兰州交通大学毕业设计（论文）- I -摘 要滚动轴承是机械设备中最常见、应用最广泛的零部件之一，其运行状态对整个设备的工作状态、生产过程都有直接影响因此对轴承的故障诊断具有非常重要的意义本文以机械设备滚动轴承故障诊断问题为背景，针对传统的时频分析方法难以全面反映故障信息的缺陷，探讨了 BP(Back Propagation,反向传播)神经网络技术在滚动轴承故障诊断中的应用选取滚动轴承三种故障类型（内圈故障、外圈故障、滚动体故障）下的轴承振动数据，经小波包三层分解后得到 8 组能量特征值，作为人工神经网络的输入层的输入，然后根据神经网络的原理，设置 BP 神经网络隐含层、输出层的相关参数，设计完成神经网络的结构模型最后在 Matlab 软件平台上对所构建的网络进行训练，得到训练误差曲线，再在训练完成的神经网络上进行测试和仿真，得出仿真结果正确率通过一系列的训练、测试和仿真可以看出，本文构建的 BP 神经网络结合对隐含层神经元参数的不同设置，得到不同的训练误差曲线，均具有良好的收敛性，在测试、诊断过程中，能够根据输入值快速、准确地识别出滚动轴承的故障类型，且具有较高的正确率与传统方法相比，将 BP 神经网络应用到滚动轴承的故障诊断问题中，具有全面、快速、准确等特点，能够更全面的体现轴承的故障信息，具有显著的优越性。
关键词：滚动轴承；故障诊断；关键词：滚动轴承；故障诊断；BPBP 神经网络；能量特征值神经网络；能量特征值兰州交通大学毕业设计（论文）- II -AbstractThe rolling bearing is one of the most common and widely used components in the mechanical equipment. Its operating state has a direct impact on the entire working status of equipment and the production process. Therefore, the monitoring and diagnosis of the rolling bearing has a very important significance. The bearing fault diagnosis technology is often based on time-frequency analysis. These methods are restricted in many ways, which causes a lot of state detecting missed. This paper is based on the research of the rolling bearing fault diagnosis of the mechanical equipment, and focus on the BP neural network technology application in the problem. The rolling bearing vibration data of three fault patterns (inner-race fault, out-race fault and rolling element fault) are chosen in this paper, and it is adopted that taking eight energy components decomposed by wavelet packet as the ANN (artificial neural network) input vector. Then, according to the ANN theory, set hidden layer and output layer parameters of the BP neural network and design the structure of the neural network model for rolling bearing fault diagnosis. At last, train the network on Matlab and get the training error curve, then test and simulate the network and calculate the correct rate of the simulation results. Through a series of training, testing and simulation process, it can be seen that the BP neural network method, which is applied to the rolling bearing fault diagnosis, can get different training error curves, combined with different set of parameters of the neurons in the hidden layer. All the curves have good convergence. In the test and diagnostic procedures, the network can identify different fault patterns quickly and accurately depending on the input data, at the same time it has a higher accuracy rate. The BP neural network method is more accurate, practical and it has a higher diagnostic accuracy rate compared with ordinary methods. So it surely has broad application prospects.Key Words：：Rolling bear, Fault diagnosis, BP neural network, Energy components兰州交通大学毕业设计（论文）- III -目 录摘 要.....................................................................................................................................IAbstract ......................................................................................................................................II目 录...................................................................................................................................III1 绪论.........................................................................................................................................11.1 论文背景与意义..........................................................................................................11.2 论文研究现状..............................................................................................................11.3 论文的研究内容与目标..............................................................................................12 滚动轴承故障特征.................................................................................................................22.1 滚动轴承的基本结构..................................................................................................22.2 滚动轴承的失效形式和故障类型..............................................................................23 BP 神经网络...........................................................................................................................33.1 人工神经网络概述......................................................................................................33.2 BP 神经网络................................................................................................................33.3 BP 算法的缺陷及其改进算法....................................................................................43.3.1 BP 算法的缺陷.................................................................................................43.3.2 BP 算法的改进算法.........................................................................................44 基于 BP 神经网络的滚动轴承故障诊断..............................................................................64.1 小波包分解与故障数据筛选......................................................................................64.1.1 小波包分解方法......................................................................。