ABSTRACT
Metaheuristic algorithms are popular for solving engineering optimization problems because of their robustness and simplicity. However, they are computationally inefficient and can have difficulties finding optimal solutions in highly-constrained problems. This paper presents the benefits of a novel constraint-handling approach to improve multi-objective optimization with genetic algorithm. High-performing designs that violate constraints are repaired based on other designs created in the optimization process. Case study involves midship section of a 40,000 dwt tanker, where structural weight and deck adequacy are optimized. The approach is implemented into genetic algorithm called NSGA-II. The modified algorithm discovers more competitive designs with larger spread of the trade-off frontier. The modified algorithm outperforms the original in every iteration of the search. The benefits are pronounced especially in the beginning of the optimization, which is very useful for real-life design where optimization might be allowed to run for only a limited number of iterations.
