Optimization of nonlinear structures based on object-oriented parallel programming

This contribution presents a computational framework for simulation and gradientbased structural optimization of geometrically nonlinear and large-scale structural finite element models. CAGD-free optimization methods have been developed to integrate shape optimization in an early stage of design and to reduce the related modelling effort. To overcome the problem of an increasing numerical cost due to the large design space, the design sensitivities for objectives and constraints are evaluated via adjoint formulations. A new parallel computation strategy for sensitivity evaluation is presented which takes advantage of a completely parallelized simulation and optimization environment. Two application examples illustrate the method and demonstrate the high parallel efficiency.