Utilizing Artificial Neural Network Ensembles for Ship Design Optimization to Reduce Added Wave Resistance and CO₂ Emissions

Abstract

Increased maritime cargo transportation has necessitated stricter management of emissions from ships. The primary source of this pollution is fuel combustion, which is influenced by factors such as a ship’s added wave resistance. Accurate estimation of this resistance during ship design is crucial for minimizing exhaust emissions. The challenge is that, at the preliminary parametric design stage, only limited geometric data about the ship is available, and the existing methods for estimating added wave resistance cannot be applied. This article presents the application of artificial neural network (ANN) ensembles for estimating added wave resistance based on dimensionless design parameters available at the preliminary design stage, such as the length-to-breadth ratio (L/B), breadth-to-draught ratio (B/T), length-to-draught ratio (L/T), block coefficient (CB), and the Froude number (Fn). Four different ANN ensembles are developed to predict this resistance using both complete sets of design characteristics (i.e., L/B, B/T, CB, and Fn) and incomplete sets, such as L/B, CB, and Fn; B/T, CB, and Fn; and L/T, CB, and Fn. This approach allows for the consideration of CO2 emissions at the parametric design stage when only limited ship dimensions are known. An example in this article demonstrates that minor modifications to typical container ship designs can significantly reduce added wave resistance, resulting in a daily reduction of up to 2.55 tons of CO2 emissions. This reduction is equivalent to the emissions produced by 778 cars per day, highlighting the environmental benefits of optimizing ship design.

Keywords

• artificial neural networks (ANNs)
• CO₂ emissions
• ship design
• optimization
• added wave resistance
• hull shape parameters