Automated condition classification of a reciprocating compressor using time–frequency analysis and an artificial neural network

Abstract

The purpose of this study is to develop an automated system for condition classification of a reciprocating compressor. Various time–frequency analysis techniques will be examined for decomposition of the vibration signals. Because a time–frequency distribution is a 3D data map, data reduction is indispensable for subsequent analysis. The extraction of the system characteristics using three indices, namely the time index, frequency index, and amplitude index, will be presented and examined for their applicability. The probability neural network is applied for automated condition classification using a combination of the three indices. The study reveals that a proper choice of the index combination and the time–frequency band can provide excellent classification accuracy for the machinery conditions examined in this work.