粗集神经网络在建筑物震害预测中的应用

Application of Seismic Damage Prediction to Buildings Based on Rough Set and Neural Network

摘要:

将粗糙粗集理论和神经网络原理结合起来,建立了基于粗集-神经网络的建筑物震害预测模型。首先运用粗糙集理论,根据原始样本建立决策表进行属性离散化、属性重要性排序、属性约简和分类规则的提取;然后将所提取的关键成分作为神经网络的输入训练模型。实例研究表明,基于粗集-神经网络的多层砖房震害预测结果与实际震害基本吻合。该模型简化了神经网络结构,提高了训练速度和分类精度,还能对各因素对房屋震害的影响度进行分析。

Abstract:

Rough set theory and artificial neural network are integrated into a model of seismic damage prediction for buildings. First the rough set theory is used to acquire the knowledge of classification, which includes the decision table construction, attribute discretization, attribute importance ranking, attribution reduction and rule abstract. Then the key components are extracted as the input of the neural network. The method reduces the structure of neural network model, and raises efficiency of training and accuracy of prediction. The importance ranking of these factors to earthquake - resistance performance can be obtained by this model. The research shows that the prediction results agree with actual seismic damage of multistory masonry building.