轧机伺服液压缸内泄漏故障诊断研究  

Study on Leakage Fault Diagnosis in Servo Hydraulic Cylinder of Rolling Mill

作　　者：刘琥铖 陈新元 杨哲 郭媛 LIU Hu-cheng;CHEN Xin-yuan;YANG Zhe;GUO Yuan(Key Laboratory of Metallurgical Equipment and Control Technology(Wuhan University of Science and Technology),Ministry of Education,Hubei Wuhan430081,China)

机构地区：[1]冶金装备及其控制教育部重点实验室(武汉科技大学),湖北武汉430081

出　　处：《机械设计与制造》2021年第1期41-44,共4页Machinery Design & Manufacture

基　　金：武汉科技大学冶金装备及其控制省部共建教育部重点实验室开放基金项目:高压液压缸内泄漏在线检测报警装置的设计与开发(2016B04)。

摘　　要：针对目前轧机伺服液压缸故障诊断过程中,故障特征提取困难,信号非线性变化,数据量大的问题,提出了一种基于深度置信网络的轧机伺服液压缸故障诊断的方法。根据轧机系统工作原理,建立轧机系统仿真模型,对轧机内泄漏故障状况进行模拟。利用深度置信网络在智能故障诊断的优越性,将信号归一化处理后放入深度置信网络进行训练,然后通过反向传播学习,优化网络各参数,提高诊断精度。深度置信网络模型由多层玻尔兹曼机以及顶层BP神经网络组成。与传统BP神经网络方法进行比较,结果表明,在训练样本数据足够的条件下,深度置信网络模型在伺服液压缸内泄漏故障诊断具有更高的诊断精度。A method of fault diagnosis of servo hydraulic cylinder of rolling mill based on deep confidence network is proposed to solve the problem of difficult fault feature extraction,non-linear signal change and large data volume in current fault diagnosis process of servo hydraulic cylinder of rolling mill.Based on the working principle of the rolling mill system,a simulation model of the rolling mill system is established to simulate the leakage failure in the rolling mill.Taking advantage of the advantages of deep confidence network in intelligent fault diagnosis,the signal normalization process is put into deep confidence network for training,and then the network parameters are optimized through reverse propagation learning to improve the diagnostic accuracy.The deep confidence network model is composed of the multi-layer boltzmann machine and the top layer BP neural network.Finally,compared with the traditional BP neural network method,and the results show that the deep confidence network model has higher diagnostic accuracy in the servo hydraulic cylinder leakage fault diagnosis under sufficient training sample data.

关 键 词：伺服液压缸 深度置信网络 系统建模 内泄漏故障诊断 仿真验证 

分 类 号：TH16[机械工程—机械制造及自动化] TH137