An image-based pattern recognition approach to condition monitoring of reciprocating compressor valves

This work presents the development of a vibration-based condition monitoring method for early detection and classification of valve wear within industrial reciprocating compressors through the combined use of time-frequency analysis with image-based pattern recognition techniques. Two common valve related fault conditions are spring fatigue and valve seat wear and are seeded on the crank-side discharge valves of a Dresser-Rand ESH-1 industrial compressor. Operational data including vibration, cylinder pressure, and crank shaft position are collected and processed using a transformed time-frequency domain approach. The results are processed as images with features extracted using 1st and 2nd order image texture statistics and binary shape properties. Feature reduction is accomplished by principal component analysis and a Bayesian classification strategy is employed with accuracy rates greater than 90%.