Neural Network Based Classification of Unbalances in Rotating Machinery

Abstract

Health monitoring for rotating machinery such as aircraft engines, motors provide economic benefits and operational efficiencies in terms of reduced downtime. In this paper we present a methodology of using artificial neural networks (ANN) and frequency-domain vibration analysis to detect and classify common types of unbalances in rotating machines. Frequency domain features are used to train an artificial neural network. The artificial neural network is trained using back-propagation algorithm with a subset of experimental data obtained from a real-world rotating machine, Machinery Fault Simulator (MFS), for known types of unbalances. The trained artificial neural network is then used to classify various types of unbalances such as static unbalance and couple unbalance. The effectiveness of the neural network to classify these different types of unbalances is tested using the remaining set of data. The advantage of this procedure is that it can be used not only to diagnose unbalance but also to identify the type of unbalance in rotating machines.