Slope Displacement Prediction Based On Variational Mode Decomposition And Improved Extreme Learning Machine

The displacement of slope is the key factor to ensure its stability.The effectiveness of displacement prediction is closely related to the performance of prediction model.Due to the complex factors affecting the slope displacement change at the mouth of high-speed railway tunnel,the monitoring data of slope displacement is nonlinear and non-stationary.Therefore,it is of certain theoretical and practical significance to build a scientific and effective prediction model and improve the prediction accuracy of the slope displacement prediction model.In view of the existing problems in model establishment and prediction accuracy of commonly used prediction models,this paper adopted variational mode decomposition(VMD),extreme learning machine(ELM),Grey Wolf optimization algorithm(GWO)and particle swarm optimization algorithm(PSO),and took the slope of Limin tunnel entrance section of the Harbin and Mudanjiang passenger dedicated line as the research object,the slope displacement prediction model is studied.(1)Firstly,in view of the nonlinearity and complexity of the slope displacement monitoring data,VMD was used to process the slope displacement monitoring data in order to improve the smoothness of the data and extract the detailed features of the data.Based on the adaptive decomposition characteristics of EMD,the optimal decomposition number k of VMD was determined,and the dispositions of trend term,periodic term and fluctuation term were obtained,which laid the groundwork for the establishment of the displacement prediction model in the following paper.(2)Secondly,to solve the problem of poor prediction accuracy caused by random generation of input weights and hidden layer neuron thresholds of the traditional ELM model,particle swarm optimization algorithm is used to optimize the grey wolf algorithm(PSOGWO)for ELM parameters,combining the better global search ability of PSO algorithm and the better local search ability of GWO algorithm.The input weight and hidden layer neuron threshold of ELM model are improved,and the displacement prediction model based on PSOGWO-ELM is established.(3)Finally,VMD and PSOGWO-ELM prediction model were combined to predict each sub-sequence obtained after VMD decomposition,and then the prediction results of each subsequence were superposed to obtain the final results of displacement prediction.The results of example verification show that the average absolute error of VMD-PSOGWO-ELM model is0.3011 mm,the root mean square error is 0.3771 mm,the average absolute percentage error is0.9931%,and the goodness of fit is 0.9943.The model has higher prediction accuracy and generalization performance.