Multilayer-perceptron-based prediction of sand-over-clay bearing capacity during spudcan penetration

A method combining a multilayer perceptron and Large-Deformation Finite-Element (LDFE) analysis is proposed for addressing large errors in conventional formulations and time-consuming numerical simulations for predicting the bearing capacity of sand over clay during spudcan penetration. The method yielded accurate bearing capacity curves, agreeing well with the LDFE and centrifuge test results. The method can be used for quickly and accurately predicting the bearing capacity curves and for generating various data samples for theoretical studies of the effects of spudcan penetration into sand over clay on the bearing capacity. The critical depth of the bearing capacity curve was revised based on several samples generated using the method, and the result was used for fitting the bearing capacity curve. The fitted equations in this study yielded high-accuracy predictions for different scenarios. •An MLP-LDFE method is proposed for fast and accurate prediction of bearing capacity.•The effect of optimizer type and number of training sets on prediction accuracy is investigated.•The effect of return sand on the bearing capacity was studied.•The critical depth of the bearing capacity curve is modified.•The prediction equation of the bearing capacity curve is presented.