基于粗集的改进对向传播网络结构损伤识别  

Structural damage identification based on rough set and revised counter-propagation network

作　　者：姜绍飞[1] 林杰[1] 

机构地区：[1]福州大学土木工程学院,福州350108

出　　处：《振动与冲击》2011年第6期1-4,14,共5页Journal of Vibration and Shock

基　　金：国家自然科学基金(50408033;50878057);福建高校优秀人才计划(XSJRC2007-24);高等学校博士点基金(20093514110005)

摘　　要：为了有效地利用结构健康监测系统冗余、互补、不确定的信息进行健康状况评估,提出一种将粗集和改进对向传播神经网络(RCPN)有机地结合在一起的损伤识别新方法。它先用粗集进行数据处理以降低数据的不确定性和空间维数,然后用RCPN进行损伤识别。为了验证所提方法的有效性,对1个框架结构的单损伤和多损伤模式进行了识别,并重点研究了噪声、神经网络模型、不同数据处理方法的影响。研究发现,所提方法不仅可以降低数据的空间维数,减少神经网络的训练与检验时间,而且具有较好的损伤识别精度和鲁棒性。In order to make full use of redundant,complementary and uncertain information and thus assess the structural health states with a structural health monitoring system,a new damage identification method by integrating rough set with revised counter-propagation network（RCPN） was proposed.In the method,rough set was used to deal with data so as to reduce their uncertainties and spatial dimensions,the current CPN model was revised to improve the capabilities of processing uncertainties and carring on classification,and then the RCPN model was used for damage identification.To validate the proposed method,single-and multi-damage patterns of a frame were identified as an example,and some important factors,such as measurement noise,neural network models and data processing technologies,were investigated emphatically.The results show that the proposed method can effectively reduce the spatial dimension of data and is of preferable damage identification accuracy and robustness.

关 键 词：粗集 改进对向传播神经网络 损伤识别 属性约简 

分 类 号：TU317.5[建筑科学—结构工程] TP274[自动化与计算机技术—检测技术与自动化装置]