Unbalance Identification in a Rotor Supported by Active Magnetic Bearing

This paper presents a formulation for unbalance fault detection in flexible rotors supported by active magnetic bearings, AMB. The model-based procedure makes use of the correlation equations, through the matrix formulation of Ljapunov for stationary linear systems along with artificial neural networks. This procedure only uses measured state variables. Through the correlation of the output variables, a group of relations involving the physical parameters of the system together with the matrices of correlations of the measured variables is generated. Unbalance changes are detected through the monitoring of variation of physical parameters related to unbalance and comparison of theoretical and estimated correlation functions. Artificial neural networks are used to map correlations involving states that are not measured. The proposed method is applied in a flexible rotor model composed of four rigid disks, a pair of active four pole magnetic bearings with feedback control. The unbalance change is applied in individual planes and in several planes simultaneously.