摘要
为了提高炼化厂循环水对管道腐蚀预测的精度,选取8种常规监测数据作为样本标准库,在此基础上考虑各指标之间信息叠加的影响,引入核主成分分析(KPCA)和广义回归神经网络(GRNN)腐蚀速率预测模型,通过KPCA对原始数据进行预处理,提取影响管道腐蚀的主要因素,应用GRNN建立管道腐蚀速率预测的数学模型,通过分析影响循环水腐蚀的关键因素,建立了循环水腐蚀预测指标体系。结果表明,将样本监测数据的维数由8降至5,可得出各个影响因素的贡献率,提取出包含原始信息95.84%的5个变量,且基于KPCA-GRNN的算法对监测管道腐蚀速率的平均相对误差为0.033,优于误差反向传播算法(BP)的0.056。因此,基于KPCA-GRNN算法建立的循环水碳钢腐蚀速率预测模型,能够获得更准确的预测结果,拓宽了循环水腐蚀速率预测方法的研究思路。
In order to improve the refinery circulating water pipeline corrosion prediction accuracy,eight kinds of routine monitoring data are selected as sample standard library,and on the basis of which,the influence of the superposition of information between the indexes is considered,and kernel principal component analysis (KPCA) and generalized regression neural network (GRNN) corrosion rate prediction model are then introduced.The original data is preprocessed by KPCA to extract the main factors influencing the pipeline external corrosion,and GRNN is applied to establish the mathematical model of pipeline corrosion rate prediction,by analyzing the key factors affecting the corrosion of circulating water,the prediction index system of circulating water corrosion is established.The results show that by reducing the dimension of sample monitoring data from 8 to 5,the contribution rate of each influencing factor can be obtained,and 5 variables containing 95.84% percent of original information can be extracted.Moreover,the average relative error of KPCA-GRNN algorithm for monitoring pipeline corrosion rate is 0.033,which is better than 0.056 of BP.Therefore,the corrosion rate prediction model of circulating water carbon steel based on KPCA-GRNN algorithm can obtain more accurate prediction results and broaden the research idea of corrosion rate prediction method of circulating water.
作者
段春莲
罗东浩
杨剑锋
陈良超
刘晓晨
安延海
DUAN Chunlian;LUO Donghao;YANG Jianfeng;CHEN Liangchao;LIU Xiaochen;AN Yanhai(School of Mechanical and Electrical Engineering,Beijing University of Chemical Technology,Beijing 100029,China;Automation College,Harbin Engineering University,Harbin,Heilongjiang 150001,China;Liaoyang Petrochemical Branch of PetroChina,Liaoyang,Liaoning 111003,China;Liaohe Petrochemical Branch of PetroChina,Panjin,Liaoning 124022,China)
出处
《河北工业科技》
CAS
2019年第5期346-351,共6页
Hebei Journal of Industrial Science and Technology
基金
国家重点研发计划项目(2018YFC0809004)
关键词
材料失效与保护
核主成分分析
广义回归神经网络
循环水腐蚀
腐蚀速率预测
material failure and protection
nuclear principal component analysis
generalized regression neural network
circulating water corrosion
corrosion rate prediction
