Neural network based modeling of audible noise for high frequency injection based position estimation for PM synchronous motors at low and zero speed

In this paper, the relationship between injected voltages, audible noise and position estimation error is investigated for low speed high frequency injection based position sensorless control of PM synchronous motors. The modeling of noise is done using feed-forward neural network. The model is capable of predicting the audible noise. The proposed model can be used to perform optimization studies for sensitive applications where proper trade off studies between noise and speed/position estimation error is required.