A Comparison of Global Optimization Methods with Application to Ship Design

Aim of this work is to present and compare Global Optimization algorithms for design optimization, focusing on their effectiveness and efficiency, in the view of their possible use in ship design. Algorithms have been chosen from different classes of global optimization methods, i.e decomposition methods (Diagonal Rectangular Algorithm for Global Optimization, DRAGO), evolutionary (Particle Swarm Optimization, PSO) and genetic algorithms (GA). These algorithms are initially tested on a single objective function problem, namely the reduction of the Response Amplitude Operator's heave peak of a high speed ship, advancing in head seas. As a second test, a new multiobjective version of a modified PSO algorithm and a GA are compared on a set of well-known algebraic test functions, to assess the capability of these two approaches in solving real life design problems. The goal is to have an accurate description of the Pareto front (which identify all the non-dominated optimal solutions) using the minimum number of function evaluations, envisaging the use of computationally expensive simulations in their computation.