云南红土抗剪强度指标的神经网络模型研究

以水泥加固红土的直剪试验结果为基础,运用人工神经网络理论建立起云南红土抗剪强度指标的神经网络模型。该模型的输入层向量包括水泥加入比例和试样养护时间两个影响因素,输出层向量为粘聚力和内摩擦角两个抗剪强度指标,隐层传递函数确定为正切函数,输出层传递函数确定为对数函数,隐层神经元数目确定为5。预测结果表明,该神经网络模型能够反映水泥加入比例和试样养护时间对红土抗剪强度的影响,预测的抗剪强度指标与实测值接近,预测效果较好。

Study on Neural Network Model of Shear Strength Index in Yunnan Laterite

Based on the results of direct shear test about cement reinforced laterite,the artificial neural network model of shear strength index in Yunnan laterite is set up using the theory of artificial neural network.The input layer vector of the model includes the adding proportion of cement and samples curing time,the output layer vector includes two shear strength parameters which are cohesion and internal friction angle,the hidden layer transfer function is determined for tangent function,the output layer transfer function is determined for logarithmic function,the number of hidden layer neurons is determined as 5.The predicted results show that the neural network model can reflect the influence of adding proportion of cement and samples curing time on the shear strength of laterite, the prediction results of the shear strength index are close to the measured values,and the predicted effect is good.