Prasad Shinde completed Predictive Maintenance: Industrial ML for Fault Detection

Built a compact rolling-bearing fault detector end to end: visualized vibration data, engineered RMS/kurtosis/crest-factor style features, used envelope-spectrum defect bands, trained a shallow decision tree on training bearings only, and evaluated it on held-out healthy and faulty bearings. The final memo justified a 0.60 alarm threshold with the team’s own catch-rate, false-alarm rate, and cost figures, and identified the inner race as the likely fault location.

• Leakage-Free Feature Engineering
• Honest Held-Out Evaluation
• Cost-Based Decision Making
• Interpreting Fault Signatures