Abstract
This study presents a deep neural network (DNN) model to estimate wave-added resistance using ships main particulars like length, beam, draft, block coefficient, and slenderness ratio. The training dataset is generated from strip-theory calculations across various hull geometries and operational conditions. The model provides fast and reliable wave resistance predictions, outperforming simplified empirical methods which often used in the industry for evaluation of ship performance. The DNN model delivers robust results when validated against direct numerical calculations from the DTU Solver. Additionally, we apply the model within a vessel performance framework to predict propulsion power and analyze biofouling for Ultrabulk ships.
Key factors like hyperparameter tuning and dataset selection greatly impact the performance of deep learning models. In this research, the numerical calculations, rather than experimental data, form the training dataset, allowing a more generalized and versatile approach. By optimizing learning rates and other parameters, the model overcomes issues like local minima. The addition of more hull data points and input features could further enhance the accuracy of the DNN model.
The model's integration into a ship performance framework enables operators to estimate delivered engine power and compare it with noon reports. This assists in evaluating hull and propeller efficiency, leading to more cost-effective voyages and energy-efficient ship operations.
The developed tool has potential for further refinement into a more accurate ship performance monitoring system. This system could improve ship performance through maintenance optimization and quantify the impact of energy-saving retrofits or advanced anti-fouling coatings. These efforts directly reduce greenhouse gas emissions and improve fleet environmental ratings, such as the Energy Efficiency Operational Indicator (EEOI) and Carbon Intensity Indicator (CII).
Key factors like hyperparameter tuning and dataset selection greatly impact the performance of deep learning models. In this research, the numerical calculations, rather than experimental data, form the training dataset, allowing a more generalized and versatile approach. By optimizing learning rates and other parameters, the model overcomes issues like local minima. The addition of more hull data points and input features could further enhance the accuracy of the DNN model.
The model's integration into a ship performance framework enables operators to estimate delivered engine power and compare it with noon reports. This assists in evaluating hull and propeller efficiency, leading to more cost-effective voyages and energy-efficient ship operations.
The developed tool has potential for further refinement into a more accurate ship performance monitoring system. This system could improve ship performance through maintenance optimization and quantify the impact of energy-saving retrofits or advanced anti-fouling coatings. These efforts directly reduce greenhouse gas emissions and improve fleet environmental ratings, such as the Energy Efficiency Operational Indicator (EEOI) and Carbon Intensity Indicator (CII).
| Original language | English |
|---|---|
| Publication date | 2025 |
| Number of pages | 1 |
| Publication status | Published - 2025 |
| Event | The 25th DNV Nordic Maritime Universities Workshop - Technical University of Denmark, Kgs. Lyngby, Denmark Duration: 30 Jan 2025 → 31 Jan 2025 |
Workshop
| Workshop | The 25th DNV Nordic Maritime Universities Workshop |
|---|---|
| Location | Technical University of Denmark |
| Country/Territory | Denmark |
| City | Kgs. Lyngby |
| Period | 30/01/2025 → 31/01/2025 |
