LTSA和深度置信网络的行星齿轮箱故障诊断  

Planetary Gearbox Fault Diagnosis for LTSA and Deep Belief Network

作　　者：王建国[1] 刘冀韬 WANG Jian-guo;LIU Ji-tao(School of Mechanical Engineering,Inner Mongolia University of Science&Technology,Inner Mongolia Baotou 014000,China)

机构地区：[1]内蒙古科技大学机械工程学院,内蒙古包头014000

出　　处：《机械设计与制造》2022年第1期5-8,共4页Machinery Design & Manufacture

基　　金：国家自然科学基金(51865045)。

摘　　要：针对行星齿轮箱振动信号维度高,传统故障诊断方法识别精度低的问题,提出一种基于局部切空间排列算法(Local Tangent Space Alignment,LTSA)和深度置信网络(Deep Belief Network,DBN)的行星齿轮箱故障诊断方法。首先,利用PCA算法预估高维数据的内在维度,确定目标数据的内在维数;其次,根据目标数据的内在维数结合LTSA算法对高维数据集进行约简,并划分测试集和训练集;最后,利用训练集训练DBN模型参数,获得行星齿轮箱故障辨识模型,并将测试集输入辨识模型实现行星齿轮箱故障辨识。实验结果表明,所提方法实现高维数据降维的同时,也提升了智能诊断模型的分类精度。Aiming at the problem that the planetary gearbox vibration signal has high dimension and the traditional fault diagnosis method has low recognition accuracy,a planetary gearbox fault diagnosis method based on local tangent space array algorithm LTSA and deep belief network is proposed.First,the PCA algorithm is used to estimate the intrinsic dimension of high-dimensional data,and the intrinsic dimension of the target data is determined;then,according to the intrinsic dimension of the target data,the LTSA algorithm is used to reduce the high-dimensional data set and divide the test set and the training set;Finally,using the training set to train the DBN model parameters,the planetary gearbox fault identification model is obtained,and the test set input identification model is used to realize the planetary gearbox fault identification.The experimental results show that the proposed method achieves high dimensional data dimensionality reduction and improves the classification accuracy of the intelligent diagnosis model.

关 键 词：行星齿轮箱 深度置信网络 局部切空间排列 状态辨识 故障诊断 

分 类 号：TH16[机械工程—机械制造及自动化] TH132.425