Undrained sinkhole stability of circular cavity: a comprehensive approach based on isogeometric analysis coupled with machine learning

An innovative approach that combines isogeometric analysis (IGA), limit analysis, machine learning, and the multivariate adaptive regression splines (MARS) model is presented in this paper to investigate sinkhole stability of circular cavity. The upper bound limit analysis method using IGA and second-order cone programming (SOCP) is employed to analyze circular cavity stability. Based on Bézier extraction, the non-uniform rational B-spline (NURBS) is used to generate a set of NURBS surfaces that define the boundary of the soil domain. The proposed approach is validated through comparative analysis with previous studies, demonstrating its effectiveness in accurately predicting soil stability. A large dataset consisting of 5000 randomly generated runs is used to train the machine learning algorithm that is integrated with the MARS model. The results show high accuracy, with a small mean squared error of 10 –3 , in predicting the undrained stability of circular cavities. The integration of IGA, limit analysis, machine learning, and the MARS model contributes significantly to advancing computational techniques for assessing soil stability. The proposed approach offers a comprehensive and precise tool for engineers and researchers, providing an accurate design formula for evaluating the undrained stability of circular cavities.