摘要
结合河北唐山地区土样样本,以地下水位埋深(hw)、地下水头(h)、标准贯入锤击数(N63.5)、土的动强度(R)及地震力(L)为评价指标建立了BP神经网络和RBF神经网络的预测模型。通过实例结果比较分析,表明RBF神经网络和BP神经网络判断砂土液化的精度都较高,但对于用埋深hs,地下水位深度h,标准贯入锤击数N63.5,土的动强度R和地震力L作为参数指标时,RBF神经网络在砂土液化的判别方面优于BP神经网络。通过对金坛石桥枢纽进行建模预测,进一步证明了以上结论,并说明了BP神经网络和RBF神经网络对于砂土基础液化的预测是普遍适用的。
The forecast model of BP and RBF neural network has been established using the soil samples from Tanshan Area in Hebei Province, based on the assessment system including the factors of underground water level depth, ground water head, blow count during standard penetration test, dynamic strength of soil and seismic force. Through the comparative analysis, RBF neural network and BP neural network has achieved high precision in estimating sand liquefaction. But the parameter indices of the underground water level depth, ground water head, blow count during standard penetration test, dynamic strength of soil and seismic force show that RBF neural network outgoes BP neural network in distinguishing the sand liquefaction. The modeling forecast of Jintanqiao Project has proved the above conclusion, and both BP and RBF neural network can be applied universally to the sandy soil foundation liquefaction forecast.
出处
《人民珠江》
2008年第5期34-37,共4页
Pearl River
