What Mugunthan demonstrated

Built a working rolling-bearing fault detector end to end: loaded and visualized vibration data, engineered time-domain and envelope-spectrum features, localized the fault to the outer race via BPFO, and trained a compact decision tree within an embedded budget. Evaluated on a held-out bearing set with reported catch-rate and false-alarm rate, then defended a cost-based alarm threshold in a maintenance memo using the candidate’s own numbers.

• Leakage-Free Feature Engineering
• Honest Held-Out Evaluation
• Physics-Anchored Fault Localization
• Cost-Based Decision Making