【作者】 陈荣刚；

【导师】 张颖；

【作者基本信息】 东北石油大学 ， 化工过程机械， 2011， 硕士

【摘要】 储罐底板腐蚀状态的严重程度是指导储罐维修的主要指标,然而储罐底板的特殊运行环境决定了它难以检测的特点。目前国际上主要采用声发射检测技术对储罐底板腐蚀状态进行在线检测。该技术在我国已有近十年的发展历史,广泛应用于石油化工行业的常压储罐底板腐蚀状态检测上。但是目前应用声发射在线检测技术对储罐底板腐蚀状态进行评价主要依赖于检测人员的经验,成为该技术推广与发展的瓶颈。本文利用声发射在线检测信息,结合领域专家经验,确定了与储罐底板腐蚀相关的声发射因素和外观检查因素。应用多元统计理论中的相关分析计算各因素与储罐底板腐蚀状态之间的相关性,并采用双尾检验对相关性的显著性进行验证。针对传统启发式贝叶斯网络结构搜索算法的局限性,本文应用遗传算法改进贝叶斯网络结构搜索算法,并采用最优保存策略设计遗传适应性函数,应用贝叶斯参数学习算法对所建立的贝叶斯网络拓扑结构进行参数学习,利用联合树推理算法对实例进行评价分析,从而建立基于检测信息的储罐底板腐蚀状态智能评价模型。利用外观检查信息,根据相关分析结果,分别建立基于外观检查因素和基于相关显著外观检查因素的储罐底板腐蚀状态评价模型。通过对测试储罐数据的评价,模型准确率分别为84%和86%;利用声发射检测信息,根据相关标准,分别建立基于标准用声发射因素和基于声发射因素储罐底板腐蚀状态评价模型。通过对测试储罐数据的评价,模型准确率分别为70%和84%。综合前述研究成果,将显著相关的外观检查因素与声发射因素结合,建立基于在线检测信息的储罐底板腐蚀状态智能评价模型。通过对测试储罐数据的评价,该模型的评价准确率为96%。评价结果表明,该评价模型能有效的利用在线检测信息,对储罐底板腐蚀状态进行智能评价,具有一定的工程应用价值。更多还原

【Abstract】 The condition of tank bottom corrosion is the main indicators to guide the maintenance of tank, however, the special operating environment of tank bottom make it difficult to detect. At present, the acoustic emission testing is used as the major testing technology to detect the tank bottom corrosion international. The technology has a history of nearly a decade in China, it is widely used in the testing of tank bottom corrosion petrochemical industry. But the current application of acoustic emission online testing technique to evaluate the tank bottom corrosion mainly depends on the experience of test people, it has become the bottleneck in development and promotion of the technology.This paper combines acoustic emission testing information with experience of experts of areas, the acoustic emission factors and appearance inspection factors correlated with tank bottom corrosion are determined. Using correlation analysis of multivariate statistical to calculate the correlation between factors and the tank bottom corrosion, and testing the correlation by two-tailed test. For the limitations of traditional heuristic Bayesian networks structure search algorithm, this paper improves Bayesian networks structure search by genetic algorithm, designs the genetic fitness function by optimal preservation strategy, learns the parameters of the established of Bayesian networks topological structure by Bayesian parameters learning algorithm, evaluates and analyses the examples by using joint tree algorithm. To establishing intelligent method in evaluation of the tank bottom corrosion based on-line information.Using the appearance inspection information, according to the results of correlation analysis, establish two evaluation methods of the tank bottom corrosion based on appearance inspection factors and significant correlation appearance inspection factors. Comparing with the result of acoustic emission online testing, the accuracies of the evaluation models are 84% and 86%; Using acoustic emission testing information, according to the testing standard, establish two evaluation methods of the tank bottom corrosion based on standard acoustic emission testing factors and acoustic emission testing factors, Comparing with the result of acoustic emission online testing, the accuracies of the evaluation models are 70% and 84%. Synthesize the above research results, combining the significant correlation appearance inspection factors with the acoustic emission testing factors, establish intelligent method in evaluation of the tank bottom corrosion based on online testing information. Comparing with the result of acoustic emission online testing, the accuracy of the evaluation model is 96%. This result shows that the evaluation model can effectively use the online testing information, and evaluate tank bottom corrosion intelligently, and can be applied in engineering applications.更多还原