摘要
在某一确定荷载作用下,结构不同部位的失效概率是不同的。为此提出了基于失效树理论进行人工神经网络样本采集的方法。在此基础上,采用人工神经网络和遗传算法进行结构动力损伤定位和损伤程度的估计。算例分析表明:基于失效树理论可以有效地减少训练样本的大小,从而有效的进行结构多损伤情况下的损伤检测分析。
Under a specified loading, the failure probability of different element is distinct. So we can get the training sample from the least reliable elements, based on the failure tree of the structures. Then we can apply the ANN and GA to identify the location and the severity prediction of damage. The numerical results of several examples show that failure tree-based sample is feasible for multiple damage detection of large-scale structures.
出处
《金属世界》
2009年第C00期34-38,共5页
Metal World
关键词
失效树
人工神经网络
遗传算法
损伤检测
多位置损伤
failure tree
artificial neural networks
genetic algorithms
damage detection
multiple damage
