BP神经网络在地基土压缩指数预测中的应用

Application of BP neural network in prediction of compression index of soil

为了寻求基于多个常规物理参数间接得到土变形参数的途径，根据几个实际工程中的土工试验数据，利用BP神经网络方法对土压缩指数进行预测。选取土塑性指数、含水量、孔隙比、密度这4个常规物理参数作为影响土压缩指数的主要因素，得出土压缩指数的BP神经网络预测模型。结果表明：训练BP神经网络时，49组自变量数据中土压缩指数的BP神经网络拟合值与实测值的相对误差为-3．5139380％-1．5704225％，相对误差绝对值的平均值为0．91548％；10组自变量数据中土压缩指数的BP神经网络预测值与实测值的相对误差为-1．8055210％～6．0124173％，相对误差绝对值的平均值为3．32940％。可见，本文建立的基于4个物理参数的土压缩指数BP神经网络预测模型是可行的。

In order to acquire deformation parameters based on many routine physical parameters, a prediction of compression index was carried out with BP neural network based on the testing data of soil in several engineering sites. Taking plasticity index, water content, void ration and density of soil as primary influence factors, the prediction model of compression index based on BP neural network was obtained. The results show that the relative error of fitting value of compression index compared with the observed value for 49 groups of independent variables training BP neural network model is from -3.513 938 0% to 1.570 422 5%, and the average value of absolute value of relative error is 0.915 48%. And the relative error of fitting value of compression index compared with the observed value for 10 groups of independent variables validating BP neural network model is from -1.805 521 0% to 6.012 417 3%, and the average value of absolute value of relative error is 3.329 40%. Therefore, the prediction model of compression index with BP neural network based on 4 routine physical parameters is doable.