A post-doctorate engineer at the University of Wisconsin, Madison, has created a series of computer models, inspired by Charles Darwin's rules of evolution, to design high-performance diesels of the future.

Peter Senecal devised algorithms for computer models that sort through billions of combinations of parameters to calculate engine performance. The process is too big for conventional computer simulations. Mr. Senecal claims that his algorithms help improve emissions without sacrificing fuel economy.

He says that engine designers usually concentrate on solving one problem and must make major tradeoffs in other characteristics. Mr. Senecal used a Silicon Graphics supercomputer at UW-Madison's Engine Research Center to create a diesel engine design that reduces oxides of nitrogen (NOx) by three-fold and soot by 50% over today's state-of-the-art technology. The design also increased mpg by 15%.

Mr. Senecal studied six engine performance parameters, including: fuel injection timing, injection pressure and amount of exhaust-gas recirculation. He then reproduced the computerized simulations in a real diesel at the ERC. “We found that the agreement was excellent between what was measured in the lab engine and what the computer predicted,” Mr. Senecal says.

His data will appear in an upcoming issue of International Journal of Engine Research and he also presented his findings to the SAE International meeting in Paris. Caterpillar Inc. is funding the research. It is one of the diesel engine manufacturers facing tough new pollution-control regulations that take effect in 2002. Caterpillar is specifically funding Mr. Senecal's work that focuses on improving the geometry of engines.

Mr. Senecal says he developed the genetic algorithms recently while seeking to solve other engineering problems in bridge design and airplane wings. “I kind of stumbled onto this in the literature and wasn't sure if it work for something as complex as engine design,” he says. The evolutionary process starts when Mr. Senecal starts out with what he calls five “individuals.” He defines these as one distinct set of engine parameters.

Mr. Senecal randomly selects four of the individuals and the fifth is the baseline, or best-known set of parameters. Then he uses a computer model to weed out the best parameters in the first group. The two fittest “parents” then are allowed to reproduce a new generation that is formed, complete with “mutations” that are improvements over the previous generation. He uses the process through successive generations until the computer finds the most “fit” member.

The evolutionary process shaves the field of potentially one billion computer calculations to 200 to 250 of the best possibilities. The supercomputer can do the calculations in weeks instead of the decades it would normally require. Current studies focus on fuel injection and air intake. Mr. Senecal's studies of engine hardware are in an incipient stage. Professor Rolf Reitz, Mr. Senecal's thesis advisor, says the model can be used for all types of engines.

Also, if enginemakers want a more powerful or more durable engine, researchers can program the genetic model to find those traits. “If you want your children to be long jumpers, high jumpers or sprinters, you can specify these attributes with this program,” Mr. Reitz says.