Study On Deformation Prediction And Stability Evaluation Of Tailings Dam Based On Intelligent Algorithm

Tailings and industrial waste residues excluded from mining in tailings reservoirs are major hazards with high potential energy.It is of great economic value and social significance to monitor the safety of tailings dam and adopt scientific and effective methods to predict deformation and evaluate stability.The deformation of tailings dam is a highly complex and non-linear system.On the one hand,it shows the macro-variation trend of displacement under the influence of its own dam-building mode and geological conditions.On the other hand,due to the changes of reservoir water level,rainfall,seepage,saturation line and other factors,the deformation sequence also shows certain periodicity and random fluctuation.How to decompose the deformation sequence effectively and establish the prediction model is the key to affect the results of deformation prediction.Thus,this paper proposes a new intelligent algorithm,Extreme Learning Machine(ELM),as the main research means,and incorporates the theoretical knowledge of robust filtering of modern surveying and mapping data,variational mode decomposition,kernel function,phase space reconstruction,multi-factor entropy weight analysis and limit equilibrium method.The Baiyan tailings reservoir project in Guizhou Province is taken as the research object,which is obtaining monitoring data of tailings reservoir and deformation sequence of tailings dam by automation and real-time on-line monitoring system.Aiming at "deformation prediction " and "stability evaluation " of tailings dam,a deformation prediction based on new intelligent algorithm is constructed.The main research contents and achievements are as follows:(1)Application of robust Kalman filter in data processing of tailings dam deformation monitoring.The horizontal and vertical displacements of dam surface monitoring points are obtained by GPS on-line monitoring technology.Due to the influence of external environment and other factors,gross errors inevitably exist in the deformation sequence.Based on Kalman filter and the idea of robust estimation,robust Kalman filter is applied to pre-process the deformation sequence of dam.The validity of robust Kalman filter is verified,and the purpose of detecting and eliminating gross errors is achieved.(2)The displacement prediction of tailings dam based on intelligent algorithm AGA-KELM.The VMD algorithm is applied to the decomposition of displacement of tailings dam,and the displacement sequence with different deformation characteristics is obtained.Based on each displacement component,the phase space is reconstructed by C-C method,and the improved AGA-KELM prediction model of each component isestablished.Taking Danba landslide and Baiyan tailings dam as examples,the simulation analysis is carried out.The prediction results are more accurate and the prediction results are better than other models.(3)Establishment of AGA-KELM prediction model considering entropy weight.The influence of accumulated rainfall in one month,the elevation of reservoir water level in one month,the average elevation of reservoir water level in one month,the change elevation of saturation line in one month,the average elevation of saturation line in one month and the increment of displacement in the previous month on dam displacement are comprehensively considered.And based on the maximum information entropy theory,the response analysis of various influencing factors on dam displacement is carried out,and different weights are given to each other.The AGA-KELM displacement prediction model based on entropy weight is constructed to realize the prediction of cumulative displacement of tailings dam,which improves the scientificity and accuracy of prediction.(4)Calculation of dam safety coefficient based on AGA-KELM model of principal component analysis.Through sample learning,the AGA-KELM safety coefficient prediction model based on principal component analysis is established.The comprehensive index extracted from the principal component is used as the input vector of the network,which reduces the complexity of the model and improves the prediction accuracy.Taking Baiyan tailings dam,Baimachong tailings dam and Luwanjing tailings dam as examples,the safety factor is calculated,and the calculation results are compared with the limit equilibrium method.The calculation is simple and comparisons show that the predicted results can be can be well applied in practical engineering.