摘要
地震动参数的衰减规律在地震工程中有着重要的地位,随着地震动参数观测记录的积累,追求较为全面准确的衰减规律也是一个重要的基本问题。本文以美国西部的地震动观测资料为基础,分别建立了三维地震动参数衰减规律的径向基函数人工神经网络模型和BP网络模型。RBF模型对进行训练的数据拟合良好,却在其它点上的误差较大;BP模型虽然对训练的数据的拟合效果不如RBF模型,但在整个预测空间中有较好的连续性。最后,还把BP模型的预测结果与传统的衰减公式进行了对比。
The ANN for attenuation rule of 3D seismic parameters is studied in the paper. Based on the seismic data of West American, both the RBF model and BP model are constructed for the attenuation rule. The seismic data is got from 63 times of seismic records for strong earthquakes. In total of 186 items of the peak acceleration records is obtained. The BP model is not good in data fitting, but it possesses a better continuity in the predictive space. At last, the BP model is adopted for the study. The comparison of the predictive results between the BP model and the other formula is also made.
出处
《土木工程学报》
EI
CSCD
北大核心
2003年第5期1-4,16,共5页
China Civil Engineering Journal
