数据分析与BP神经网络相结合的乙烯装置智能故障诊断系统  

RESEARCH ON INTELLIGENT FAULT DIAGNOSIS SYSTEM OF ETHYLENE PLANT BASED ON DATA ANALYSIS AND BP NEURAL NETWORK

作　　者：常亚娜 武锦涛 代玉强[1] Chang Yana;Wu Jintao;Dai Yuqiang(School of Chemical Engineering,Dalian University of Technology,Dalian,Liaoning 116024)

机构地区：[1]大连理工大学化工学院,辽宁大连116024

出　　处：《石油炼制与化工》2023年第6期97-104,共8页Petroleum Processing and Petrochemicals

摘　　要：针对化工生产流程中各单元具有强关联性、连续性、复杂性的特点,以某乙烯生产装置为对象进行Aspen仿真,提取数据并进行灵敏度分析,发现设备故障可以通过分馏塔再沸器和冷凝器的热负荷变化来反映;进而,以脱甲烷塔、脱乙烷塔和乙烯精馏塔为例建立动态模型,把引起热负荷变化的温度、压力、甲烷含量和进料流量4种可观测参数数据集进行故障分类编码,再利用3层BP神经网络建立热公用工程(主要指蒸汽)智能诊断系统模型。对不同可观测参数及其组合数据各选取10000组数据样本进行训练和验证,在获得阶段性诊断结果后,综合设备热负荷变化得到完整的装置故障情况。测试集样本验证结果显示,所建BP神经网络诊断模型的准确率较高,可达到99.75%。实际应用结果表明,该故障诊断系统在实际操作中能够快速有效地判断出设备故障诱因。In view of the characteristics of the strong correlation,continuity and complexity of each unit in the chemical production process,the Aspen simulation was carried out on an ethylene production unit.The data were extracted and the sensitivity analysis was carried out.It was found that the equipment failure could be reflected by the change of heat load of reboiler and condenser in fractionating column.Then,a dynamic model was established by taking the demethane column,deethane column and ethylene rectifying column as examples.The data sets of four observable parameter,temperature,pressure,methane content and feed flow,which caused the change of thermal load,were classified and coded,then the intelligent diagnosis system model of thermal utility(mainly steam)was established by using three-layer BP neural network.10000 samples of different observable parameters and their combined data were trained and verified.After the periodic diagnosis results were obtained,the complete fault condition of the integrated equipment was determined.The results of test set sample verification showed that the accuracy of BP neural network diagnosis model was high,reaching 99.75%.The practical application results showed that the fault diagnosis system could quickly and effectively judge the cause of equipment failure in the actual operation.

关 键 词：公用工程 神经网络 乙烯装置 故障诊断 

分 类 号：TQ221.211[化学工程—有机化工] TP183[自动化与计算机技术—控制理论与控制工程]