基于GoogLeNet的多级液压缸故障诊断方法  

Fault Diagnosis Method of Hydraulic Telescopic Cylinder Based on GoogLeNet

作　　者：曹秋涵 赵俊鹏 石健 薛子良[3] 吴鹏辉 曹向荣 CAO Qiu-han;ZHAO Jun-peng;SHI Jian;XUE Zi-liang;WU Peng-hui;CAO Xiang-rong(Tianmushan Laboratory,Hangzhou,Zhejiang 310023;Beijing Institute of Space Launch Technology,Beijing 100076;School of Automation Science and Electrical Engineering,Beihang University,Beijing 100191)

机构地区：[1]天目山实验室,浙江杭州310023 [2]北京航天发射技术研究所,北京100076 [3]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《液压与气动》2024年第6期33-43,共11页Chinese Hydraulics & Pneumatics

基　　金：国家自然科学基金青年科学基金(52202456)。

摘　　要：针对多级缸故障模式复杂且难以实现精准诊断的问题,提出一种基于GoogLeNet神经网络的多级缸故障诊断方法。以多级缸伸出动作时的工作原理为出发点,通过对其建立动力学模型,搭建包含多种故障模式的仿真模型,获得不同状态下的多级缸故障信号。提取关键故障特征,并采用GoogLeNet神经网络构建故障诊断模型,实现多级缸故障诊断与故障定位。仿真和试验结果表明,所建立的多级缸仿真模型与实际相契合,且据此提出的故障诊断方法能够精准识别多级缸的不同故障模式,从而为多级缸维护维修工作的开展提供重要依据。Aiming at the problem that the fault modes of hydraulic telescopic cylinder are complicated and it is difficult to realize accurate diagnosis,a fault diagnosis method is proposed based on GoogLeNet neural network for hydraulic telescopic cylinder.The working principle of hydraulic telescopic cylinder during the stretching process is taken as starting point to establish its dynamic model.Then,a simulation model containing multiple fault modes is constructed by which fault signals are obtained in different states.On this basis,some key fault features of hydraulic telescopic cylinder are extracted,and the GoogLeNet neural network is applied to establish a fault diagnosis model to realize fault diagnosis and fault localization.The simulation and experimental results show that the simulation model of hydraulic telescopic cylinder is compatible with the actual situation.In addition,the proposed fault diagnosis method is effective in accurately identifying different fault modes of hydraulic telescopic cylinder,thus providing an important guidance for the maintenance and repair.

关 键 词：多级液压缸 升降系统 GoogLeNet 故障诊断 

分 类 号：TH137[机械工程—机械制造及自动化]