DETECTION OF VALVE LEAKAGE IN RECIPROCATING COMPRESSOR USING ARTIFICIAL NEURAL NETWORK (ANN)

摘要

In the present work, Artificial Neural Networks (ANN) techniques are being applied for detection of valve leakage in reciprocating compressor. It has been experienced that replacement of defective valves before they cause further damage can greatly reduce maintenance and production costs. In the past, valve problems were unnoticed until process flow instruments showed a reduction in flow or when compressors became noisy or overheated. These symptoms usually did not occur until the very last stages of valve degradation. By this time the compressor frequently was damaged because of valve parts were being ingested into the cylinder and thus causing piston ring or liner damage. The application of Artificial Neural Networks technique identifies the most practical yet sensitive form of signature to use for trend monitoring and will additionally help the system to accept the fact that a compressor is malfunctioning without the aid of additional instruments capable of establishing credibility.