Research On Subsidence Prediction Of Subway Based On Particle Swarm Optimization Support Vector Machine

With the accelerating rate of urbanization in our country,a large number of people have flooded into cities,increasing the pressure on urban resources and environment,especially traffic,and the development of infrastructure such as underground rail transit has become the only way to solve these problems.Due to the influence of construction conditions,construction technology and human factors,the structural deformation during the construction of underground rail transit is unavoidable.To ensure the safety of life and property of construction personnel,it is particularly important to predict and forecast the deformation monitoring data of subways.This paper takes building settlement monitoring as the research object.In order to improve the accuracy of forecasting results,the phase space reconstruction of monitoring data is selected by selecting the appropriate embedding dimension and time delay.Taking into account the selection of different kernel functions and related parameters will have a great impact on the accuracy of the support vector machine,we propose a particle swarm optimization algorithm using support vector machines with fewer adjustment parameters,less training volume,faster calculation speed and global search capability.The relevant parameters were optimized and selected,and the support vector machine model was improved.Through the analysis of settlement monitoring data of Chaoyangcun Station of Metro Line 2 in X City,combining with matlab2014a platform,Microsoft Visual C++2012 Professional is used as a compiler and extended programming is based on libsvm toolbox.The minimum reconstruction error method is used to reconstruct the phase space of the subsidence monitoring data.The reconstructed data are trained and predicted by using the improved SVM model,the traditional support vector machine model and the BP neural network model respectively.Through comparative analysis of the prediction accuracy results of the three models,it is shown that the improved SVM model is superior to the other two models in comparison of accuracy.Experiments show that the particle swarm optimization support vector machine model has strong ability of learning and generalization,has high prediction accuracy,stability and adaptability,can reflect the overall change information of metro settlement data,and has a certain degree in practical engineering applications.The promotion value has certain research significance.