基于BP神经网络的高山草甸区路基冻融特性预测研究

Research on Prediction of Subgrade Freezing and Thawing Characteristics in Alpine Meadow Area Based on BP Neural Network

文中依托西藏昌都地区贡觉至芒康公路改扩建工程,针对路基土的级配组成,设计了路基土室内冻融试验.基于室内试验所得数据,建立了路基土冻融变形预测的神经网络模型,利用该模型预测了路基土的冻胀、融沉变形,得到实际工程中所需路基填料的级配范围.结果表明:在初试含水率为9%的情况下,细粒组含量在11%~16%与17%~18%范围内的路基土样表现为“Ⅱ级弱冻胀”与“Ⅲ级冻胀”;细粒组含量在11%~18%范围内的土样表现为“Ⅱ级弱融沉”;综合考虑细粒组含量对路基土冻胀变形、融沉变形及压实效果的影响,建议将路基填料中的细粒组含量控制在16%的范围内,并可将细粒组含量为13%时的配比作为最佳级配进行考虑.

Based on the reconstruction and expansion project of Gongjue-Mangkang Highway in Changdu,Tibet,the indoor freeze-thaw test of subgrade soil was designed and made according to the gradation composition of subgrade soil.Based on the data obtained from laboratory tests,a neural network model for predicting the freezing-thawing deformation of subgrade soil was established.The frost heave and thaw settlement deformation of subgrade soil were predicted by using this model,and the gradation range of subgrade filler needed in practical engineering was obtained.The results show that,when the moisture content of the initial test is 9%,the subgrade soil samples with fine-grained components in the range of 11%~16%and 17%~18%show“weak frost heaving of Grade II”and“frost heaving of Grade III”.The soil samples with the content of fine-grained components in the range of 11%~18%show“Grade II weak thawing settlement”.Considering the influence of fine particle content on frost heave deformation,thaw settlement deformation and compaction effect of subgrade soil,it is suggested that the fine particle content in subgrade filler should be controlled within 16%,and the proportion of fine particle content at 13%can be considered as the best gradation.