Download files
Abstract
The important role of initial imperfections on decreasing the buckling load of imperfection sensitive thin-shelled structures, commonly used on launch vehicles such as the Ariane family, has been widely studied since the 1930s. Research in this area has resulted in a variety of analytical and empirical design procedures to account for the diminished buckling loads of such shells. Some notable design guidelines which were formulated in the 1960s and are currently being used are the NASA-SP 8007 for imperfection sensitive cylinders and the NASA-SP 8019 for truncated cones; both of which are based on the lower-bound curve of empirical data. The guidelines have since been determined to be conservative and therefore unsuitable for composite structures, as they were developed for shells constructed from isotropic materials.
Numerical stochastic methods may provide a cheap and rapid solution to the calculation of knock-down factors for composite shells through their ability to incorporate a wide range of complicated imperfection types intrinsic to composites. Currently, there is no method that is able to stochastically replicate the full range of realistic imperfections for a full account of possible buckling loads.
A procedure which aims to improve the stochastic modelling of thickness and material imperfections in imperfection sensitive composite cylindrical and conical shells is investigated in this thesis. Monte Carlo simulations of axially compressed cylinders and truncated cones are performed to show that the proposed stochastic methods are able to capture the scatter in buckling loads introduced from the imperfections. When paired with other imperfection types, the simulations are able to replicate the range of buckling loads seen in test situations. The stochastic approaches developed here are also used to compare and validate the accuracy of the knock-down factors calculated by a newly proposed method called the ‘Single Perturbation Load Approach’ (SPLA), for a variety of conical geometries and lay-ups. Comparison with experimental results of a truncated cone supports the assumptions used in the simulations.