Fault tree based approach for system fault diagnostics

With the ever increasing complexity and functionality of systems the task of identifying and correcting faults is itself a complex issue. Although often designed for reliability, at times components will fail, leading to a decline in system performance or at worse complete system failure. When faults do occur it is imperative they can be diagnosed and ultimately rectified as quickly as possible, minimising the effects of such a failure. In the case of an aircraft system efficient diagnosis can optimise the time to return the aircraft to service. For an unmanned air vehicle diagnosis of the status of the system can mean that missions can be altered or aborted given the faults detected. Many system failures are not usually the result of one single fault, therefore the ability to diagnose multiple faults is vitally important. A method of finding faults or combinations of faults as they occur is the subject of this paper. The approach uses sensor readings to assess the state of the system. Fault trees, which traditionally provide a diagrammatic description of the causes of system failure, are used to develop causes of a system symptom, as exhibited by the sensor readings. The method diagnoses the faults by considering deviations in the sensor readings from the expected system state. The primary research has shown the applicability of using a fault tree based approach for system diagnosis. Both coherent fault trees (considering failure events only) and non-coherent fault trees(considering failure and functioning events) have been used to evaluate deviations for an example system. Where combinations of faults result from the diagnosis importance measures can be used to rank the contribution of the failure events and hence highlight the most likely cause of the fault.