Research On Multi-source Information Fault Warning Method For Reciprocating Compresso

Reciprocating compressor is the core equipment of the national industrial development,and has been widely used in petroleum,chemical industry and other fields.Research on fault warning method of reciprocating compressor is of great practical significance for improving the stability of reciprocating compressor operation and ensuring the safety of industrial production.In this paper,the valve components of a reciprocating compressor on a petrochemical platform are taken as the research object.By analyzing the working mechanism of the reciprocating compressor,and combining with data processing analysis and time series prediction algorithms,a multilevel fused LSTM network parameter prediction model based on CEEMD-FCM data processing algorithm and a multi-source information fused air valve fault warning strategy are presented.Achieve early warning of valve failure and improve the operating efficiency of the equipment.To solve the problem that the parameter operation data of reciprocating compressor contains noise signal and characteristic information,which makes it difficult to extract useful feature information directly from the original data,this paper uses CEEMD-FCM algorithm to perform multi-scale mode decomposition and fuzzy cluster fusion on multi-parameter and multi-point operation data of reciprocating compressor,such as vibration and pressure.Accurate characterization of data detail feature information is achieved to reduce the complexity of subsequent parameter prediction modeling.The CEEMD-FCM&MLF-LSTM network for predicting parameters of reciprocating compressor is constructed to solve the problems of low prediction accuracy and poor robustness in the prediction model of parameters of reciprocating compressor.The network includes data processing layer,information learning layer and prediction output layer.In the data processing layer,the CEEMD-FCM algorithm is used to process the multi-parameter running data and extract the feature information.In the information learning layer,MLF-LSTM is used to learn the temporal and spatial characteristics of fuzzy modes and their changing rules,and to reconstruct the data.In the prediction output layer,the prediction results are output after the parameter prediction results are normalized.The experimental results show that the CEEMD-FCM&MLF-LSTM network can predict the operating parameters of different types of compressors with high prediction accuracy and good robustness.A fault early warning strategy based on multi-source information fusion is presented for the complex parameter types of reciprocating compressors and the timeliness of fault early warning.This strategy combines the multi-parameter change information of reciprocating compressor equipment into valve health value,and then synthesizes the health status of the device according to warning threshold and warning restrictions,so as to realize the valve failure warning of reciprocating compressor.The experimental results show that the fault alarm method proposed in this paper has good warning performance in the aspects of accuracy,stability and sensitivity.Based on the algorithm mentioned in this paper,a set of software for fault warning system of reciprocating compressor is designed.The software system mainly includes model warning and updating,data collection and collation,and device status visualization function modules.The theoretical research results of fault warning for reciprocating compressor based on multi-source information fusion are applied to industrial production.