摘要
将神经网络集成引入到结构损伤识别领域中,并利用灰色聚类技术对获得的全部个体神经网络模型进行聚类,将得到的差异较大的部分神经网络进行集成,以提高神经网络间的差异性和增强网络的泛化能力。损伤识别实验结果表明,基于灰色聚类的神经网络集成方法不仅可行,而且其损伤识别效果优于传统的神经网络模型。
The neural network (NN) ensemble was adopted into the field of the structual damage detection and the grey clustering(GC) technique was employed to cluster all individual NN models and integrate the part of the NNs obtained with bigger differences to enhance the differences and the generalization ability of the NN. The experiment results show that the GC based NN ensemble method is feasible and superior to the traditional NN models.
出处
《吉林大学学报(工学版)》
EI
CAS
CSCD
北大核心
2007年第2期438-441,共4页
Journal of Jilin University:Engineering and Technology Edition
基金
高等学校博士学科点专项科研基金资助项目(20030183025)
关键词
计算机应用
神经网络集成
结构损伤识别
灰色聚类
泛化能力
computer application
neural network ensemble
structural damage detection
greyclustering
generalization ability