Outer race bearing fault detection in induction machines using stator current signals

This paper discusses the effect of the operating load as well as the suitability of combined startup and steady-state analysis for the detection of bearing faults in induction machines, Motor Current Signature Analysis and Linear Discriminant Analysis are used to detect and estimate the severity of an outer race bearing fault. The algorithm is based on using the machine stator current signals, instead of the conventional vibration signals, which has the advantages of simplicity and low cost of the necessary equipment. The machine stator current signals are analyzed during steady state and start up using Fast Fourier Transform and Short Time Fourier Transform. For steady state operation, two main changes in the spectrum compared to the healthy case: firstly, new harmonics related to bearing faults are generated, and secondly, the amplitude of the grid harmonics changes with the degree of the fault. For start up signals, the energy of the current signal frequency within a specific frequency band related to the bearing fault increases with the fault severity. Linear Discriminant Analysis classification is used to detect a bearing fault and estimate its severity for different loads using the amplitude of the grid harmonics as features for the classifier. Experimental data were collected from a 1.1 kW, 400V, 50 Hz induction machine in healthy condition, and two severities of outer race bearing fault at three different load levels: no load, 50% load, and 100% load.