Optimum Design of Liquified Natural Gas Bi-lobe Tanks using Finite Element, Genetic Algorithm and Neural Network

A comprehensive set of ten artificial neural networks is developed to suggest optimal dimensions of type ‘C’ Bi-lobe tanks used in the shipping of liquefied natural gas. Multi-objective optimization technique considering the maximum capacity and minimum cost of vessels are implemented for determining optimum vessel dimensions. Generated populations from a genetic algorithm are used by Finite Element Analysis to develop new models and find primary membrane and local stresses to compare with their permissible ranges using PYTHON coding. The optimum design space is mathematically modeled by training ten artificial neural networks with design variables generated by the Taguchi method. The results are compared with actual design data and the 93% achieved accuracy shows the precision of the developed design system.