Adaptive dynamic prediction model of mining subsidence aided by measured data

Underground mining-induced surface subsidence adversely affects both the surface environment and the structures located above it. Accurately predicting the dynamic subsidence and deformation caused by underground mining is crucial when employing maintenance and remediation methods to mitigate these adverse effects, as it directly impacts the selection of maintenance strategies, timing, and volume assessments. In response to the limitations of traditional time function and parameter models in adapting to the dynamic changes of actual underground mining activities-resulting in low subsidence prediction accuracy-this paper presents an adaptive prediction model for dynamic subsidence supported by measured data and developed through programming. This model utilizes historically measured data on surface subsidence to derive optimal parameters for each historical period. By analyzing the trends in these parameters, it dynamically adjusts the parameter value for subsequent predictions, achieving high-precision prediction of the surface dynamic subsidence. Engineering case study results indicate significant variations in the optimal time function parameter values throughout the mining process. The estimated parameter values obtained through the extrapolative prediction method, supported by measured data, align closely with the optimal values. The average relative RMSE of predicted dynamic subsidence for each period is 4.3%, markedly lower than the 9.1% achieved by traditional prediction models. This enhancement significantly improves the accuracy of dynamic subsidence predictions due to underground mining and provides robust technical support for the maintenance and remediation of structures.