【作者】 王文杰；

【导师】 宫俊；

【作者基本信息】 东北大学 ， 系统工程， 2015， 硕士

【摘要】 近年来随着我国经济的发展,我国炼油业发展迅速,与此同时由于设备腐蚀而带来的安全问题也日益严重,因腐蚀造成的泄露和爆炸时有发生,因此准确有效地预测炼油设备的腐蚀速率和剩余寿命具有十分重要的意义。本文首先介绍了国内外腐蚀预测问题的研究现状,比较了不同预测方法的优缺点。考虑到影响腐蚀的因素众多且相互影响,采用传统数学建模的方法进行腐蚀预测存在一定的难度和不足,于是采用基于支持向量机回归的智能学习算法,从数据分析角度对腐蚀速率进行预测,同时给出了问题描述和预测模型。其次,本文重点研究了支持向量机参数(C,ε,g)的选择,给出了基于区间估计法的参数优化方法和基于GA的最优参数设计。并就区间估计法和GA优化后的支持向量回归算法同BP神经网络和加权移动平均模型的预测结果进行比较,结果表明基于GA参数优化的支持向量回归预测结果优于其他三种方法。选择GA_SVR算法作为系统实现的主算法。然后,由于设备的剩余寿命主要受腐蚀速率的影响,根据预测得到的腐蚀速率采用蒙特卡罗法进一步对设备失效概率进行统计分析,得到设备的剩余寿命。最后,对炼油设备腐蚀预测系统进行了系统需求分析和设计,实现了基于支持向量回归算法的腐蚀速率预测以及基于蒙特卡罗方法的剩余寿命预测。更多还原

【Abstract】 In recent years,with the development of our country’s economy,China’s oil refining industry is developing rapidly,at the same time due to the equipment corrosion,the security problem is increasingly serious.Leakage and explosion happened from time to time.So accurately and effectively predict the corrosion rate of the refining equipment and remaining life is of great significance.First,this paper summarizes the research status of corrosion prediction,compares the advantages and disadvantages of different methods.Considering the influence factors of corrosion is complex,the traditional mathematical modeling methods has certain difficulty and insufficiency.So from the aspect of data analysis,the SVR method is proposed to address the corrosion predict,and the problem description,the model formulation are provided orderly.Secondly,the key part of this paper is to obtain the reasonable and optimal parameters for the SVR model.Firstly,the trial and error method is proposed to get the ranges of parameters(C,ε,g)and GA SVR is proposed to get the best parameters.Moreover,the comparison with the BP neural network and AR methods are done to illustrate the advantage of GA_SVR.Choosing GA SVR as the main algorithm of system implementation.Then,because the remaining life is affected by the corrosion rate,according to the prediction of corrosion rate,using the monte carlo method testing the failure probability of the equipment and obtain the remaining life.Finally,the system requirements analysis and design is developed for the equipment corrosion prediction system.Implements the corrosion rate prediction based on SVR method and remaining life prediction based on monte carlo method.更多还原