Introducing Model-Based Residual Spectrum Analysis for a Practical Improvement in Induction Motors Fault Diagnosis

Motor current signature analysis (MCSA) for induction motors (IM)s fault diagnosis has been applied in industries for years. However, its practical implementation is challenging because most of the fault-characteristic frequencies appear in a very low fault signal-to-noise ratio (SNR) and the magnitudes of fault-characteristic frequencies are masked by the current harmonics and noises. In this study, a model-based residual spectrum analysis (MRSA) is proposed to improve the practical implementation of IM fault diagnosis. The MRSA uses the IM state-space model to filter out the domination of current harmonics and noises which yields residual signals with higher fault-characteristic frequency SNR. A straightforward statistical fault threshold is implemented to evaluate the fault-characteristic frequency energy that minimizes the need for an expert in fault diagnosis evaluation. The validation experiments are performed in an industrial wastewater centrifugal pump driven by a 17 kW direct online (DOL)-driven IM and in a 0.75 kW laboratory test-bed variable-frequency drive (VFD)-driven IM. The proposed MRSA performance is compared to MCSA and extended Park's vector modulus (EPVM). The experimental results show that the practical implementation of MRSA demonstrates better sensitivity and performance compared to MCSA and EPVM.