Michael Keller, a 44-year-old electrical engineer, swung his arms up and down and made booming and crunching noises as he tried to recreate the motions of metal-stamping machinery in the living room of his Alexandria home.

Keller was explaining how his innovation could detect a malfunction in such machinery and shut it down before costlier damage occurs.

Holding up two partially formed plates for an automobile shock absorber bumper system, he said one complex set of tools takes about eight seconds to form a plate in eight operations. If the tooling that shears the outlines of the blank breaks, part of the tooling probably will get stuck in the blank and break the next piece of tooling when it tries to bend the blank. The sequence could continue through each of the remaining operations, with each piece of tooling in the machinery breaking.

The cost of replacing a piece of tooling could be between $25,000 and $30,000, while all eight pieces of tooling could cost $125,000, Keller estimated. He said that the tooling used in progressive metal stamping can break as often as once a week, with each break raising the possibility of more tooling being wrecked if the machine isn't stopped in time.

Keller's patented impact-sensing detector, on the other hand, would prevent such costly problems at a cost of between $10,000 and $15,000, he said.

The detector would convert the shock and vibration caused by the metal-stamping operation into what Keller calls a "signature" -- an identifiable electrical signal. A special computer would learn to recognize the signature generated when the machine is operating correctly. It then would compare the signatures from subsequent strokes of the metal-stamping machinery with the recorded one, and shut down the machinery if there were an unacceptable deviation from the recorded standard.

According to Keller, monitors now in use work differently than his would: They measure the total machine load, or force exerted to perform the operations. Keller said his system is better because some of the metal-stamping tooling most prone to failure contributes very little to the total load, and their failure could go unnoticed. His system would compare signatures on a point-by-point basis, in some instances indicating which part of the metal-stamping machinery malfunctioned.

As the sensitivity required to detect minor tooling malfunctions increases, so does the rate of false alarms -- a problem that Keller is giving a lot of attention to by trying to match hardware and software. To gather raw material for this effort, he is recording metal-stamping machine shocks and vibrations at various plants, including facilities owned by M. S. Willet Inc. in Cockeysville, Md., and General Motors Corp. in Detroit.

Keller is seeking venture capital from major companies that are potential customers, including GM and Chrysler Corp. He would like to have the funding lined up by the end of the year, and predicts construction of a prototype could begin six months later and take about 10 months.

Meanwhile, Keller is doing consulting work for defense electronics companies, helping them with their business development and reviewing proposals.

Hanging around factories is nothing new for Keller.

He grew up in Franklin, N.J., a mining town, "with not a heck of a lot to do," he recalls. His first job, at age 16, was "sweeping the floors in a New Jersey plant that stamped out parts for ballpoint pens."

"When I got out of the service in 1964, I was looking for what to do with my life," Keller said.

He returned to the ballpoint pen plant to see if he could improve the electrical sensing and registering system that was supposed to make sure that pen parts were moving from one stamping station to another. The owner "gave me a month to build a prototype" for an improved system, and Keller succeeded.

He formed Keller-Nace Associates with college roommate David Nace, an electrical engineer who now works at the Goddard Spaceflight Center. They marketed the monitoring system, calTed Multigard, but "many of the features we offered didn't sway the industry at large," Keller said. In 1966, Keller switched to part-time hours with Keller-Nace and went to work for the Marine Corps as a civilian employe. He folded Keller-Nace in the mid-'70s.

Keller's experiences as an electrical engineer for the military, a consultant and an inventor have given him what he calls a warm-bread theory of inventing and entrepreneuring:

"When bread is in the oven, it smells awful good and you can get excited about it," he said. "When it comes out of the oven, you can get burned. The analogy is that new technology can burn you if you touch it too soon. An idea has its time frame; a technology has its time frame."