Im Rahmen der Dissertation wird ein modulares Flugzeugentwurfssystem vorgestellt, welches auf einer verteilten Rechnerarchitektur und den Analysewerkzeugen disziplinärer Experten basiert. Zur automatisierten Entwurfsoptimierung wird ein effizientes Optimierungsverfahren entwickelt, welches die unterschiedlichen Modellhierarchien diese Systems ausnutzt. Es wird dargelegt, dass das Entwurfssystem gute Ergebnisse im Vergleich zu etablierten Methoden für konventionelle Flugzeuge liefert und dass das angewandte Optimierungerfahren für dieses System erhebliche Einsparungen des Berechnungsaufwands im Vergleich zu gängigen MDO-Verfahren (multidisciplinary design optimization) ermöglicht.
This thesis is devoted to the development and proof of applicability of a collaborative design approach for preliminary aircraft design as well as to the identification of suitable multidisciplinary design optimization strategies for such design methodology. To sustainably meet the growing demand for transport aircraft, new aircraft configurations and new system technologies are required. Detailed information in the first phases of the product design process are mandatory, to accurately assess the potential of such novel design solutions to contribute to future carbon-neutral growth. However, this implicates a shift of complexity in the design process, as in order to provide sufficiently detailed data early in the process, simulation models of higher fidelity need to be introduced. Now, the more intricate the models become, the less a single designer can provide detailed explanations for all disciplinary results. Hence, the collaborative aircraft design process ADS is proposed, which builds on a distributed computing approach and on higher fidelity analysis modules, which are usually not employed in the early design phase. It is presented that the proposed approach provides similar quantitative and qualitative results compared to actual aircraft data as well as to an established aircraft synthesis code, while providing for more detailed analysis capabilities. Due to the exploitation of the underlying distributed computing architecture, the runtime for single analyses with the ADS is in the same order of magnitude as established approaches. Simulation-based optimization approaches become increasingly important in product design. However, it is shown that employing established non-linear optimization methods to distributed high-fidelity design processes can become prohibitively expensive. To retain the response time for design processes, an optimization strategy is proposed for the ADS, which uses the best of both worlds - the low computational expense of low-fidelity models and the more detailed results of increased accuracy of high-fidelity models. The approach aims at exploiting the levels of model hierarchy of the system by incorporating both models into a single optimization framework. In this automated adaptive scaling approach the optimization is then performed using primarily the scaled low-fidelity conceptual design module in lieu of the high-fidelity ADS, which promises to reduce the overall computational expense. The benefit of such approach is that only comparably few high-fidelity evaluations are interspersed into the overall process to update the scaling function, which calibrates the low-fidelity model. It is presented that the developed multifidelity optimization approach significantly reduces the required computational expense for design optimization with the ADS. For the employed example aircraft design optimization problem, significant reductions in overall high-fidelity evaluations of up to 81% can be achieved, when compared to established MDO approaches. It is further demonstrated that the framework always converges to the reference optimal point. However, the automated multifidelity framework relies on the scaling function for calibration, which needs to be initialized prior to the start of the optimization process. Thereby, the obtainable reduction of computational expense is strongly influenced by the quality of the initial scaling function and a set of fixed framework parameters. The original contributions made in this research are the development and the assessment of the collaborative workflow ADS as well as the investigations on an appropriate multidisciplinary design optimization methodology, which significantly lowers the computational expense for design optimization in comparison to conventional approaches.
Model hierarchy exploitation for efficient multidisciplinary design optimization in a distributed aircraft design environment
2013
149 Seiten, 39 Bilder, 15 Tabellen, 103 Quellen
Theses
German
Optimierungsverfahren , Rechnerarchitektur , Flugzeug , Entwurfstechnik , Luftfahrzeug , Design-Optimierung , Produktionsprozess , Transportflugzeug , Anlagentechnik , Simulationsmodell , Designer , HiFi (naturgetreue Wiedergabe) , Aktualisierung , Produktdesign , nichtlineare Optimierung , Eichen (Abgleichen) , Berechnungskosten
Multidisciplinary Design Optimization in a Collaborative Distributed Aircraft Design System
British Library Conference Proceedings | 2012
|