Prediction Of Pile-net Composite Foundation Settlement In Lakeside Area And Development Of A New-style FBG Settlement Sensor

At present,the highway built in the soft soil area has the phenomenon that large settlement or uneven settlement leads to the destruction of the road surface,which seriously affects the driving comfort and operation safety.Therefore,it is a technical problem to correctly predict and timely monitor the settlement of soft foundation after construction in the field of highway engineering.Based on the engineering background of Jiujiang beltway expressway project,BP neural network,genetic algorithm,GA-BP neural network and other intelligent inversion methods combined with finite element numerical simulation are used to back calculate the parameters to predict the settlement of Pile net composite foundation after construction.In addition,a new-style FBG sensor which can monitor the cross section settlement of soft soil foundation was developed by using the fiber bragg grating sensing technology.The indoor simulation experimental results verifies its feasibility for monitoring settlement.Mainly following aspects of the work:1.According to the research results of pile-net composite foundation,the definition and mechanism of pile-net composite foundation are expounded.The basic algorithm of pile-net composite foundation is summarized.2.A parameter back calculation model of pile-net composite foundation based on BP neural network is established.The results show that many training data can be provided by the orthogonal design method.BP neural network algorithm can map out the complex nonlinear relationship.Based on the engineering data,BP neural network can calculate the "equivalent parameter" of the foundation.The average error of prediction is 15.32%.3.A parameter back calculation model of pile-net composite foundation based on genetic algorithm is established.The results show that the combination of and genetic algorithm can be realized by Fortran language.Genetic algorithm is used to calculate the "equivalent parameter" of the foundation by 87 iterations.The average error of this model is 7.65%.4.A parameter back calculation model of pile-net composite foundation based on GA-BP neural network is established.The results show that the genetic algorithm can optimize the weights and thresholds of the BP neural network.The "equivalent parameters" of the soil are calculated by the GA-BP neural network model.The average error of this model is 5.26%.Compared with the three intelligent methods,GA-BP method has highest prediction accuracy.5.A new-style FBG sensor which can monitor the cross section settlement of soft soil foundation is developed.Based on the fiber grating sensing technique and beam bending theory and differential algorithm,a fiber grating sensor which can monitor the deformation of the whole section of the soft foundation is developed.The feasibility of the new-style FBG is verified by the indoor model test.