Adaptive friction compensation of servo mechanisms

Abstract

Friction exists in all machines having relative motion, and plays an important role in many servo mechanisms and simple pneumatic or hydraulic systems. In order to achieve high precision motion control, accurate friction modeling and effective compensation techniques have to be investigated. In this chapter, we shall present a systematic treatment of adaptive friction compensation techniques from both engineering and theoretical aspects. Firstly, a comprehensive list of the commonly used classical friction models and dynamic friction models is presented for comparison and controller design. Then, by considering the position and velocity tracking control of a servo mechanism with friction, adaptive friction compensation schemes are given based on the LIP static friction model and dynamic LuGre model respectively. Finally, extensive simulation comparison studies are presented to verify the effectiveness of the proposed methods.

Intelligent control of mechatronic systems

“…Because of the complexity and difficulty in modelling friction, neural networks may be used to generate input-output maps using the propeny that a multi-layer neural network can approximate any function, under mild assumptions, with any desired accuracy. A general friction model f(z, i) was proposed [29] There are also other friction models in the literalre although they cannot be conveniently or approximately expressed in the LIP form but may be able to describe the friction phenomenon more accurately such as the Lucre model Based on the LuGre model, several nice model-based controllers have been developed. Under the assumption of known system parameters and functions, a model-based controller was presented in [31] with the unmeasurable friction state being estimated by an observer which is driven by the tracking error.…”

“…Remark I: Based on the above discussion, a more complete classical model consisting of the following components: Stiction, Coulomb, viscous, drag friction and square root friction was proposed [29] fm(Z,z) = f t~6 ( i ) + feSPn(z) + f v ? + f d z b l +fv(lil)"2 S g n ( 4 (25) where S ( z , 5 ) is a vector of known basis functions, and P is of unknown parameters.…”

“…In any given circumstance, some features may dominate over others and some features may not be detectable with the available sensing technology, but all these phenomena are present all the time [29].…”

“…The discontinuities modeled by stiction and Coulomb friction are actually observations at the macro level. Accordingly, friction could be approximated by neural networks [29]. 7) Neural Network Friction Model; Neural networks offer a possible tool for the nonlinear mapping approximation.…”

Adaptive control of hydrodynamic loads in splash zone

“…Neural networks has also been embedded in the overall control strategy for modelling and compensation purposes in [7], [8], [9] and [10]. Some in-depth developments in neural networks for modelling and control purposes have been made in [11], [12] and [13]. The resulting scheme is non-model based and does not require the exact hydrodynamic load model which is difficult to obtain in practice.…”