Machine friction estimation for modeling, diagnostics, and control

This paper describes experiments whose objective is to estimate dynamic friction without relying on measured friction force or a structured friction model. The extended Kalman-Bucy filter based estimator relies on an accurate model of system dynamics and measured motion to extract friction force. The paper motivates and describes the estimation method and reports results of estimating friction force during unidirectional and bi-directional motion of a sliding contact. Filter-based friction estimation addresses characterization of nonstationary friction due to uncontrollable factors such as wear, normal force variation, and lubricant condition, factors that are difficult to capture with a constant parameter friction model. The method aims to estimate friction force for three purposes: friction modeling, friction-based diagnostics, and friction compensation