Green Silicon

Stefan Wild

(Page 3 of 3)

Computationally expensive models involve as many as dozens of parameters, or variables.  For example, in assessing the best way to clean contaminated groundwater in on-site remediation using wells, pumps and purifying techniques, the parameters include the rate of pumping, the changing levels of contamination, and groundwater movement in response to the pumping.

Some computationally expensive models can involve a hundred or more parameters.  The more parameters, the more expensive and complex the model becomes, since each parameter requires its own algorithm, or part of the simulation code that must interact with all the other parts.  This boosts the time it takes to run the model on a computer, often making it impractical to use.

There can be conflict between optimization and numerical conditioning of the resulting set. Click image for larger version and more information

“What we’ve done,” Wild says, “is to build a faster mathematical surrogate that can replace key bottlenecks in expensive models and make them computationally inexpensive to evaluate.”

Wild and Moré’s optimization algorithm is performing well in a Matlab implementation, and the two are continuing to finesse the code to solve implementation details and make it user-ready for addition to TAO.

“It’s already beating the competition,” Moré says.

For Wild, it’s an important step toward bringing applied math to the environmental engineering community.  He says one of the most rewarding aspects of the practicum at Argonne was realizing the enormous impact of optimization codes.

“Argonne’s Toolkit for Advanced Optimization is a facilitator for important science around the world, and it felt great contributing to this,” he says.

Nowhere is this truer than in the realm of environmental engineering. There are tens of thousands of contaminated groundwater sites in the United States, from EPA-designated Superfund sites to smaller ones, involving pollutants from radioactive wastes to pesticides and petroleum residues.  Remediating these sites often costs tens of millions of dollars and involves decades of work. Optimizing the cleanup approach before starting can save years of work and millions of dollars.

When Wild presented his preliminary results at the Society for Industrial and Applied Mathematics (SIAM) meeting on Computer Science and Engineering in September 2006, he was approached by scientists from the United States Geological Survey interested in applying the optimization techniques to their models.

It was a heartening response — one that told Wild he had the right recipe for long-term success.

« Previous       1   |   2   |   3   |   Print