear acceleration will impact the normal work of the vibrator. So how to control the vibrator under centrifuge force has been concerned by many researchers.This study tried to realize an active control of a compounded system's random vibration using the adaptive inverse method. The statistic characters, analytic method and traditional experiment of a random vibration were introduced firstly. Then the main idea of the adaptive inverse control was discussed and several key parts of the whole process were studied in detail, such as employing the shaking signal and BP neural network to establish the vibrator's model and it's inverse model respectively, and applying feedforward system to cancel the unexpected noise. A new way to form the sample data for network training was suggested additionally.At last, based on the discrete state equation, two kinds of adaptive inverse control system were set up to control the acceleration of the countertop under the random excitation. One system utilized LMS filter as inverse controller while the other system had an inverse controller achieved by BP neural network. Seen from the simulation results, both of them perform well, except for the condition they adapt to is different.