Identification And Control Of Nonlinear Systems With Hysteresis Characteristics Based On Support Vector Machine

Support vector machine (SVM) is a brand-new machine learning technique based on statistical learning theory. It is an ideal facility for modeling of various nonlinear systems. It is an implementation of structure risk minimization principle in the statistical learning theory. Its basic idea is mapping the input data into a high dimensional feature space in which an optimal separating hyper plane or function regression is done. It has been proved that SVM presents a lot of advantages in dealing with the small samples, nonlinear and high dimensional pattern recognition, as well as other machine-learning problems such as function regression.Hysteresis phenomena are common in actuators and sensors, such as gears and saturation, which would undermine the stability of system and the accuracy of control badly. A support vector machine based modeling approach and a support vector machine based controlling approach are presented to analyze and control nonlinear systems with hysteresis.Several hysteresis modeling methods are presented, and a hysteresis model based on Preisach model is developed. In this paper, a support vector machine based approach for modeling of systems with hysteresis is proposed, and an improved version is developed. The developed identification approaches are numerically implemented in Matlab simulation program, and the improved version is proved to be effective and more accurate than BP neural networks when being used for modeling of systems with hysteresis.Then, an inverse modeling controller is presented for nonlinear systems preceded by unknown hysteresis nonlinearity. The inverse modeling of hysteresis systems is accomplished by a new development of support vector machine. A control impression comparison between classical PID controller and support vector machine inverse modeling controller is provided, and the effectiveness of the new approach is illustrated through Matlab simulation.