基于Pareto分层混合优化算法的土石坝渗透系数多目标反演方法OA

In the inversion of permeability coefficients for earth-rockfill dams,conventional multi-objective optimiza-tion algorithms often suffer from insufficient convergence,poor diversity of Pareto solution sets,and low computa-tional efficiency due to their difficulty in balancing global exploration and local search.To overcome these limita-tions,this study proposes a hybrid framework integrating the strengths of NSGA-Ⅱ and MOPSO algorithms through a Pareto-dominance based hierarchical search strategy,enabling efficient multi-objective inversion of permeability coefficients.By utilizing monitoring data of both pressure head and seepage rate,a surrogate model based on SSA-XGBoost was developed to characterize the relationship between permeability parameters and seepage responses.A hybrid NSGA-MOPSO algorithm,guided by Pareto-dominance hierarchy,was subsequently employed to identify the optimal permeability coefficient set,thereby reconstructing the current seepage state of the dam.Case study results demonstrate that the proposed SSA-XGBoost surrogate model significantly outperforms RBF,ELM,and RF models in predictive accuracy,achieving superior performance across four evaluation metrics:R2,RMSE,MAE,and MAPE.By employing this surrogate model,the inversion process not only maintains solution diversity but also reduces computational time by approximately 39％.Compared to standalone NSGA-Ⅱ and MOPSO algorithms,the hybrid algorithm shows performance improvements of 19.5％and 17.7％for the pressure head objective(f1),and 72.3％and 62.8％for the seepage discharge objective(f2),respectively.These results validate the enhanced predic-tive accuracy and the reliability of the inversion outcomes achieved by the proposed NSGA-MOPSO hybrid optimiza-tion algorithm.