Robust sensorless tuning of a PMSM drive with 2D experimental designs

This paper presents an original application of the experimental designs method to a PMSM drive control without mechanical sensors. The position, velocity and back emf observer used is a redundant observer structure. This observer includes different gains and filters that have to be tuned optimally for the best convergence performance. The proposed method is based on the statistic analysis of parameter influence on a defined performance criterion. The most influent parameters and their optimal levels are identified. A new two dimensional design is also carried out for a robust tuning of the observer coefficients, in order to reject the influence of the parameter variations. This analysis is called cross product plan and highly reduces the number of required experiments.