ETRI’s AI sensor sniffs out false fire alarms

Anyone who has ever seared a steak knows that smoke detectors have a serious trust issue. But while a false alarm at home is just annoying, on a national scale, it is an expensive disaster.

Researchers at South Korea's Electronics and Telecommunications Research Institute (ETRI) just developed an AI-powered sensor that can distinguish between real fire smoke and non-fire aerosols, and it is already knocking on the door of commercialization.

Smoke detectors have a massive trust issue

Traditional photoelectric smoke detectors are shockingly basic.

They house an infrared light source and a photodiode. When any airborne particles enter the chamber and scatter that light beyond a certain threshold, the alarm screams. It does not care whether those particles came from burning wood, kitchen steam, cigarette smoke, or everyday dust.

That limitation causes staggering waste. According to South Korea's National Fire Agency, fire engines were dispatched 258,220 times between 2021 and July 2022. A wild 96.6% of those dispatches were false alarms caused by malfunctioning detectors.

ETRI Director Kang Bok Lee estimated that these false dispatches burn through roughly 20 billion KRW every single year in wasted firefighting resources.

Giving aerosol particles an optical fingerprint

ETRI solved this by looking at how particles scatter different wavelengths of light.

By shining multiple wavelengths onto airborne particles, the team discovered that smoke from actual combustion scatters light differently than dust or moisture. They built a database of these wavelength-dependent scattering patterns and trained an AI model on it.

Instead of blindly sounding the alarm whenever light scatters, the new sensor analyzes the optical fingerprint first, deciding whether particles are genuinely from a fire before triggering the siren.

ETRI plans to deploy this AI sensor first into aspirating smoke detectors. These systems actively draw air in using fans to catch smoke quickly, but they are notorious for tripping over moisture and dust.

Because of that sensitivity, aspirating detectors are currently confined to hyper-controlled environments like semiconductor clean rooms and server rooms. Most of them are also pricey imports lacking any real intelligence, giving ETRI's domestically developed sensor a sharp competitive edge.

The institute noted that this scattering spectrum measurement could also branch out into cosmetics, medical testing, and environmental monitoring, and commercial discussions with industry partners are already underway.

Why it matters

For facilities like chip fabs and data centers, a false alarm can trigger automated gas suppression systems or costly downtime, making this upgrade an immediate no-brainer.

In the broader sensing industry, ETRI took a remarkably pragmatic route. Rather than introducing complex chemical sensors that degrade quickly, or surveillance cameras that invite privacy headaches, they stuck with optical scattering but upgraded the intelligence.

We think this approach is clever because it leverages established physical principles while letting lightweight AI do the heavy lifting of classification.

The real test will be long-term hardware durability. In real-world environments with oily grime and uneven dust buildup over years, maintaining precise multi-wavelength spectral data will be the true hurdle before this tech can replace legacy alarms worldwide.