Optimization of Ship Structural Design by Genetic Algorithm

In general, ship basic design is based on design-spiral approach where variable design parameters are defined through repetition of trial and error. This kind of approach is a process of satisfying a lot of complicated design criteria empirically, rather than optimizing some object function. However, in order to realize optimum or rational design for a new type of structural concept with complicated design criteria, it is necessary to model the design problem as strictly as possible and accomplish more highly numerical optimization.As for ship structural optimization, the object function (hull steel weight or total cost including material, fabrication and paint cost) is closely related to principal dimensions and general arrangement. Therefore, in order to obtain better design in the optimization process, it is desirable to search broader design space with many design variables such as principal dimensions, arrangement, frame spaces, stiffener spaces, scantlings and material.This type of optimization is considered as a combinatorial optimization problem, and such optimization method that is robust and quickly converging to global optimum is required. In this paper the authors apply the genetic algorithm which simulates the law of the survival of the fittest to structural optimization problems.At first, the authors apply the genetic algorithm to minimum weight problem of a hatch cover with 6 independent variables, and compare several approaches to the genetic algorithm through numerical experiments. As a result of the examination the authors propose the genetic algorithm which attaches importance to vicinity search as the method appropriate for structural optimization problems.Then, the authors apply the genetic algorithm to the optimum design problem of a double hull tanker with 40 independent variables. It is shown that the proposed method is effective for the optimization with a variety of independent variables of different types. Using this method it is possible to obtain more accurate optimum design of ship structure by enhancing the accuracy of the estimation of object functions and many appropriate constraints.