Structural design and optimization using neural networks and genetic algorithms of a tanker vehicle

Abstract

The aim of this article is to introduce a structural optimization tool based on the combination of neural networks and genetic algorithms, applied to optimize a cryogenic tanker vehicle. To use this tool, it is necessary to introduce an equation that defines the aspects that are going to be optimized (weight, resistance, rigidity, etc.) of a part of the vehicle (the parking zone). Sometimes, this equation is difficult or impossible to model mathematically and then the neuronal networks allow approaching this equation from some particular solutions of the equation and then we will be able to find the solution at any point in the search space on the basis of a set of values of different solutions to the equation. For the tanker vehicle, the neuronal network is used to know the maximum Von Mises stress that is impossible to model mathematically. The genetic algorithms are used to optimize the equation and allow knowing the optimal design of the vehicle tanker. The symbiosis of both techniques are very useful because it will enable us to find the solution to the equation at any point of the search space on the basis of a discrete set of solutions, following which we will be able to optimize the equation in the search space. This allows us to optimize an equation about which the only knowledge we have is a finite set of solutions in the search space. At the end with a Pareto multi-objective set, the effectiveness of the methodology has been verified, and with an extensometrical analysis, the numerical results have been correlated too.