Failure Probability Evaluation of Turbogenerator Coil Retaining Rings based on LCF Experimental Data and Local States of Load

Turbogenerator coil retaining rings (CRRs) are shrunk fitted onto the rotor over the coils, in order to restrain them against the centrifugal force. They are typically subjected to low cycle fatigue (LCF): a cycle is completed at every machine switch-on and switch-off, for a total amount of 10,000-15,000 cycles in the whole life. Incidents caused by unexpected failures may lead to very serious consequences, consisting in fire and explosion. For this reason, the conventional deterministic calculations must be combined to probabilistic simulations for the estimation of the probability of failure at the main life stages. The widely applied Monte Carlo method has the drawback of being computationally expensive in the estimation of very low probabilities: alternative numerical methods, such as the AMV, are available instead. However, there are few papers dealing with the analytical issues of these approaches, moreover no applications regard the turbogenerator field. The object of this paper is to show a suitable methodology, based on the AMV, to determine the failure probability of a CRR: its estimation is based on the experimental evaluation of material static, cyclic and fatigue properties and on the knowledge of the entity of cyclic loads. The determined value, about 10-10 at the last stage of the machine life, is compatible with reference values for structures under fatigue in the aeronautical field.