文摘
The prediction of groundwater inflow in karst collapse pillar has an important impact on safety underground mining, and the occurrence of these groundwater disasters is likely to be controlled and decreased via establishing an accurate groundwater inflow prediction system. However, the relationship between groundwater inflow and the factors such as geological structure, hydrogeology, aquifer, groundwater pressure, groundwater-resisting layer, mining damage and so on can be highly nonlinear, so it is difficult to establish a suitable model using traditional mechanics methods to predict the groundwater inflow and inrush risk using time series data. It is appropriate to consider modelling methods developed in other fields in order to provide adequate models for rock behaviour on groundwater inflow, nonlinear grey Bernoulli model (NGBM) is a new grey prediction model which is a simple improvement of GM(1,1) together with Bernoulli differential equation. This paper presents a new parameter optimization scheme of NGBM using the genetic algorithm (GA). The power index r of Bernoulli differential equation and production coefficient of the background value are considered as decision variables, and the prediction mean absolute percentage error is taken as the optimization objective. Parameters optimization of NGBM was formulated as the combinatorial optimization problem and would be solved collectively using GA. The model can be optimized once the GA finds the optimal parameters of NGBM. NGBM with this parameter optimization algorithm is then applied in time series groundwater inflow prediction system. Results of long-term groundwater inflow prediction show that GA is an effective global optimization algorithm suitable for the parameter optimization of NGBM and the NGBM-GA model is suit for the groundwater inflow and inrush risk prediction. After grouting reconstruction, the groundwater inflow is decreased and there is no groundwater inrush risk in the process of mining.