Boston, Mass. -- Researchers at Boston University's College of Engineering received funding to develop innovative techniques for managing such complex systems as modern manufacturing facilities, global communication networks, and world-wide economic systems. The new tools, which will draw on advanced computational techniques and today's unprecedented computing power, will also help scientists solve complex problems in computational physics.
The research will be supported by a $1.2 million award from the National Science Foundation's (NSF) Knowledge and Distributed Intelligence (KDI) initiative. The initiative seeks to foster interdisciplinary research and "change the way scientists collaborate and the way they prepare to examine the world as they seek new frontiers for discovery," said NSF Director Rita Colwell. Boston University's project is one of just 40 to be funded out of 850 projects submitted.
"Manufacturing today necessitates decision-making in large stochastic systems -- systems that include many random variables," says Michael Caramanis, professor of manufacturing engineering and the project's principal investigator. "In a manufacturing system, the complexity arises from the fact that lots of decisions have to be made involving multiple time scales. They range from decisions about building new plants and ordering new machinery -- which may be made only once in ten years -- to the minute-to-minute decisions made on the production line."
Cellular manufacturing technology, introduced by the Japanese in the 1960s, and widely adopted by American industry is based on small factories, or cells, within a larger factory. "Managing such a process effectively is extremely complex and if you are slow in making the right decisions you may be left with too many outdated, semifinished products in the pipeline, driving up costs," says Caramanis.
"Making the right decisions requires an understanding of the dynamics, or changing requirements of the process, and the time scale of these dynamics is incredibly varied. In addition you must factor in uncertainty -- machine breakdowns, absent workers, changes in the economy," Caramanis continues.
Most businesses rely on a worst-case analysis to deal with problems that arise in an uncertain environment -- generally a costly approach. "The BU team's new computational approaches will enable accurate prediction and flexible, efficient decision support," says Caramanis. "We already have a good indication that the tools we are developing apply directly to problems in computational physics -- like understanding how fluid flows through a rough conduit, and we will be testing how we can apply them in other complex systems. We will also be testing the manufacturing management tools in the factories of our two industrial collaborators, ALCOA and Pratt Whitney."
In addition to Caramanis, the research team includes Christos Cassandras and Yannis Paschalidis, of BU's department of manufacturing engineering; Francis Alexander, Boston University's Center for Computational Science; Dimitri Bertsekas and John Tsitsiklis, MIT's department of electrical engineering and computational science; and Yannis Ioannides, of the Tufts University's economics department.