Ashutosh Undure completed Predictive Maintenance: Industrial ML for Fault Detection

Built a compact bearing-fault detector end to end: plotted a faulty vibration snapshot, identified periodic impact behavior, selected kurtosis as the strongest time-domain indicator, and trained a shallow decision tree on training bearings only. Evaluated on held-out healthy and faulty bearings, then defended a 0.60 alarm threshold with catch-rate, false-alarm rate, and cost-based maintenance reasoning tied to an outer-race diagnosis.

• Leakage-Free Model Training
• Held-Out Evaluation
• Cost-Based Operational Decision Making