Self-sensing method for a switched reluctance motor using delta-sigma modulators and neural networks

Abstract

A self-sensing method for switched reluctance motors (SRM) for low speed and stand-still is introduced in this paper. The approach utilizes current slopes excited by voltage pulses that are injected into idle phases. To obtain an undisturbed signal for rotor position estimation, a simple integration of the bit-stream output of the delta-sigma modulators in the current measurement path is applied. Thereafter a neural network approximates the actual rotor position from the integrated currents of the idle phases. In comparison with look-up-table based approaches multiple input signals for the approximation are used. The results are evaluated on a test-bench with an 8/6 SRM in open loop operation.