An intelligent approach to prediction of tailings dam displacement safety using dynamic preventive control modelingOA

Tailings dams are a critical infrastructure for mining enterprises,and their safety directly affects production security and environmental protection.However,owing to the loose nature of dam materials and their unique geological structures,traditional slope stability assessment models have limited applicability to tailings dams.Dam displacement serves as a key indicator for evaluating the stability and identifying potential developmental issues during operation,making it essential for safety monitoring.Therefore,developing reliable displacement prediction methods is crucial for early warning and mitigation of disasters.This study proposes a“feature derivation-decomposition forecasting-model optimization”approach for predicting displacements in tailings dams.First,the IDBO-VMD(Improved Dung Beetle Optimizer-Variational Mode Decomposition)decomposition algorithm is employed to separate dam displacement into the trend and periodic components.Subsequently,the trend and periodic displacements are predicted using the DBN(Deep Belief Network)and IDBO-TCN(Temporal Convolutional Network)-BiGRU(Bidirectional Gated Recurrent Unit)-self-attention models,respectively,with linear weighting applied to enhance feature representation.The final displacement prediction is obtained by superimposing the predicted components.The method was validated using the tailings reservoir of the Dayi Company in Lueyang County.The results showed that the predicted cumulative landslide displacement closely matched the measured values,achieving a correlation coefficient of 0.995 and a mean absolute error(MAE)of 0.092 mm.Specifically,the trend component prediction yielded an R value of 0.996 and MAE of 0.065 mm,whereas the multi-algorithm coupled IDBO-TCN-BiGRU-self-attention model achieved higher overall precision for the periodic component,with an MAE of 0.132 mm and R of 0.984.These results demonstrate that the proposed model provides a novel framework for intelligent early warning of tailings dams and can accurately predict stagewise variations in displacement.