Fault diagnosis in induction motor installation using Discrete Wavelet Energy and low-cost sensors

In industries, three-phase induction motors (TIMs) are crucial elements in production lines. Consequently, faults in these machines are closely linked to huge losses in productivity. Therefore, predictive fault detection methods are valuable tools in this field. Within this framework, the correct installation of the motor is the first step to avoiding flaws. The key procedures are leveling, alignment, and tightening. However, over time, the TIM's fixing bolts can become loose. This phenomenon leads to other types of mechanical failure, damaging the machine. Therefore, this work studied the application of piezoelectric sensors and the Discrete Wavelet Energy Technique (DWET) to identify loose bolts in the base of three-phase induction motors. The four mounting bolts were tested during the experiments, and after the signal processing, they could be individually diagnosed as tight or loose. The fault classification was achieved by using 3D classification maps. The clusters related to each bolt condition were well defined and spatially far from each other. Also, different wavelet levels were tested, and their efficiency was compared through silhouette and precision statistical indexes. Piezoelectric sensors were applied as transducers to acquire the vibration of the motor due to their low cost and availability. Several experiments were carried out with different conditions to ensure the efficiency of the proposed system. Finally, the results showed that the new low-cost system successfully diagnosed and classified loose bolts in TIMs.