Details

Autor(en) / Beteiligte

• McDonald, Geoffrey Lyall

Titel

Vibration Signal-Based Fault Detection for Rotating Machines

Ort / Verlag

ProQuest Dissertations & Theses

Erscheinungsjahr

2011

Quelle

ProQuest Dissertations & Theses A&I

Beschreibungen/Notizen

• Fault detection in rotating machines from vibration data is a difficult task and is important for maintenance planning and preventing equipment damage or failure. The aim of this thesis is to improve upon existing vibration signal methods for detecting rotating machine faults in gears, bearings, and rotors. Faults manifesting in impulse-like vibration signals are focused on, which includes faults such as rotor-to-stator rubbing, bearing inner/outer race failures, and gear tooth faults. Towards this goal, two novel techniques for detecting these faults are proposed in this thesis and experimental data from gear tooth crack, gear tooth chip, and suspected turbine rotor-to-stator rubbing is analysed. An adaptive sum-of-sinusoids model is presented and compared to the widely accepted autoregressive model approach. The results indicate that the proposed method performs better on experimental gear tooth crack data, requires no data fitting, and is of similar computational cost. This method is particularly suitable for equipment with changing rotational speed. A deconvolution-based approach is presented as a periodic extension upon the minimum entropy deconvolution method. The proposed method aims to deconvolve periodic impulses, which is the vibration signal nature of many rotating machine faults, as opposed to the single impulse deconvolved by minimum entropy deconvolution. Performance of the deconvolution technique are shown to be strong on simulated and experimental gear tooth chip data, and an online implementation is proposed for the automated monitoring of equipment. Both proposed methods are applied to turbine proximity sensor data, and a sensor fault plus two suspected rotor-to-stator rubbing faults are identified.