Bearing Fault Detection on IM Using MCSA and Sparse Representation

Nowadays, the monitoring of rotatory machinery is an essential task in the industry, due to its help in specific maintenance programs, increasing the life of the machine, reducing interventions, and getting off costs in spare parts. The bearings are essential pieces of rotatory machinery; thus, monitoring and fault detection are important tasks to maintain their proper functioning. In this paper, a methodology for bearing fault detection on induction motors based on Motor Current Signature Analysis, the Sparse Representation, and Dictionary Learning techniques are presented. The fault detection is possible due to the Sparse Representation is developed with dictionaries trained with specific faults each one, obtaining a minor error with the dictionary related to the same fault of the analyzed signal. In this way, the detection of related to localized defects and distributed defects are possible. Bearings of three-phase and 1-HP induction motors with two load conditions were used to testing. The proposed methodology presents an accuracy greater than 96%.