Robot to the rescue


The US Naval Research Laboratory (NRL) has built a robot to pull, bend and twist samples of the composite materials used to build F/A-18s and other aircraft. Dr John Michopoulos leads the project.

With a machine that can, as he says, “measure so much more than anything else”, and some very advanced math, he can “create a theory that is consistent with all these experiments that we made, and works for all scales”. He predicts how the materials will perform when made into large structures and used over many years.

Sitting in his lab, Michopoulos picks something the size of a luggage tag off his desk. “This is the material that was used for the original versions of the F/A-18, for the skin of the airplane,” he says. It’s a sample of advanced composite, made of resin reinforced by carbon epoxy fiber. “That little thing that’s so light is actually stronger than steel.”

The F/A-18 Hornet became the US Navy and Marine Corps’ first strike fighter in 1978; the median age of today’s active aircraft is 22-23 years old. As F/A-18s continue to age beyond their design lifetime, showing structural stress corrosion cracking and wing panel composite skin abnormalities, engineers have had to do extensive analysis to develop repairs. “So,” explains Michopoulos, “the need for certifying a new material comes in and we ask, ‘How are we going to compare a new material and, if we start using it, have confidence it’s going to behave the same as or better than the old material?'”

With the multi-axial robot, Michopoulos’s lab has tested thousands of samples like this one and asked, ‘How are you behaving, when you see all possible loads that you may see when you are part of a structure?’ Today, the Department of Defense (DoD) uses a building-block approach from 1999, as set out in the Composite Materials Handbook-MIL 17. The approach starts with testing fibers and matrix materials. Then the tests get, as the handbook states, “increasingly more complicated”, until reaching the level of structural subcomponent (or higher).

It’s easy to see that this is time-consuming and expensive. “I can tell you that qualifying the system for the F/A-18 took about 13,000 specimens and about 18 years,” says Michopoulos. “Engineers are forced to conduct tests at multiple scales because they do not have a theory to connect the behavior across multiple scales.” As private companies and the military continue to look to advanced composites for new aerospace and other applications, NRL’s robot could help get aircraft from factory to fleet faster.

Up to 72 loading paths

Michopoulos realized that, to accurately characterize the behavior of a material, he would have to capture the behavior of the composites in every possible combination of loading – as opposed to traditional approaches, where scientists take only simple loading cases, such as tension or compression. No technology existed to do that.

“So we said, to heck with it, let’s invent it.” Named NRL66.3, the robot is a multi-axial loading machine. It has six devices that apply linear movement, termed actuators, in a hexapod configuration. While a material sample is held by a fixed grip from one end, the actuators move a grip that holds the other end of a material sample, moving it in any combination of up to three translations and three rotations. “This means the robot can apply combinations of tension or compression, bending and torque simultaneously,” explains Michopoulos.

Like the old jukeboxes with an arm to load the next record, NRL66.3 is fully automated. Assisted by two other robots, it will take 72 specimens of, for example, the material used in part of an F/A-18, and apply 72 loading paths in that six-dimensional space. The robot loads each specimen until it snaps, then quickly moves onto the next. A custom-developed machine vision system, with four cameras, captures digital images of what’s happening in real time.

While the experiment is going on, the scientists use custom-developed full-field measurement algorithms that Michopoulos’s group has now patented to take those digital images and analyze them, and convert them to displacement and strain fields.

What the experiments do is to very quickly capture what might happen to an advanced composite in the real world. Advanced composites age in a very particular way. “The resin that’s between the fibers starts developing micro-cracks,” says Michopoulos. These micro-cracks can cause the resin to separate from the fibers or the fibers to break. “A continuous accumulation of micro-cracking, which leads to a softening of the material, can be used as a metric for material degradation assessment.”

The group has, over the past 20 years, used various robots to test more than 150 material systems, with potential applications for ballistic missiles to rocketry to automobile manufacturing. “So we do have a very rich database,” says Michopoulos.

Computational prediction

A snapped composite specimen is one thing; a theory of how that material will behave when formed into a jet wing is another. But that’s what Michopoulos’s group has done. “It’s highly computationally driven,” he says. “You cannot write on a small piece of paper a single equation that encapsulates how composites behave.”

From 2008-2012, the Cooperative Research Centre Advanced Composite Structures (CRC-ACS) of Australia provided specimens to NRL to test and characterize, while it did its own tests using more traditional methods. Universities in Australia and the Massachusetts Institute of Technology also participated in the project, with support from the Office of Naval Research (ONR). “We tested 1,152 specimens in 12 days; that has never really happened anywhere anytime before,” notes Michopoulos. Additionally, as Michopoulos explains, “ONR said, ‘CRC-ACS is also going to create specimens that you’re not going to test, but instead they are going to test. But you will tell us ahead of time, in a blind prediction, how they’re going to behave.'”

Michopoulos’s group ran predictions on the tests that they would make, which were really bigger specimens with holes or with stiffeners and so on. “We came within a maximum error of 3% on anything they asked us to predict,” he says. Without having tested these specimens experimentally at NRL, they were still getting the same results as CRC-ACS because of all the data they’d captured with NRL66.3.

“For me,” he says, “it really was kind of fun to see this going on before my eyes, and having all the collaborators experience it.”

NRL collected 12TB of data during the testing period. “Just to give you an idea of how much richness there is in this data,” he says, “out of the 72 loading paths we applied, an MIT student based a dissertation on just one of them.”

Michopoulos wants to offer design engineers a simulation environment based on actual data. “Instead of asking the question, ‘Is this good or bad in a general sense?’ – which again, is never really what they mean – it is better to ask, ‘Is it good for a particular use for me at this moment for my current needs?'”

Best of times

While Michopoulos is interested in continuing to run experiments on composites, he refers to it as a third-best time activity. “Characterization of composite materials,” he says, “is a domain that I’m fortunate to be able to express – as an industrialization process, an application.”

But, he says, “If done correctly and efficiently, we have some leftover time to reinvest in developing the tools we want in the future.” That time he calls second-best time. “If you look at the roots of development of technology, you see that the real revolution and progress has come from the time people spend developing their tools.”

What Michopoulos wants to build in the future is a self-configuring, self-organizing machine: “That’s the area I want to go. A robot that has a process to optimize its own performance based on how well it collects data for material characterization.” Such a robot could give more information, and do it more quickly and with fewer specimens.

The robot he has today can apply one loading path per specimen. Instead, he envisions a robot that could “follow a different path, a zig-zag, or a curvy path” – maybe even have actuators that move around to change the machine’s shape. “We have initial analysis that shows that, indeed, we can have the machine decide where it wants to go to get the best possible data for characterizing the material in real time.”

But then, he says, “There’s first-best time, which is really developing your mind. It’s that time that’s the most inefficient and the most painful – but also the biggest fun, and therefore the most exciting.”

Michopoulos has a PhD in theoretical and applied mechanics and applied mathematics from the National Technical University of Athens and pursued postdoctoral research in multiphysics, fracture mechanics, and applied mathematics at Lehigh University. When he tells how he came to be at NRL in 1986, he describes what his future mentor, Dr Phillip Mast, asked in the interview: “He said ‘If you imagine there is a line connecting Socrates with Bertrand Russell, where are you on this line?’ Instead of asking me an engineering and mathematics question, he asked me that question. So then I knew I had to stay.”

Kyra Wiens is a public affairs specialist at the US Naval Research Laboratory

Share this story:

About Author

Comments are closed.