American Institute of Aeronautics and Astronautics 1 A BI-LEVEL OPTIMIZATION APPROACH FOR TECHNOLOGY SELECTION Aditya Utturwar * , Sriram Rallabhandi † , Dr. Daniel DeLaurentis ‡ and Prof. Dimitri Mavris § Aerospace Systems Design Laboratory Georgia Institute of Technology, Atlanta, GA 30332 Abstract Technology selection is a crucial step in the process of aircraft design. If the performance and economic requirements are not fulfilled for any combination of the design variables, new technologies need to be infused in the design. Typically, the designer has a pool of technology options. The technologies to be infused in the new design are to be selected from this pool so as to achieve improvements such as increased performance, reduced risk, reduced cost etc. Thus, it is critical to be able to perform a quick and accurate assessment of the available technologies in the early stages of the design process. However, if the set of available technologies is large, the designer runs into a huge combinatorial optimization problem. To tackle the technology selection problem, a systematic approach called Technology Identification, Evaluation and Selection (TIES) has been developed to choose the best set of technologies and arrive at a feasible and viable design solution. However, the issue of dealing with large combinatorial problems still remains. A new approach for tackling the same problem of technology selection was inspired from the TIES methodology and is discussed in this paper. This approach is based on identifying an optimal point in an intermediate variable space, which later on serves as the target point for technology selection. The new approach, called ‘Bi-level approach’ provides additional insights and expedites technology selection, thus rendering efficiency to the preliminary design process. After describing the bi- level approach, its application to an aircraft design problem is presented. Introduction and Background One can think of the technology selection process in an abstract sense as being an exercise in constrained combinatorial optimization. The objective is to select an optimal set of technologies from a list of possible technology choices. Optimal in this context means those technologies that represent the best fit to a given set of conflicting requirements and program objectives. This situation is encountered by aircraft manufacturers during the process of designing an aircraft to meet a new * Graduate Student. Student Member, AIAA † Graduate Research Assistant. Student Member, AIAA ‡ Research Engineer. Member, AIAA § Associate Professor. Senior Member, AIAA Copyright © 2002 by Aditya Utturwar et al. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. requirement. The production company’s objective is to find an aircraft design that will be sufficiently advanced to offer a significant competitive advantage but which can also be developed within allowable time and budget constraints. A typical aircraft manufacturer may have dozens of technology concepts under development at any given time, but only a few of these technologies are likely to be suitable for a given set of program objectives. Some of these technologies are incompatible with others or are dependent on others in complex ways. Finally, the number of permissible technology combinations grows exponentially with the number of technologies available. The computational analysis involved in such a case is prohibitively time consuming making it difficult to investigate every possible technology combination. It therefore becomes imperative for the manufacturing company to find a means of efficiently searching technologies in order to locate those few that are the best fit for the given requirements combinations while taking the compatibility constraints into account. Roth, Mavris et al 1, 2 have shown that use of genetic Algorithms (GA) within the Technology Identification, Evaluation, and Selection (TIES) method is an effective means of solving the constrained combinatorial optimization problem. The method works by using TIES techniques to create a compact, generic model to represent the impact of any given technology in terms of certain kappa-factors 3 , which later on are used to evaluate the system level responses. The functionality of the TIES method can be depicted as in Figure 1. Figure 1: Technology Evaluation in the TIES Method The GA uses this technology impact model to assess the fitness of the new design. The GA works by creating a pool of technology combinations and evaluating them with the TIES model to estimate the performance of the whole system. These combinations are then compared to one another and the superior ones are kept in the pool while the inferior are discarded. The surviving combinations are then used as ‘parents’ to create a new generation of combinations. This process is repeated over many generations until the population converges to an optimized set of technology combinations. The surviving technology combinations are Technology Combination Vector of Kappa factors System Level Responses RSE Metamodel Technology Impact model