文档介绍：华中科技大学硕士学位论文
摘要
我的提高,极大地促进了我国电力事业的发展,
而核电由于其经济、安全可靠、清洁的特性,更是得到了重点关注。预计到 2020
年,我国核电发电量将占到总发电量的 4%。
在核电企业的生产经营过程中,对设备的依赖程度很高。作为设备正常维护检
修和应急处理的保障性物资,备件的重要程度不言而喻。
本文首先分析了现有的分类法存在的问题:分类方法过于简单、引入了过多定
性的主观判断、过于复杂或者需要大量数据等。然后,介绍了支持向量机(SVM)
分类理论,以及该分类理论的优点:对有限样本情况下的数据分类问题,具有全局
优化、训练时间短、泛化性能好、算法复杂度与特征空间维数无关等优点。
接着,分别介绍了核电企业的两种备件——常用备件和不常用备件,并在分析
不常用备件特点,如可用性要求高、专用性强、单价高、生产周期长、使用频率较
低且规律性不强、寿期不确定等的基础上,建立了备件分类的指标体系。
然后,在研究 SVM 多类别分类方法的基础上,选择了决策有向无环图法作为
多类别分类方法,设计了基于 SVM 的备件分类模型。
最后,本文以某核电企业的不常用备件数据为背景,进行了 SVM 核函数选择
及参数调整的应用研究,并验证了本文设计的分类模型的正确性和有效性。
关键词:支持向量机备件分类模型
I
华中科技大学硕士学位论文
Abstract
panied with the development of the economy and the improvement of the life
quality of people, the electricity industry is developing rapidly. The nuclear power
industry is playing a more and more important role in the electricity industry, with the
attribute of economy, safety and cleanness. It is predicted that by the year 2020, the
percentage of nuclear power industry will raise to 4% of the whole electricity industry.
During the operation of the nuclear power station, reliability of machines is the key
point. As a backup of the machines, the spare parts play an important role in maintenance
jobs.
First of all, the paper analyzes the problems existing in the current inventory
classification, the prehensive coverage of indicators, the great subjectivity, the need
of a large training data set etc. After that, the basic theory of support vector machine
(SVM) and the advantages of SVM are introduced, such as simple structure, faster
classification speed, better generalization ability and global optimized etc.
Secondly, two kinds of spare parts, the frequently used spare parts and the
intermittent demand spare parts, are introduced. A system of indicators in classification of
nuclear power station spare parts is established, based on the attributes of the intermittent
demand spare parts, such as high price, long lead-time and short l