Research On Slope Deformation Prediction Based On Wavelet Support Vector Machine

With the rapid development of economy,there are more and more types and quantities of slopes.Meanwhile,landslides,collapses and other disasters caused by slope instability due to various reasons often occur,so it is necessary to monitor slope deformation and predict its deformation trend.However,due to the randomness and complexity of slope deformation monitoring data,traditional SVM model cannot achieve high accuracy,so a wavelet kernel support vector machine(WSVM)model optimized by improved particle swarm optimization(IPSO)is proposed for slope deformation prediction.Firstly,three improvement strategies are proposed for the shortcomings of standard PSO algorithm.On the one hand,in order to improve the quality of the initial particle population and the diversity of the population,the population initialization is carried out in a uniform distribution way instead of the completely random way of the standard PSO algorithm,and the initial population is evenly distributed in the solution space while considering the randomness of the initial population.On the other hand,the method of linear decreasing inertia weight in the standard PSO algorithm has some limitations.Therefore,the cosine function is introduced to improve it,which better balances the global optimization ability and local optimization ability of the algorithm.What's more,the standard PSO algorithm is easy to fall into the local optimal,so a particle elimination mechanism is proposed.After each iteration,a small number of particles with poor fitness are re-initialized.While ensuring the optimal path of particles with good fitness,the ability of the algorithm to jump out of the local optimal is greatly increased.Finally,the optimization experiment is carried out by the test function,and the experimental result shows that the optimization accuracy of IPSO algorithm is better than the standard PSO algorithm,and the convergence speed of the algorithm is improved.Next,established a WSVM slope deformation prediction model based on IPSO algorithm.Aiming at the shortage of commonly used kernel function at present,based on the allowable conditions of SVM kernel function and wavelet transform theory,the wavelet kernel function is constructed,which is taken as the kernel function of SVM,and the optimal value of SVM model parameters is solved by IPSO algorithm,so as to build the IPSO-WSVM slope deformation prediction model.Considering the practicability and usability of the model,the slope deformation prediction system is realized based on the software development knowledge,which includes the functions of data import,model prediction and output evaluation.Finally,the IPSO-WSVM slope deformation prediction model is applied in an engineering example.Taking the vertical displacement data of Huapingzi slope and Gantianba slope of Xiluodu project as an example,the slope displacement was predicted by IPSO-PSVM,IPSORSVM and IPSO-WSVM model,and the average absolute error(MAE),average absolute percentage error(MAPE)and root mean square error(RMSE)were used for model evaluation.The experimental result shows that the three evaluation indexes of the IPSO-WSVM model are the smallest for the Huapingzi slope.Compared with the IPSO-RSVM model with the worst accuracy,MAE,MAPE and RMSE decrease from 0.371,0.694 and 0.415 to 0.216,0.402 and 0.288,respectively.For the slope of Gantianba,the three evaluation indexes of IPSO-WSVM model are all the smallest.Compared with the IPSO-PSVM model with the worst accuracy,MAE,MAPE and RMSE decrease from 0.256,0.369 and 0.309 to 0.123,0.177 and 0.139,respectively.Thus,IPSO-WSVM slope deformation prediction model has a high prediction accuracy,which can provide scientific basis for slope safety monitoring.