Detection of high-resistance motor connections using symmetrical component analysis and neural network models

High resistance electrical connections in the power wiring of induction motor circuits can result in excessive heating, with unsafe operating temperatures and consequent equipment damage. A predictive maintenance diagnostic system that could analyze electrical waveforms to detect the presence of such faulty connections in a motor circuit would be potentially valuable innovation. This paper describes preliminary analytical and experimental work in the detection of high-resistance connections. A neural network model is trained to characterize the behavior of "healthy" circuits in terms of the symmetrical components of three-phase voltage and current. Subsequent system operation, with additional resistance in the cabling due to faulty electrical connection, will exhibit a variation between the measured behavior and the expected "healthy" behavior. Experimental results show that this deviation is roughly proportional to the added resistance, and can be useful as a figure of merit for diagnosing the state of motor wiring.