摘要
为了更好地识别噪声与时变转速条件对变工况齿轮箱的故障,开发了一种通过深度小波自动编码器(DWAE)与门控循环单元神经网络(GRUNN)相结合的变工况齿轮箱故障识别方法,其能够从含噪样本自主提取得到鲁棒故障特征;通过Adam与Dropout方法进行训练,通过Softmax分类器对待诊样本的变工况齿轮箱运行状态进行了准确识别。研究结果表明,采用该模型识别齿轮故障时,能够达到有效分离齿轮的6种故障状态,从而满足齿轮状态聚类的优化功能;该模型能够提取出DWAE的鲁棒特征参数,也可以发挥GRUNN以实现消除梯度的效果。当训练样本数增加,待诊样本的准确率也发生了明显提升。样本数超过200后,测试待诊样本可获得稳定准确率,通过DWAEGRUNN方法识别得到的准确率最高。针对变转速工况,该模型可以保持很好的准确率。
In order to better identify the faults of gearbox under varying working conditions caused by noise and time-varying speed conditions,a fault identification method of gearbox under varying conditions is developed by combining DWAE and GRUNN,which could extract robust fault characteristics from noisy samples.The Adam and Dropout methods are used for training,and the softmax classifier is used to accurately identify the operating state of the gearbox under different working conditions.The results show that the six fault states of gears can be separated effectively when the model is used to identify gear faults,so as to meet the optimization function of gear state clustering.This model can extract the robust characteristic parameters of DWAE,and it can also use GRUNN to eliminate the gradient.When the number of training samples increased,the accuracy of waiting samples also improved significantly.When the number of samples is more than 200,stable accuracy could be obtained by testing the waiting samples,and the accuracy is the highest by dWAe-Grunn method.The model can maintain good accuracy under variable speed working conditions.
作者
刘传慧
陈晓静
侯晓晓
朱强
董勇
Liu Chuanhui;Chen Xiaojing;Hou Xiaoxiao;Zhu Qiang;Dong Yong(School of Automotive Engineering,Yellow River Transportation Institute,Jiaozuo 454950,China;School of Mechanical Engineering,Henan Polytechnic University,Jiaozuo 454003,China;Henan Dongtai Gear Co.,Ltd.,Xinxiang 453400,China)
出处
《机械传动》
北大核心
2022年第2期155-159,共5页
Journal of Mechanical Transmission
基金
河南省教育厅科技研究课题(ZD2018020)。
关键词
变工况齿轮箱
故障识别
深度小波自动编码器
门控循环单元神经网络
准确率
Variable working condition gearbox
Fault identification
Deep wavelet autoencoder(DWAE)
Gated cyclic unit neural network(GRUNN)
Accuracy
作者简介
刘传慧(1967-),女,河南滑县人,硕士,高级工程师,主要从事于机械设计与机械制造。
