Model-based Intermittent Fault Detection

Development of reliable fault diagnostics for intermittent failure modes are an important tool to adequately deal with realistic failure behavior within complex systems. A large proportion of previous work utilizing model-based fault diagnostics has focused on persistent faults and often neglects the case where the system intermittently switches between a faulty and non-faulty behavior at discrete random intervals. Such intermittent behavior complicates the diagnostics task, with difficulties in detecting and isolating intermittent faults, which occur with low frequency but yet at high enough frequency to be unacceptable. Accurate assessment of intermittent failure probabilities is critical to diagnosing and repairing equipment and requires the development of models to describe the dynamics of the intermittent failure. This paper presents an overall framework for detecting sensor faults, through the use of nonlinear unknown input observers which are applicable to both persistent and intermittent faults. The work presented demonstrates the detection capabilities of the approach through the use of robust residuals insensitive to system uncertainties and the application of adaptive thresholds.