UA_Engineer_25_FINAL_spreads (1) - Flipbook - Page 18
“
ABOVE LEFT: Dr. Samit Roy stands inside his aerospace engineering and mechanics lab, where his team develops intelligent self-healing
materials. ABOVE CENTER: A researcher examines a composite panel embedded with smart sensors designed for autonomous damage
detection and repair. ABOVE RIGHT: Dr. Roy speaks with his research team about integrating damage sensing, prognosis and self-healing
into a single system.
Led by Dr. Samit Roy and supported by co-investigator
Dr. Sameer B. Mulani, the project has achieved what was
once science 昀椀ction: an intelligent autonomous system
that can detect damage, diagnose its severity and heal the
material in real time. Funded by the Air Force Of昀椀ce of
Scienti昀椀c Research and NASA, the project’s implications
are already shaking up how scientists and engineers
think about maintenance, inspection and survival in
hostile environments.
At the center of the innovation is an intelligent system
composed of three tightly integrated modules for damage
sensing, damage prognosis and self-healing activation.
What makes this system so revolutionary is that it does not
rely on human intervention. Instead, the material “knows”
when it has been compromised and how to mend itself.
Traditional aerospace maintenance relies on routine
manual inspections. But these inspections cannot catch
all types of damage, especially internal microcracks that
slowly grow over time. Even worse, they also do not help
mid-昀氀ight.
“The current procedure is you have to do regular
inspections every few hundred hours for a particular
aircraft,” Mulani said. “Even if there is no problem,
you still have to do it. So that’s a problem with the cost.”
That downtime ends up costing money, and in some cases,
greater logistical complications.
“The concept is inspired by the human body,” Roy said.
“You get a cut, and it heals itself. Can we get structures
to do that? If we can extend the life of an aircraft from 20
to 40 years, we’re saving lives and money and reducing
maintenance costs. In space, it’s even more critical,
where repairs aren’t an option.”
The self-healing system uses a network of smart
sensors, speci昀椀cally, piezoelectric devices called macro
昀椀ber composites, that are embedded into the surface of
aerospace structures. These sensors constantly monitor
vibration patterns within the material. The sensors are
about 14 by 28 mm. Each patch can heat a zone about
twice its size.
“It’s similar to a heart monitor,” Roy said. “If your pulse
18 | CAPSTONE ENGINEER | FALL 2025
jumps to 200, it tells you something is wrong because
it should be 72. Then the question becomes: What’s the
problem? Is it a crack? Is it wear? How big is it? One
sensor can’t answer that, so we use multiple sensors.
These sensors provide voltage signals—lots of data, which
by itself would be meaningless. You need something to
read and interpret it.”
Roy’s lab uses higher-level simulations, reduced-order
models and arti昀椀cial intelligence to diagnose the size
and location of the damage. Not all damage is the same,
however. Once it reaches a critical size, the structure will
soon fail. Their goal is to catch it well before it reaches
that point.
“The sensors pick up vibration in the structure,” Roy
said. “Since you’ve tuned it to a baseline, any change from
that baseline tells the prognosis module: ‘Hey, there’s a
problem.’” Once a crack or delamination is detected, the
next step is determining the extent of the damage. Some
structural 昀氀aws can wait, while others demand immediate
action. But how is an aircraft, satellite or spacecraft able
to make the call on what is a priority?
“We use actuators and sensors in a pitch-catch
con昀椀guration,” Roy said. “You place one actuator on top
and another on the bottom. One goes into tension, the
other into compression, and that 昀氀exes the structure. If
there’s a delamination or a crack in between, the signal
changes profoundly. That gets picked up, and based on
baseline data, we can tell how big the damage is.”
Using high-resolution simulations, like 昀椀nite element
models that predict how materials respond to forces,
Roy’s lab trained a machine learning system to make
those snap decisions. The machine learning model is
physics-informed, and its training is like training a dog
to understand and quickly perform a command.
“That’s what we want: speed without sacri昀椀cing
accuracy,” Roy said.
Roy and Mulani use 昀椀nite element analysis: a high昀椀delity simulation tool used widely in aerospace. It can
model nonlinear material behavior and actuator responses
but takes hours to run.
