Computationally efficient Anisotropy-Identification based on a Square-Shaped Injection Pattern

Abstract:
This paper proposes a discrete-time anisotropy-based sensorless control approach that compared to literature causes only a fraction of the computational load. While all known methods consist of two computation stages with vector transformations and/or matrix inversions, the proposed method identifies the anisotropy in a single stage with only 6 subtractions. The basis for this simple evaluation is a square-shaped injection pattern and a current control cycle of two PWM periods. The paper provides the mathematical derivation, a comparison to other methods and the experimental validation of the proposed approach.