基于信息融合技术的故障诊断模型和方法研究

【作者】 张远；

【导师】 罗大庸；

【作者基本信息】 中南大学 ， 交通信息工程及控制， 2003， 硕士

【摘要】 故障诊断系统利用诊断对象系统运行的各种状态信息和已有的各种知识，进行信息的综合处理，最终得到关于系统运行状况和故障状况的综合评价，因此信息融合技术是进行故障诊断的一种有效的方法。本文在详细分析了故障诊断技术和信息融合技术的基本原理以及它们之间关系的基础上，提出了一种新的基于信息融合技术的故障诊断模型和方法。该模型包括信号采集和处理、特征提取、信息分配、局部诊断和决策融合诊断等过程。在信号处理部分采用小波分析方法提取信号的特征；在信息分配过程，采用特征的广义模糊熵作为评价准则，利用相关分析法降低特征维数，从而实现特征信息的分配；局部诊断网络由多个神经网络实现；全局决策融合网络，采用证据理论和模糊积分方法进行决策融合。本文提出了应用证据理论进行全局融合诊断时基本概率赋值的一种构造方法，提出了应用模糊积分进行故障诊断时的模糊测度的一种构造方法。与其他的故障诊断方法相比较，本文提出的模型和方法在原理上从人类的神经生理结构入手，模拟人类对故障判断的一般过程，因此方法上更接近人类对事物的认识过程。通过对故障征兆的融合，可以给出精度更高和更可靠的诊断结果，避免了诊断过程中的信息丢失；证据理论和模糊积分融合方法可以减少决策过程中的不确定性，大大的提高了决策的精度。 将本文提出的基于信息融合技术的故障诊断方法应于汽车发动机的故障诊断，对发动机气阀机构的故障进行诊断。诊断结果表明，该方法的故障诊断结果与实际情况是非常吻合的，验证了方法的可行性和正确性。该方法还可以进一步推广应用于其他复杂系统的故障诊断。 最后本文在理论分析的基础上设计了一个离线的汽车发动机故障诊断系统，并探讨了基于嵌入式处理器的在线故障诊断系统的实现方法。本文进行的是一些理论上的基础性研究工作，要在故障诊断中真正应用本系统，还需要进一步的测试、修改和完善。更多还原

【Abstract】 The fault diagnosis and detection of system is an important methodology to insure the equipment turn over normally. Information fusion is a suitable methodology for fault diagnosis. In this paper, the model and methodology of fault diagnosis with information fusion based on the depiction of the basic principle of information fusion and fault diagnosis and the relationship of them is expounded. In this model, the acquisition , processing and allocation of information , the partial and the all-round diagnosis is included. The wavelet is used in the processing of message; in the allocation of message, the general fuzzy entropy is regard as the evaluation criterion and the correlation analysis is used to the feature selection and allocation; the partial diagnosis is realized by neural network; the evidence theory and fuzzy integral is used to decision in decision fusion net, in this section, this paper give a texture methodology of basic probability and fuzzy integral. Compare with other methodology, this methodology enter upon from the structure of the neuron of capita in theory, so this methodology is closed to the acquaintance course of people. By the fusion of fault symptom, higher accurate and more reliable diagnosis result can be obtain and the lose of information is lesser than others. The evidence theory and fuzzy integral can abate the incertitude and improve the accuracy of decision.The methodology that is given by this paper is used to the fault diagnosis of automobile motor. The result of emulation declaration that the result is according to the matter of factor, which shows this methodology, is workable and correct. This methodology can also be used in others complication system.This paper design a system to diagnose the fault of motor based on the research of theory and discuss the way to achieve the fault diagnosis online. The work that in this paper is the research of basic theory, it must be monitor , modify and integrated if it want to be used in practice.更多还原