摘要
基于一致性数据融合算法和"能量-损伤"结构特征提取原理,提出一种结构多损伤模式识别方法,能够充分利用结构不同位置的不同状态信息,更好地对结构损伤做出判断。对一致性数据融合算法进行改进,使得改进后的算法能够克服一致性算法中两传感器在测量精度不同时置信距离不同的缺点,并对支持矩阵进行模糊化处理,避免人为定义阈值而产生的主观误差。利用"能量-损伤"特征提取技术,构造融合后多传感器测量数据的特征向量,以ART2神经网络作为模式识别工具进行结构多损伤识别。五层框架结构数值计算结果表明,提出的方法能够用于结构多损伤识别,且具有较强的鲁棒性、稳定性和适应性。
A structural multi-damages identification method was presented based on consensus data fusion,eigenvalue extraction by energy-damage and powerful pattern recognition function of ART2.The method can make full use of structural status information in different sites and recognize structural damages better.The traditional consensus data fusion algorithm was improved.The improved algorithm can overcome the shortcoming of the traditional consensus algorithm with two sensors,which has different confidence distance for different measuring precision.The supporting matrix was fuzzified,which can avoid the subjective error in determining the threshold value.Eigenvectors of multi-sensor measured data after fusion were constructed using eigenvalue extraction technology of energy-damage.ART2 network was adopted as the pattern recognition tool to identify structural multi-damages.The numerical simulation results of a five-layers frame structure show that the method can identify structural multi-damages and it is more robust,stable and adaptive.
出处
《振动与冲击》
EI
CSCD
北大核心
2010年第8期120-123,共4页
Journal of Vibration and Shock
基金
国家自然科学基金(50708013)
关键词
数据融合
一致性算法
特征提取
模式识别
data fusion
consensus algorithm
feature extraction
pattern recognition