3D printing robot using AI machine learning

In a laboratory at Boston University's School of Engineering, a robotic arm drops small plastic objects into a box placed flat on the floor and catches them as they fall. One by one, he fills the boxes with small, feather-light, cylindrical structures less than an inch tall. Some are red, blue, purple, green, or black.

Each object is the result of an experiment in robot autonomy. Robots are learning to search for and create objects with the most efficient energy-absorbing shapes that have ever existed.

To do this, the robot creates a small plastic structure with a 3D printer, records its shape and size, and transfers it to a flat metal surface, equivalent to an adult Arabian horse standing in quarters. crush it with pressure. The robots then measure how much energy the structure has absorbed, how its shape has changed after being crushed, and record every detail in a vast database. The crushed object is then dropped into the box and the metal plate is wiped clean, ready for printing and testing the next piece. This is slightly different from its predecessor, whose design and dimensions were fine-tuned by his algorithms on the basis of all past experiments, the so-called Bayesian optimization of the robot's computer. With each experiment, the 3D structure becomes better at absorbing the impact of being crushed.

These experiments are possible because keith brownENG Associate Professor of Mechanical Engineering and his team KAB Lab. The robot, named “MAMA BEAR,” is an abbreviation for its long formal title, “Mechanics Bayesian Experimental Autonomous Researcher for Additive Manufacturing Architectures,” and has been around since it was first conceptualized by Brown and his lab in 2018. It has evolved. By 2021, the lab will have installed this machine on a robot. We seek to create shapes that absorb the most energy, a property known as mechanical energy absorption efficiency. This current iteration he has been running continuously for more than three years and includes dozens of boxes filled with his more than 25,000 3D printed structures.

Why are there so many shapes? There are countless uses for anything that can efficiently absorb energy. Examples include cushioning materials for sensitive electronic equipment shipped around the world, and knee pads and wrist guards for athletes. “This data allows him to leverage the library to make better car bumpers or packaging equipment, for example,” Brown says.

To function ideally, the structure must be perfectly balanced. It shouldn't be so strong that it damages what it's protecting, but it should be strong enough to absorb the impact. Before MAMA BEAR, the best structure observed to date had an energy absorption efficiency of about 71%, Brown says. But on his chilly January afternoon in 2023, Brown's lab witnessed the robot achieve his 75% efficiency, breaking the known record.The results were announced the other day Nature Communications.

“When we started, we didn’t know if it would be in this record-breaking shape,” says Kelsey Snapp (ENG’25). PhD student in the Brown lab The person who oversees MAMA BEAR. “Slowly but surely we continued to move forward step by step and made the breakthrough.”

This record-breaking structure appears to be something researchers weren't expecting. It has four points shaped like thin petals and is taller and narrower than earlier designs.

“We're excited that we have so much mechanical data here that we can use to learn lessons about more general design,” Brown says.

Their vast amount of data has already been applied in real life for the first time, helping to design new helmet pads for U.S. Army soldiers.Brown, Snap, Project Collaborator Emily WhitingAn associate professor of computer science in the BU College of Arts and Sciences, he worked with the U.S. Army to conduct field tests to ensure that helmets with patent-pending padding were comfortable and provided adequate protection from impact. His 3D construction used in the padding differs from record-breaking items by being softer in the center and lower in height for increased comfort.

MAMA BEAR isn't Brown's only autonomous research robot. His lab has other “BEAR” robots that perform a variety of tasks, including Nano BEAR, which uses a technique called atomic force microscopy to study how materials behave at the molecular scale .Mr. Brown also cooperates. Jörg WernerENG, assistant professor of mechanical engineering, has tested thousands of thin polymer materials and Works best with batteries.

“These are all research robots,” Brown said. “The philosophy is that by using machine learning and automation in combination, we can significantly increase the speed of research.”

“It’s not just fast,” Snapp adds. “It allows you to do things you wouldn't normally do. It allows you to achieve structures and goals that you wouldn't have been able to achieve otherwise because it would be too expensive and time consuming.” He said the experiment began in 2021. We have since worked closely with MAMA BEAR to give the robot visual capabilities known as machine vision and the ability to clean its own test plates.

KABlab wants to further demonstrate the importance of autonomous research. Brown hopes to continue collaborating with scientists from a variety of disciplines who need to test an incredible number of structures and solutions. Although they have already broken the record, “we don't have the ability to know if they have reached maximum efficiency,” Brown said, meaning they could break the record again. So while Brown and his team consider what other uses the database could be useful for, MAMA BEAR will continue to run and push the boundaries even further. They are also studying how the more than 25,000 shattered pieces can be rewound and reloaded into a 3D printer, allowing the material to be recycled for further experiments.

“We will continue to study this system because, like many other material properties, mechanical efficiency can only be accurately measured experimentally,” says Brown. as soon as possible. ”

Republishers are reminded to maintain journalistic integrity by including a direct link to the original source URL and providing appropriate credit.