Due to changes in the environmental load and deformation of rockfill bodies, cracks likely form and develop in the face slab of concrete face rockfill dams (CFRDs), affecting the working behaviour of face slabs and the safe operation of dams. Considering that the uncertainty of construction quality leads to the randomization of rockfill material parameters, the probability function and coefficient of variation in the parameters of rockfill materials were determined based on field test data, and then, parameter samples were obtained by Latin hypercube sampling (LHS). LHS and the long- and short-term memory (LSTM) algorithm were applied to obtain the response surface function.

In order to avoid the phenomenon of over fitting and low prediction accuracy of conventional time series model due to not considering the non-linear and random environmental quantity, the integrated empirical mode decomposition (EEMD) method is used to decompose the actual seepage flow monitoring value of water measuring weir, and multiple groups of stable eigenmode functions (IMF) and residual quantities (R) are generated, and then the correlation vector machine (RVM) is used to decompose some of them The IMF series and R are fitted and predicted.