概率神经网络在车辆齿轮箱典型故障诊断中的应用

Application of Probabilistic Neural Network to Typical Fault Diagnosis of Vehicle Gearbox

为提高汽车齿轮箱典型故障的诊断效率和准确性,提出一种基于概率神经网络的齿轮箱故障诊断方法。通过对某型齿轮箱的实验采集齿轮箱在正常状态、齿根裂纹和断齿状态下的振动信号,经过数据处理得到样本数据后输入概率神经网络模型,通过交叉验证并与BP神经网络对比的结果表明:概率神经网络能准确地识别出齿轮箱典型故障,且与BP神经网络相比,诊断准确率更高、诊断速度更快。

In order to enhance the diagnosis efficiency and accuracy of the typical faults of automotive gearbox,a fault diagnosis method of gearbox based on probabilistic neural network(PNN)is proposed.The vibration signals of the gearbox under normal state,tooth root cracks and broken tooth are collected through an experiment on a gearbox.After data processing,the sample data are obtained and input into PNN model.The results of cross-validation and comparison with the BP neural network(BPNN)show that PNN can accurately identify the typical faults of the gearbox and has higher diagnostic accuracy and faster diagnostic speed compared with BPNN.