System fault detection through modeling and estimation techniques

Major failure detection methods are presented. Both unmeasurable parametric quantities and state variable approaches are considered. In the fault detection process, mathematical models (whether parametric or state-space, state variables models) play a prominent role. Parameter estimation techniques and state estimation techniques, recursive and nonrecursive methods are all used, and the appropriate approach in a given situation is directly related to the particular case, the measurements available, the constraints (physical and others), and many other factors that have to be analyzed appropriately. In many situations, accurate models are necessary, and only well-known processes are amenable to such fault detection schemes. On the other hand, if estimates of very high accuracy are a requirement, then only low-order linear time-invariant models can be utilized efficiently. There are tradeoffs that need to be evaluated concerning accuracy, simplicity, practicality, implementation, and safety