Research And Application On Reciprocating Compressor Fault Diagnosis Expert System

Reciprocating compressor is the key equipment of petrochemical industry。Once the reciprocating compressor breaks, it will bring serious damage to theproduction, or it will threat the people's life and property security around.therefore, some effective detection means must be used to prevent andtroubleshoot faults. Because reciprocating compressor itself has complicatedstructure, many vibration sources, etc, and the later start of domesticreciprocating compressor fault diagnosis, research on reciprocatingcompressor fault diagnosis expert system has important practical significanceIn the domestic, research on reciprocating compressor intelligentdiagnosis started late. Taking centrifugal compressor expert system successfulcases as reference, contrasting the advantages and disadvantages of theintelligent diagnosis methods that have successful experience, reciprocatingcompressor expert system that based on rules and case was chosen as thereciprocating compressor intelligent diagnostic methodIn this paper, the common faults are listed and summarized, includingfault symptoms and fault reason. In the knowledge acquisition, more than10kinds of the failure and the fault mechanism on the reciprocating compressorfive failure positions are summarized. Combining with monitoring signals, thefault signal characteristics are summarized, and the failure and the signalcharacteristics are corresponded. In the structure design, a modular design waywas adopted to design reciprocating compressor fault diagnosis exert system.ID naming rules were established to name each module's content. In faultsymptom reasoning to failure, using failure monitoring site as classificationstandard, combining with actual diagnosis experience, a fault reasoning logicdiagram was made and verified.Research shows that reciprocating compressor fault diagnosis expertsystem plays an important role in actual operation monitoring. It can also findearly fault, forecast fault that has not occurred.