Automated Structural Design Optimization of Composite High-Pressure Vessels Considering Manufacturing Constraints

Pressurized hydrogen storage is a key technological element to carbon-neutral transportation systems. When designing pressure vessels for novel hydrogen aircraft or launchers, different tank geometries and configurations must be compared in the context of the overall design. However, off-the-shelf solutions are generally not available for a specific set of requirements. To address this challenge, this study presents a method that quickly generates suitable tank layups for filament-wound high-pressure vessels. The winding and thickness buildup of the layers is simulated by the tool muWind, which creates the axisymmetric shell FEM models for the structural analysis of the vessel. To apply the optimizer to thick-walled pressure vessels, an analytical model of thick-walled composite tubes was employed to correctly scale the through-thickness hoop stress distribution in the cylinder region, which cannot be captured by shell elements. The developed optimization strategy uses a layer-by-layer approach to optimize the layup of the tank. For each layer added, the winding angle is chosen by minimizing a target function that consists of the maximum Puck fiber failure criterion across all layers and a penalty function for excessive steepness or concavities in the contour. This ensures a manufacturable design for filament winding. After each added and optimized layer a postprocessing step is performed, where the stacking sequence is reordered to improve the design efficiency and the winding angles are adjusted to the closest valid winding pattern. The optimization routine is tested by comparing its results to a manually created reference design. For the same boundary conditions and safety factors, the optimization procedure is able to generate vessels with similar gravimetric efficiency as the reference design. The stress distributions for the generated vessels are also compared to higher fidelity models. The stresses in the cylinder region match closely. The stresses in the inner layers of the dome region, where no analytical scaling could be used, are generally underestimated in the shell model. As an application, the optimizer was used in the design of a small hydrogen aircraft. Through a comparison of different design points for the same hydrogen mass, but differing geometries and design parameters, the implications for mass-efficient pressure vessels are presented.