Failure diagnosis for a large-scale, nonlinear and time-varying plant based on a new identification method of small variations of parameters

A failure diagnosis system is proposed which can deal with large-scale, nonlinear, and time-varying plants. It is based on a novel identification method of small variations of parameters, which is named the SANQ (simulator-based approach to neat and quick system identification) method. The method does not require enriching input signals for the purpose of identification; with this system, one can estimate both the location and the injury of the failure neatly and quickly. In order to deal with large-scale plants, the plants are divided into small blocks. This reduces the number of parameters to be estimated at one time. The SANQ method provides estimates of the state variables of plants as well. Consequently, a model reference adaptive control system combined with the SANQ method will be a very useful scheme in recovering a faulted plant. The overall system consisting of the functions of diagnosis of a failure, estimation of dynamic property after failure, and appropriate control which leads to the ideal state will raise the reliability of the plant operation