What Taimoor demonstrated

Built a working rolling-bearing fault detector end to end: engineered RMS, kurtosis, crest factor, and envelope-spectrum defect-band features; localized the fault to the outer race via BPFO; trained a compact depth-capped decision tree on training bearings only; and evaluated it on held-out bearings with reported catch-rate and false-alarm rate. The final memo justified the alarm threshold using the student’s own cost numbers and translated the diagnosis into maintenance language.

• Leakage-Free Feature Engineering
• Honest Held-Out Evaluation
• Physics-Based Fault Localization
• Cost-Based Thresholding