基于DWAE-GRUNN算法的齿轮箱早期故障智能诊断研究  

Fault diagnosis of gearbox based on DWAE and GRUNN

作　　者：张力丹[1] 袁晓华 李峰 张国强 ZHANG Lidan;YUAN Xiaohua;LI Feng;ZHANG Guoqiang(School of Computer Engineering,Shangqiu University,Shangqiu 476000,Henan,China;Basic Science Department,Ordos Vocational College,Ordos 017010,I nner Mongolia,China;Department of Mechanical Engineering,Henan Polytechnic University,J iaozuo 454000,Henan,China;Department of Design,Henan Shenzhou Precision Manufacturing Co.,Ltd.,Zhengzhou 450064,Henan,China)

机构地区：[1]商丘学院计算机工程学院,河南商丘476000 [2]鄂尔多斯职业学院基础部,内蒙古鄂尔多斯017010 [3]河南理工大学机械工程系,河南焦作454000 [4]河南神州精工制造股份有限公司设计部,河南郑州450064

出　　处：《中国工程机械学报》2023年第1期85-89,94,共6页Chinese Journal of Construction Machinery

基　　金：河南省高等职业学校青年骨干教师培养计划资助项目(2019GZGG034)。

摘　　要：为了提高在时变转速条件下对齿轮箱故障进行识别的能力,综合运用门控循环单元神经网络(GRUNN)与深度小波自动编码器(DWAE),开发了一种齿轮箱故障诊断模型。通过Adam优化算法与Dropout处理技术对模型进行了训练,采用经过训练处理的模型并通过Softmax分类器实现对样本齿轮箱运行状态的识别。研究结果表明:运用DWAE和GRUNN模型获得了良好的诊断效果,对齿根裂纹、齿面磨损、断齿的故障识别准确率都达到96%以上。本模型的齿轮箱故障识别具备DWAE鲁棒特征提取能力,表现出比其他方法更高的待诊样本准确率。在逐渐增加训练样本数量的过程中,获得了更高的待诊样本准确率。该方法具有较强的抗噪能力和时变转速适应能力,易于在同类机械传动设备的故障诊断领域应用推广。In order to improve the ability of gearbox fault identification under the condition of timevarying speed,a gated recurrent unit neural network(GRUNN) and deep wavelet automatic encoder(DWAE) fault diagnosis model was developed. The model was trained using the Adam optimization algorithm and Dropout processing technology. The trained model and Softmax classifier were used to identify the running state of the sample gearbox. The results show that DWAE and GRUNN models have obtained excellent diagnosis effect,and the fault identification accuracy of tooth root crack,tooth surface wear and broken teeth is above 96%. The gearbox fault identification model has the ability of DWAE robust feature extraction,showing higher sample accuracy than other methods. In the process of gradually increasing the number of training samples,a higher accuracy of waiting samples was obtained. This method has strong anti-noise ability and time varying speed adaptability,which is easy to be applied in the fault diagnosis field of similar mechanical transmission equipment.

关 键 词：齿轮箱 故障识别 小波 神经网络 准确率 

分 类 号：TH137[机械工程—机械制造及自动化]