Neural Net Approach
Fault injection can be exploited and used to train a neural-net to detect hazardous states and were never contemplated by system designers
Two approaches
- Radial Basis Function Net (1% misclassification using 1000 training sets from fault injection)
- Back-Propagation Net