What Arham demonstrated

Built a compact rolling-bearing fault detector end to end: engineered time-domain features, used envelope-spectrum energy at supplied defect frequencies to localize the fault, trained a shallow decision tree on training bearings only, and evaluated it on a held-out set with catch-rate and false-alarm metrics. The candidate also defended a 0.40 alarm threshold with explicit cost reasoning and translated the result into a clear maintenance recommendation.

• Leakage-Free Model Training
• Honest Held-Out Evaluation
• Physics-Informed Fault Interpretation
• Cost-Based Operational Decision-Making