基于MIV特征筛选和BP神经网络的滚动轴承故障诊断技术研究

发布时间：2019-07-08 17:36

【摘要】：滚动轴承是旋转机械中比较常见的而且易损坏的部件,轴承的好坏对机器的工作状况影响很大。它的运行状态直接影响整台机器的性能,一旦发生故障就成为引起机械设备失效的重要原因,据统计旋转机械的故障有30%是由轴承引起的。因此,对滚动轴承的状态监测和故障诊断是机械设备故障诊断的重要研究内容。在设备故障诊断中,构成故障的特征空间是非常复杂的非线性关系,很难建立数学模型。人工神经网络能够映射任意复杂的非线性关系,具有自学习、自组织和自适应等特征,被广泛地应用于机械故障诊断中。本文提出了MIV算法,时/频域分析法和BP神经网络算法对滚动轴承做了故障诊断。主要研究了以下内容：首先,介绍了滚动轴承的结构特征以及产生的故障原因,以及人工神经网络的基本知识,选取BP神经网络做滚动轴承故障诊断的原因。设计了滚动轴承的信号采集系统,采集并处理了正常轴承、内圈故障轴承、外圈故障轴承、滚动体故障轴承的振动数据,作为基于BP神经网络的滚动轴承故障诊断实验仿真的数据样本。其次,结合BP神经网络和MIV算法对常用的时域特征参数和频域特征参数做了特征筛选,根据仿真结果即MIV值的大小,得到8个敏感故障特征参量,即他们更能反映故障特征。之后选用时域、频域分析法,分别分析这8个特征参量的对故障特征的敏感程度,得到方差、均方根值、峰值、裕度因子、总功率谱和以及峭度系数对故障最为敏感,并确定为BP神经网络的输入。再次,设计了对滚动轴承做故障诊断的BP神经网络的结构。确定了BP神经网络的结构参数,分析了BP神经网络的结构参数的设置对网络性能的影响。主要包括：激活函数、学习算法、数据预处理方法、网络初始权值以及期望误差的选取,输入神经元、隐层节点数、网络学习率和训练系数对网络性能的影响情况。最后,利用前文的实测的滚动轴承的数据,以及设计好的BP神经网络故障诊断系统,对滚动轴承做了故障诊断仿真实验。确定了训练样本和预测样本,完成了对轴承的故障诊断和分类,并求取了分类误差率。仿真研究结果表明了诊断系统的有效性,验证了BP神经网络对滚动轴承做故障诊断的可行性,诊断效果良好。
文内图片：

图片说明：传感器
[Abstract]:The rolling bearing is a common and easy-to-damage component in the rotating machinery, and the quality of the bearing has a great influence on the working condition of the machine. The operation state of the machine directly affects the performance of the whole machine, and once a failure has become an important cause of the failure of the mechanical equipment, according to the statistics,30% of the fault of the rotating machine is caused by the bearing. Therefore, the state monitoring and fault diagnosis of rolling bearing is an important research content of the fault diagnosis of mechanical equipment. In the fault diagnosis of equipment, the characteristic space that constitutes the fault is a very complex non-linear relationship, and it is difficult to establish a mathematical model. The artificial neural network can map any complex non-linear relationship, and has the characteristics of self-learning, self-organization and self-adaptation, and is widely used in the mechanical fault diagnosis. In this paper, an MIV algorithm, a time/ frequency domain analysis method and a BP neural network algorithm are proposed for fault diagnosis of rolling bearing. The following contents are mainly studied: Firstly, the structural characteristics of rolling bearing, the cause of the failure and the basic knowledge of the artificial neural network are introduced, and the original BP neural network is selected as the original of the fault diagnosis of rolling bearing. The signal acquisition system of the rolling bearing is designed, and the vibration data of the normal bearing, the inner ring fault bearing, the outer ring fault bearing and the rolling body fault bearing are collected and processed, as the data sample of the rolling bearing fault diagnosis experiment simulation based on the BP neural network. Secondly, combining with the BP neural network and the MIV algorithm, the time-domain characteristic parameters and the frequency-domain characteristic parameters are screened, and 8 sensitive fault characteristic parameters are obtained according to the simulation result, namely, the magnitude of the MIV value, that is, they can be more reflected. and then selecting a time domain and a frequency domain analysis method to analyze the sensitivity degree of the eight characteristic parameters to the fault characteristic respectively, obtaining the variance, the mean square root value, the peak value, the margin factor, the total power spectrum and the kurtosis coefficient to be most sensitive to the fault, and determining the BP neural network Input. Again, the BP neural network for fault diagnosis of the rolling bearing is designed. The structure of the network is determined. The structure parameters of the BP neural network are determined, and the structure parameters of the BP neural network are analyzed. It mainly includes: activation function, learning algorithm, data pre-processing method, network initial weight and expected error selection, input neuron, hidden layer node number, network learning rate and training coefficient to network performance and finally, using the measured data of the rolling bearing and the designed BP neural network fault diagnosis system, the rolling bearing is The simulation experiment of the obstacle diagnosis is carried out. The training samples and the prediction samples are determined, the fault diagnosis and classification of the bearing are completed, The classification error rate is obtained. The simulation results show the effectiveness of the diagnostic system, and verify the feasibility of the BP neural network for the fault diagnosis of the rolling bearing.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH133.33;TH165.3

【相似文献】