Improved observer for sensor fault diagnosis of a power plant

The paper presents some results on parametric identification of linear systems applied to robust Fault Diagnosis schemes. In our approach, an equation error model is derived from input--output data. In particular, the error term takes into account disturbances (non measurable inputs), nonlinear and time--variant terms, measurement errors, etc. In this manner, state--space realization of the equation error model leads to define a disturbance distribution matrix related to the error term, and, thus, well--known eigenstructure assignment results for robust fault detection can be successfully applied. The proposed procedure has been tested on a industrial gas turbine prototype model in which a sensor fault is simulated. Results from this simulation campaign are also reported.