Computational & Technology Resources
an online resource for computational,
engineering & technology publications
PROCEEDINGS OF THE SECOND INTERNATIONAL CONFERENCE ON PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING
Edited by: P. Iványi and B.H.V. Topping
Finite Element based Structural Optimization using Object-Oriented Parallel Programming
H. Masching, M. Fischer, M. Firl and K.-U. Bletzinger
Chair of Structural Analysis, Technische Universität München, Germany
H. Masching, M. Fischer, M. Firl, K.-U. Bletzinger, "Finite Element based Structural Optimization using Object-Oriented Parallel Programming", in P. Iványi, B.H.V. Topping, (Editors), "Proceedings of the Second International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 67, 2011. doi:10.4203/ccp.95.67
Keywords: structural optimization, finite element based optimization, parameter-free optimization, parallel programming, object-oriented programming, finite element method, C++.
The field of computational structural analysis is gaining more and more importance in product design. In order to obtain an impression about possible approaches of design improvement already in an early planning phase, an efficient optimization tool is required that needs little modelling effort. This demand can be satisfied by a combined analysis and optimization tool working on the same model. For this purpose the finite element based optimization method is an excellent approach, leading to the highest possible diversity within the optimization process .
The advantages of an object-oriented and module-like structured finite element code including an optimization toolbox are presented. Additionally different approaches of parallel optimization strategies are compared and a short insight into a domain decomposition based parallel simulation and optimization is given, including the HALO approach in order to limit computational overhead due to inter-process communication.
In order to overcome the problem of increasing computational cost arising from the large design space an efficient method of semi-analytic sensitivity analysis  using an adjoint formulation is presented.
Beside programming and numerical topics the necessity for the consideration of nonlinear kinematics in analysis and optimization  is shown using a practical example. In this example the potential of nonlinear optimization in combination with the huge design space provided by a parameter-free optimization approach becomes obvious and the excellent parallel performance of a fully parallelized analysis and optimization tool is shown using standard office computers.
purchase the full-text of this paper (price £20)