PITTSBURGH -- Distinguishing fine wine from plonk is usually left to connoisseurs and winemakers, who rely on their senses, rough chemical measurements and the whims of nature to produce an exceptional tipple.

But a Carnegie Mellon University professor, working with industry scientists in Chile, is hoping that computer models will identify the traits of good wine to help vintners produce more of it.

Lorenz "Larry" Biegler, who teaches chemical engineering at the university, is working on mathematical formulas to automate the fermentation process, adjusting ingredients and conditions to ensure robust flavors and higher yields from grape harvests.

Scientists don't fully understand the delicate mix of compounds that emerge during fermentation and why they create such pleasing sensations for wine drinkers. Biegler's research focuses on yeast, which consumes sugar and produces alcohol.

"We would like to come up with a reasonably good model of how this yeast cell behaves ... then control this fermentation process so we can make better-quality wines," he said.

Much of the work is being carried out at an "aroma lab" at Pontifical Catholic University in Santiago, Chile, where industry-sponsored researchers are trying to isolate chemicals that produce desirable fragrances and flavors, Biegler said.

One goal is to help vintners avoid "stuck batches" -- batches that spoil and are thrown away when fermentation stalls, leaving too much sugar.

Biegler hopes to make winemaking more efficient, consistent and, ultimately, profitable. Similar computerized systems are routinely used at chemical plants, oil refineries and pharmaceutical manufacturing facilities, he said.

The researchers have been collaborating for more than two years and are studying only white wines, since reds are more complex and contain solids that make them difficult to analyze.

Stephen Menke, an oenologist at Pennsylvania State University's Cooperative Extension, said sorting out wine ingredients is difficult "because it's a natural product and we don't really do much processing."

Researchers are trying to refine more traditional methods of assessing a wine's quality -- tasting it, sniffing its bouquet -- and monitoring production, Menke said.

Scientific efforts to unlock wine's secrets have generally met with little success, said Mark Chien, a wine grape specialist also at Penn State.

"I wish it were that easy," he said. "In a textbook sense, it should be, but it never quite is. It's a fascinating scientific exercise, but nobody's been able to prove you can do something like that."

That has not stopped some from trying.

One company, Enologix, of Sonoma, takes juice samples from grapes, analyzes them and, using proprietary software, recommends how to make wines that please leading critics.

Although technology has the potential to shape the winemaking process, "there's still an art to it in terms of preference and taste and some of those things we can't describe by listing the compounds in the wine," said Mario Mazza, an oenologist at his family's North East, Pa., winery.

For Biegler, controlling fermentation is a start. "It's the dream of seeing whether we can do that," he said.