Broken Signal Detection
Industrial AI / Binary Classification
A classifier system for detecting broken or abnormal measurement signals in automotive testing.
Overview
Automotive testing produces measurement channels that may be broken or abnormal and need to be identified before downstream use.
The problem
The system must preserve domain-relevant signal behavior while optimizing detection of unacceptable channels.
My contribution
Compared classical and deep-learning architectures, applied CFC filtering, and tuned decision thresholds for NOK recall.
Technical highlights
- 01CFC low-pass filtering per SAE J211 standards
- 02More than 10 model architectures compared
- 03Threshold tuning for NOK recall optimization