Research And Application Of Foundation Pit Settlement Prediction Based On SAPSO-SVM Model

In today's economic globalization,my country is urban construction has been accelerating.Urban construction is a trend of developing from the air and taking from the underground.Therefore,the excavation of foundation pits has gradually increased.Due to the influence of construction conditions,construction technology and human factors during the excavation of the foundation pit,structural deformation during the construction of underground rail transit is inevitable.In order to ensure the safety of construction personnel and property,monitoring and data processing of the deformation of the foundation pit is essential.It seems particularly important.This article takes the underground garage project of Wuxi Yugang Middle School as an example.First,according to the the improved false domain method and the mutual information method,the appropriate embedding dimension and optimal delay time are selected to reconstruct the phase space of the maximum settlement monitoring point(JD28)data,and according to the "accuracy rule" and polynomial The fitting method performs singular value checking and interpolation on the data;secondly,the SAPSO-SVM model is established based on the reorganized data,a method of alternate iteration using simulated annealing and particle swarm algorithm,and the penalty parameter C and kernel function of the support vector machine The parameter g is optimized,which overcomes the shortcomings of the basic particle swarm algorithm that is easy to fall into local extreme values,and enhances the global optimization capability.At the same time,it shows many unique advantages in terms of nonlinear data processing.Finally,compare and analyze with SVM regression algorithm and PSO-SVM algorithm.The results show that the three algorithms have good feasibility in the prediction of foundation pit settlement,but the SAPSO-SVM model has a residual sum of squares(SSE)1.412mm,a root mean square relative error(MSE)0.213mm,and an average absolute error.(MAE)The value of 0.171mm is smaller than the other two models,and it has higher prediction accuracy.Experiments have proved that the SAPSO-SVM model has strong learning and generalization capabilities,high prediction accuracy,good stability and adaptability,can better reflect the change trend of settlement data,and is more suitable for application in settlement prediction.In order to verify the reliability of the conclusion,another monitoring point(JD30)was selected for the second largest settlement,and the same three models were used for processing analysis,further verifying that the SAPSO-SVM model is superior to the other two models.This paper realizes simulation prediction through the MATLAB program written,and the research of this subject has certain reference value for similar project construction and disaster prevention.Figure[23]table[26]reference[82]...