基于核函数的SOM网络在齿轮箱状态监测中的应用  

APPLICATION OF KERNEL SELF-ORGANIZING MAPS IN GEARBOX CONDITION MONITORING

作　　者：廖广兰[1] 史铁林[1] 张桂才[1] 

机构地区：[1]华中科技大学机械科学与工程学院机械电子信息工程系,武汉430074

出　　处：《机械强度》2003年第4期369-372,共4页Journal of Mechanical Strength

基　　金：国家重大基础研究项目基金资助 (G1 9980 2 0 32 0 ) ;湖北省自然科学基金资助 (2 0 0 0J1 2 5)~~

摘　　要：基于支持向量机用内积运算实现非线性变换的思想 ,提出基于核函数的自组织映射方法。该方法借助于核函数在原输入空间构造非线性映像空间的自组织竞争评价函数及权值调整方法 ,从而更好地解决输入向量的聚类问题。研究该方法在齿轮箱状态监测中的应用 ,分析表明 ,它可以清楚地将齿轮箱不同状态区分开 ,并且特征数据映象点在网络输出层的轨迹变化趋势直观反映齿轮箱工作状态的变化 ,便于及时监测到齿轮箱早期故障及变化趋势 ,与标准self or ganizingmaps(SOM)相比 ,该方法性能更稳定 。Based on the idea of realizing non-linear transformation with inner product in support vector machine, a new technique, kernel self-organizing maps (KSOM), is presented. By use of the kernel function in the input space, which constructs self-organizing competitive discrimination criterions and adjusts neurons weights in the mapped high-dimensional space, the KSOM method clusters the input data more efficiently. The application of the proposed method in industrial gearbox condition monitoring is studied. The analysis results show that using the method the gearbox-operating condition with fatigue crack or broken tooth compared with the normal condition is identified clearly. Moreover, with the trajectory of the image points for the feature data in the output-layer of the network, the variation of gearbox conditions is observed visually, and the development of gearbox early-stage failures is monitored in time. Compared with the standard SOM, the proposed method performs better.

关 键 词：齿轮箱 状态监测 自组织映射 核函数 评价标准 SOM网络 

分 类 号：TH132.41[机械工程—机械制造及自动化] TH17