基于经验模态分解和遗传神经网络的轨道车辆轴承故障诊断研究

发布时间：2018-05-24 15:08

本文选题：轨道交通车辆 + 轴承故障诊断　；参考：《南京理工大学》2013年硕士论文

【摘要】：滚动轴承作为轨道车辆走行系的最重要部件之一,其运行状态对于保障车辆的行驶安全具有重要意义,因此对轨道车辆滚动轴承的故障进行准确、高效的诊断是一个亟需解决的问题。本文在总结和吸取前人研究成果的基础上,提出将经验模态分解与遗传算法优化的RBF神经网络相结合实现轨道车辆滚动轴承的故障诊断。首先,介绍了滚动轴承故障诊断的机理、故障形式及产生原因和振动模型,并讨论了故障特征信息提取的常用方法和各自的特点。重点对小波包分析和EMD方法进行了实例仿真,证明了小波包分析和EMD可以有效用于轴承故障特征信息的提取。其次,在故障模式识别上,采用基于神经网络的故障模式识别方法,选择BP和RBF两种典型神经网络,结合小波包和EMD两种特征提取方法,分别建立了小波包-BP、小波包-RBF、EMD-BP、EMD-RBF四种故障诊断模型。利用Benchmark数据对各模型进行仿真实验,结果证明EMD在故障特征提取上相比小波包有优势,RBF神经网络比BP神经网络有更好的故障识别性能。再次,提出采用遗传算法优化RBF神经网络参数,进一步提升RBF神经网络的故障识别性能,结合小波包和EMD分别建立小波包-GA-RBF和EMD-GA-RBF两种故障诊断模型。利用Benchmark数据对各模型进行仿真实验,结果证明基于遗传算法优化的RBF神经网络在故障识别精度上有了很大提高。最后,以上述研究分析为基础,采用小波包-GA-RBF和EMD-GA-RBF两种轴承故障诊断模型,利用实测的轨道车辆滚动轴承故障数据进行仿真实验,结果证明本文提出的基于EMD分解结合遗传算法优化的RBF神经网络可以用于轨道车辆的轴承故障诊断。
[Abstract]:Rolling bearing is one of the most important parts of rail vehicle running system. Its running state is of great significance to ensure the safety of vehicle, so the fault of rolling bearing of rail vehicle is accurate. Efficient diagnosis is an urgent problem to be solved. On the basis of summing up and absorbing the previous research results, this paper proposes to realize the fault diagnosis of rolling bearing of rail vehicle by combining empirical mode decomposition with RBF neural network optimized by genetic algorithm. Firstly, the fault diagnosis mechanism, fault form, causes and vibration model of rolling bearing are introduced, and the common methods of fault feature information extraction and their respective characteristics are discussed. The simulation results of wavelet packet analysis and EMD method show that wavelet packet analysis and EMD can be used to extract bearing fault feature information effectively. Secondly, in fault pattern recognition, the neural network-based fault pattern recognition method is adopted, two typical neural networks BP and RBF are selected, and two feature extraction methods, wavelet packet and EMD, are combined. Four kinds of fault diagnosis models, wavelet packet -BP and wavelet packet RBFN EMD-BPU EMD-RBF, are established respectively. The simulation results of each model using Benchmark data show that EMD has better fault identification performance than wavelet packet in fault feature extraction. Thirdly, genetic algorithm is used to optimize the parameters of RBF neural network to further improve the fault identification performance of RBF neural network. Combined with wavelet packet and EMD, two fault diagnosis models of wavelet packet -GA-RBF and EMD-GA-RBF are established respectively. The simulation results of each model based on Benchmark data show that the RBF neural network based on genetic algorithm has greatly improved the accuracy of fault identification. Finally, based on the above research and analysis, two kinds of bearing fault diagnosis models, wavelet packet -GA-RBF and EMD-GA-RBF, are used to simulate the rolling bearing fault data of rail vehicle. The results show that the proposed RBF neural network based on EMD decomposition and genetic algorithm can be used for bearing fault diagnosis of rail vehicles.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：U279.3;TH165.3;TP183

【参考文献】