基于一维AlexNet卷积神经网络的铁路货车轴承故障识别

Fault Identification of Railway Freight Train Bearings Based on One-dimensional AlexNet Convolution Neural Network

铁路货车是我国货物运输的一种重要交通工具,轮对及滚动轴承是铁路货车的关键部件之一,与其他普通轴承相比,铁路货车轴承由于结构复杂、承载力大、故障特征频率难于精确计算,一般基于信号处理的特征提取加分类器的传统智能诊断算法,对专家经验要求高。在此基于AlexNet神经网络,提出基于改进一维卷积神经网络智能诊断模型,通过引用改进的卷积核增强神经网络的非线性表达能力。以353130B铁路货车轴承为研究对象,使用实验室轴承实验机分别对正常轴承、外圈故障、内圈故障、滚子故障4种类型信号进行采集识别研究,研究结果显示该网络对轴承故障的识别率可达99%,具有优秀的故障识别特性,表明该模型具备良好的泛化价值,对铁路货车轴承故障识别检测有重要的研究价值。

Railway freight trains are an important means of transportation for freight transportation in China.Wheel sets and rolling bearings are one of the key components of railway freight trains.Compared with other ordinary bearings,railway freight trains bearings are difficult to calculate accurately due to their complex structure,large bearing capacity and fault feature frequency.Generally,the traditional intelligent diagnosis algorithm based on signal processing feature extraction and classifier requires high expert experience.Based on the AlexNet neural network,an intelligent diagnosis model based on the improved one-dimensional convolution neural network is proposed.The improved convolution kernel is used to enhance the nonlinear expression ability of the neural network.Taking the 353130B railway freight trains bearing as the research object,the laboratory bearing experimental machine was used to collect and identify four types of signals,namely,normal bearing,outer ring fault,inner ring fault and roller fault.The research results showed that the recognition rate of the network for bearing fault could reach 99%,and it has excellent fault identification characteristics,indicating that the model has good generalization value,and has important research value for railway freight trains bearing fault identification and detection.