Robust Fault Diagnosis in Dynamic Systems

Many schemes for fault detection in dynamic systems have been introduced in recent years which use analytical or functional redundancy [1,2,3,4]. Fault tolerance especially those in which faults could result in a catastrophic event is usually achieved through hardware redundancy. To enable normal operation to continue in the presence of a fault, critical hardware, such as sensors and computers, are repeated in triplex or quadruplex redundancy schemes. Repeated elements are usually distributed spatially around the system to provide protection against localized damage. Such schemes operate by comparing outputs from repeated hardware in majority voting so that software/hardware faults may be detected and isolated. The major problems encountered with hardware redundancy, and those problems which have led to investigations into other forms of fault detection, are the extra weight and the space required to accommodate this, which could be used for example for more mission-orientated equipment.