Fault identification by means of model-based techniques, both in frequency and time domain, is often employed in diagnostics of rotating machines, when the main task is to locate and to evaluate the severity of the malfunction. The model of the fully assembled machine is composed by the submodels of the rotor, of the bearings and of the foundation, while the effect of the faults is modelled by means of equivalent force systems. Some identification techniques, such as the least squares identification in frequency domain, proposed by the authors, have proven to be quite robust even if the submodels are not fine-tuned. Anyhow, the use of a reliable model can increase the accuracy of the identification. Normally a supporting structure is repre...