Induction machines performance evaluator 'torque speed estimation and rotor fault diagnostic'

This paper proposes a new DSP based tool for evaluating the performance of induction motors based on the data extracted from the stator current. In the proposed algorithm, a pattern recognition technique according to Bayes minimum error classifier is developed to detect incipient rotor faults such as broken rotor bars and static eccentricity in induction motors. Also, part of the algorithm is based on the acceleration method presented in the IEEE Std. 112. It helps to calculate the motor's torque using two line currents and voltages. The use of linear and quadratic time-frequency representations is investigated as a viable solution to the task at hand. Speed information is vital in this approach, so an algorithm to track the speed related saliency induced harmonics from the machine's line current spectrogram is presented. Capturing the harmonics gives the rotor speed that can also be used to extract the feature vector for diagnostic. The implementation of the algorithm on TMS320C6000 family of DSP chips is currently underway. The complete algorithm is then be used to obtain the induction motor's performance curves. This is a complete stand-alone panel mounted induction motor diagnostic tool currently being developed in their lab. This package will be used in conjunction with a drive system (inverter) for online performance monitoring and preventing unwanted shutdown of the induction motor. The difficulties encountered, including a limited dynamic range and the presence of cross terms, are addressed and the suggested solution is provided. Experimental results corroborating the proposed algorithm are presented, and a discussion of the advantages and disadvantages of such an approach is touched upon.