| dc.description.abstract | Currently, ship designers are under increased pressure to provide energy-efficient designs and reduce greenhouse gas emissions. For a long time, ships were mostly designed and optimized for operation in calm water. However, ships should be designed to operate in real sea states, meaning that performance under different operational conditions such as wind, waves, and currents, as well as different transit speeds, should be considered. To reach this goal, there are several challenges. First of all, there are typically uncertainties to be considered during the design phase. Also, it is important to select (or develop) appropriate design approaches for optimization, and tools to enhance the accuracy of design models before developing a design optimization method. Developing reliable and practical methods that are sufficiently accurate, robust, and computationally effective in real sea states in the presence of uncertainties is also a big challenge. Furthermore, accurately estimating the added resistance in waves is also necessary to be able to design ships with improved performance at sea. Therefore, the main objectives of this research are to develop and demonstrate practical methods to optimize ship designed for operation in real sea conditions, taking into consideration the quantified uncertainties of the environment and performance predictions. These objectives are addressed in this thesis through the following research:
In the first study, a new method for optimizing ship design in real sea states is suggested, aiming to ensure that the ship performs optimally over the long term under different operational conditions. The purpose of that research is to present a probabilistic approach for optimizing ship design for different levels of propulsive power and operational conditions. A long-term analysis of all propulsion power and environmental condition combinations is conducted with the objective of optimizing the average ship speed over the long term. A Post-Panamax container ship operating on the route from Busan to Hamburg ports is used to test the efficiency of the proposed approach. In comparison with traditional methods, the proposed method results in a 14.28% increase in average ship speed over the long term. Better performance of the ship designed based on the adopted method on high seas is also reported. A trade-off analysis is also suggested to determine whether ships should be designed for a seaway or calm water. According to the results, the sea margin of 15% of the engine power, which is a typical value for sea margin, is not sufficient to satisfy the engine requirements of the studied ship.
In another study, a new method (called the “Combined method”) combining two existing empirical methods is suggested to estimate added resistance in waves. It is demonstrated that the proposed method performed better when it came to estimating added wave resistance at high waves, resonance frequencies, arbitrary waves, and low speeds.
The third research proposes a new method for predicting ship propulsion power at sea. In order to achieve this, a normalized root mean square error (NRMSE) is defined by comparing the results of the power prediction model with the ship in-service data of a similar ship in order to reach more accurate power prediction models. Afterward, a tuning surrogate-based optimization problem is developed to minimize the NRMSE in order to find the best combination of existing methods and their input parameters (calm water resistance, added resistance in wind and waves, wetted surface, wake, fouling, etc.), resulting in the lowest NRMSE and thus the best power prediction accuracy. In theory, the proposed method can be applied to different ship types and application scenarios. In that study, it is tested on a general cargo ship operating between Italy and Norway. Based on the results, the proposed approach is effective with an 86% percentage difference between the NRMSE of the best and worst power prediction models. The model developed based on the adopted approach showed robustness in different ship voyages, missions, sampling time intervals, and time history lengths compared to a well-known machine learning method like the artificial neural network (ANN) method. This method also performs well when compared with traditional power prediction models.
In the final study, a two-stage method is proposed for designing ships under uncertainty, operating at different speeds under real sea conditions, addressing both aleatory uncertainties due to weather and epistemic uncertainties resulting from the model and methodology. As a test case, a general cargo ship is considered. Up to 26.91% improvement in the ship performance compared to traditional methods is reported. Also, the effect of a ship design based on a slow steaming operational profile is explored. For all cases studied, the adopted approach proved effective in reducing epistemic and aleatory uncertainties. Verification of the suggested approach with in-service data also reaffirms the effectiveness of the proposed approach in reducing the effects of both epistemic and aleatory uncertainties in the studied ship design problem. | en_US |
