Discrete optimization of ship structures from the viewpoint of practical design

In practical design of ship structures, it is obvious that some structural elements, for instance, longitudinals, must be chosen from a given set of standard discrete values being available commercially. Furthermore, some integer parameters such as the number of transverse web frames in the midship segment should be considered as the design variables to be optimized. In view of those facts, we investigate especially how to deal with the above kinds of discrete design variables in this paper. A cargo ship with large hatch opening is taken as the numerical example. The structural construction cost of hold length between two transverse bulkheads is chosen as the objective function to be minimized by genetic algorithms. Besides the conventional variable design parameters, some particular parameters such as the serial number of standard shaped steel for longitudinals and for transverse frames, the number of longitudinals in a given length, the number of transverse web frames in the midship segment, etc., are also selected as the design variables from the viewpoint of practical design. For each design variable, we describe it by a specific length binary string with a corresponding relationship to its available discrete values. Through the numerical analyses, it shows that the present method is a simple yet efficient method to deal with the discrete optimization of ship structures. The search efficiency of some genetic strategies such as the elitist preserving selection, the multi-point crossover and the uniform crossover is also investigated in this paper.