Research On Prediction Of Bearing Capacity Of CFG Pile Composite Foundation Based On HPSO-SVM

The bearing capacity of composite foundation is one of the important parameters to evaluate the construction quality of CFG Pile Composite Foundation,and whether it can reach the design requirements is also the precondition of the next main structure construction.Static load test is the most accurate and reliable in-situ testing method for foundation bearing capacity,but the time and economic cost of field test is large,and the test belongs to destructive test,it is difficult to get the characteristic value of bearing capacity comprehensively and accurately.Because the design of the bearing capacity of composite foundation is relatively conservative,the design value obtained according to the specification is far below the field measured value,resulting in the waste of the bearing capacity of the pile material and the natural Foundation.In order to make a more comprehensive prediction on the bearing capacity of the composite foundation on the basis of the original static load test,a prediction method with low cost and high precision is proposed in this paper.this paper firstly analyzes the mechanism of CFG Pile Composite Foundation,extracts 12 factors from physical characteristics of soil body and parameters of pile,and establishes index system of influencing factors.Secondly,the particle velocity and position updating method of the nonlinear inertial weighting particle swarm algorithm(ULWPSO)is improved by introducing the metropolis criterion in the simulated annealing algorithm(SA),and the hybrid particle swarm algorithm(HPSO)is formed.Based on the optimization ability of HPSO,the kernel function ?,penalty factor C and the insensitive loss function ? of support vector Machine(SVM)are optimized,and the bearing capacity prediction model of HPSO-SVM CFG Pile Composite Foundation is established,In order to further predict the bearing capacity of CFG pile composite foundation.This paper selects the related data of CFG Pile Composite Foundation in WK Real estate project as the bearing capacity prediction sample,and introduces the principal component Analysis(FCA)to remove the correlation of the influence factors of the sample.Comparing HPSO-SVM with PSO-SVM,the results show that the former has obvious superiority in the selection of SVM parameters and prediction precision.And the HPSO-SVM,PSO-SVM,GA-BP andlogistic regression were tested 100 timesbythe leave-one-out cross-validations.The result shows that the average of mean square error(MSE)of the HPSOSVM model is 35.51312,which is significantly lower than the latter three(44.79668,45.20706,54.86805),and further validates the more accurate prediction effect of the HPSO-SVM prediction model,and the value of engineering application in the prediction of bearing capacity of CFG pile Composite Foundation.