eduction of the numerical resource requirement for the operational use of multidisciplinary optimization of vehicle structures
In the vehicle development process, many CAE departments work in parallel but also sequentially on the structural design of a vehicle. They cannot make design decisions independently of each other without restrictions, as their decisions may not only affect the requirements of their own responsible discipline (e.g. crash, stiffness and durability). Multidisciplinary Optimization (MDO) is a tool to capture this complexity and interdisciplinary interaction. It searches for designs that meet all discipline-specific and cross-discipline requirements in a targeted and automated (algorithm-based) way. The outcome is a best possible compromise and thus full vehicle- rather than discipline-oriented. MDO promotes development quality and speed. However for operational use, it requires an approach for an efficient use of CPU resources and for a targeted integration of all disciplines and thus departments relevant to the specific problem. This Dissertation discusses different strategies that serve the needs mentioned for operational use. These strategies can be grouped into the following areas: the choice of efficient optimization algorithms and strategies, the selection of relevant load cases and sensitive design variables as well as the reduction of the finite-element calculation times of computationally intensive crash analyses (FE-submodels). For all these focal points, a best practice of existing solution proposals and methods is elaborated, and in some cases new approaches are established. By assembling these elaborations in a clever way, a novel, adaptively controllable MDO process based on approximation models is developed. The three following features make this process special: First, a module called adaptive complexity control gradually reduces the complexity and dimensionality of the optimization problem. Second, with the help of an implemented local, heuristic and thus on all approximation model types computable prediction uncertainty measure, the prediction quality (deviation prediction to finite-element value) can be estimated and used in the optimization. Third, the global sensitivity matrix supports the coordination of all disciplines and thus helps planning and executing a MDO. All investigations are carried out based on a complex full vehicle example and compared with the standard procedure of a MDO. It is shown that the developed efficient approach not only causes a significant reduction in numerical resource requirements but also a significant improvement in the quality of results compared to the standard approach.
