Research And Application Of Early Fault Intelligent Warning Key Technology For Reciprocating Compressor

Reciprocating compressor is the key equipment in the process industry such as oil refinery, chemical industry and so on. It contains two periodic motion types both rotating and reciprocating, the parts are influenced by alternating load during operation process and compressed gases are mostly hydrogen, natural gas and other inflammable gases. It leads to serious accident frequently. Since monitoring system does not play its true value yet although many units have already installed monitoring system and operators are lack of diagnostic knowledge, fault warning and diagnosis of reciprocating compressor relies on human currently and it will appear omissions inevitably. The parameters in traditional warning system are simple and other factors affecting judgment have not been considered, which results in a high error rate.In order to detect anomaly timely and comprehensively, to obtain specific reasons of anomaly information, that is equipment run/stop state and accurate diagnostic conclusion after a quick judgment, to master equipment operating status information comprehensively and manage the equipment efficiently, this paper studies three aspects:(1) Study the deterioration regular pattern of the big end bearing shell wear fault. The movement mechanism dynamics model of reciprocating compressor is built based on the theory of multi-body dynamics. Wear in different degree is simulated by changing the clearance between the crankshaft and connecting rod. Analyze the changing characteristics of related parameters with increasing of the big end bearing shell wear degree and reveal the variation of parameter during the fault deterioration process.(2) Taking into account that fault mechanism of reciprocating compressor is complicated and fault signals are coupling, non-linear and non-stationary, anomaly detection system which is provided with two functions is constructed according to parameter extraction based on fault case diagnostic experience. The two functions contain both signal type automatic identification and alarm threshold self-learning and automatic setting. The system is able to detect anomaly comprehensively and accurately.(3) Warning key technologies such as run/stop state automatic judgment, sensor anomaly self-test and unit fault decision model construction based on state subspace are in-depth studied when detecting the anomaly information. It can draw an accurate conclusion and then obtain output result automatically after a quick judgment. The warning system can achieve early warning for unit fault successfully, reduce previous error rate significantly and protect the equipment running in a safe and steady condition after verification.