A joint optimization by the combination of BPNN and NSGA-III for the polyester mooring system

Abstract

The design of a mooring system includes multiple variables such as the line’s span and number, orientation, segmentation, materials, counterweights, load conditions, and the requirements like dynamic performance, structural safety, and economic costs. Traditional optimization methods, relying on extensive analysis with professional tools, usually performed inefficiently after large scale computation and long-time post processing. In this study, a surrogate-assisted framework combining Back Propagation Neural Network (BPNN) models and the Non-dominated Sorting Genetic Algorithm III (NSGA-III) is proposed to the optimization of polyester mooring systems. An optimization model considering the common requirements of mooring system design was established and three independent surrogate models were developed for each objective among maximum tension, platform displacement and cable lines’ weight. Then, the NSGA-III was integrated with the surrogate models to select excellent combinations within the feasible sample space. The joint optimization by the combination of BPNN and NSGA-III was validated through a design case of polyester mooring system for a Floating Production Unit (FPU). As a result, it showed reliable prediction accuracy with errors below 5 % and time saving with 80 % less than normal operations. The results show that the proposed framework offers an efficient and accurate solution for mooring system optimization.