|Volume 7, Issue 1 -
Optimization Group Fosters Knowledge Transfer through Collaboration
Researchers in the optimization group of the CRPC have generated significant interest their ongoing research within industry and academia. Bob Bixby has been involved for several years with the development and distribution of industrial strength linear programming software through CPLEX, a company he helped form. More recently, he has collaborated with scientists at Bellcore and AT&T Bell Labs on parallel solutions of very large integer programming problems and has worked with scientists at DIMACS, an NSF Science and Technology Center. This group has made significant progress in the important class of scheduling problems that include scheduling of airliners and their crews.
In collaboration with Boeing, the robust derivative-free nonlinear optimization algorithm developed by John Dennis and Virginia Torczon is being used by the company in its flexible manufacturing facility, and Torczon's implementation has been distributed to 100 industrial sites during the past year. Shell researchers, for example, have noted how easily they had translated it into C. Dave Tolle, Research Manager of Emerging Technologies at Shell Development, anticipates future collaborations with CRPC researchers in this field.
The optimization group played an important part in the CRPC involvement with MADIC, the Multidisciplinary Analysis and Design Industrial Consortium, and they plan to continue their involvement during the development of USMADE, the MADIC multidisciplinary design optimization, or MDO, environment.
Commenting on a recent CRPC proposal entitled "Multidisciplinary Design Optimization in Aeronautics," Gregory Shubin of Boeing Computer Services in Seattle, WA said that the CRPC is bringing a "fresh approach" to multidisciplinary design by putting more emphasis on optimization and computer science. "I firmly believe that the marriage of analysis codes and optimization will lead to better aircraft design and a significant reduction in the design cycle time," said Shubin. The CRPC Optimization Group is playing a vital role in this process by making the first systematic investigation of alternative solution methods to multidisciplinary design optimization (MDO).
Shubin and colleagues at Boeing are actively involved in the MDO project, planning to provide model problems to explore in aerospace design. They will also implement the most promising MDO methodologies in real Boeing applications. Shubin's group had previously collaborated with Dennis and Robert Michael Lewis on a 1992 AIAA Multidisciplinary Design Optimization Conference paper entitled, "On Alternative Formulations for Multidisciplinary Design Optimization." The paper describes several powerful methods for formulating MDO problems.
A key technology for solving MDO problems that will be explored is the use of automatic differentiation (AD) to obtain accurate sensitivity information from analysis codes. ADIFOR, an AD tool developed by CRPC researchers at Rice University and Argonne National Laboratory, will play an essential role in these sensitivity calculations. "ADIFOR is one of the most practical AD tools we have seen," noted Shubin.
Visits to the CRPC have also resulted in inter-university collaboration on optimization research. Richard Tapia has been very productive in work with CRPC visitors. He has written joint papers on interior point methods for continuous optimization with researchers from the University of Iowa, the University of Maryland, the Federal University of Brazil, and two Japanese universities. Xiadong Zhang, a computer science professor at the University of Texas at San Antonio, collaborated with John Dennis after an extended visit to Rice University. The collaboration involved developing efficient parallel methods and software for solving large- scale sparse nonlinear systems of equations.
For more information on CRPC optimization research, see the "Research Focus: Optimization Group Explores New Areas in Multidisciplinary Design" article.
Table of Contents