Multi-step optimization of large composite structures involves formulating the stiffness distribution in terms of lamination parameters that provide a continuous design space, followed by a second optimization in the discrete design space to obtain equivalent stacking sequences. The retrieval of stacking sequences is often challenging as a result of additional manufacturing constraints such as blending in the discrete design space which are not factored into the initial continuous optimization. By subjecting the continuous optimization to blending constraints, a better match between the discrete and continuous design spaces is achieved which in turn improves the retrieval of stacking sequences with equivalent stiffness and mechanical behaviour.
An open-source Python application capable of multi-component optimization was developed in this thesis in order to revisit the application of lamination parameter matching objective functions in the discrete step. By subjecting these objectives to mechanical constraint penalties, the vicinity of reference continuous optimum could be explored more effectively to steer the discrete solutions towards feasible design regions. A comparison was made between discrete optimization according to blended and unblended continuous designs points, where it was found that the former translated to improvements in the retrieved stacking sequences with a considerable decrease in weight versus target continuous references without blending constraints of an aircraft wing.