Research Feature
July 1, 2004

The picture at right comes from calculations done recently at TACC by Victor Calo, a graduate research assistant in the group of Thomas J. R. Hughes, Professor of Aerospace Engineering at The University of Texas at Austin (UT Austin). Calo's research examines how a fluid running over a flat plate suddenly becomes turbulent. "The smooth flow seen at the left-hand side has low-amplitude fluctuations," Calo says. "These interact with the smooth, laminar boundary layer of the fluid, and as the energy of the fluctuations is convected downward into the boundary layer, we see a sudden explosion of the entire flow into turbulence." The visualization was made from Calo's calculations by undergraduate researcher Gregory P. Johnson, who worked with Dr. Kelly Gaither, director of the ACES VisLab.

The calculations are of interest for several reasons. Calo's work with Professor Hughes has applications to turbomachinery design, aerodynamics, and electronic cooling devices. The code they use, called a variational multiscale large-eddy simulator, is computationally intensive and very scalable to increasing numbers of processors. It can stretch the capacities of Lonestar, TACC's recently upgraded Cray/Dell Linux cluster, which is one of the fastest supercomputers available in academia.

"Victor has developed one of the most advanced codes to date," Hughes says, "and we are expecting to be able to detect the critical features in the data and visualizations that mark the transition to turbulence. This is a problem that has been under study by many groups for a very long time, and we believe our approach will be a significant advance." Funding for the study came from the NASA Ames Research Laboratory.

In working to extend the methods developed by Tom Hughes, Calo is the latest in a long line of students and postdoctoral researchers who have had the privilege of working with one of the founders of the field of computational mechanics.

Tom Hughes was a leading professor of mechanical engineering at Stanford University before being lured to UT Austin in 2002 to join the Aerospace Engineering Department and the Institute for Computational Engineering and Sciences (ICES). "In addition to the opportunity to work closely with excellent colleagues like Tinsley Oden and Graham Carey, one of the compelling reasons for my transition," Hughes says, "was the planned growth of TACC into a major academic supercomputer center."

When Hughes arrived as a graduate student at Berkeley in 1969, the development of computer implementations of the finite-element method was in its infancy. He began making contributions to the field immediately, and by 1974, when he received his doctorate in Engineering Science, he was already planning courses in "computational mechanics" to be given at Berkeley and elsewhere. They may have been the first such courses given anywhere; engineering departments had been treating computation as a dreary adjunct to other methods, principally of use in aircraft engineering. Large-scale computers were rare, they did not have much memory, and computational procedures were cumbersome and time-consuming.

But mechanical engineering itself is a fundamentally creative endeavor, no less creative than music, art, or literature. In architecture and in solving the myriad problems of society's built environment--what is often called our "infrastructure"-- the finite-element method in its infinite variations has ultimately turned the computer into the engineer's crucible of creativity.

How far the field has come, owing in large part to Hughes's innovations and those of his students, was abundantly evident in April 2004, when more than a hundred practitioners of computational mechanics gathered at Rice University in Houston for a conference celebrating Tom Hughes's sixtieth birthday. Called "Advances in Computational Mechanics," the conference was sponsored by Rice and Northwestern University and organized by the chairman of the Rice department of mechanical engineering, Tayfun E. Tezduyar--who had been a student of Hughes's at Caltech in the late 1970s.

One after another, the 36 speakers from institutions all over the world built a picture of the dramatic changes in the scope and applicability of finite-element computational methods. Robert L. Taylor, professor of civil and environmental engineering at Berkeley, pointed to early contributions made by Hughes to the problem of dynamic contact--impact and release--that arises in modeling mechanical systems. He noted that Hughes is still making major contributions in this area.

Robert Ferencz from the Lawrence Livermore National Laboratory pointed to work done by Hughes and colleagues in the early 1980s on modeling "shells"--not the seashore kind, but the structural elements of many parts of the built environment, from ship hulls to cooling tower chimneys. Tezduyar reviewed Hughes's early contributions to the finite-element modeling of fluid flows and their relation to current problems in which both a fluid and something contained within it are moving--such as a parachute falling through the air.

The three-day conference moved from discussions of general methods for solids, fluids, and multiphysics problems through methods of handling problems that span orders of magnitude in space and time to methods of modeling very complex and irregular objects. New computational methods for handling flows of various kinds occupied most of the second day, and these included modeling of cardiovascular flows. These investigations are a part of Hughes's current research program, and he is working to unite a community of biomechanical modelers to bring mathematical rigor and clarity to the systematic investigation of blood-flow problems.

Other applications discussed at the conference included modeling reacting flows (think of rocket fuel), materials science problems (fault-tolerant elasto-plastic composite structures), and ways to determine the sensitivity of the models themselves to changes in method (what happens when the timesteps are smaller?). Myriad new methods to bridge scales in space and time are under development, in part because the miniaturization of components has brought different engineering regimes into close proximity. Among the speakers were Hughes's UT colleagues Ivo Babuska, Mary Wheeler, and Tinsley Oden.

Every speaker pointed to a paper or group of papers by Hughes that has influenced methodological developments across all of these constantly exfoliating and radiating subfields. And all, without exception, spoke warmly of their meetings with Tom Hughes, as his students or at conferences. As a current student, Victor Calo echoes the sentiments of the meeting participants. "After three years I am even more impressed than ever," he says. "I'll bring him something I'm working on, and he will immediately simplify the problem, leaving me in awe. Or he'll come to me a week later and go over it again in a way that makes me understand the implications of whatever I said earlier, much more fully."

Using the resources at TACC, Hughes is applying his wealth of experience and insight to "fresh fields and pastures new." In addition to organizing cardiovascular modelers, he is working on a new set of multiscale methods particularly well adapted for the study of the nano- to micro-regime of electronic devices. He describes these as "locally atomistic and globally continuous" methods that will have wide applicability. The modeling done by his student Victor Calo is an example of progress made in this work. Hughes is also working with the Office of Naval Research on new methods for analyzing the noise profile created by turbulent wakes around ships.

"TACC and its visualization capabilities are going to be more and more valuable in all of this work," Hughes says.