Self-directed project
Predictive Maintenance: Industrial ML for Fault Detection
Condition-Monitoring Engineer · AI / ML · August 2026
A ProoV case study · educational project, not employment
Built a working rolling-bearing fault detector end to end: engineered time-domain and envelope-spectrum features, trained a compact depth-2 decision tree on training bearings only, and evaluated it on a held-out set with an 83% catch-rate and 25% false-alarm rate. The memo defended a 0.50 alarm threshold using the candidate’s own numbers and correctly localized the fault to the outer race via dominant BPFO energy.
Graded against
- Feature engineering (signal to honest numbers)25%
- Held-out evaluation (honest measurement)25%
- Cost-based threshold decision + memo25%
- Compact model + embedded-budget justification15%
- Plain-language explanation10%
Passed · pass mark 60/100
What stood out6
- Used the provided envelope_spectrum helper and correctly found BPFO as the dominant defect band
- Trained a depth-capped decision tree on training bearings only and stated depth 2 with 3 nodes against the embedded budget
- Reported a held-out confusion summary with both catch-rate (0.83) and false-alarm rate (0.25)
- Wrote a maintenance memo that ties the threshold choice to concrete operational costs and identifies the outer-race fault
- Leakage-Free Feature Engineering
- Honest Held-Out Evaluation
The work I submitted14 tasks
7 tasks · 8 lines · Python · 7 written answers · 344 words
- Initial Prediction62 characters
- Time Features143 characters
- Freq FeaturesPython · 1.2k characters
- Train Classifier363 characters
- Evaluation Heldout288 characters
- Decision Memo1.6k characters
- Jupyterlite Code2 lines
- Initial Prediction9 words
- Time Features10 words
- Freq Features101 words
- Train Classifier31 words
- Evaluation Heldout24 words
- Decision Memo168 words
- Jupyterlite Code1 word
Summarised on purpose — the submitted code and writing stay private so this page cannot be reused as an answer key. The full submission sits behind the verified certificate.