六维腕力传感器的动态非线性-控制理论与控制工程专业论文.docxVIP

RESEARCH RESEARCH ON DYNAMIC NONLlNEARITY OF THE SIX一^XlS WRIST FORCE SENSOR ABSTRACT According to the dynami c caliI)ration data of the step response experiments。the dynamic models of the six—axis wrist force sensor are set up for i ts 6 paths respectivelY by tile system identification method．The performance indexes in the frequency domain are determined．The dynamic characteristics of the sensor are described comprehensivelv． The nonlinear dynamic models in the time domain are set up using the modilied bi+”diagonal least——square method and the functional 1 inking neural network method．When compared with the 1inear dynamic models．the nonlinear dynamic models describe the dynamic characteristics of the sensor more aecuratelY． The problem of the nonl inear dynamic compensation is presented．The functional 1inking neural network method is used to design the compensating devices．The impulse and step responses of the sensor are compensated respectively，and good results are obtained and more desirable than those of 1inear compensation．But this method has its limitations． The sensor with the nonlinear dynamics may be divided into two parts： one is a nonl inear static part and 1he other is a 1inear dynamic part．A new kind of nonlinear dynamic compensating method of sensor is presented． The nonlinear static part and the 1inear dynamic part are obtained through the static calibration of the sensor。The output signals of the sensor are handled using the nonlinear static correcting method，then are processed by the linear dynamic compensating method．In order to test the validity of this method，we have designed a DSP based simulator of the sensor and a nonlinear dynamic compensating system．The experimental results show that this method is valid and universal． The constant dynamic decoupling method is modified and the static coupl ing error is removed． Key words six-axis wrist force sensor bi—diagonal least—square functional linking neural network dynamic nonl inear compensat i