Global Optimization Methods for Ship Design Problems

Most of recent papers on ship design are based on local optimization algorithm. Nevertheless, a real break-through can be only obtained with a Global Optimization (GO) approach. The purpose of this paper is to compare different GO methods on a realistic ship design problem, to reduce the heave motion of a ship advancing in head seas. The objective function is evaluated by simulation code and hence its first derivatives are not available. Moreover, our aim is to locate a global solution avoiding the local ones. Thus, we have considered derivative-free global optimization methods. As first step, we have performed a numerical comparison on a class of well known test-problems among some of these methods which are considered suitable for solving optimal design problems. The method which proved to be the best on the test set is based on the use of a filled function. Then, on the basis of this comparison the most promising methods have been further investigated. In particular, a new filled function has been defined and its theoretical properties have been investigated. Finally, we report the numerical results obtained by employing the best methods on the optimal design problem.