| Geotechnical mechanical parameters are an important basis for the design and construction of tunnel engineering.In order to obtain geotechnical mechanical parameters efficiently and accurately,this paper first proposes a multi-dimensional limit learning machine model(M-ELM)optimized by the improved slime fungus algorithm(ISMA).Due to the inevitable disturbance of surrounding soil during shield construction,resulting in ground settlement,when the ground settlement is too large,it will seriously affect the normal use of surrounding buildings,and even cause engineering accidents.Therefore,in order to reliably predict the ground settlement caused by shield construction,this paper selects the limit learning machine model(ELM)optimized by Peck formula and Harris Eagle algorithm(HHO)to study the laws and prediction methods of ground settlement.The main research work of this article is as follows:(1)Parameter inversion method research: Based on the multi-dimensional limit learning machine model(M-ELM),the slime mold algorithm(SMA)of natural simulation is used to optimize the weight and threshold parameters of M-ELM.Sine chaotic mapping,reverse learning,and adaptive t-distribution mutation strategies are introduced to improve the efficiency,accuracy,and stability of the algorithm.A new displacement inverse analysis method is proposed combining with the FLAC 3D threedimensional finite element numerical model.Firstly,the compression modulus and Poisson’s ratio of peat soil are selected as the parameters to be inverted,and the learning sample and test sample schemes of the model are constructed according to orthogonal tests and uniform tests.Secondly,a three-dimensional tunnel model is established using FLAC 3D finite element software to simulate the shield construction process,and 34 monitoring points are set according to the actual working conditions.Finally,based on the measured data of shield tunneling in Kunming Metro Line 5,the applicability of the model is tested through algorithm performance analysis,single dimensional and multidimensional inverse analysis,and settlement result analysis.The results show that the improved slime fungus algorithm can improve the prediction accuracy of the multidimensional limit learning machine model,improve the inversion effect of geotechnical mechanical parameters,and reduce the running time of the model.(2)Research on Peck Formula Settlement Law and Value: Based on a large amount of measured data of Kunming Metro Line 5,combined with the actual settlement on site,47 groups of sections on the left line and 53 groups of sections on the right line are selected for Peck formula fitting,and the settlement trough width i and formation loss rate of each typical section are obtained using the displacement back analysis method ηValue.Considering the shield and soil quality information of lacustrine sedimentary strata,respectively study η The influence rules of relevant factors such as tunnel burial depth z,weighted compression modulus E*,and grouting rate g.According to the principle of maximum and relatively conservative values,the width coefficient k and ηBased on the statistical analysis and the Peck formula superposition principle,a prediction formula for surface subsidence of double-track tunnels suitable for Kunming lake facies sedimentary strata is obtained.Finally,the rationality of the correction formula is checked based on the two cross sections of a certain line.The results show that the Peck formula has high applicability in the sedimentary strata of Kunming lake facies η Both have certain distribution rules,and the upper limit curve of the modified two-line prediction formula can roughly predict whether the settlement exceeds the engineering warning,which can provide a reference for similar tunnel construction.(3)Research on prediction methods of maximum ground settlement: In order to reliably predict the maximum ground settlement caused by shield tunneling,a limit learning machine(ELM)prediction model based on principal component analysis(PCA)and Harris Eagle optimization algorithm(HHO)was proposed.In geological,geometric,and shield parameters,14 impact factors are initially selected,and five principal component variables are separated and extracted from a 14-dimensional array using PCA as input to the model.The input layer weight and hidden layer threshold parameters of the ELM model are optimized using HHO to obtain the optimal solution of the prediction model.Based on the monitoring data of the Jinhui section of Kunming Rail Transit Line5,simulation verification was conducted,and the model was compared and analyzed with BP neural network,RBF neural network,and the unoptimized ELM model.The results show that the Harris Eagle optimized limit learning machine model can reliably predict the maximum ground settlement caused by shield construction. |
PDF Full Text Request