Torque and Angular Velocity Estimation of an Electric Machine Through Neural Networks Based Approach

In this work, two Feedforward Neural Networks are presented for the estimation, respectively, of the angular velocity and torque of a permanent magnet electric motor. The inputs for both networks were statistical indices directly calculated from accelerometer signals sampled for various operating conditions of the system. The study of relative errors and root mean square errors calculated for the test phase demonstrated the predictive reliability of the two networks. The purpose of this work is to provide, through a preliminary study, an alternative to the current techniques for estimating and measuring the angular velocity and torque delivered by an electric motor.