Reduction of the Numerical Resource Requirements for Multidisciplinary Optimization
- PDF / 903,346 Bytes
- 6 Pages / 595.276 x 790.866 pts Page_size
- 65 Downloads / 177 Views
© Porsche AG | Büttner
Reduction of the Numerical Resource Requirements for Multidisciplinary Optimization The simulation-based multidisciplinary optimization offers the possi bility of finding solutions for the structural design of the entire vehicle. However, to enable its operative application, various adjustments to reduce the high numerical resource requirements have to be carried out according to Porsche and University of Wuppertal.
MULTIDISCIPLINARY OPTIMIZATION AS AN OPPORTUNIT Y
Increasing complexity in the vehicle and requirement portfolio combined with decreasing availability of hardware prototypes leads to new challenges in
52
simulation-based vehicle development (stiffness, crash safety, passenger and pedestrian protection, etc.). Robust vehicle concepts have to be established in the early development phase. One tool for this are simulation-based optimization strategies. By integrating all load cases
that are important for the components to be dimensioned (sill, side members, etc.), it is possible to generate an optimization result that represents the best possible compromise taking into account all relevant discipline-specific requirements. This Multidisciplinary Optimization (MDO) is vehicle- and not disciplineoriented. It identifies complex interrelationships, promotes development quality and reduces iterative coordination. REQUIREMENTS AND CHALLENGES
To ensure the viability of a MDO, pre parations are necessary on an organiza-
tional and technical level. On the orga nizational level, discipline-specific and cross-disciplinary design criteria and variables have to be developed and coordinated. The variables relevant for each discipline must be clearly definable in each Finite Element (FE) model of each discipline. Diverging FE models require manual preparation of the models depending on the degree of difference. A synchronization of the MDO for defined milestone dates in the product development process simplifies this preparation effort. On the technical level, a fully automated process must be implemented otherwise the handling of discipline dependent solvers (simu lation), post-processors (extraction of design criteria) and computational architectures (location) is not possible with a reasonable effort. TABLE 1 shows an example of the diversity of disciplinedependent requirements. The fully automated optimization process, FIGURE 1, usually starts with a Design of Experiments (DOE) when integrating highly non-linear load cases. The design of experiments generates strategically well-distributed samples (design variable configurations) in the design space. FE models are set up automatically with these configurations and calculated by the respective FE solver on the corresponding computing infrastructure. Subsequently, the design criteria values are determined either directly from the out-
AUTHORS
put files or via the load case dependent post-processors. The design variable con figurations with the corresponding values for each design criterion are summarized in tabular fo
Data Loading...
