Uncertainty estimation and quantification applied to fault models of rotating machines

Abstract

The usage of rotating machines in the industry is vastly. The knowledge of its demeanor is important and the faults that this equipment can suffer as well. The parameters of these faults should be determined, so a better model of the rotor system can be done and the proper maintenance can be made. The parameter has inherent uncertainties and the model of some considered is unknown. Therefore, the principle of maximum entropy and the bayesian inference can be used, combined or separately, to determine these distributions, so they can be propagated to the rotor system dynamic response. Moreover, the accuracy analysis is also considered through an acceptable variation of the system response, and the bayesian inference application is used as an inverse problem method. The stochastic rotor system dynamic response can be estimated by stochastic Galerkin method or stochastic collocation method; both determine projections of the differential equation system in a polynomial space. The uncertainties are expanded in polynomial chaos and, then, the methods mentioned before are applied to solve the stochastic differential equations. (AU)