基于BP神经网络的重塑黄土冻融过程渗透特性研究

冻融循环导致黄土渗透特性的变化规律十分复杂,传统单一因素评价方法难以确定冻融过程黄土渗透系数与多因素之间的量化统计关系。基于此,首先对西安Q3重塑黄土进行冻融过程三轴渗透试验,得到不同干密度、含水率、围压及冻融次数下的渗透系数;然后采用BP神经网络算法对试验数据进行学习训练,建立各因素与渗透系数间的预测模型。研究结果表明:重塑黄土渗透系数变化规律,随围压增大,其值逐渐减小,且减小幅度先快后慢;随干密度和初始含水率增大,其值先增大后减小;随冻融次数增加,其值逐渐增大,且上升幅度先急后缓。冻融过程黄土渗透系数神经网络模型预测值和试验值之间相对误差较小,表明该方法具有较好的预测精度,能够综合描述诸因素与渗透系数的量化关系。

Permeability of Remolded Loess During Freezing -Thawing Process Based on BP Neural Network

The variation regularity of loess permeability caused by freezing and thawing cycles is very complicated .How-ever ,the traditional method based on only one factor is difficult to determine the quantitatively statistical relationship be-tween the permeability coefficient and multi factors .This paper firstly carried out triaxial permeability test to get the per-meability coefficient index data of Xi'an Q3 remolded loess under different dry density ,moisture content ,confining pres-sure and freeze-thaw times .Then a prediction model for the relationship between permeability coefficient and multi fac-tors was obtained by training testing data by using BP neural network .The results show that the loess permeability varia-tion coefficient decrease with the increasing of confining pressure ,and decreases soon after the first slow ;with the dry density and initial water content increased ,the value increased first and then decreased ;with the freeze-thaw cycles in-creases ,its value increases gradually ,and rise after the first emergency relief .The relative error of prediction value of permeability coefficient compared with experimental data is little ,which indicates that BP neutral network forecasting method has better accuracy and can describe the quantitative relationship between permeability coefficient and factors .