Vibration model and response-based method to detect and localise potential faults in bolted rotor systems

Abstract

Bolted rotor systems in industrial rotating machines like aero engines and gas turbines are easily subjected to bolt-loosening faults due to variable operating conditions, which may cause connection failures as well as large vibration magnitudes. To diagnose loosening faults in bolted rotor systems as early as possible, a reliable method is proposed in this paper through detailed theoretical and numerical studies. Firstly, a typical simply supported rotor system with two offside bolted disks is considered. A finite element model is used to characterize the system, in which bolt-loosening faults are modelled as nonlinear stiffness elements. Taking into account the limited measurable responses in practice, a reduced order frequency response function (FRF) matrix is obtained. Then, fault-induced nonlinear restoring forces in two orthogonal directions are used to define related diagnostic index accordingly. The new index and developed method are then verified through a series of cases on a numerical rotor system with bolt-loosening faults at one/two bolted disks. Diagnostic results illustrate that proposed index and related method can not only detect the existence of single/multiple bolt-loosening faults but also give their correct positions in the bolted rotor system. Meanwhile, results of different measured points show that output responses have no effect on the calculation of diagnostic indexes, and results of different rotating speed show that the rotating speed should be selected around the critical speed to obtain larger diagnostic indexes.