中文摘要：

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.

英文摘要：

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.