SOM神经网络在复合材料损伤监测中的应用  

Application of Self-organizing Feature Map Neural Network for Destructive Test of Composite Material

作　　者：赵进昌[1] 原思聪[1] 张满意[1] 刘道华[1] 耿素花[1] 

机构地区：[1]西安建筑科技大学机电工程学院,西安710055

出　　处：《煤矿机械》2009年第4期201-203,共3页Coal Mine Machinery

摘　　要：总结研究了自组织特征映射(SOM)神经网络的结构及学习算法,提出了利用SOM神经网络对输入样本的"聚类"作用及MATLAB神经网络工具箱来实现对故障模式的分类,通过U矩阵图对其分类结果进行仿真与分析的新方法。结果表明:该网络对复合材料损伤监测的诊断故障能够准确识别和分类,与一般可视化界面相比,此方法可视化界面更简单直观,故障识别率高,应用于材料无损检测是有效可行的。The construction and algorithm of a self - organizing ieature map neural network is researcneu. A new fault diagnosis method based on SUM neural network＇ s ＂clustering＂ effect on the importation of samples and MATLAB neural network toolbox, and classification and analysis of simulation results with U -matrix. The results show that this network model can be accurate identification and classification for fault diagnosis. By comparing the general visual method, the use of Unified Matrix is more simple ＆ intuitive. This method can accurately identify the fault and is feasible and effective in applying to the nonde- structive testing composite material damage

关 键 词：神经网络 U矩阵 复合材料 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]