Search papers, labs, and topics across Lattice.
This paper introduces Physics-Informed Kolmogorov-Arnold Networks (PI-KANs) to predict vessel shaft power and fuel consumption, addressing the limitations of purely physics-based and data-driven models. PI-KANs integrate interpretable univariate feature transformations with a physics-informed loss function and a leakage-free chained prediction pipeline. Experiments on operational data from five cargo vessels demonstrate that PI-KANs outperform polynomial methods and neural network baselines in terms of MAE, RMSE, and R^2, while also rediscovering physically consistent relationships between operational parameters.
By rediscovering known physics like cubic speed-power relationships, PI-KAN shows that hybrid models can achieve both accuracy and interpretability in vessel performance prediction, outperforming black-box neural nets.
Accurate prediction of shaft rotational speed, shaft power, and fuel consumption is crucial for enhancing operational efficiency and sustainability in maritime transportation. Conventional physics-based models provide interpretability but struggle with real-world variability, while purely data-driven approaches achieve accuracy at the expense of physical plausibility. This paper introduces a Physics-Informed Kolmogorov-Arnold Network (PI-KAN), a hybrid method that integrates interpretable univariate feature transformations with a physics-informed loss function and a leakage-free chained prediction pipeline. Using operational and environmental data from five cargo vessels, PI-KAN consistently outperforms the traditional polynomial method and neural network baselines. The model achieves the lowest mean absolute error (MAE) and root mean squared error (RMSE), and the highest coefficient of determination (R^2) for shaft power and fuel consumption across all vessels, while maintaining physically consistent behavior. Interpretability analysis reveals rediscovery of domain-consistent dependencies, such as cubic-like speed-power relationships and cosine-like wave and wind effects. These results demonstrate that PI-KAN achieves both predictive accuracy and interpretability, offering a robust tool for vessel performance monitoring and decision support in operational settings.
