1. Introduction
It is crucial that modelling and identification of a system is tailored to meet the requirements of the intended application. For the purpose of analyzing the system behaviour under various types of inputs, and the noise and disturbances affecting the output, and for designing a controller, a state space, or input output model might be employed. The model may be identified using the conventional identification techniques including the generalized least-squares method, instrumental variable method and the subspace identification method. However, when the intended application is fault diagnosis, or controller tuning, where a diagnostic parameters vary, the conventional model and the identification scheme must be generalized to capture not only the input output behaviour but also the faults.