基于BP神经网络软粘土循环强度研究

A STUDY ON CYCLIC STRENGTH OF SOFT CLAY BASED ON BP NEURAL NETWORK

以软粘土动三轴试验结果作为样本集,以静应力、循环应力和确定位移破坏标准下的循环破坏振次作为输入层,建立基于BP神经网络软粘土循环强度的预测模型.通过循环扭剪试验结果对模型进行检验.结果表明,该模型稳定性良好,其稳定性不受试验应力状态的影响,可以用来预测一般应力状态下软粘土循环强度,预测精度满足岩土工程的要求.

A prediction model of cyclic strength of soft clay is built on the basis of BP neurat network with results of dynamic triaxial test for the soft clay used as the sampling set. The static stress and cyclic strss, as well as the number of cycle failures, determined according to the displacement failure criterion, are taken as the input layer. Performance of the model is checked through cyclic torsion tests. Results show that the model has good stability which can hardly be affected by the testing stress states. Therefore,it can be applied to the prediction of cyclic strength of soft6 clay in general stress states. Precision of the prediction can meet the requirements of geotechnical engineering.