Model-based fault detection on a non-linear rotor system

This paper describes a model-based fault detection method for nonlinear rotor systems based on a linearised model. The rotational system on which the faults are tested is a non-linear simulation model. The model-based fault detection procedure is developed on the basis of a linearised rotor model, in which the non-linearity is considered as unknown input. In this paper, we consider a rotor system which is supported by means of active bearings with piezoelectric actuators. For simulation purposes, the rotor is modeled with the help of finite element methods based on Timoshenko beam theory. The non-linear behavior of the piezo-actuators are described using the Preisach model. In order to reduce the complexity of the fault detection procedure, this non-linear system has been linearised. The effect of non-linearity is approximated as an unknown input in the linearised model. The linear model with the additional unknown input is intended to deliver the same system output as the non-linear rotor system, and thus can be used for the model-based fault detection and isolation (FDI). The faults which are detected in this paper are point-mass unbalances. The faults are detected on the basis of parity equations and observers. An unknown input observer (UIO) equivalent to parity equations is constructed to separate the disturbances caused due to the unknown inputs on fault detection. The results show that the methods are very effective in detecting the faults in a non-linear system.