Co-authors​ :

 Gaspard LANDRON (FRANCE), Alain RASSINEUX , Christian FAGIANO (FRANCE), Emmanuel BARANGER (FRANCE), Christophe BOVET , Martin HIRSEKORN (FRANCE), Denis ESPINASSOU (FRANCE) 

Predicting the burst pressure and damage onset localization is a challenge for high-pressure hydrogen pressure vessels design. These structures are made of a plastic liner with a cylindrical part closed by two domes with metallic nozzles. Nowadays, a thermoset composite shell made of winded tapes ensures the load bearing. Since 2013, Cetim has been developing a laser-assisted filament winding process for thermoplastic tapes for industrialization purposes. The tape placement with contact on the laying surface changes the geometry of overlaps in the structure compared to the thermoset composite winding process. Current composite tank finite element (FE) models are axisymmetric. For each layer, thickness variation and stacking sequence evolution along the dome are described by analytical functions. These models offer the advantage of a limited number of degrees of freedom appropriate for non-linear behavior modeling. However, they neglect the interwoven architecture induced by the manufacturing process. For thermoplastic tanks, porosity levels are higher in the domes than in the cylindrical part, affecting the damage onset in these areas. Cetim has observed that the burst pressure of thermoplastic overwrapped composite pressure vessels cannot be predicted using current approaches. Consequently, a new modeling strategy is needed to take into account the winded architecture.
In this study, a mesh generation tool at tape scale has been developed based on actual laying trajectories. Elements are generated by elevation from a quadrangle grid to guarantee mesh conformity between overlapping tapes. However, this method generates a voxelised description of tape edges (Fig 1). Crossovers are represented by disoriented hexahedrons, generating voids at every intersection. Tape width and thickness are supposed to be constant in the whole structure. The generated geometry needs to be validated to ensure the quality of the related FE model. Consequently, Cetim has built a composite shell with ten layers of glass fiber thermoplastic tapes. Cetim has scanned the surface between each layer to take into account the influence of new layers on previous ones. These scans provide a measurement of the thickness and its variations around the vessel due to the winding patterns for each layer. A comparison between these scans and the mesh allows a better understanding of tape slippage, neglected by trajectories. Besides, the shell geometry changes when it cools down, making validation at layer scale on the final geometry necessary. Plastic separators have been added between layers to ensure each one can be distinguished on a tomography with a voxel size of 17 µm. The separators allow layer segmentation to compare shapes between the mesh and the tank for each layer. Tape segmentation allows orientation and dimensions extraction. Orientations extraction is used to verify stacking sequence. Crossover geometries and porosity levels of the tank and its 3D mesh are compared.