Diagnosis of Intermittent Faults with Conditional Preferences

Diagnosis of intermittent faults is significantly different from diagnosis of permanent faults, especially when selecting one or several preferred diagnoses. In this paper, we describe a new modeling approach for intermittent faults based on Past Time Linear Temporal Logic (PTLTL), and we suggest a conditional diagnosis selection approach based on Conditional Preference theories. We describe how intermittent faults can create diagnosis instability and show how our approach can be used to mitigate it. We then describe an incremental way to compute diagnoses at each time step, and a reduction of the incremental diagnosis computation to a MaxSAT query. Finally we discuss some limitations of our diagnoser and how they can be overcome.