Multi-objective Optimization of Curtain Grouting Construction Scheme with Ensemble Residual Surrogate Model

Curtain grouting construction is crucial to the stability and seepage safety of the dam foundation. The traditional way of designing a grouting construction scheme is to rely on the limited number of test boreholes and engineering experience, failing to take into account the economic and safety factors quantitatively. We propose a multi-objective optimization scheme with an ensemble residual surrogate model for the curtain grouting construction design. Based on the dataset obtained from 3D stochastic fracture network grouting simulation, the ensemble residual surrogate model based on the Gaussian regression process is established to reduce the computational time and resource consumption significantly. The weights of residuals in the surrogate model are optimized with the jellyfish search algorithm to obtain the best prediction performance. The construction scheme is optimized afterwards considering the objectives of economic and safe construction, with construction parameters as design variables. The dynamic constraint non-dominated sorting genetic (DCNSGA-III) algorithm is used for the multi-objective optimization. The proposed method is applied to the Baihetan dam in southwestern China to optimize the curtain grouting construction of a riverbed section. Compared with the existing construction scheme designed with engineering experience, the permeability of the optimized curtain is increased by \(71.50\%\) $71.50\%$ , and the cost is reduced by \(11.51\%\) $11.51\%$ , which verifies the effectiveness of the proposed model.