【作者】 宋斌；

【导师】 李小平；

【作者基本信息】 西安电子科技大学 ， 电路与系统， 2009， 硕士

【摘要】 本学位论文以数据挖掘、虚拟仪器和信息融合为理论基础,对复杂系统的故障诊断方法及其应用进行了深入的研究。该研究对保证大型复杂系统高效、安全、稳定、可靠地运行,提高设备生产效率和管理水平,最大程度恢复故障设备并挽回故障造成的经济损失,满足国民经济又好又快的发展等方面有着十分重要的现实意义。本文设计完成了基于虚拟仪器的故障数据采集实验,为后续研究提供了充足数据。针对复杂系统的实际情况,重点研究了粗糙集算法在属性约简和特征提取中的应用,决策树算法在挖掘故障规则中的应用以及聚类算法在增量数据中挖掘新故障规则的应用。通过改进粗糙集属性约简算法,解决了其在属性约简中效率较低的问题;通过利用嵌入式SQL直接对故障数据库进行高效的数据查询与处理,大大提高了ID3算法的效率和可实现性;通过利用ART2算法与K-means算法相结合的方法,有效抑制了ART2聚类中心的漂移。同时研究了故障诊断中的信息融合,通过对诊断数据融合,使得诊断数据更加全面;通过对诊断方法的融合,使得诊断结果更加快速、准确和可靠。最后以此三种算法为核心,以液压系统故障诊断为应用背景,设计实现了一个故障测试诊断系统。通过各功能模块的综合使用,实现了故障数据的采集、存储、预处理、规则挖掘、故障诊断和报表生成打印等一系列功能。通过对设备故障实际诊断的实验,证明该系统可以准确地进行故障诊断,同时具有快速自学习发现新规则的能力,能较好的满足实际故障诊断的需要。更多还原

【Abstract】 This thesis taking data mining, virtual instrument and information fusion as the theoretical bases, has studied deeply on complicated system’s fault diagnosis methods and its application. The researches ensure the large-scale complicated system running effectively, safely, stably and reliably, raise the equipment production efficiency and the management level, restore the fault equipment and retrieve the economic loss caused by the fault to the greatest extent, and have the very vital practical significance on ensuring sound and rapid growth of the national economy and other aspects.This paper designs and completes fault data acquisition experiment based on the virtual instrument, which provides the sufficient data for the following research. For actual situation of complicated system, this thesis mainly studies the application of attributes reduction and feature extraction based on rough set, the application of fault rules mining based on decision tree algorithm, and the application of the new fault rules mining based on clustering algorithm in the incremental data. An improved algorithm of rough sets is presented, which solves the problem of low efficiency in the attribute reduction. An improved ID3 algorithm is presented by querying and processing database directly and efficiently using the embedded SQL, which improves the efficiency and feasibility. An improved clustering algorithm is presented by combination ART2 algorithm with K-means algorithm, which restrains the ART2 cluster center drifting effectively. At the same time, this paper studies information fusion in the fault diagnosis, through the data fusion, making the data more comprehensive, through the method fusion, making diagnosises more rapid, accurate and reliable.At last, taking the three algorithms as the core and fault diagnosis of the hydraulic system as the application background, a fault testing and diagnosis system has been designed and realized. Through various functional modules’synthesis use, it realizes a series of function of data acquisition, storage, pretreatment, fault rules mining and fault matching, the report generation printing . Through actual experiment of fault diagnosis, it has proved that this system can diagnose accurately and discover the new fault rules quickly by self-learning, which satisfies the actual need of fault diagnosis.更多还原