Berkeley Engineering FOREFRONTS 1994
by Nancy Bronstein
Photo by Peg Skorpinski
Some sample outputs of the GTRAN2 code Nuclear Engineer Jasmina Vujic and her colleagues across the country are monitoring the political climate in Washington D.C. this year with more anxiety than usual. The federal government's support for research and development in nuclear science and engineering is in jeopardy.
"This definitely affects us," says Vujic, who left Argonne National Laboratory to join the Berkeley Faculty in 1992. "The only specifics President Clinton mentioned in one of his early speeches on the deficit targeted cuts in nuclear engineering research and development. It could mean that thousands of engineers and scientists in the national laboratories and the nuclear industry would lose their jobs. If the national; labs are cut, then the universities will have to cut back too."
Political decisions such as this one could directly affect Vujic's long-standing work to refine and unify various mathematical models that describe the chain of events inside a nuclear reactor core related to neutron transport. Vujic's research, now nearly complete, has already significantly altered the nuclear engineering playing field.
Reactor vendors such as General Electric and General Atomics are already using her methodology. Others, like Westinghouse, have expressed strong interest in using Vujic's methods and codes, particularly in the design of advanced nuclear reactors, including those needed to replace aging reactors in the United States, Russia, and Europe.
Until a few years ago, Monte Carlo codes were the only codes that could accurately model neutron behavior in complex geometries, such as those found within a reactor core. Developed in association with World War II and the atomic bomb, Monte Carlo codes calculate neutron transport probabilities within a nuclear reactor core by following one neutron at a time to determine when it might collide, scatter, or become absorbed.
"With Monte Carlo, you play games with your neutrons in order to calculate the probabilities," says Vujic. "it's like flipping a coin and asking 'what will happen to my neutron at this point?'"
Flexible in their ability to account for complex geometries, Monte Carlo codes are immensely time-consuming and expensive to use when applied to reactor cores. For a new generation of nuclear engineers such as Vujic, this is not good enough. She and others began searching for new ways to model neutron transport accurately in complex geometries using more computationally efficient methods.
"To get answers, we have to follow millions of neutrons, a process that could take in some cases months or years on supercomputers using the Monte Carlo codes," says Vujic. "I wanted to combine methods for flexibility, accuracy, and speed and come up with a single method that is fast, that is accurate and as geometrically flexible as Monte Carlo, but that is not Monte Carlo."
The capability came with Vujic's invention of GTRAN2 (pronounced gee-tran-2), the General Geometry Transport Theory Code in Two Dimensions. Six years in production, GTRAN2 software will be patented this year, and Vujic expects to complete a three-dimensional counterpart, GTRAN3, next year.
According to Vujic, nuclear reactors have barely been tapped for their energy-producing potential in the United States, currently supplying about 20 percent of U.S. electrical power. In France, nuclear reactors supply more than 70 percent of the country's electricity. Vujic would like to see advanced design nuclear power plants ordered and on the drawing boards by the year 2000. To get there, says Vujic, researchers first have develop the means to stimulate and predict neutron behavior accurately within the reactor core at all times.
"My methods can do that," says Vujic. "My methodology and codes have a dual purpose - to design advanced reactors and to improve the analysis of currently operating reactors by accurately modeling what's happening inside the reactor core. With small modifications, the same methodology can even be used to model neutral particle transport in other fields, such as cancer therapy."
The process of creating energy begins with a fission event - bombarding uranium atoms with neutrons to split the atom nuclei, which releases kinetic energy of fission products, neutrons, and gamma rays. This kinetic energy is transformed into thermal energy, which in turn is converted into electricity. Like free spirits in a spectacular but erratically choreographed ballet, neutrons travel in straight lines until they collide with the atom nuclei of the fuel pellets, their protective cladding, moderator, coolants, and other core materials. After a collision, neutrons may scatter and change direction and energy, or may be absorbed.
To predict the net outcome of all the neutron collisions, Vujic uses a technique called ray tracing. Rather then following the track of each neutron, as is done with Monte Carlo methods, Vujic uses ray tracing to cover the domain of interest with a large number of rays (pseudo neutron tracks) and then calculate the probabilities of collision anywhere along that track. Once calculated, these probabilities can be reused many times. Using Monte Carlo codes, on the other hand, researchers would have to track each neutron (out of several million) and repeat the ray tracing over and over again.
In Berkeley's New Advanced Nuclear Engineering Computational Lab, which she helped organize last fall, she uses a cluster of workstations to execute her code in parallel. By using more than one workstation, she can analyze larger and larger parts of the reactor core.
"Each workstation can be given one or more reactor core assemblies to calculate out of several hundred of them in the reactor core," she says. "It seemed impossible even a year ago, but now we hope to use this very accurate method to analyze the entire reactor core.
"Before, you had to use approximative methods to come up with a simple model of a reactor core because the computer power wasn't that large," says Vujic. "Nuclear engineers and scientist had to come up with a clever ideas to obtain an accurate answer starting from an approximative model. Now we can analyze large parts of the reactor core while taking into account its full complexity"
Standing on the brink of what appears to be technological revolution in reactor core simulation, Vujic looks to Washington not only to let it happen, but to understand and actively support it.
"With continued research, a new generation of reactors can be used to solve the problems of nuclear waste, burn the waste generated from other operating reactors, and even produce new fuel for old reactors," says Vujic. "Instead of stockpiling the weapon-grade uranium and plutonium, we can get rid of it by using it as fuel for advanced nuclear reactors and even produce electricity. It would be pity to lose our opportunity to develop such advanced technology."