【作者】 苗伟；

【导师】 马建伟； 范波；

【作者基本信息】 河南科技大学 ， 检测技术与自动化装置， 2009， 硕士

【摘要】 目前,交流传动系统的故障诊断技术受到广泛关注,在电机的可靠性解决后,变频器的故障诊断研究成为重点。故障诊断已进入智能化的诊断阶段,智能诊断方法无需对象的数学模型,引起极大的重视和研究。但是目前已开发的方法都存在局限性,融合多种智能诊断方法已成为当今故障诊断的一个研究热点。本文在研究智能故障诊断理论的基础上,提出了两种集成智能故障诊断方法,一种是融合FTA和BAM的故障诊断方法,一种是基于BP神经网络和D-S证据理论的故障诊断方法,分别对变频器及变频器最易失效的逆变电路部分进行故障诊断研究。在研究FTA和BAM神经网络在故障诊断中应用的基础上,提出了一种融合FTA和BAM的故障诊断方法。利用FTA得到系统所有的故障模式,进而由故障模式和根据经验得到的故障分析归纳出BAM的学习样本,即故障模式和故障分析之间的对应。BAM通过联想记忆矩阵并行联想,得到诊断结果,扩展综合故障诊断能力。用上述方法对变频器进行故障诊断仿真分析,结果表明该方法用于解决此类问题是有效的。根据BP和D-S证据理论的特点,使BP神经网络和D-S证据理论优势互补,构建基于BP和D-S推理的综合诊断模型。各测点建立各自的BP,其信号由自己的BP先诊断,完成单测点的局部诊断。然后利用子BP的局部诊断结果构造D-S证据理论的可信度函数,完成多测点证据信息的融合,最终得到综合诊断结果。通过对变频器的关键部件-逆变电路故障诊断的仿真实例分析,验证了该方法的可行性和有效性。更多还原

【Abstract】 At present, the fault diagnosis technology of the ac transmission system is widely concerned, and the research on diagnosis is focusing on the inverter’s fault diagnosis after the reliability of the motor is solved. The fault diagnosis has entered the phase of intelligent diagnosis. The intelligent fault diagnosis methods have been paid much attention because they don’t need the math model of the object. However, in terms of the various intelligent methods which have been developed at present, there are limitations,so the technology of integrating multiple intelligent diagnosis has become a hotspot..In this article, on the basis of studying about the theory of intelligent fault diagnosis,two methods of integrated intelligent fault diagnosis were proposed, one was the integration of FTA and BAM, the other one was based on the BP network and D-S evidence theory, specificly, those two intelligent methods were applied to the inverter and the inverter’s most vulnerable part -the inverter circuit. Based on the study about the application of FTA and BAM network in fault diagnosis, a method of fault diagnosis based on the fuse of FTA and BAM was proposed. All of the fault modes were obtained by FTA, then the learning sample ofBAM were summarized which were the corresponding relations between the fault sign(the status of the monitoring points)and the happening or not of the bottom events in fault tree. The results of diagnosis were associated parallel by the associative memory matrix, thus expanding the general ability of fault diagnosis. An emulation of the realization of fault diagnosis with the method mentioned above was performed, and analytical results showed that the approach is valid to fault diagnosis.According to the characteristics of BP network and D-S evidence theory, a diagnosis model combined the advantages of BP and D-S theory was constructed. Each measuring point established its sub-BP network, finishing partial diagnosis of single measuring point. Then the credibility functions of D-S were constructed by using each BP’s partial diagnosis result, completing the fusion of multiple evidence information. Finally, the comprehensive diagnosis results were got. The feasibility and effectiveness of this method was verified by the simulation analysis of fault diagnosis in inverter’s key components-the inverter circuit.更多还原