Design and Analysis of a Simple Adaptive Observer for the Torque-Control of a PMSM for Traction Application 2021-26-0160

In a traction drive, feedback of the torque developed on the motor shaft is necessary for proper implementation of torque-control. However, measuring torque using a torque-sensor for feedback is not practically feasible due to high levels of vibration in vehicles and high sensor cost. Therefore, an accurate run-time estimation of the instantaneous value of the developed electromagnetic torque is essential for robust control. Online estimation of torque is a challenging task due to non-linear variation in motor parameters such as self and mutual inductances. Traction motors are designed to achieve high levels of power and torque density. This causes wide variations in magnetic saturation levels in the rotor and the stator core. This results in significant variation in inductance values with change in loading condition and during field-weakening operation of these motors. This paper presents the design and analysis of a torque observer that works on the principles of a Model Reference Adaptive System (MRAS). A dq-frame model of permanent magnet synchronous motor (PMSM) is used as the reference model to estimate d and q-axis currents. Thereafter, those estimates are compared with measured values of i_d and i_q to generate error which then is used by a suitably designed adaption mechanism to alter the parameters of the reference model to better represent the actual plant. The reference model thus keeps adapting itself to the parameter variations in the motor and hence, can estimate the developed torque accurately. A simple adaption mechanism is proposed in this paper that does not require high computational resource for implementation. The performance of this approach has been verified by simulation studies performed on a torque controlled PMSM drive. To analyse the effects of inductance variations and other uncertainties a high-fidelity mathematical model of a PMSM drive is created using data generated from electromagnetic finite element analysis (FEA). This model accurately captures the variations in self and mutual inductances of all phases with changing load, field-weakening, and rotor position. Results of an extensive simulation study indicate that the proposed adaptive mechanism is quite effective in estimation and control of developed torque.