Seakeeping metamodel of fast ro-ro ships: RSM or ANN technique?

Abstract

Current design practice in selecting main geometrical characteristics of new ships does not give due consideration to seakeeping yet, particularly at conceptual design phase. A more rational approach should be taken when designing for seakeeping. The very idea is to develop an efficient seakeeping prediction model to be included in a comprehensive multiattribute optimisation procedure for optimisation of conceptual design. To facilitate this strategy, an extensive database of modern fast ro-ro vessels was built with the aim of investigating the effect of different hull forms on seakeeping behaviour in rough sea. Synthesis of seakeeping predictions is carried out with particular regard to vertical accelerations in actual operating conditions, after decoupling the dynamic characteristics of the vessel from the wave spectrum properties. Thus, it was possible to develop low-order regression equations accurate enough as descriptors of seakeeping properties, which are the basis to develop meta-models using either response surface methodology or artificial neural networks. The pros and cons of the two approaches have been compared. They look very promising for the purpose of conceptual ship design.