为了高效准确地对网壳在强震下的失效模式进行分类和判别,以抛物线型单层柱面6网壳为研究对象,在 Taft 和El-Centro三向地震波作用下,运用ANSYS软件对其进行动力全过程分析,全面考察网壳结构的荷载-最大节点位移全过程曲线、塑性发展情况、破坏时的变形情况等,并归纳总结最大节点位移、截面12.5%以上进入塑性杆件比例、截面100%进入塑性杆件比例、平均塑性积分点值4个特征指标.在Matlab软件环境下运用有导师神经网络和无导师神经网络对81个算例进行分类判别.运用的有导师和无导师神经网络分别是BP神经网络和自组织竞争神经网络.研究结果表明:2种神经网络都有准确的分类判别和预测结果,自组织竞争神经网络有直观的结果输出,BP神经网络则可以很好地看出数据中稍微偏离常规情况的数据.%In order to classify and discriminate the failure modes of reticulated shells under strong earthquake, this paper took parabolic single-layer cylindrical reticulated shells as the research object. Under the action of Taft and El-Centro three-dimensional seismic waves, the dynamic whole process analysis was carried out by the ANSYS software. The Load-the maximum displacement of nodes curve, the development of the plastic, the deformation of the damage, and etc. were comprehensively studied. The four characteristics of indicators were summed up such as the maximum displacement of the nodes, the proportion of more than 12.5% cross-section into plastic members, the proportion of full-section into plastic members, and the value of average plastic integral point. Based on the Matlab platform, the instructor and non-instructor neural network were used to classify 81 examples. The application of the instructor and non-instructor neural networks were BP neural network and self-organizing competitive neural network respectively. The results show that the two neural networks have accurate classification and prediction results, and the self-organizing competitive neural network has intuitive results. Self-organizing competitive neural networks have an intuitive result output. BP neural network can be very good to see the data slightly deviate from the conventional situation of data.
展开▼
