Sensorless neural network speed control of permanent magnet synchronous machines with nonlinear stribeck friction

In this paper, a sensorless neural network speed control strategy of permanent magnet synchronous machines (PMSMs) is introduced as an alternative to conventional control techniques. The control strategy achieves accurate tracking by making use of artificial neural network (ANN) learning capabilities to approximate the machine's nonlinear dynamics. The ANN controller's output is then fed to a Space Vector Pulse Width Modulation (SVPWM) to produce duty cycles for the inverter. On the other hand, a second ANN is used as an observer to estimate rotor speed and the rotor position is obtained by direct integration to reduce the effect of the system's noise. Unlike other sensorless control strategies, no a priori offline training, weights initialization, voltage transducer or mechanical parameters knowledge is required. Simulation results for different situations highlight the performance of the proposed control approach in transient, steady-state, and standstill conditions