【作者】 屈胜男；

【导师】 顾慧萍；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2008， 硕士

【摘要】 信息融合技术是将来自多传感器的信息和数据进行综合处理,从而做出正确、可靠的判断和决策,近年来在许多领域得到了广泛的应用和研究。在机械故障诊断中,可利用的信息很多,充分利用有用的信息对设备的故障进行诊断才能提高故障诊断的精度和准确性,因此信息融合技术是进行机械故障诊断的一种有效方法。本文以齿轮箱为研究对象,分析了齿轮箱的故障机理,具体研究了齿轮箱中齿轮和轴承的故障,为齿轮箱故障诊断提供了理论依据。针对模糊理论和神经网络在故障诊断中存在的不足和互补性,构建了一种结合两者优点的改进的模糊神经网络,并推导了相应的算法,建立了相应的故障诊断框架。利用改进的模糊神经网络对齿轮箱中的齿轮进行故障诊断,并与BP神经网络的诊断结果作对比,结果表明该方法的学习速度快、诊断精度高。针对故障诊断中的不确定性,采用D-S证据理论进行故障诊断。在对D-S证据理论的基本概念和融合推理方法深入研究的基础上,建立了故障诊断框架,并提出应用改进的组合规则处理故障诊断中的冲突信息。算例证明经过融合可以有效的提高故障诊断的精度,而改进的D-S证据理论能有效的处理相冲突的信息。为了提高诊断精度,本文设计了一种将模糊神经网络和D-S证据理论相结合的综合故障诊断方法,该方法是将模糊神经网络的初步诊断结果进行处理转化为基本概率赋值,然后利用D-S证据理论进行信息融合。对齿轮箱中齿轮和轴承故障诊断的结果验证了这种方法能够有效的提高故障诊断的精度。更多还原

【Abstract】 Information fusion technology is used for deal with multi-sensor information by synthesis to make correct and reliable judgment, it has been researched and applied in many fields recently. There is much available information in mechanical fault diagnosis, only when the available information is used sufficiently, can the precision and incredibility be improved. So information fusion technology is a useful method to implement mechanical fault diagnosis.The gear box is made as the research object, and the fault mechanism of gear box is analyzed in this dissertation. The fault of gear and bearing in gear box is studied in-depth, so the theory assists for the gear box fault diagnosis is provided.For the shortage of fuzzy theory and neural network in fault diagnosis, a new fuzzy neural network structure which combines the merits of fuzzy theory and neural network is constructed and study arithmetic is showed in detail. Through the application in gearbox fault diagnosis and compared with the result from BP neural network, it has been proven that the fuzzy neural fault method is valid. It can be applied valuable in engineering field.Aiming at the uncertainty information in fault diagnosis, this dissertation applies D-S evidence theory to fault diagnosis field. On the basis of research on the essential concept and fusion reasoning methods in D-S evidence theory, this dissertation establishes the framework of fault diagnosis, and uses the modified D-S evidence theory to deal with the conflicted information. The result of examples proves that this method can improve the precision of fault diagnosis effectively, and the modified D-S evidence theory can handle with conflicted information.In order to improve the accuracy of fault diagnosis, an integrated fault diagnosis method that is based on fuzzy neural network and D-S evidence theory is presented in this dissertation. First it converts the preliminarily diagnosis result from fuzzy neural network to basic probability assignment, then use the D-S evidence theory to implement final fusion. The result of gear box fault diagnosis shows that this method can improve the accuracy of fault diagnosis effectively.更多还原