Multi-objective optimization of ship structures: Using guided search vs. conventional concurrent optimization

Structural optimization regularly involves conflicting objectives, where beside the eligible, weight reduction, increase in e.g. safety or reliability is imperative. For large structures, such as ships, to obtain a well-developed Pareto frontier can be difficult and time-demanding. Non-linear constraints, involving typical failure criteria, result in complex design space that is difficult to investigate. Evolutionary algorithms can cope with Such problems. However they are not a fast optimization method. Here we aim to improve their performance by guiding the search to a particular part of Pareto frontier. For this purpose we use a genetic algorithm called VOP, and use it for optimization of the 40 000 DWT chemical tanker midship section. Beside weight minimization, increase in safety is investigated through stress reduction in deck structure. Proposed approach suggests that in the first stage one of the objectives is optimized alone, preferably more complicated one. After obtaining satisfactory results the other objective is added to optimization in the second stage. The results of the introduced approach are compared with the conventional concurrent optimization of all objectives utilizing widespread genetic algorithm NSGA-II. Results show that the guided search brings benefits particularly with respect to structural weight, which was a more demanding objective to optimize. Salient optimized alternatives are presented and discussed.