基于改进受限玻尔兹曼机的滚动轴承健康因子构建方法  

Construction method of rolling bearing health indicator based on enhanced restricted Boltzmann machine

作　　者：孙世岩[1] 张钢 梁伟阁[1] 佘博[1] 田福庆[1] SUN Shiyan;ZHANG Gang;LIANG Weige;SHE Bo;TIAN Fuqing(College of Weaponry Engineering,Naval University of Engineering,Wuhan 430033,China)

机构地区：[1]海军工程大学兵器工程学院,湖北武汉430033

出　　处：《系统工程与电子技术》2023年第9期2979-2985,共7页Systems Engineering and Electronics

基　　金：国家自然科学基金(61640308);海军工程大学自然科学基金(20161579)资助课题。

摘　　要：针对传统方法构建的健康因子各类性能指标不高、信息冗余的问题,提出一种基于改进受限玻尔兹曼机(restricted Boltzmann machine,RBM)的滚动轴承健康因子构建方法。首先,提取滚动轴承振动监测信号时域、频域特征组成物理健康因子集。其次,将RBM隐藏层节点数随时间变化斜率引入到正则化项中,提取物理健康因子集中的趋势性特征。最后,利用滚动轴承全寿命周期试验验证所提方法的有效性。实验结果表明,相对于主成分分析(principal component analysis,PCA)法、传统RBM虚拟健康因子构建方法,基于改进RBM构建的虚拟健康因子单调性分别提高178.0%和33.3%,趋势性分别提高126.8%和16%,鲁棒性分别提高60%和6.02%。Aiming at the problems of low performance indicators and information redundancy of health factors constructed by traditional methods,a method for constructing health factors of rolling bearings based on the improved restricted Boltzmann machine(RBM)is proposed.Firstly,the time-domain and frequency-domain features of rolling bearing vibration monitoring signals are extracted to form a physical health factor set.Secondly,the slope of the number of nodes in the hidden layer of the RBM with time is introduced into the regularization term to extract the trend characteristics of the physical health factor set.Finally,the effectiveness of the proposed method is verified through full life cycle testing of rolling bearings.The experimental results show that compared with the principal component analysis(PCA)method and the traditional RBM virtual health factor construction method,the monotonicity of the virtual health factor based on the improved RBM is increased by 178.0%and 33.3%respectively,the trend is increased by 126.8%and 16%respectively,and the robustness is improved by 60%and 6.02%respectively.

关 键 词：滚动轴承 健康因子 受限玻尔兹曼机 正则化 评估准则 

分 类 号：TN911.72[电子电信—通信与信息系统]