【作者】 李常镱；

【导师】 潘宏侠；

【作者基本信息】 中北大学 ， 车辆工程， 2012， 硕士

【摘要】 发动机作为典型的往复式机械设备，在整个机械系统中占有重要的地位，然而其结构的复杂性使得发动机的故障呈现出多样性，另外由于工作环境噪声、信息采集系统精度、数据处理和故障诊断方法的影响，使得发动机故障诊断中不确定性尤为突出，因此解决不确定性是目前国内外相关学者的主要研究内容。以贝叶斯理论为基础的贝叶斯网络通过实践积累可以随时进行在线学习以改进网络结构和参数，提高故障诊断能力，并且基于网络结构的概率推理算法，可以随时更新网络中的概率信息。可见贝叶斯网络非常适合解决复杂的、不确定性的问题。本文首先阐述了本课题国内外的研究现状，以及贝叶斯网络解决发动机故障诊断的可行性，其次重点研究了贝叶斯网络的建立，并将粒子群优化算法应用于贝叶斯网络参数优化上，给出粒子群优化的贝叶斯网络结构。通过对柴油机进行试验得到振动信号，然后用小波能量谱对振动信号进行特征值的提取。用训练样本训练贝叶斯网络，用测试样本对贝叶斯网络的训练结果进行测试。得到贝叶斯网络对柴油机故障诊断的结论，对比结论表明贝叶斯网络能够很好的解决柴油机故障诊断中的不确定性。且发现此方法可以对发动机多故障并存的故障类型进行故障诊断，这也为以后学者研究多故障并存提供了方向。更多还原

【Abstract】 Engine as a typical reciprocating machinery and equipment, mechanical systems occupyan important position, but the complexity of its structure showing a diversity of the enginefailure, as a result of environmental noise, the accuracy of the information collection system,data processing and fault diagnosiseffects, resulting in engine failure diagnostic uncertainty inparticularly prominent, so to solve the uncertainty is the main contents of the current domesticand foreign scholars.Bayesian network based on Bayesian theory practice has accumulated at any time, onlinelearning to improve the network structure and parameters, to improve the fault diagnosiscapability, and probabilistic inference algorithm based on network structure, you can alwaysupdate the probability of information in the network. Be seen, the Bayesian network is verysuitable to solve the problem of complexity and uncertainty.This paper first describes the current situation of this subject at home and abroad, as wellas the feasibility of Bayesian networks to solve the engine fault diagnosis, followed by focuson the establishment of the Bayesian network, and particle swarm optimization algorithm isapplied to the Bayesian network parameter optimization, particle swarm optimization of theBayesian network structure.. We can get the vibration signal by diesel engine test, extract the eigenvalue of thevibration signal by Wavelet power spectrum. Training samples used to train Bayesiannetworks, and test results of the test samples on the training of the Bayesian network. Theconclusion of the Bayesian network for fault diagnosis of diesel engine, the contrastconcluded that the Bayesian network can be a good diesel engine fault diagnosis ofuncertainty. And found that this method can be diagnosed by the coexistence of multi-faulttype of engine failure, it also provides a direction later scholars of multiple faults coexist.更多还原