Artificial Intelligent Based Friction Modelling and Compensation in Motion Control System

The interest in the study of friction in control engineering has been driven by the need for precise motion control in most of industrial applications such as machine tools, robot systems, semiconductor manufacturing systems and Mechatronics systems. Friction has been experimentally shown to be a major factor in performance degradation in various control tasks. Among the prominent effects of friction in motion control are: steady state error to a reference command, slow response, periodic process of sticking and sliding (stickslip) motion, as well as periodic oscillations about a reference point known as hunting when an integral control is employed in the control scheme. Table 1 shows the effects and type of friction as highlighted by Armstrong et. al. (1994). It is observed that, each of task is dominated by at least one friction effect ranging from stiction, or/and kinetic to negative friction (Stribeck). Hence, the need for accurate compensation of friction has become important in high precision motion control. Several techniques to alleviate the effects of friction have been reported in the literature (Dupont and Armstrong, 1993; Wahyudi, 2003; Tjahjowidodo, 2004; Canudas, et.al., 1986). One of the successful methods is the well-known model-based friction compensation (Armstrong et al., 1994; Canudas de Wit et al., 1995 and Wen-Fang, 2007). In this method, the effect of the friction is cancelled by applying additional control signal which generates a torque/force. The generated torque/force has the same value (or approximately the same) with the friction torque/force but in opposite direction. This method requires a precise modeling of the characteristics of the friction to provide a good performance. Hence, in the context of model-based friction compensation, identification of the friction is one of the important issues to achieve high performance motion control. However, as discussed in the literatures, several types of friction models have been identified (Armstrong et al., 1994; Canudas et. al., 1995; Makkar et. al., 2005) and classified as static or dynamic friction models. Among the static models are Coulomb friction model, Tustin model, Leuven model, Karnop model, Lorentzian model. Meanwhile Dahl model, Lugre model, Seven parameters model, and the most recent Generalized Maxwell-Slip (GMS) model, are among the dynamic friction models (Tjahjowidodo, 2004). The static friction model is simple and easy in the identification process, however using such model