Designing Ships Using Constrained Multi-objective Efficient Global Optimization

摘要

A modern ship design process is subject to a wide variety of constraints such as safety constraints, regulations, and physical constraints. Traditionally, ship designs are optimized in an iterative design process. However, this approach is very time consuming and is likely to get stuck in local optima. Not only does this optimization problem have complex constraints, it also consists of multiple objectives like resistance, stability and cost. This constrained multi-objective optimization problem can be dealt with much more efficiently than through the traditional approach. In this paper, we propose a novel global optimization algorithm that explores the design space with the help of integrated software tools that are capable of simultaneous evaluation of the ship objectives and constraints. The optimization algorithm proposed uses the S-Metric-Selection-based Efficient Global Optimization (SMS-EGO) in combination with constraint handling techniques from an algorithm called Self-Adjusting Constrained Optimization by Radial Basis Function Approximation (SACOBRA). Since the evaluation of these ship designs is expensive in terms of computational effort, it is crucial for the algorithm to find feasible near-optimal solutions in as few evaluations as possible. In this paper, it is shown that the proposed Constrained Efficient Global Optimization (CEGO) algorithm can significantly improve ship designs by automatic optimization using a small evaluation budget.