Fault Detection Using Reduced Order Models

Based on the reduction of dynamic system models, an algebraic redundancy method is proposed which consists of fault detection through the redundant variables and fault diagnosis using a diagnostic matrix. This diagnostic matrix approach, which is an extension of the influence matrix method, is capable of identifying different combinations of failure modes. It is shown that this method can produce a correct diagnosis for the failure of sensors and actuators and can also detect changes in the system parameters. A simulation of a three-shaft gas turbine is used to demonstrate the potential of the method.