Deformation Prediction And Safety Assessment Of Foundation Pit Based On SSA Elman And Grey Cloud Model

With the rapid construction of urban comprehensive pipe gallery and other infrastructure,the monitoring in the process of foundation pit construction is becoming more and more important.Through the monitoring of foundation pit,we can obtain the changes of multiple indicators in time and judge the state of foundation pit.How to carry out data mining,forecast the monitoring index data,and comprehensively consider multiple monitoring indexes to evaluate and grade the comprehensive condition of foundation pit is very important.Based on the actual monitoring data of Panjin pipe gallery in Liaoning Province,this paper studies the single index deformation prediction and multi index safety assessment of foundation pit,the main work is as follows:(1)The machine learning method Elman dynamic neural network for foundation pit prediction and the grey theory and cloud model method for comprehensive evaluation of foundation pit are introduced;This paper summarizes the types of accidents that may occur in the foundation pit,and analyzes the actual monitoring data of the project site.(2)Elman dynamic neural network has adaptive time-varying characteristics due to the existence of bearing layer,which is more suitable for the prediction of foundation pit monitoring data.The sparrow search algorithm SSA,which has better search ability,is used to optimize the initial weights and thresholds of Elman network,and the optimized network is used to predict the monitoring data in the stable deformation period.Compared with Elman network and Elman network optimized by genetic algorithm,the results show that SSA Elman has better prediction performance;In view of the monitoring data which fluctuates violently during the construction period,complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)is used to denoise and decompose the original data,and then each component is predicted by SSA Elman and then superimposed.Compared with the three models which are predicted directly without data decomposition,the accuracy is better;Considering the relationship between the monitoring indexes,multiple monitoring index data are selected as the input,and the average impact value MIV algorithm is used for correlation screening and input variable dimensionality reduction,from which the input item with the largest and relatively independent relationship with the surface subsidence is selected,and then the prediction is carried out.The results show that compared with MIV Elman network MIV、GA Elman and SSA Elman,the MIV SSA Elman network after data screening has higher prediction accuracy and fitting degree.(3)Combined with grey whitening weight function clustering evaluation model and cloud model,based on the actual monitoring data,the grey cloud model is constructed,and the response of each monitoring index to the overall condition of foundation pit is considered for comprehensive evaluation.According to the technical standard of foundation pit engineering monitoring(GB 50497-2019),the evaluation system and grey classification are determined,and the standard cloud of each index classification is formed by using the positive cloud generator;Decision making and laboratory method DEMATEL is used to determine the subjective weight,critical method is used to determine the objective weight combined with the measured data,and the optimal combination weight is obtained based on the sum of squares of deviations;The x-condition cloud generator is used to calculate the whitening value of gray cloud,which increases the fuzziness and randomness.The comprehensive model of foundation pit evaluation is constructed,and the quantitative and qualitative analysis of multiple sections of foundation pit is carried out.