Ice-Seabed interaction analysis in sand using a gene expression programming-based approach

Arctic subsea pipelines are usually buried for physical protection against the ice-induced scours. Determination of the maximum horizontal deformations for guaranteed operational integrity and cost-effective design is probably the most challenging aspect of subsea pipelines in ice-prone areas. The large uncertainties associated with the design of ice-protected subsea pipeline using the existing empirical equations, advanced experimental and sophisticated numerical studies with significant time and cost impacts are usually preferred to ensure the sufficiency and the cost-effectiveness of the pipeline design against the ice attack. This has caused the industry to keep looking for more effective and less-costly solutions for modeling the ice impact on buried pipelines. In this study, a Gene Expression Programming (GEP) model representing the Artificial Intelligence (AI) approaches was used for the first time to simulate the subgouge soil deformation in the sand. A database was constructed using some of the published experimental studies identifying the key input parameters including soil depth, bearing pressure, the maximum vertical extent of subgouge deformation, attack angle, and dilation index. Subsequently, six GEP models were developed and validated by using a K-fold cross-validation method. The performance of the GEP method was compared with an Artificial Neural Network (ANN) model, and uncertainty analysis (UA) along with a partial derivative sensitivity analysis (PDSA) was conducted to assess the influence domain of the key parameters. The study showed that the evolutionary numerical methods could be used as an accurate and cost-effective alternative for modeling of the ice-induced subgouge deformations.