Work

Physics-guided machine-learning surrogate for AIS-based ship fuel-consumption inventory—part a: Framework development and diagnostic evaluation

Year: 2026

Type: article

Abstract: Accurate fuel-consumption inventories are essential for greenhouse-gas monitoring, regulatory reporting, operational benchmarking, and evaluating maritime decarbonisation measures. Reliable ship fuel-consumption inventories require models that capture complex operational, environmental, and hydrodynamic interactions while remaining computationally efficient. However, existing methods often rely on oversimplified relationships that incompletely account for these coupled factors. This study presents a physics-guided machine learning surrogate for ship fuel estimation. Using a vessel-specific synthetic operating space spanning operational and environmental conditions, physics-based fuel labels were generated via a resistance-to-power chain, incorporating calm-water, aerodynamic, biofouling, and added-wave resistance, to train a multilayer perceptron surrogate. Evaluated across five real voyages of a 197 m bulk carrier using AIS, ERA5, and noon-report data, the proposed model achieved an aggregate voyage-level error of −3.22%, against STEAM2 (+5.04%), the IMO method (+8.50%), and STEAM (+13.56%) for the investigated vessel and voyages. It also recorded the lowest median and mean absolute errors at reporting-interval scales. Diagnostic analyses of residual, encounter wave, sea-state, and ablation study revealed that remaining errors stemmed mainly from restricted directional coverage and underrepresented added-wave resistance, demonstrating that surrogate performance strictly depends on embedded physics fidelity. This framework initiates a computationally efficient surrogate approach for vessel-specific AIS-based fuel estimation and a diagnostic pathway for integrating higher-fidelity hydrodynamic models.

Source: Results in Engineering

DOI: 10.1016/j.rineng.2026.112968

Authors: Hamed Majidiyan, Gholam Reza Emad, T M Rabiul Islam, Mohan Anantharaman

Institutions: University of Tasmania, Australian Maritime College

Language: English

Topic: Maritime Transport Emissions and Efficiency

Subfield: Environmental Engineering

Field: Environmental Science

Domain: Physical Sciences

SDGs: Climate action, Affordable and clean energy

Open Access status: gold

Estimated APC paid: $1,940