Input-Output Subspace-Based Fault Detection*

Subspace-based fault detection method using input-output information is developed in this paper. In some practical applications, the environment noise is the only input that excites the system. Although the statistical properties of the noise might be estimated, the value of the noise is not usually available at each time instance. The traditional subspace fault detection is already developed for such situations. In several other applications, measured inputs are applied to the system or even the stochastic noise might be measurable. While it is still possible to use the traditional output-only detection method, it is reasonable to expect that the application of extra input information together with the output data improves the detection. Several computation issues are discussed in this paper to include input data in the detection method, correctly. Simulation results show the efficiency of using the input information to improve the quality of fault detection.