Genetic algorithms based parameter identification for nonlinear mechanical servo systems

Parameter identification for mechanical servo systems with nonlinear friction term is very difficult, and linear identification techniques are not adoptable because that the parameters can not be linear parameterized as well as the local minimum problem. Based on genetic algorithms, this paper presented a two-step offline method for the parameter identification of mechanical servo embedded with LuGre friction model. In the first step, four static parameters were estimated through the Stribeck curve, and in the second step, two dynamic parameters were obtained by the typical limit cycle output of the system. Genetic algorithms with different control parameters and objective functions were used in both steps to minimize the identification errors. At last, the simulation are developed for a typical nonlinear mechanical servo systems, and the results have shown that the convergence of identified friction parameters are robust and not affected by the coupling property between the dynamic parameters and static parameters.