Influence of material configurations in the design of filament-wound composite pressure vessels with combined load investigated using deep learning evolutionary algorithms

This study focuses on the analysis and optimization of filament-wound composite pressure vessels loaded with a combination of mechanical and thermal loading, which occurs during the curing process and subsequent cooling to the ambient temperature. The state of stress in the pressure vessel is determined using the analytical description based on the classical lamination theory supplemented by netting theory. The analytical description is used for generating robust input data for the data-driven deep learning evolutionary algorithm EvoDN2, to carry out machine learning followed by many-objective optimization, to compute the optimal tradeoff between the nominal pressure and weight of the pressure vessel, maintaining a high level of safety. The influence of material configurations on the design of pressure vessels is also studied, where four material configurations of E-glass/DA 4518U, T300/N5208, AS4/3501-6, and Kevlar 49/CYCOM 919 are considered. The results indicate that the usage of aramid fiber in the design has its limits, whereas the materials based on carbon-epoxy show an impressive performance and great strength-to-weight ratio.