GA- LVW算法在化工过程故障诊断中的应用  

Application of a genetic algorithm-based improved Las Vegas wrapper (GA- LVW) algorithm in chemical process fault diagnosis

作　　者：耿志强[1,2] 李俊 曹原[1,2] 韩永明 GENG ZhiQiang;LI Jun;CAO Yuan;HAN YongMing(College of Information Science and Technology;Engineering Research Center of Intelligent Process Systems Engineering,Ministry of Education,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]北京化工大学信息科学与技术学院,北京100029 [2]北京化工大学智能过程系统工程教育部工程研究中心,北京100029

出　　处：《北京化工大学学报（自然科学版）》2022年第6期101-109,共9页Journal of Beijing University of Chemical Technology(Natural Science Edition)

基　　金：国家自然科学基金(21978013);中央高校基本科研业务费(XK1802-4);贵州科技计划项目重大专项(Guizhou Branch [2018]3002)。

摘　　要：针对田纳西-伊斯曼(Tennessee Eastman, TE)化工过程中故障特征难区分、诊断方法易陷入局部最优等问题,提出一种基于遗传算法(genetic algorithm, GA)改进的拉斯维加斯包裹式(Las Vegas wrapper, LVW)特征选择方法GA-LVW。利用GA算法改进LVW算法搜索盲目性的缺点,使得特征组合能够快速有效地收敛到近似最优,进而集成机器学习的分类器对TE过程进行过程监控,发现异常状态,从而实现故障诊断。通过TE化工过程的故障诊断实验验证,将GA-LVW算法与未改进的LVW算法及未进行特征选择的分类算法进行对比,结果表明所提GA-LVW方法提高了LVW特征选择的稳定性和寻求近似最优解的迭代速度,从而提升了分类器故障诊断发现异常状态的准确率。In the Tennessee Eastman(TE) chemical preocess, fault features are difficult to distinguish and diagnostic methods tend to fall into a local optimum. A genetic algorithm(GA)-based improved Las Vegas wrapper(LVW) feature selection method(GA-LVW) has been developed to address these problems. The GA is used to improve the shortcomings of the blindness of the LVW algorithm search, so that the combination of features can converge to the approximate optimum quickly and efficiently, and then integrate the machine learning classifier for process monitoring of the TE process to find abnormal states, thus enabling fault diagnosis. GA-LVW algorithm was compared with the unimproved LVW algorithm and the classification algorithm without feature selection in experimental fault diagnosis in the TE chemical process. The results showed that the proposed GA-LVW method improves the stability of LVW feature selection and the iteration speed of seeking the near-optimal solution, thus improving the accuracy of classifier fault diagnosis and detection of abnormal states.

关 键 词：故障诊断 遗传算法 拉斯维加斯算法 化工过程 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]