Accuracy of fault detection in real rotating machinery using model based diagnostic techniques

An experimental validation of a model based identification technique is presented in this paper. The validation is performed on both real machines, like large steam and gas turbogenerators, and test rigs. The aim is to show that the method is able to locate the actual fault, to evaluate its severity and to discriminate among faults that have similar symptoms. A quantitative index, called residual is introduced to evaluate the accuracy of the performed identification. The actual developing faults taken into account are some of the most common on rotating machines such as unbalances, thermal bows, fatigue cracks and radial and angular misalignments of couplings.