Discriminate intermittent from permanent faults based on fault event evaluation and diagnoser

The inefficiency to discriminate intermittent faults (IFs) from permanent faults (PFs) of systems, possibly resulting in early removal of components, may result in a scenario of increased can not duplicate, no trouble found, and retest OK. To address this problem, a novel fault model which includes both IFs and PFs is constructed. Thereafter, a fault discrimination approach based on diagnoser is derived. Since fault events are usually unobservable, it is difficult to discriminate IF from PF events captured in succeeding sensors, by treating environmental stresses (ESs) as fault events, the association model between ESs and fault is built. And then an algorithm is given to identify the fault events by evaluating the level of the correlative ESs. Finally, an example of aeronautic gyroscope is presented to demonstrate the proposed approach, and the analysis results show the approach is effective and feasible.