Residual Generation for Fault Diagnosis of Systems Described by General Linear Differential-algebraic Equations

An algorithm for designing linear residual generators is presented. A main result is that the algorithm is able to design residual generators for any model described by general linear differential-algebraic equations. Previous algorithms have been restricted to models on transfer function, state space, or descriptor form. The presented algorithm can handle all these types of models, but also more general model descriptions not handled by previous algorithms. This is important since more general linear differential-algebraic equations models are often the result of object-oriented equation-based modeling. Also included is an extension to the stochastic case. Since the algorithm is based on well studied algebraic manipulations of polynomial matrices, it will have good numerical performance. The algorithm is demonstrated on a linearized model of a three-link robot.