Application Study On Online Condition Monitoring And Failure Analyzing System Of K-201 Centrifugal Compressor

There have two sets of aromatic complex units in Chemical Plant of TPCC, one unit with a production capacity of 80000t per year was built in the end of 70's, 20th century. The other with a production capacity of 200000t per year was built in the end of 90's, 20th century. Centrifugal compressor is one of essential large-scale equipments in aromatic complex units and its operation condition has an important effect on the whole plant.In order to guarantee a long period, highly effective rate and stable operation of the centrifugal compressor units, it has became an important safeguard and essential technology for normal production to use a condition monitoring and analytical system with the capacity of monitoring the running status of a large-scale centrifugal compressor unit, analyzing with corresponding monitor theory and the statistical data and forecasting operational accidents.Taking the circulating centrifugal compressors in the K-201 BPX unit of Chemical Plant of TPCC as the research target and considering the actual conditions of the bigger aromatic hydrocarbon device of chemical plant, the present paper describes the dynamic characteristic of the centrifugal compressor in detail and reviews various automatic online monitoring and forecasting methods for the operational condition of mechanical equipments. On this basis, using the S8000 online condition monitoring system which develops more completed nowadays Bayesian failure analyzing system, corresponding software systems and hardware systems, the online accident analysis and trend prediction of rotating mechanical equipments are achieved through the combination of online condition monitoring system and failure analyzing system.In addition, on the basis of analyzing the existing problem of the reform compressor trouble diagnose (the comprehensive utilization of uncertainty and many source information of trouble diagnose), Bayesian failure analyzing system is set up. Furthermore, a multi- step forecast method based on the recursive neural network is proposed through the combination of BP algorithm and recursive neural network. According to the prediction results of peak value of K-201 vibration, the forecasting system for the key unit has good multi-step prediction ability.