粒度分析与渗透系数关系的BP网络分析

On BP network analysis of relationship between grading analysis and penetration coefficient

针对不同颗粒粒径分布对渗透系数影响的非线性特征,利用BP人工神经网络模型,对14个典型样本进行学习并预测4组不同级配的粒组样本的渗透系数,结果证实该模型能很好地反映出分选性和粘粒含量对渗透系数的影响,运用神经网络预测的渗透系数充分考虑到土体本身的结果。因此,预测值应较现有的计算公式得出的计算值更接近于自然状态下的真实值。

According to the nonlinear features of the influence of various particle size distribution on the penetration coefficient, the paper adopts the BP manual neural network model, studies the 14 typical samples and predicts the penetration coefficient of the soil fraction samples of differ- ence grading of four groups, and proves by the result that the model can better reflect the influence of the sorting and clay content on the penetra- tion coefficient, indicates the soil should be considered in applying the neural network to predict the penetration coefficient, so the calculated value bv the predicted value is closer to the factual value in natural status, comnared with the current calculatln~ formula.