摘要
目前对于土层地震液化问题,多关注于砂土液化现象及判别方法研究,而砂砾土液化问题研究较少。基于砂土震害资料建立的剪切波速液化判别方法不能满足砂砾土液化判别的需要,往往会给出偏于危险的判别结果。在分析砂砾土液化主要影响因素的基础上,将神经网络引入砂砾土液化判别领域,建立了一种基于剪切波速和神经网络的砂砾土地震液化判别模型。将汶川地震中获取的砂砾土液化样本数据用于模型的训练和测试,实现了砂砾土液化与各影响因素之间的非线性映射。将所建立的模型用于汶川地震砂砾土液化判别分析,并和已有方法进行了对比,表明所提出的砂砾土液化判别方法具有判别准确率高的特点。
For seismic soil liquefaction problems, most researches focused attention on sandy soil liquefaction phe- nomenon and related discrimination techniques, and less effort was made on gravel soils liquefaction problems. The liquefaction evaluation methods based on shear wave velocity can not satisfy the requirements of gravel soil liquefac- tion prediction, because they generally give unsafe results. In this paper, the main factors that cause liquefaction of gravel soils are studied, and neural network is introduced for liquefaction evaluation of gravel soils. A new liquefac- tion discrimination model for gravel soils is established on the basis of artificial neural networks with shear wave ve- locity. The data collected from 2008 Wenchuan earthquake are adopted to train and test the proposed neural net- work model. Then the model is applied to liquefaction prediction of gravel soils in Wenchuan earthquake, and it is also compared with the existing methods. Results show that the proposed liquefaction discrimination model for grav- el soils on the basis of artificial neural networks with shear wave velocity is more effective than other methods.
出处
《地震工程与工程振动》
CSCD
北大核心
2014年第1期110-116,共7页
Earthquake Engineering and Engineering Dynamics
基金
国家自然科学基金项目(51109028)
国家留学基金资助项目(201208210208)
中央高校基本科研业务费专项资金资助项目(DUT11RC(3)38)
关键词
砂砾土
液化判别
神经网络
剪切波速
汶川地震
gravel soil
back-propagation neural networks
liquefaction discrimination
shear wave velocity
Wen-chuan earthquake
