基于BP神经网络冻土强度预测模型研究

人工冻结法已作为苏州地铁连通道施工的辅助工法之一,但目前尚缺乏与设计配套的苏州地区冻土强度归一化参数,以人工神经网络为基本工具,利用其强大的非线性映射能力,综合考虑温度、含水率、干密度等影响因素指标对苏州地铁典型冻土抗压强度的影响,建立了抗压强度预测模型。结果表明,BP网络具有较高的求解能力,预测值与实测值接近,其精度远高于其他方法,且简单实用,这为获取试验中无法实现的参数值提供了新的途径。

Study on Prediction Model of Frozen Soil Strength Based on Back-propagation Neural Network

Artificial freezing method has been used in Suzhou subway interconnecting channel construction as an auxiliary technique.However,there is no normalized parameter about frozen soil strength in Suzhou area for design at present.Therefore,a prediction model is established in this paper considering the influence of temperature,water content and dry density on typical frozen soil strength in Suzhou subway using artificial neural network to map the non-linearity among these influencing factors.The result shows th...