This article focuses on predicting the instant of failure in a real scale component of complex geometry and loading using a ductile damage model calibrated exclusively on small-scale laboratory specimens of relatively simple shape. The ductile behaviour of a strain hardened aluminium alloy AA1050, formed into a thin-walled component, is modelled by a coupled ductile fracture locus model presented in a recent study (Baltic et al., 2020). The component is exposed to high internal pressure and has a safety vent designed for safe pressure handling. The extensive plastic deformation in the safety vent leads to localised ductile failure occurring at a limit load. The pertaining material parameters were calibrated solely from basic ductile fracture experiments in the preceding work (Baltic et al., 2020), where the bottom section of the thin-walled component was machined into notched and shear samples to characterize different states of stress and to construct a well-defined fracture locus. Although the calibrated material model relies on the local fracture strain measurements, it involves a regularization as a function of the length scale defined as a width of the observed localisation band from Digital Image Correlation (DIC) analysis. In the current study the calibration on small-scale specimens is complemented by a large-scale specimen to determine the length scale correction crucial for capturing the correct width of the localisation band in the analysed structure. This is necessary because the failure initiation zones of the calibration specimens and the real size structure, i.e. their gauge lengths where the localisation band appears, vastly differ in size. Finite element (FE) model results are compared to measurements of the deformation of the aluminium component under pressure and maximum load prior to failure. The numerical and experimental results show an excellent agreement and consistent fracture predictions for various mesh discretizations.