Unbalance localization through machine nonlinearities using an artificial neural network approach

Abstract

Excessive levels of unbalance in rotating machinery continue to contribute to machine downtime and unscheduled and costly maintenance actions. While unbalance as a rotor dynamic fault has been studied in great detail during the last century, the localization of unbalance within a complex rotating machine is today often performed in practice using little more than ‘rules of thumb’. In this work, unbalance faults have been localized through a data driven approach applied to a rotor dynamic test rig fitted with multiple discs. Sub-synchronous nonlinear features in the frequency domain have been identified and studied as a method of aiding the localization of unbalance faults, particularly in situations where sensor placement options are limited. The process of automating the localization has been achieved using an artificial neural network (ANN), and the addition of rub and misalignment faults in the study have been used in order to validate the performance of the system. The results of the study are discussed from the perspective of next-generation integrated vehicle health management (IVHM) systems for rotating machines.