Structure-Based Connectionist Network for Fault Diagnosis of Helicopter Gearboxes

Abstract

A new method of diagnosis is introduced for helicopter gearboxes that relies on the knowledge of the gearbox “structure” and characteristics of the “features” of vibration for component fault isolation. Both the structural knowledge and featural knowledge in this method are defined as the fuzzy weights of a connectionist network that maps each sampled set of vibration features, obtained from a signal analyzer, into fault possibility values associated with individual gearbox components. The structural weights in this network are defined to represent the influence of gearbox component faults on the overall vibration sensed by each accelerometer, and the featural weights are defined to denote the influence of components faults on individual vibration features. Given the extremely complex structure of helicopter gearboxes which prohibits accurate modeling of the effect of faults on their vibration, the structural weights in this method are defined based on the root mean square value of the frequency response of a simplified lumped-mass model of the gearbox. The experimental evaluation of the method based on vibration data from two different gearboxes is included in a separate paper.