Genetic Algorithms WSJ Article

Genetic Algorithm Wall Street Journal Article

The Front Lines Column by Thomas Petzinger Jr.

14 July 1995

At Deere They Know A Mad Scientist May Be A Firm's Biggest Asset

Celebrate your eccentrics.

They are under utilized corporate assets, too often dismissed as unmanageable loners who traffic in peculiarity. Yet with the slightest encouragement, eccentrics display a gift for conceiving the unforeseen - which explains why Deere & Co. is applying the science of sexual reproduction to the assembly of farm equipment.

Meet Bill Fulkerson, staff analyst. A rumpled man who one taught math at Central Missouri State, he toiled anonymously in Deere's engineering labs for nearly two decades. He has received precisely one promotion over the years.

But today he is a hero, recognized as a problem-solver extraordinaire by Deere's top executives. "They know who Bill Fulkerson is now," says Robert Wismer, Deere's corporate engineering director.

The story involves planters, ungainly contraptions that farmers use to drop seeds into the soil. Some farmers want four-row planters, others 24-row planters or something in between. Some planters apply liquid fertilizer, others no fertilizer. A farmer can order a planter form Deere according to more that a million permutations.

In bygone days, Deere filled orders from inventory, shipping planters in pieces for assembly by local dealers. But to enhance quality and cut inventories, Deere in 1992 began assembling at the factory. Workers were organized into self-directed teams, each devoted to a particular system on the planter with its endless combinations.

Imagine an assembly line that produces Yugos, Jaguars and school buses each month, one after the other. Deere's scheduling challenge was no smaller. Half-assembled machines were soon bunching at one station while another remained idle. With incentive pay at stake, worker teams began bickering. In the scheduling department, frenzied staffers compiled spreadsheets by hand trying to smooth the flow of machines and reconcile it with the dispatching of delivery trucks.

Deere had re-engineered itself into a corner. 'It became obvious we had an ongoing problem," says engineer Dick McKinnon. "I made some phone calls." One of the calls went to Bill Fulkerson.

Purely by chance a short time earlier, a Deere executive had stopped by Mr. Fulkerson's cubicle. Looking at the mathematician's notoriously messy desk, the boss made a comment about a book on the emerging science of chaos theory, then wished him a Merry Christmas and walked away. That was in December 1992.

Mr. Fulkerson responded by spending the holidays absorbed in the book He became enthralled with chaos theory, and soon the adjunct speciality of "complex systems" research. During a fact-finding mission to the University of Illinois, Mr. Fulkerson listened to a professor expound on complex-systems research involving genetics.

Eureka! "It was kind of a visceral response," Mr. Fulkerson recalls. "It just clicked."

As species reproduce, those with the best characteristics tend to live long enough to reproduced themselves, gradually reducing the number of bad characteristics in the population. Scientists had begun modeling this process with formulas called genetic algorithms.

Perhaps, Mr. Fulkerson thought, the algorithms could compile planter schedules. Instead of having people (or computers) randomly test schedules, perhaps the genetic algorithms could use an improvement in any one schedule as a basis for seeking a further improvement, until a near perfect schedule emerged.

Mr. Fulkerson was energized. He had often felt discouraged at Deere, relegated to supporting other people's projects. He was 50 years old. "If I'm ever going to do anything on my own," he told himself, "now's the time."

This was 1993, long before the Internet had been hijacked by teenagers and voyeurs. Mr. Fulkerson found a site where scientist exchanged information about genetic algorithms. He posted a note.

Did anyone, he asked , know anything about scheduling a production line? The answer came a week later.

A high-tech defense contractor, Bolt Beranek & Newman Inc., had used genetic algorithms to schedule work at a U.S. Navy lab. The engineers on the project were planning to establish their own software company to develop commercial applications. They visited Moline and told Mr. Fulkerson they'd be delighted to make Deere their first commercial client.

Though Mr. Fulkerson encountered skepticism within Deere, it was no match for the overall eagerness to cure the bottlenecks at the seeding division. A pilot system was developed on a PC adjacent to the loading dock. By last year, a former shop worker named Larry DeClerck was cranking up the software before going home each day, leaving the machine to "breed" more than 600,000 schedules, each an improvement over the last. Planters now flow smoothly through the assembly line, with monthly output up sharply. Overtime has nearly vanished.

Mr. Fulkerson, meanwhile, has turned his attention to ant colonies and other subjects of "swarm" theory and what promise it might hold for manufacturing. His business card ways "project manager, adaptive manufacturing systems," but it's an invented position, he admits, a substitute for his official title, which remains staff analyst. "I haven't been promoted," But, he says, 'I'm having more fun."

Return to Recreational Boat Building Industry Home Page