Creep parameter inversion for high CFRDs based on improved BP neural network response surface method

The creep parameters of rockfill materials obtained from engineering analogy method or indoor tests often cannot accurately reflect the long-term deformation high Concrete-Faced Rockfill Dams (CFRDs). This paper introduces an optimized inversion based on multi-population genetic algorithm-improved BP neural network and response surface (MPGA-BPNN RSM). used for are determined by parameter sensitivity analysis statistical orthogonal test method. MPGA-BPNN RSM, validated root-mean-square error, mean absolute percentage squared correlation coefficient (R2), etc., completely reflects between settlement calculation values finite element (FEM). MPGA objective function to obtain optimal parameters. results show that Xujixia CFRD calculated FEM using has great consistency with monitored both in size distribution, suggesting model introduced feasible effective. can improve efficiency prediction accuracy applications.