Forecasting Special Section Subgrade Settlement Of Mountainous Freeway By Single/Combined Method

In this paper,the soft soil subgrade and high fill subgrade of Zhangsang and YiLou freeway in Hunan Province are studied,their engineering characteristics and fill mechanism are analyzed.On this basis,through the field subsidence monitoring and investigation,the typical section is selected.The detailed analysis of the settlement monitoring data is made in the construction period and the natural settlement period,and several typical methods for the prediction of roadbed settlement are analyzed,the limitation of different forecasting methods are compared.Finally,GM(1.1)Model,hyperbola method,Poisson curve method and s?lgt method are used to predict the settlement of soft soil and high fill subgrade respectively.It is found that the single model GM(1.1))has the highest prediction accuracy,and it is not universal for soft soil settlement prediction.On this basis,the post-construction settlement of soft soil and high-filled subgrade is calculated.The results show that the post-construction settlement of soft soil subgrade is larger.In this paper,two different combinations of models are established based on four kinds of single prediction methods.A combination model based on error inverse proportional distribution method is proposed to overcome the difficulties of calculating variable weight,and the weight is determined by traditional entropy method.The results of the combination model are compared.The results show that the accuracy of the variable weighting model is higher than that of the entropy method,and the results of the combined model are better than the entropy method,It can be seen that this combination model(entropy method to determine the weight of the combined model)is not practical significance.In order to further improve the accuracy of prediction,based on the variable weight combination model,the neural network is introduced to make the error compensation for its prediction result.The results show that the proposed neural network can further optimize the prediction results and improve the accuracy of the prediction results on the basis of the combined model prediction.This method has a great advantage in some high precision prediction.