【作者】 许茁；

【导师】 王国峰；

【作者基本信息】 大连海事大学 ， 控制理论与控制工程， 2005， 硕士

【摘要】 本文以大连海事大学自动化研究所与瓦轴集团轴承试验测试中心合作的轴承疲劳寿命二期技术改造与增容项目为选题背景,结合当前国内外关于轴承寿命研究的发展现状,针对数据库和数据仓库技术在故障诊断领域的广泛应用以及现在各工厂基本上都对重要设备实施了实时监控,由传感器不间断的传回试验机组运行状态的各种数据及其参数,已经形成大型数据库或数据仓库这一事实,提出将数据挖掘技术应用于轴承的故障诊断领域,找出典型故障数据中的潜在知识,完成典型故障的识别问题,以推动故障诊断技术的不断向前发展,同时,对轴承寿命的预测做了探讨性的尝试。 决策树、关联规则等数据挖掘方法能产生显示的规则,并能有效解决海量数据中知识的发现问题。决策树是一个类似于流程图的树结构,主要用途是提取分类规则,进行分类预测;关联规则挖掘寻找给定数据集中项之间的有趣联系,并产生规则。 本文选用C++Builder自带数据库作为存储设备状态信号的数据库管理系统,建造用于存储设备数据状态的简易数据库。在WINDOWS98开发平台上,使用C++Builder开发工具采用面向对象的程序设计思想和模块化程序设计方法,对给予数据挖掘的轴承故障诊断系统进行软件实现。 鉴于将数据挖掘方法应用于轴承的故障诊断上,毕竟是一种新方法的尝试,若采用单一的分类方法未必会取得良好的效果,本文就是采用上述两种数据挖掘的方法来处理实际问题的。实践证明,根据具体问题,有针对性的采用几种方法结合使用,可提取出很有价值的轴承故障规律。 另外,文章最后调用SPSS软件包,应用回归分析原理中的曲线估计对轴承的峰值指标进行简单的曲线拟合,预测轴承疲劳寿命。实验结果表明,建立了较准确的预测模型,就可以较准确地预测轴承寿命。更多还原

【Abstract】 This paper is backgrounded on the project of The Second Technology Alteration and Add-content for the Test of Bearing’s Fatigue and Life, which is cooperated by Dalian Maritime University Automation Institute and the Bearing Test Inspiring Center of WangFangDian Bearing Group. Combined with the current development status of the research on bearing’s life at home and abroad, aiming at the fact that data-base and data warehouse technology have been applicated in fault diagnose field widely and the reality of the installation of on-line and off-line monitoring system to the significant equipments and large-scale databases and data warehouses forming in the field of fault diagnosis, a new method is presented that is using Data Mining technology in the field of bearing’s fault diagnosis to find out the latent knowledge in the typical fault data and finish the identification of representative faults. At the same time, we make an attempt at the forecast on the bearing’s life.The two Data Mining methods of Decision Tree and Association Rule can produce visible rules and solve the discovery of knowledge in the magnanimity data effectively. The Decision Tree has a tree construction like flow chart, whose main usage is to pick up classified rules and predict respectively. The Association Rule is to find out interesting rules among the data items and generate rules.In this paper C++ Builder’s own database is chosen to be the database management system for storing the status signals of the equipments and build simple database for saving the data status of the equipments. Adopting the object oriented program design and modularization program design method by using C++ Builder empolder tool in Windows 98, the software of Bearing’s Fault Diagnosis system based on Data Mining is realized.Due to Data Mining application in bearing’s fault diagnosis is a new method and it’s not high-point by only using one of them, the two Data Mining method of Decision Tree and Association Rule are chosen to resolve factual problem. Facts have proved that using multi-methods and combined with each other can get valuable rules of bearing’s fault according to material problem.In addition, SPSS is applied in the last part of this paper. Curve estimation in the Regression Analysis Principle is used to simulate and compose peak curve to predict the bearing’s fatigue and life. The final experimental result shows that we can forecast the bearing’s life more accurately as long as an exact prediction model is built.更多还原