A fault detection, isolation, and identification technique for complex MISO linear systems

This paper presents a method, based on a generalized likelihood ratio test (GLRT), for combining fault detection, isolation, and identification in complex multi input-single output (MISO) linear systems. By "complex"" it is meant that the system is comprised of a number of subsystems, connected through series, parallel, feedback, etc. relations. The method applies to systems corrupted by multiple noise terms with additive faults entering the system at various locations, and requires measurement of the process output and inputs. In addition to detecting faults, the method determines which subsystem experienced the fault, determines the type of fault (if more than one type is hypothesized), estimates the time of occurrence of the fault, and estimates the magnitude of the fault (shown to be both unbiased and efficient). A simulation example illustrating the performance of the GLRT method is also provided.