September 16, 2002

he next time you are flying on an airplane, take a moment to look out of the window and check the wing to see if it is vibrating and still attached to the plane.

The durability and strength of the intersection where the wing connects to the fuselage is a topic that has occupied Thomas Peters for two decades. That joint - a geometric intersection - presents the kind of design problem Peters, an associate professor of computer science and engineering, has been asked to resolve for the automotive and aerospace manufacturers each time a new product is under development.

Now, with a $715,000 grant from the National Science Foundation, Peters and his team will develop new mathematics and algorithms that could solve the problem once and for all, with potential savings to both industries of more than a billion dollars annually. The grant is part of a joint venture between the NSF and the Defense Advanced Projects Research Agency.

Thomas Peters, associate professor of computer science and engineering, with a drawing representing the attachment of a wing to the body of an airplane. With a new three-year grant from NSF, PEters hopes to improve the design of these intersections. Photo by Peter Morenus

Peters is leading a group of scientists from Purdue University, the Massachusetts Institute of Technology, and Boeing, which is creating an efficient and, more importantly, precise way of designing intersection algorithms. The project is called I-TANGO for "Interval Approach using Topology and Accurate Numerics for Geometric Objects".

"For decades, engineers using computer-aided design have been forced to use outdated mathematical computations - algorithms developed primarily for linear construction - which are not adequate for the complex surface intersections common in modern industrial design," explains Peters, who holds a joint appointment in mathematics.

Computer-aided design, better known as CAD, was originally used to help people visualize an object, replacing the traditional engineering design paradigm of draftsmen using pencil and paper to sketch designs. At the time, there was a great deal of human intervention, particularly by engineers who interpreted the design before any manufacturing or testing was done.

This process and CAD software evolved over the years, to the point where much of this work is now automated. But, as Peters points out, the mathematics used for the underlying algorithms did not keep up with the technology.

"While the design may look perfect on the computer, it is impossible to detect tiny cracks with the naked eye. So the aerospace industry and others have had to resort to costly and time-consuming physical tests," he says.

Once a new plane, for example, is designed, a full-scale model is built and then the wing is broken off to see how much stress it was able to endure. Engineers extrapolate that information and take into account environmental conditions the aircraft may encounter before design changes are made. Even the slightest change in geometry can mean a huge difference and ultimately, error, Peter says.

"By making the geometry better to begin with," he says, "we could avoid the tedious test and repair cycle." One of his goals is to have the geometry and stress analysis be sufficiently accurate so these costly destructive tests are no longer needed or required by the Federal Aviation Administration.

Winning this highly competitive three-year grant is a particularly sweet victory for Peters. For six years he had been trying to get major research funding for this initiative and was under pressure, he says, to go after smaller, more attainable projects instead. He credits two factors in securing the backing this time: the relationship he developed with Boeing and the interdisciplina ry team of experts he assembled. The team represents a number of related fields in mathematics, computer science, and engineering.

Three UConn students, one graduate and two undergraduates, will be working with Thomas and his team.