Competitive assessment of aerospace systems using system dynamics

Aircraft design has recently experienced a trend away from performance centric design towards a more balanced approach with increased emphasis on engineering an economically successful system. This approach focuses on bringing forward a comprehensive economic and life-cycle cost analysis. Since the success of any system also depends on many external factors outside of the control of the designer, this traditionally has been modeled as noise affecting the uncertainty of the design. However, this approach is currently lacking a strategic treatment of necessary early decisions affecting the probability of success of a given concept in a dynamic environment. This suggests that the introduction of a dynamic method into a life-cycle cost analysis should allow the analysis of the future attractiveness of such a concept in the presence of uncertainty. One way of addressing this is through the use of a competitive market model. However, existing market models do not focus on the dynamics of the market. Instead, they focus on modeling and predicting market share through logit regression models. The resulting models exhibit relatively poor predictive capabilities. The method proposed here focuses on a top-down approach that integrates a competitive model based on work in the field of system dynamics into the aircraft design process. Demonstrating such integration is one of the primary contributions of this work, which previously has not been demonstrated. This integration is achieved through the use of surrogate models, in this case neural networks. This enabled not only the practical integration of analysis techniques, but also reduced the computational requirements so that interactive exploration as envisioned was actually possible. The example demonstration of this integration is built on the competition in the 250 seat large commercial aircraft market exemplified by the Boeing 767-400ER and the Airbus A330-200. Both aircraft models were calibrated to existing performance and certification data and then integrated into the system dynamics market model. The market model was then calibrated with historical market data. This calibration showed a much improved predictive capability as compared to the conventional logit regression models. An additional advantage of this dynamic model is that to realize this improved capability, no additional explanatory variables were required. Furthermore, the resulting market model was then integrated into a prediction profiler environment with a time variant Monte-Carlo analysis resulting in a unique trade-off environment. This environment was shown to allow interactive trade-off between aircraft design decisions and economic considerations while allowing the exploration potential market success in the light of varying external market conditions and scenarios. The resulting method is capable of reduced decision support uncertainty and identification of robust design decisions in future scenarios with a high likelihood of occurrence with special focus on the path dependent nature of future implications of decisions. Furthermore, it was possible to demonstrate the increased importance of design and technology choices on the competitiveness in scenarios with drastic increases in commodity prices during the time period modeled. Another use of the existing outputs of the Monte-Carlo analysis was then realized by showing them on a multivariate scatter plot. This plot was then shown to enable by appropriate grouping of variables to enable the top down definition of an aircraft design, also known as inverse design. In other words this enables the designer to define strategic market and return on investment goals for a number of scenarios, for example the development of fuel prices, and then directly see which specific aircraft designs meet these goals.