基于GHSOM网络的故障识别  

Fault recognition using growing hierarchical self-organizing map

作　　者：廖广兰[1] 史铁林[1] 刘世元[1] 轩建平[1] 

机构地区：[1]华中科技大学数字制造装备与技术国家重点实验室,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2008年第7期105-107,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家重点基础研究发展计划资助项目(2003CB716207);国家自然科学基金资助项目(50775091);教育部新世纪优秀人才支持计划资助项目(NCET-06-0639)

摘　　要：提出了一种基于生长型分级自组织映射(GHSOM)网络的故障识别方法,给出了方法的基本原理,并将该方法应用于汽轮机组与齿轮的故障数据分析.研究结果表明,GHSOM能根据数据特征无监督地对故障进行正确聚类和识别,并且具有动态增长及分层特性,能解析出数据内在的层次结构,实现由粗到精的聚类识别,该方法可以扩展应用于机械故障的诊断与识别.A novel technique based on growing hierarchical selforganizing map（GHSOM） for fault diagnosis and its basic principle were introduced.Experiments with clustering based on GHSOM were implemented on a turbine machine and gearbox data.The analysis results prove that the GHSOM model can cluster and recognize machine faults correctly with little prior experiences according to the characteristics inherent in the data.Furthermore,the model can dynamically grow architecture evolving into a hierarchical structure of self-organizing maps,resolve the hierarchical relationships in the data,and realize pattern recognition from roughness to detail.Therefore,the GHSOM has great potential for machine fault diagnosis and recognition in the future.

关 键 词：机械故障诊断 故障识别 生长型分级自组织映射 聚类 

分 类 号：TH17[机械工程—机械制造及自动化]