Detection of rotating machine faults using vibro-acoustic techniques

Effective and accurate classification of rotating machine faults plays a significant role in the minimization or elimination of costly plant downtime and frequencies. The vibration-based faults detection methods that dominate the industry at present still rely significantly on conventional techniques that are solely driven by subjective judgments of condition monitoring experts. In an era where-by the availability of such experts is becoming more and more difficult, exploring approaches that possess the capability to facilitate automatic classification of faults under changing operational conditions is crucial. As a part of ongoing studies, the current paper explores the possibilities of integrating features from vibration and vibro-acoustic measurements from an existing laboratory scale rig so as to develop an artificial neural network (ANN) based classification approach that can detect the migration of routine condition monitoring data on a typical industrial rotating machine and hence trigger the initiation of a maintenance intervention. Since the cornerstone of this paper is based on the simplification of faults diagnosis, the features used for the proposed method are well-established time and frequency domain features so as to enhance the accurate detection of common ma-chine faults.