LOUISVILLE, Ky. (AP) – The day of racing has ended at Churchill Downs, and tractors have plowed away the evidence of thousands of hoof marks left behind on the dirt track.

This is when Mick Peterson goes to work creating a few more.

He calls his machine a robotic horse, but the metal contraption attached to the back of a van doesn’t at all resemble the graceful animal it is designed to imitate. That is, until he turns it on.

As it slams the dirt, the angle, force and especially the sound of the device’s “leg” are unmistakable. It’s just how a hoof hits the ground during the ferocious stretch run of a thoroughbred race.

Peterson, a professor in mechanical engineering at the University of Maine, has been affectionately called a mad scientist by track superintendents. But his hope is to use the simulated hoof prints and data his machine produces to detect trouble spots on racetracks, giving maintenance crews new information to avert potentially fatal accidents.

“You don’t want to say, ‘There’s a rash of injuries, let’s go fix something,”‘ Peterson said. “What you want to do is catch it before the horses get hurt.”

Catastrophic injuries at thoroughbred tracks in the United States are not uncommon, although the death of filly Eight Belles last year was the first in 134 runnings of the Kentucky Derby. While her accident has not been attributed to a problem with the track, Peterson is convinced injecting science into track maintenance can significantly curb fatality numbers.

While this is the first year Peterson has been using his device at Churchill, it has been employed for five years at racetracks in California, where it did catch a drainage problem.

The idea for the device came more than a decade ago when Peterson was on the faculty at Colorado State and was working with equine orthopedic surgeon Wayne McIlwraith on a graduate student’s research project.

That project dealt with the correlation between horse exercise and injuries, but Peterson wondered whether the data could ever really be valid unless there were uniform standards for how racetracks and training surfaces should be maintained. As it turned out, there weren’t.

Besides working on the device, Peterson and McIlwraith have teamed up this year to create a laboratory in Maine that will analyze samples from synthetic and dirt tracks to determine the risk of injuries to horses. Churchill and several other tracks have contributed to the startup funds for the lab.

“We’re still on a learning curve, particularly for synthetics, but you adjust the track to accommodate,” McIlwraith said. “Sometimes it may not be changing material. It may just be adding water.”

As for the robotic hoof in use at Churchill and elsewhere, it produces two sets of numbers measuring vertical and horizontal force.

The first reading measures the “load,” or how well the surface can withstand the slamming of a leg onto the ground. The other measurement is the “slide,” the horizontal force of the hoof once it is in contact with the ground.

The wrong load can potentially break a horse’s leg while slide problems can create soft tissue injuries.

Finding a track with the perfect balance of load and slide is the goal, but Peterson says the device isn’t designed to answer the long debate about whether dirt, grass or synthetic tracks are safest. Rather, he’s using the machine on a test basis at 27 tracks, giving supervisors information to improve whatever surface they have.

“We’re looking for a way we can scientifically judge what we’re doing,” said Butch Lehr, track supervisor at Churchill Downs. “We’re really new to it, but we’re also excited about what we’re going to find.”