Sliding mode sensorless position observation

Abstract

Rotor position is required for commutation of Brushless DC Motors (BLDCMs) and it is often desirable to eliminate the traditional position sensor. Despite a decade of considerable research effort there is still no consensus on the best approaches to adopt for sensorless control of BLDCMs with most proposals suffering from limitations in operating conditions, parameter sensitivity or quality. The principles of sliding mode have also been the subject of much research but few have applied the technique to the problem of observer design[1] and just such a system holds much promise for sensorless control due to its inherent robustness. This paper seeks to evaluate sliding mode observers particularly with respect to parameter insensitivity, stability and accuracy with a view to further enhancing the system by coupling an artificial neural network. The methodology adopted was to develop an implementation of an existing scheme in order to characterise sliding mode observers and subsequently to develop an improved, integrated, observation system. The results show good accuracy and exceptional stability in the face of parameter uncertainty.