Case Study: Low-Frequency Ultrasound ML
Real-time classification for grinding wheel state and process quality.
Project Snapshot
- Role: ML Solution Lead
- Domain: Industrial process monitoring
- Stack: Python, signal processing, supervised ML, real-time decision logic
- Timeline: Research-to-production deployment
>60% Defect-Crisis Reduction
Analytics-led quality programs using this ML-first operating style reduced major defect crises by more than 60% (public-shareable resume metric).
5% → 1% Nuisance Defects
Targeted defect classes dropped from 5% to 1% when detection, intervention timing, and process control were tightened.
>10% Yield Improvement
Model-informed root-cause loops helped drive more than 10% yield improvement in a one-year process optimization horizon.
Quantified Outcomes (Public-Shareable)
- >60% defect-crisis reduction in analytics-led quality programs aligned to this ML operating style.
- 5% to 1% nuisance defect-rate reduction where detection and intervention timing were improved.
- >10% yield improvement delivered in a one-year process optimization context with model-informed control loops.
Problem
Grinding performance drifted over time, and manual detection lagged actual process behavior. Teams needed a reliable way to interpret non-obvious signals before defects or downtime escalated.
Approach
I combined low-frequency ultrasound signal processing with supervised machine learning to classify wheel condition in near real time. The solution was tuned for practical deployment, not just lab accuracy.
Outcome
The resulting model improved process visibility and operational control, helping teams intervene earlier and maintain consistent quality under production constraints.