Abstract

This thesis studies the effects of temperature in friction modeling for a harmonic drive. A mathematic model of how friction varies with temperature change is built based on the exponential friction model. The mathematic model is then used in building the nonlinear dynamic model of the harmonic drive. A Cascaded Fuzzy Model of friction with temperature is proposed so that the friction experienced in the harmonic drive can be modeled as a fuzzy combination of linear models at different operating temperature and velocity. The fuzzy model is then integrated to the dynamic model of the harmonic drive to build the fuzzy TS model which is used in the controller design. The parameters in both mathematic model and the fuzzy model are estimated. The estimation is formed into an optimization problem of minimizing the quadratic cost function of estimation errors. Since the cost function is highly nonlinear, a new algorithm of Evolutionary Parallel Gradient Search is applied to help the search escaping from local minima. The optimal controller for fuzzy TS system is proposed in the form of LMI based optimization with new constraints and applied to the harmonic drive. In simulation, the closed loop system shows the capability of accurately tracking large range of reference signals even when the temperature is changing, either continuously or drastically