Fluid storage tank structures have been of interest for a long time, because these have unique characteristics due to the interaction between fluid and structure. This interaction which causes the sloshing, can be a severe problem in vehicle stability and control. In this thesis, the conventional methods for analysis of the partially filled fuel tanks have been enumerated and then the fuel tank of the locomotive ER 24 has been considered as a real analyses model. Stress analysis has been done based on both the Computational Fluid Dynamics ( CFD ) and structural models. In the structural modeling, the fuel tank was assumed to be fully filled with fuel and therefore, the greatest forces will be generated. Then, the fuel tank has been modeled and analyzed using finite element method under suggested loading conditions by the DIN EN 12663 standard, in railway applications and real boundary conditions. In six different loading conditions, high stressed areas have been identified. Also, in the CFD modeling it is considered that the locomotive had been moving in a certain speed and suddenly stopped at a constant acceleration. Next, the behavior of the fuel and air during the braking time were investigated by using a two-phase model. The distribution of the pressure on the surface of baffles caused by sloshing was extracted. Finally, by comparing the distribution of pressure caused by sloshing phenomena and suggested standard loading conditions, optimal design of the fuel tank was taken into consideration. After identifying the critical areas of the stress distribution, optimization of the primary design of the fuel tank has been studied. The objective function was assumed as the mass and manufacturing cost of the fuel tank. The manufacturing cost consists of the cost of welding, cutting and bending in the process of manufacturing of the baffles. The geometric properties of the baffles including the thickness, angle of bent, radius of the circular holes and the number of transversal baffles were considered as the design parameters. This multi-objective optimization problem was solved by employing Non-dominated Sorting Genetic Algorithm ( NSGA-II ) and Artificial Neural Networks ( A ). Two Multi-Layer Perceptron ( MLP ) networks were used to approximate mass and maximum Von-Mises equivalent stress that later have been used in the optimization process. Each network has one hidden layer with 6 neurons and a single neuron in its output layer. The networks were trained using 44 data sets generated using a normally distributed set of input vectors across the design space. It has been shown that the optimized fuel tank not only provides the safety conditions, but also, compare with the primary design, the mass and manufacturing cost are reduced by %39 and %73, respectively. Keywords : Optimal design, Multi-Objective Optimization, Sloshing, Artificial Neural Networks, Fuel Tanks, Baffles.
