On the existing test dataset of isotropic cylindrical metal shells under axial compression and the design of modern metal civil engineering shells

In the structural engineering limit state design philosophy of the Eurocodes, a target reliability is achieved by using partial safety factors on both the loading and the resistance that are, in principle, to be calibrated based on test data. The partial factor that is currently used for the buckling limit state of metal shells has been adopted from the knowledge base on other structural elements and has been retained for reasons of historical continuity and to maintain a close relationship between all Eurocode steel design standards. However, the mechanics of the behaviour of thin-walled metal shells gives strong reasons to believe that the partial factors for buckling should be dependent on the shell geometrical form, slenderness, load case and quality of fabrication for the target reliability to be consistently achieved. None of these are currently considered in defining the partial factor for resistance. The most ubiquitous thin-walled metal shell structures are imperfection-sensitive cylinders under uniform axial compression. A dataset of many hundreds of these test results has thus been accumulated over many decades, though it is of variable quality and sparsely documented. This paper shows that this dataset is an entirely inappropriate basis on which to calibrate the safety level of full-scale metal civil engineering shells. Indeed, the professional community should face the uncomfortable reckoning that an experimental test dataset suitable for the reliable calibration of the safety level of design relationships for full-scale metal civil engineering shells may likely never come into existance, and that the bolder approach of extensive computational simulation must instead be embraced.