Aymen Rebhi completed Predictive Maintenance: Industrial ML for Fault Detection

Built a bearing-fault detection workflow that distinguished healthy from faulty vibration, identified kurtosis as the strongest time-domain fault indicator, and evaluated the detector on a held-out set with reported catch and false-alarm rates. The candidate also wrote a maintenance memo that justified a 0.48 alarm threshold using explicit cost trade-offs and correctly localized the fault to the outer race from BPFO evidence.

• Leakage-Aware Fault Interpretation
• Honest Held-Out Evaluation
• Cost-Based Operational Decision-Making