转子故障数据分类方法研究与实验台测试信息系统开发

发布时间：2018-02-28 08:28

本文关键词： 自动化测试反馈控制智能诊断数据集降维数据分类　出处：《兰州理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着信息技术、控制理论和人工智能等领域的快速发展，旋转机械故障诊断技术已逐渐呈现出自动化、实时化、智能化等信息化技术水平偏低，难以满足先进工业生产实际应用需求的问题。因此，将传统机械设备拓展为具有控制自动化、监测实时化和诊断智能化等功能特性于一体的机械装备信息化装置，，将成为机械故障智能诊断研究的前沿。其中故障数据分类的科学问题研究，已成为提高故障模式辨识技术水平的核心内容。本项研究以旋转机械核心部件－转子系统为研究对象，综合利用常用统计特征与数据挖掘方法，对故障特征数据的分类问题重点开展了研究，并采用虚拟仪器技术，开发了一套双跨转子实验台振动实验测试与反馈控制系统。开展的具体研究工作内容和获得的主要结果如下： 1)在分析常用统计特征测度的基础上，筛选并从信号中提取出具备表征转子不同运行状态能力的特征向量，以此建立的故障特征数据集表现出维数过高与可分性差的问题。在此基础上，总结出经典数据集降维方法存在着非线性因素干扰、降维准则不足的问题。 2)通过公式推导，归纳出主成分分析（PCA）法和费歇判别分析（FDA）法的本质与内在联系。以上述理论分析为前提，提出了一种偏费歇判别分析（BFDA）的数据集降维方法，并进行实例验证，结果表明，BFDA法在达到FDA法降维性能的基础上，具有更低的算法复杂度。 3)在提出将核主成分分析（KPCA）法与费歇判别分析法相结合实施数据集降维的基础上，推导出以费歇准则建立粒子群优化算法适应度函数中存在的等价关系。并进一步提出将KPCA法与BFDA法相结合的数据集降维方法，提出基于偏费歇准则的粒子群优化方案。将提出的两种降维算法应用于特征数据集降维，并将降维结果输入至设计的分类器进行分类验证，分类效果均较显著。 4)利用虚拟仪器技术，开发了一套双跨转子实验台振动实验测试与反馈控制的软件平台，全面拓展了现有转子系统的软硬件功能，且该平台具有人机交互界面直观，开发、维护便捷，功能易扩展等特点。并在常规控制与监测功能实现的前提下，对状态信息监测功能中存在的若干难点进行了研究与开发。研究表明，特征数据集中包含了大量能够反映转子运行状态的信息，故如何在数据集挖掘算法研究中获得新突破，及如何将取得的研究成果合理地嵌入于自动化测控系统，将是故障诊断领域开展研究工作的重要方向。
[Abstract]:With the rapid development of information technology, the field of control theory and artificial intelligence, fault diagnosis technology of rotating machinery has been gradually showing the automation, real-time and intelligent information technology level is low, it is difficult to meet the practical application requirements in advanced industrial production. Therefore, the traditional mechanical equipment to expand with automatic control and real-time monitoring. Intelligent diagnosis and other features in one of the mechanical equipment information device, will become the frontier research of intelligent mechanical fault diagnosis. The scientific researches on fault data classification, has become the core content of technology level of fault mode identification.
This study is based on core components of rotating machinery rotor system as the research object, the comprehensive utilization of common statistical features and methods of data mining, the key problems of classification of fault data is studied, and the virtual instrument technology, developed a set of double span rotor experimental platform vibration test and feedback control system. The specific research work the content and the main results are as follows:
1) based on the analysis of characteristics of commonly used statistical measure, screening and extract the feature vector representation with different running state of the rotor ability from the signal, in order to establish the fault feature data set showed high dimension andbad separability problems. On this basis, summed up the classic data set reduction method is nonlinear factors interference reduction rule is insufficient.
2) by formula derivation, summed up the principal component analysis (PCA) method and Fisher discriminant analysis (FDA) method the essence and internal relationship. Based on the above-mentioned theoretical analysis foundation, proposed a partial Fisher discriminant analysis (BFDA) data set reduction method is verified, the results show that the base of BFDA method of reducing dimensional performance in reaching FDA method and has lower complexity.
3) in the kernel principal component analysis (KPCA) method and Fisher discriminant analysis method combining implementation based on dimensionality reduction data set, according to Fisher criterion of particle swarm optimization algorithm to adapt to the existence of equivalent relation function is derived. And further put forward by combining KPCA method and BFDA method for dimensionality reduction of data set this method, partial Fisher principle of particle swarm optimization scheme based on two dimension reduction algorithm proposed was applied to feature data dimensionality reduction and classification will be reduced to the design of the verification result of dimension input classifier, classification results were more significant.
4) the use of virtual instrument technology, developed a set of double span rotor experimental platform vibration test and feedback control software platform, comprehensive development of the hardware and software function of the existing rotor system, and the platform has intuitive man-machine interface, convenient maintenance, easy development, function expansion and other characteristics. The premise and the realization in the conventional control with the function of monitoring, some problems of state information of the monitoring function of research and development.
Research shows that the feature data set contains a large number of information can reflect the running state of the rotor, so how to make a breakthrough in the data mining algorithm, and how to apply the research achievements reasonably embedded in automation control system, will be an important direction of fault diagnosis research.

【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【引证文献】