一种基于Stacking算法的乙烯裂解炉炉管结焦智能诊断方法  

An intelligent diagnosis method for coking of ethylene cracking furnace tube based on Stacking algorithm

作　　者：彭心怡 熊建斌 PENG Xinyi;XIONG Jianbin(College of Mathematical Sciences,South China Normal University,Guangzhou,Guangdong 510631,China;College of Automation,Guangdong Polytechnic Normal University,Guangzhou,Guangdong 510655,China)

机构地区：[1]华南师范大学数学科学学院,广东广州510631 [2]广东技术师范大学自动化学院,广东广州510655

出　　处：《石河子大学学报（自然科学版）》2021年第6期786-792,共7页Journal of Shihezi University(Natural Science)

基　　金：国家自然科学基金面上项目(62073090,61473331)。

摘　　要：为提高炉管结焦诊断的准确度,提出一种基于Stacking算法的炉管结焦智能诊断方法。该方法首先利用多重层次聚类算法实现了对炉管外表面温度的精准计量;然后融合相关性度量和网格搜索算法,实现了基学习器和次级学习器的最优组合;最后构建了基于支持向量机、朴素贝叶斯、径向基函数神经网络、逻辑回归和随机森林的炉管结焦诊断模型。对比实验表明,基于Stacking算法的炉管结焦诊断模型的准确率(99.82%)、有效性和稳定性比使用单一的基于学习器SVM、朴素贝叶斯、RBF和LR算法训练的结焦诊断模型均有一定程度的提高。因此,本文方法可为乙烯生产过程中裂解炉炉管结焦程度的判断提供可靠依据。In order to improve the accuracy of furnace tube coking diagnosis,an intelligent methodbased on Stacking algorithm is proposed in this paper.This method firstly uses a multi-level clustering algorithm to realize the accurate measurement of the outer surface temperature of the furnace tube;then integrates the correlation measurement and grid search algorithm to achieve the optimal combination of the base learner and the secondary learner;finally constructs a furnace tube coking diagnosis model based on support vector machine(SVM),naive Bayes,radial basis function(RBF)neural network,logistic regression(LR)and random forest.The comparison experiment showed that the furnace tube coking diagnosis model based on Stacking algorithm had higher accuracy(99.82%),validity and stability than that based on single learning machine SVM,naive Bayes,RBF and LR algorithms.Therefore,the method in this paper can provide a reliable basis for the determination of coking degree of cracking furnace tube in ethylene production.

关 键 词：乙烯裂解炉炉管 Stacking算法 结焦诊断 

分 类 号：TB791[一般工业技术—真空技术]