An optimal design of marine systems based on neuro-response surface method

Abstract

The shapes of marine system have an effect on various performances; hydrodynamic, structural, propulsion. In case of engineering structure including the marine system, most performances are determined during initial design stage. However, the prediction of system performance is very difficult because performance analysis is time consuming. Thus, if the design techniques considering the performance were developed, it will be very useful design tool in the initial design stage. The major objective in this research is establishing a design methodology for marine system in the initial design stage. For this purpose, we proposed a framework for optimal design based on the Neuro-Response Surface Method (NRSM). The constructed framework is composed of three parts: the definition of the geometry, the generation of the design space using NRSM, and an optimization process. The optimization algorithm of the constructed framework uses a Back-Propagation Neural Network (BPN) and Non-dominated Sorting Genetic Algorithm-II (NSGA-II). The BPN is used for generating the design space, and the optimization process is done using NSGA-II. Through case study on a 5MW TLP-type wind turbine and ultimate strength of ship stiffened panel, we have confirmed the usefulness of the constructed framework in view of hydrodynamic performances and structural performances.