Comparative Analysis of Machine Learning Methods for Predicting Structural Responses in Ship Hull Monitoring

Abstract

Monitoring of ship hulls and their technical equipment is a critical component of ensuring ship safety and operational efficiency. It plays a key role in maintaining structural integrity under dynamic and often extreme loading conditions encountered at sea. Accurate prediction of structural responses not only helps in preventing structural failures but also enables the optimization of performance, reducing operational risks and costs. The structural responses of a ship subjected to wave-induced loads can be modeled through a combination of hydrodynamic simulations and Finite Element analyses. While these simulations offer deep insights into the behavior of ship structures under varying conditions, their computational intensity and complexity present significant challenges for real-time applications. To address these limitations, this research evaluates the efficiency and accuracy of machine learning methods, specifically Artificial Neural Networks and XGBoost, in approximating and predicting structural responses.