Data‐Driven Methods for Fault Diagnosis

This chapter addresses the problem of the identification of both linear and nonlinear dynamic systems for fault diagnosis. In the case of nonlinear dynamic systems, the identification will be performed by exploiting parametric nonlinear models, such as affine, piecewise-affine and fuzzy models. Using the concepts of model-based fault detection, the design of a residual generator based on a fuzzy model of a nonlinear dynamic process is addressed. The chapter also addresses the decomposition of a nonlinear identification problem into a set of locally linear models by means of product space fuzzy clustering. The identification algorithm exploited to estimate the parameters and orders of the local affine submodels is based on the well-established Frisch Scheme method for linear systems. A set of optimal parameters with respect to the model output can also be estimated from the identification data set by using the ordinary least-squares methods.