Induction Motor Bearing Fault Diagnosis Using Envelope Analysis and 1-D CNN

Rolling-element bearings are the most frequent cause of induction motor failures. A damaged bearing produces a train of small impacts at a characteristic defect frequency (BPFI for the inner race, BPFO for the outer race, 2 x BSF for a ball). This project diagnoses a healthy bearing and inner-race, ball and outer-race faults of three sizes (10 classes) from the drive-end vibration of a 2 hp induction motor (CWRU Bearing Data Center, 12 kHz, motor loads 0-3 hp). It compares the classical envelope-spectrum analysis at the bearing defect frequencies (rule-based baseline) with an SVM and a Random Forest on 25 time / frequency features and a small 1-D CNN on 2048-sample windows - and shows honestly that a random window split is leaky (99.9-100 % for every model) while changing the motor load is the real test. A GNU Octave cross-check (run), a complete MATLAB version and an OpenModelica motor model (written, not run) and a web application with five databases are included. Everything runs on a student laptop CPU in about 5-6 minutes plus a 140 MB download.

Category: Electrical and Electronics Engineering (Electrical Machines, Condition Monitoring, Vibration Analysis, Signal Processing, Machine Learning, Deep Learning, GNU Octave, MATLAB, OpenModelica, Python)

Electrical & Electronics Engineering Project EE06 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.