Every conversation you have, every passing car, every hum of a crowded room — that sound is energy, and almost all of it simply disappears.
Scientists at the Terasaki Institute for Biomedical Innovation have developed paper-thin fibrous mats that can be worn on the body and harvest ambient acoustic energy, converting it into usable electricity. To get there, they turned to artificial intelligence — and what it helped them achieve went well beyond what conventional fabrication methods had managed.
Sound as a power source — an idea whose time has come
Solar and wind dominate the conversation around environmental energy harvesting. Acoustic energy gets almost no attention — despite being constant and omnipresent. Every voice, engine, and background hum represents mechanical energy moving through the air, and nearly all of it is wasted.
Piezoelectric nanogenerators offer one path forward. These devices convert mechanical vibrations, stress, or strain directly into electrical power, which means sound waves can become a functional energy source. The physics is well established. The challenge is engineering.
Most ambient sound falls in the low-frequency range, while conventional piezoelectric devices perform best at higher frequencies — a poor match for everyday acoustic environments. Wearable acoustic energy harvesters, known as NAEHs, have emerged as a promising solution, but building one that performs reliably in real-world conditions has proven technically difficult.
Choosing the right materials: PVDF, polyurethane, and ultrathin fibers
The TIBI researchers selected polyvinylidene fluoride — PVDF — as their primary material because of its well-documented capacity to capture acoustic energy efficiently. Flexibility is critical in any wearable device; a rigid harvester worn on the body would be impractical.
The team addressed this by blending polyurethane into the PVDF solution, giving the composite nanofibers the pliability required for body-conforming use.
To produce the ultrathin fibers, electrospinning was used — a fabrication technique that draws material through an electric field to generate fibers far thinner than a human hair. The result was a PVDF/PU composite nanofiber material ready for further optimization.
Where AI entered the lab
Electrospinning involves multiple adjustable parameters, and small changes can substantially affect the final product. The team applied AI techniques to identify the optimal combination of applied voltage, electrospinning time, and drum rotation speed in their process.
Conventional optimization means running experiment after experiment, adjusting one variable at a time. AI navigates that parameter space far more efficiently. TIBI director Ali Khademhosseini put it plainly: models using AI optimization “minimize time spent on trial and error and maximize the effectiveness of the finished product.” It is an approach that has already compressed development timelines in other advanced manufacturing fields while improving outcomes simultaneously.
Building the harvester: from nanofibers to a wearable device
With optimized nanofibers in hand, the team assembled the harvester itself. The PVDF/PU material was formed into a nanofibrous mat and sandwiched between two layers of aluminum mesh, serving as electrodes, then enclosed within two flexible frames.
The resulting device is compact, body-conforming, and designed to capture low-frequency ambient sound — the range that dominates everyday environments. Testing showed it could recognize speech with high-resolution word distinction, pointing directly toward hearing aid applications where accurately capturing and converting ambient sound is the core requirement.
Performance that outpaces conventional designs
The AI-optimized NAEHs significantly outperformed their conventionally fabricated counterparts. Power density came in at more than 2.5 times higher, and energy conversion efficiency rose from 42% to 66% — a meaningful gain for this class of device.
Those results held across a wide range of low-frequency sound, consistent with real ambient noise conditions. Strong performance in a controlled lab setting is one thing; maintaining it across the acoustic range of everyday life is another matter entirely. The AI-optimized device delivered on both counts.
What this means for the future of wearable healthcare tech
Hearing aids represent the most immediate application, but the implications reach further. Flexible wearable health monitors — devices that require small, continuous power sources without bulky batteries — are another natural fit.
The broader framework the TIBI team developed may prove just as significant as the device itself. Strategic material selection paired with AI-guided fabrication optimization is a transferable model, one applicable to other energy harvesting challenges and medical device development more broadly.
Khademhosseini noted that this approach “can have far-reaching effects on the fabrication of medical devices with significant practicability.” As AI tools become more integrated into materials science and biomedical engineering, this kind of intelligence-guided design may shift from exception to standard practice.
Carlos is an engineer with strong expertise in technical and industrial topics. He previously worked at international companies such as Siemens and speaks Spanish, German, English, and Italian.
