Multiple model filter based position tracking in CNC machines

Abstract

A multiple model based Unscented Kalman Filter (MMUKF) approach for position tracking of CNC servo drives is presented in this paper. The motion controller in the CNC machine has to generate motion profile to drive servo motor based on the feed back from the Encoder. The output of the feed back is affected by measurement variations due to friction and non linear behavior of tool motion. The unscented kalman filter(UKF)gives better results in non-linear motion applications by deterministic sampling of sigma points. The MM algorithm gives good estimate by combining the individual estimates of parallel filters matched to different motion models of the CNC Tool. In this paper, an Unscented Kalman Filter is used inside the multiple model algorithm to improve the estimation accuracy of tool motion at different dynamic regions of motion profile. The motion profile of tool is simulated by including fixed velocity and turn model and the MM based UKF gives better results compared to traditional kalman filters.