Sensitivity analysis of multiple fault identification in rotor systems

The robustness, as regards measuring and modelling errors, of a model based identification method, in the frequency domain, for multiple faults in rotor systems is presented in this paper. First of all, to evaluate the robustness, several data sets representing faults on real machines, such as unbalances and bows, are generated by means of simulation. Then the identification method is applied by using these data sets and “corrupted” system models. The latter are obtained by adding noise to the stiffness and damping coefficients of the bearings. The sensitivity analysis, performed on several cases, has shown that the proposed method is robust as regards the modelling errors.