For many tenants and homeowners, smoke alarms are the bane of the household. Although sensors are meant to warn residents of dangerous fires and other threats such as carbon monoxide emissions, they often react to smoke from the stove or steam from the shower. Sometimes people become frustrated enough by such “nuisance” alarms that they remove smoke alarm batteries or otherwise disable the devices, putting their safety in jeopardy.
To address this issue, researchers led by Bruce Warmack of the Oak Ridge National Laboratory’s Electrical and Electronics Systems Research (EESR) Division developed the Smart Smoke Alarm, a more accurate but affordable detection device based on a sophisticated mathematical algorithm.
“Smoke always accompanies fires, but smoke also sometimes accompanies nuisances like burning toast or cooking pizza or steam from a shower,” Warmack said. “It’s difficult to distinguish those types of aerosols from fire aerosols.”
Although deaths from house fires have been cut in half since smoke alarms were widely implemented in 1975, thousands more lives could be saved by eliminating nuisance alarms and providing more warning time in the event of an actual emergency.
“According to one Hughes study, as many as 1,235 people per year could escape death from home fires if nuisance alarms were not there to provoke homeowners to disable their alarms or move alarms away from areas that really need to be protected,” Warmack said.
The Smart Smoke Alarm is also programmed to quickly detect smoldering fires, which progress gradually as smoke and carbon monoxide steadily rise to unsafe levels.
“With our algorithm, we can detect smoldering fires as much as 30 or 40 minutes sooner than with conventional alarms,” Warmack said.
Warmack and his colleagues used a mathematical process called linear discriminant analysis to program the smoke alarm, allowing it to operate at a high level usually reserved for advanced chemical detectors. ORNL’s Dennis Wolf of the Computer Science and Mathematics Division developed the main LDA algorithms, and Shane Frank of the EESR Division’s Sensors and Embedded Systems Group designed the electronics component.
The team ran sensor data from the National Institute of Standards and Technology in conjunction with reported nuisance data through the mathematical model, which helped to determine optimal methods of combining and analyzing the information. From the results, they developed the prototype. The changes added only a few lines of code.
“We developed our own prototype smoke alarm that had all the conventional sensors actually taken from commercial units, then we programmed that with our linear discriminant model,” Warmack said.
With funding from the US Fire Administration and the Consumer Product Safety Commission, the ORNL team took 12 prototypes to Underwriters Laboratories to undergo fire and nuisance tests in 2014. The Smart Smoke Alarm exceeded expectations, and the technology has since been licensed to a manufacturer.
The team’s technology was a finalist in the 2016 R&D 100 award competition, which recognizes exemplary technologies newly introduced to the market.
“I was very gratified because the R&D 100 nomination is a good way of signifying that other people think it’s a significant advance, but it’s also a way of getting the word out that this new technology exists,” Warmack said.
“If I’ve contributed in any way, if our team has contributed in any way to reduce the number of deaths, even by a small number, that would be very gratifying,” Warmack said. “We see the potential for a significant impact on safety.”
