RIASSUNTO
ABSTRACT
To improve the ship hull optimization efficiency and take full advantage of the non-linear fitting capability of neural networks and the fast random search capability of genetic algorithms, the Wigley hull optimization based on artificial network and genetic algorithm is investigated in the present paper. The in-house hull form optimization software OPTShip-SJTU is firstly applied to obtain a series of new hull form and to calculate these hull resistances. Then a surrogate model of 3-layer BP neural network is constructed based on the sample data and a genetic algorithm is used to optimize the design of the Wigley ship with the total resistance minimum as the optimization objective function. During the calculation of hull hydrodynamics, potential flow solver NMShip-SJTU combined with ITTC formula is adopted to efficiently obtain the total resistance of the Wigley hull. The verification is also carried out to ensure the reliability of the optimization result. The results show that the resistance performance of the Wigley hull can be improved by designing the hull form reasonably. Besides, the form of bow bulbous is essential for the decreasing of total resistance according to the parameters sensitivity analysis. The design method—artificial network and the genetic algorithm can accurately work out the minimum resistance hull form and can be taken as a practical and efficient design tool.
INTRODUCTION
With the implementation of the green ship and Ship Energy Efficiency Design Index (EEDI), how to reduce fuel consumption and carbon emission becomes the focus of the attention of shipyards and ship owners. One way to alleviate this problem is to optimize ship-shape curves. Based on the original ship, the ship-shape curves are optimized to reduce the wave-making resistance of the hull, and the ship-shape line also can be optimized with multiple objectives considering the ship's 6-DOF motion index.
The method of combining neural networks and genetic algorithm is used widely in different fields. Wang, Han, Sun, and Guo (2020) combined the elliptic basis (EBF) neural network approximation model and genetic algorithm to optimize the KP505 propeller, obtained the optimal design scheme theoretically and improved the optimization efficiency. Zeng, Ding, and Tang (2010) used the BP neural network and genetic algorithm to establish a new method for the optimal design of ship propeller based on the original map design method. Koushan (2003) used the genetic algorithm and neural networks to optimize the resistance and wave-making of a high-speed ship, and the optimization effect was obvious. Xu, Zhou, and Wang (2017) used the neural networks and genetic algorithm to optimize the ship's mooring system, and the optimization result is well. Yan, Liu, Xu, and Feng (2013) used the BP neural network and genetic algorithm to obtain the seaworthiness layout of trimaran ships with different layouts at different speeds. Wang, Lu, and Wang (2020) applied neural network and genetic algorithm to the airfoil optimization, optimized FFAW3- 301 airfoil have better aerodynamic performance. The optimization results showed that the optimization method was feasible. Lv, and Wang (2018) use the RBF neural network and genetic algorithm to optimize the strength of the ship hull after the broken. Chen and Ye (2009) firstly used the genetic algorithm to optimize the weights of the neural network, and then used the optimized neural network to predict the resistance of series 60 ship types. The neural network is simple and fast to calculate the resistance of ships, which can be applied to the calculation of ship resistance. Lin, Chen, Luo, and Wang (2019) analyzed a large number of data collected during the operation of a bulk cargo ship and used BP artificial neural network for training under the condition of considering fuel consumption. The fuel consumption rate optimization model is based on the neural network and the genetic algorithm is established. Xu (2012) used the BP neural network to optimize the layout of the trimaran with static water resistance as the target. From above all, we can realize that the BP neural network surrogate model is applied in optimization. But for hull optimization, it doesn’t been applied for the wide hull.