This paper presents a combined experimental-numerical procedure for development and calibration of macroscopic crack propagation criteria in large-scale shell structures. A novel experimental set-up is described in which a mode-I crack can be driven 400 mm through a 20(+) mm thick plate under fully plastic and controlled conditions. The test specimen can be deformed either in combined in-plane bending and extension or in pure extension. Experimental results are described for 5 and 10 mm thick aluminium and steel plates. By performing an inverse finite-element analysis of the experimental results where the simulated crack growth is forced to correspond to the experimental observations, empirical criteria for ductile crack propagation emerge very cleary. Using the experiments with edge crack specimens (ECS) in combined in-plane bending and extension, crack propagation criteria are developed for steel and aluminium plates, mainly as curves showing the critical element deformation versus the shell element size. These derived crack propagation criteria are then validated against a separate set of experiments considering centre crack specimens (CCS) which have a different crack-tip constraint. The applicability of the often-used equivalent strain criterion is discussed versus a more rationally based criterion which takes into account the stress tri-axiality. A large-scale grounding experiment is also simulated showing very good agreement with measurements. The performance of the proposed model is in general good and it is believed that the presented results and experimental-numerical calibration procedure can be of use in practical finite-element simulations of collision and grounding events with the use of shell elements. As discussed, the paper provides a clean framework for further development of macroscopic crack propagation criteria in large-scale plate structures.