Research On The Opene-pit Slope's Subsidence Rule And Predicion Based On Support Vector Machine(SVM)

Open pit mining is an important way for mineral exploitation, relative to the extraction measures else, it has many merits, for instance:low cost, high productivity, low loss, recovery is convenient. For this reason, open pit mining ratio showing a rising trend year by year. With the scope expands unceasingly, the mining subsidence caused by mining scale also will increase.lt will gradually developed from the internal to the surface part of the rock mass, which can lead to move or deformation on the surface of the earth. Open pit mining also formed many bare slope, the slope of the potential geologic hazards slowly become a danger to mine safety production and one of the most important risk factors of surrounding residents. For this reason, so as to be able to guarantee mining production order, it is necessary for slope deformation forecast.Due to the settlement deformation of slope is an internal with complicated nonlinear relationship system, the traditional forecasting theory is difficult to obtain ideal results. Based on the statistical theory of Support Vector Machine (SVM) theory which has strong ability of fitting, this paper choose Fushun power generation to Support Vector Machine (SVM) model to predict the settlement of the open pit mining slope. But because of the influence of the observation condition and other outside factors, the existing subsidence observation data contains some noise, wavelet analysis has the characteristics of multi-resolution analysis, which can be separated the noise from the signal and extract useful information from the signal. So this paper choose the wavelet transform theory to denoise the subsidence observation data, for the future studying on the subsidence and the modeling prediction lay a good foundation. Then, researching on the subsidence district the overall trend of subsidence to the factory. . Support Vector Machine(SVM) is a machine learning based on the statistical theory, and Least Square Support Vector Machine is improved theory based on it, Both of them have the very strong learning ability. The key of Support Vector Machine model's establishment is parameter optimization, in view of the limitations of traditional parameter optimization method, this paper put forward the parameter optimization method that combining the cross validation with Particle Swarm Optimization(PSO). This paper uses cross validation method, PSO, combining the cross validation with PSO to build the model of SVM and LSSVM model, then compare the accuracy of different models. Proved by experiments, LSSVM model that founded by using the parameter optimization method combining cross validation with PSO is of higher precision, is applicable to the subsidence prediction of open pit slope. Then, applicat of the model to predict the subsidence data, and analysis the forecast results of Fushun power plant, draw the future trend of subsidence, achieve the purpose of subsidence prediction of slope.