Studies on Meta-Modeling for Lazy-Wave Steel Catenary Risers

Steel Catenary Riser (SCR) design is a complex issue for the petroleum industry. The fact that it is strongly influenced by safety and cost saving factors has motivated the use of optimization techniques mostly based on metaheuristic algorithms such as Genetic Algorithms, Artificial Immune Systems and Particle Swarm Optimization. However, this particular offshore engineering problem requires high computational costs, associated to the time-domain nonlinear dynamic analyses with Finite Element (FE) models that are needed for a large number of loading cases for each candidate solution of the optimization process. This fact motivates studies the use of meta-models (or surrogate models) to replace the expensive FE analyses, leading to results with adequate accuracy and remarkably lower computational costs. In this context, this work applies as meta-models Artificial Neural Networks (ANN) and Multivariate Adaptive Regression Splines (MARS) to estimate the response of SCR risers in a lazy-wave configuration. Case studies are presented to assess the results comparing to a FE analysis.