Compensation and Estimation of Friction by Using Extended Kalman Filter

A friction exists in every mechanical system. The phenomenon of friction is complicated and difficult to measure the exact value. In this paper, the compensation and estimation of friction by using extended Kalman filter (EKF) have been introduced and investigated. The study plant is a second order system with low-pass PD controller. The actual friction of the plant is assumed to be the modified Tustin friction model. The simulations are conducted for three cases of the input signal: sinusoidal, triangular, square, and two cases of compensation: no friction compensation and with friction compensation. The performances of the system are indicated by root-mean-square error (RMS error) and peak error (peak error). The studies show that the system with the friction compensation by using EKF has the better performances than the system with no friction compensation. The RMS errors of the system are decrease by 90.4 %, 89.2 %, and 12.5 % for sinusoidal, triangular, and square input signals, respectively. Also, the peak errors are decrease by 65.9 % and 58.7 % for sinusoidal and triangular input signals, respectively. For square input signal, there is no significant decrease of the peak error due to the high overshoot of the system with friction compensation