基于全信息技术的远程诊断关键技术研究

【作者】 杨春燕；

【导师】 韩捷；

【作者基本信息】 郑州大学 ， 机械设计及理论， 2004， 硕士

【摘要】 随着机械设备的大型化、自动化、高速化和复杂化，使得设备状态监测和故障诊断技术变得越来越重要。传统的机械故障诊断大都以单通道信号分析为基础，从中提取有关机器行为的特征信息。虽然利用一种信息有时可以判断机器的故障，但在许多情况下得出的诊断结果并不可靠。转子一个截面上的振动信息一般是由相互垂直两个径向和一个轴向传感器拾取的，一个截面上的通道属同源信息，它们之间必然存在一定的关系，对这些信息在数据层次上采用全信息技术融合可以较全面的获得转子信息。然而转子的振动是一个空间的范畴，对一个截面的同源信息融合也只能最大限度反映这一个截面转子的振动信息，所以根据转子一个截面的振动信息诊断出来的结果往往也是片面的。因此，采用转子多个截面信息进行故障的诊断是必要的。 本文结合转子空间振动和现场实际，利用转子两截面信息，建立一种数据层、特征层和决策层三种层次混合融合诊断模型，全信息技术和集成概率神经网络是实现该模型的关键技术。本文把全信息技术中的全息谱、全谱和全矢谱分别应用于诊断模型，并对三种方法的诊断结果从各个方面进行了比较，最后得出结论是全矢谱是最有发展前途的一种全信息分析技术。 在全信息技术中，全矢谱由于其良好的信息处理方式，使其拓展出一系列的分析方法。本文讨论了在故障诊断中几种有意义的矢谱拓展分析方法：矢功率谱、矢倒谱和针对非平稳信号的矢Wigner分布，并把这些分析方法应用于故障诊断中，用实际信号验证了其诊断准确性。 本文在理论探讨的基础上，应用C++Builder5.0构建了基于Browser／Server模式的远程全信息故障诊断系统，系统实现了基于全信息技术及全矢谱拓展分析方法的故障诊断，诊断结论准确可靠。更多还原

【Abstract】 With the advancement of modern science and technology, the equipments tend to be larger, faster as well as more roboticized and complicated, which makes technology of state monitoring and fault diagnosis more and more important. But the traditional fault diagnosis of machinery is based on the analysis of the signal coming from single channel from which the character information about the machinery action is picked up. Although using individual information can reflect the machinery faults sometimes, in many instances the diagnosis result is irresponsible. Commonly the vibration information of one rotor’s section is picked up from two radial sensors and a axes sensor,one section’s channels belong to same dimension,so inevitably each other of them exists certain relation, accordingly these information fusion at data level through full information technology can farthest show the section’s informations the diagnosis result sometimes is unilateral on the basis of rotor’s one section. Therefore it is necessary to recognize faults according to the rotor’s multi-sections information.Combining with the rotor space vibration and the field practical situation.and making use of the rotor two ends, The thesis founds a model of data level and character level and decision level mixture fusion,and the key techniques to realize the model are the full information and integration Probabilistic Neural Network. The thesis applys Full-spectrum and Holospectrum and Vector-spectrum to the diagnosis model,and compare the diagnosis result of the three methods,finally the conclusion is that the Vector-spectrum is the best of the full information.The Vector-spectrum of full information continues a series of analysis methods because of its good information processing way. The thesis discusses several important analysis methods to fault diagnosis: Vector power spectrum and Vector-cepstrum and Vector-Wigner that used to analysis stationary signals,and combines these methods width system model to diagnose faults as well as validates the methods veracity.After the rigorous study of the theory, using the C++ Builder5.0 as the developing tool, a full information fault diagnosis system was visualized, which based on the mode ofBrowser/Server and can run on the Internet. Also the full information and the Vector-spectrum’s continued methods are visualized,and the diagnosis conclusion is exact and reliable.更多还原