A new approach for seepage parameter inversion of earth-rockfill dams based on an improved sparrow search algorithm

This paper introduces a novel seepage parameter inversion method for earth-rockfill dams. The method utilizes pore water pressure data and employs a radial basis function (RBF) neural network as a surrogate model, which is optimized with the improved chaos sparrow search algorithm (ICSSA) using a hybrid strategy. The improved surrogate model (ICSSA-RBF) establishes a nonlinear relationship between the seepage parameter and pore water pressure. Unlike the original algorithm, the algorithm proposed in this paper can avoid three principal problems: the optimization search falling into a local optimal value, the population diversity decreasing during the iteration process, and the RBF neural network being prone to overfitting. The ICSSA, which is proficient in recognizing an objective function's significant values, is also chosen for identifying the parameters. To validate the effectiveness of the proposed approach, four classical test functions, a numerical model, and an actual engineering project are considered for the comparative analysis. The study outcomes reveal that ICSSA-RBF exhibits an exceptional level of prediction accuracy, with a mean relative error of less than 5 parts per thousand. The findings also affirm the potential of the proposed approach in parameter identification and its superiority in facilitating the evaluation of seepage parameters in earth-rockfill dams.