基于BP神经网络和证据理论的离心压缩机喘振诊断方法

Centrifugal Compressor Surge Diagnosis Method Based on BP Neural Network and Theory of Evidence

离心压缩机是石油化工生产中的核心动力设备,然而运行过程中易发生喘振故障造成事故。对于喘振故障,传统方法采用时频特征分析方法诊断,而该方法通常在喘振发展到后期、信号特征明显的情况下才诊断出喘振故障。为解决该问题,提出基于BP神经网络和证据理论的诊断方法,该方法使用故障数据训练得到BP神经网络,进而对采集的数据进行初步诊断,再采用证据理论融合各初步诊断结果得出诊断结论。通过在离心压缩机实验台上模拟喘振故障,结果表明:该方法能够准确诊断压缩机喘振故障,此外与传统方法相比,采用该方法能在喘振发生初期诊断出故障,从而进行调控避免喘振发展到后期,这对实现离心压缩机防喘具有重要意义。

Centrifugal compressor is the key power equipment in petrochemical production industry. However,surge faults are prone to occur during the operation. It is traditional to use time-frequency characteristic analysis method to diagnostic fault.While,it shows that the traditional method cannot come to conclusion until when the surge has occurred and the signal characteristics are obvious. In order to solve this problem,the surge diagnosis method based on BP neural network and evidence theory has been proposed. It is characterized by using fault data to train BP neural network,then using BP neural network to analyze the collected data. And the corresponding initial diagnosis result is obtained,then using the theory of evidence draw the conclusion. Results of the simulation conducted on the centrifugal compressor show that the method is effective in surge diagnosis. In addition,compared with the traditional method,the method proposed in this paper can detect the fault before surge occurs,which takes great significance for the realization of the centrifugal compressor anti-surge.