A Simulation Study on the Performance of Jacket Type Offshore Structures Using Machine Learning Algorithms

Abstract

In this study, the behaviors of jacket-type offshore structures are numerically investigated. The examined four-legged models with a total height of 60 m have four layers and three different cylindrical element sizes are fixed to the seabed. The structures are under the effect of environmental forces, including wind and wave loads, as well as operational loads. Three different marine environments have been generated in environmental modeling. Thus, the parametric study has been performed using bidirectional fluid-structure interaction (FSI) analyses of 36 models. Structural outputs such as displacement, reaction force, and stress values are determined by numerical analyses. In the second part of the study, the implementation of machine learning algorithms, including Xgboost, Random Forest, and Support Vector regressors, is employed to automate the assessment of performance in jacket-type offshore structures. The evaluation of these machine learning models in predicting the displacement, reaction force, and stress values of offshore jacket structures is conducted, revealing Xgboost as the most promising technique, although with satisfactory overall performance across all algorithms. These findings provide empirical evidence supporting the feasibility and applicability of employing machine learning methodologies in the analysis of performance for jacket-type offshore structures.