Artificial Intelligence-Based Modelling and Optimization of Multi-Deck Passenger Ships Structural Parameters for Minimum Stress Distribution

This study presents a hybrid artificial intelligence-driven regression approach to develop predictive mathematical models for estimating the structural stress behaviour of a multi-deck passenger ship. Seven independent design parameters (side opening size, recess position, longitudinal bulkhead length, deck opening position, bulkhead height, transverse bulkhead count, and deck opening length) were analysed in relation to the structural stress. The models were trained and tested using a dataset derived from finite element simulations, with performance evaluated through the coefficient of determination (R2) and a boundedness check to ensure physical plausibility of the results. Also, to assess the model's robustness and reliability, a 10-run cross-validation was performed with randomly shuffled training and testing datasets. Among the tested models, the Absolute First Order Trigonometric Non-linear, and the Absolute Second Order Trigonometric Non-linear models emerged as the most reliable for defined loading conditions (LC1 and LC2), respectively. These models exhibited high prediction accuracy and provided stress distribution within realistic physical bounds. Optimization was conducted using the Modified Differential Evolution algorithm, resulting in significant stress reductions of approximately 33% under LC1 and 37% under LC2 when compared to the reference study. The findings demonstrate the effectiveness of the proposed methodology in structural stress prediction and optimization, offering valuable insights for the design of marine structures.