A design space exploration methodology to support decisions under evolving uncertainty in requirements and its application to advanced vehicles

Design space exploration;Multi-objective optimization;Decision-making;Conceptual design;Evolving requirements uncertainty;Robust design;Fuzzy set theory;Suborbital vehicles;Rocket engines;Safety evaluation;Life-cycle cost estimation

Frank, Christopher Pierre ; Mavris, Dimitri N. Aerospace Engineering Pinon-Fischer, Olivia J. Schrage, Daniel P. Lizy-Destrez, Stéphanie I. Kennedy, Graeme J. ; Mavris, Dimitri N.

University:Georgia Institute of Technology

Department:Aerospace Engineering

关键词: Design space exploration;    Multi-objective optimization;    Decision-making;    Conceptual design;    Evolving requirements uncertainty;    Robust design;    Fuzzy set theory;    Suborbital vehicles;    Rocket engines;    Safety evaluation;    Life-cycle cost estimation;   

Others  :  https://smartech.gatech.edu/bitstream/1853/55628/1/FRANK-DISSERTATION-2016.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

Recent technological developments have resulted in the emergence of new advanced vehicles such as suborbital vehicles and personal air vehicles that have opened up new markets. These markets are characterized by a complex multi-objective decision space, a large combinatorial space of possible configurations for which no baseline has been established, and the presence of evolving uncertainty in requirements. To support the successful development of such markets, a rigorous approach is needed that systematically and efficiently investigates the entire design space of solutions. In particular, this research aims at establishing a methodology that enables a broad design space exploration at a conceptual level to select solutions against unclear objectives and under evolving uncertainty in requirements.A four-step methodology is developed based on the generic top-down design decision support process. First, the decision criteria are established. In particular, the design objectives are clearly identified and the design constraints, modeled with time-dependent membership functions, are propagated using fuzzy set theory. Second, a new variable-oriented morphological analysis is developed to generate all feasible concepts so that they can be systematically optimized and compared. Third, a modeling and simulation environment is developed, which is capable of rapidly evaluating the performance, life-cycle costs, and safety of all types of suborbital vehicles at a conceptual design level. Finally, a new evolutionary multi-architecture algorithm based on architecture fitness is implemented that drives multi-objective optimization algorithms to simultaneously compare and optimize all configurations.The new modeling and simulation environment was developed and implemented in the context of suborbital vehicle design. By leveraging cycle-based approaches and surrogate modeling techniques, the performance of all chemical rocket engines can be evaluated with an accuracy of 3%, while dividing the execution time by a factor of 100,000 compared to current physics-based models. This environment is also the first of its sort capable of estimating the life-cycle costs of hybrid rocket engines. The application of the proposed methodology also provides decision makers with key insights into the suborbital market. In particular, it demonstrates that a wisely developed commercial suborbital program might be profitable. The methodology also quantifies the trade-offs between affordable winged air launched vehicles powered by solid engines and safe slender vehicles powered by hybrid engines. When compared with existing approaches, the proposed methodology allows decision makers to find solutions 40% more performant for the same execution time or 40 times faster for the same accuracy. By quantifying the trade-offs between risk and expected performance, this methodology also helps designers make challenging go/no-go decisions and provides them with the best program start date. In particular, it provides a robust solution that increases the probability of success by 10% compared to those generated by traditional approaches.