Sensorless control of PMSM drive with state and load torque estimation

Purpose - The purpose of this paper is to obtain a fully sensorless permanent magnet synchronous motor (PMSM) drive control algorithm used for robot arm drive with load recognition. The paper shows how to use an extended Kalman filter (EKF) instead of sensors of the mechanical quantities as well as how to adapt the model of the PMSM to the filter procedures and aims at the real time application in the field-oriented control (FOC) structure of the high-dynamic drive. Design/methodology/approach - The synthesis of the control system is based on the method of the FOC, theory of the EKF and object description in the form of state equation with suitable choice of state vector. The adequate connection of these three methodologies is a core of the approach to design. First, the control algorithm was tested by means of simulation method then the real laboratory plant was built and investigated. Findings - Owing to task-oriented formulation of the PMSM model, adequate organization of the EKF procedures and suitable choice of covariance matrices the proper control algorithm was obtained. The algorithm can be applied on DSP and gives a good result for the high dynamic drive in spite of the fact that the Kalman procedures are recognized as a time-consumed solution of the estimation problem. Research limitations/implications - The proposed algorithm can be used also in the case of another type of drive, for instance induction motor drive, although the model of the motor has a different formula. A disadvantage of the method is lack of the general rules for choosing the elements of the covariance matrices. Practical implications - The presented algorithm is written in open-programming language. Obtained results may be important for synthesis of robot arm drive, where the information about load forces is needed. Originality/value - The paper presents an original sensorless vector control of PMSM with load recognition based on EKF. The originality elements are the choice of the covariance matrix elements and the real-time realisation of the algorithm on the DSP.