砂土液化判别最优BP神经网络模型

Discussion of the Best BP Neural Network Model Using Evaluation of Soil Liquefaction

利用BP网络模型在解决砂土液化评价这类非线性问题方面的优势,选取不同的参数组合,建立不同的砂土液化判别BP神经网络模型,并根据现场实测资料进行比较分析.结果表明,以地震烈度、标准贯入点深度、地下水位深度、标贯击数、不均匀系数及地震剪应力比作为输入节点的砂土液化判别BP神经网络模型最为合理.

Taking the advantage of BP neural network in solving nonlinear problems such as sand liquefaction , different BP neural network models for liquefaction differentiation are established based on different combinations of the input neurons. By means of analyzing the observation data, the results show that the most logical BP neural network model is the model that selects earthquake intensity , the depth of standard penetration test point , underground water level, standard penetration blow- count, non- uniformity coefficient and the ratio of shearing stress as its indexes.