'Flare' For Fire Research Leads To New Detector

"One advantage of this device is that it doesn't have to be 'looking' directly at the flame to 'see' it," says Jay Gore, a professor of mechanical engineering at Purdue who specializes in combustion research.

"It can pick up reflections from a fire off the walls, so it can directly survey a multiple-room enclosure for fire from a single location. This gives it a very fast response time compared to a smoke detector, which doesn't go off until smoke has traveled to it -- a delay that can be several minutes." Because the new detector is very sensitive, Gore says, it may first be used in large warehouses, but it also will benefit the home.

Gore and Yudaya Sivathanu, a Purdue research scientist, developed the new detector with the help of a two-year research grant from the Center for Fire Research at the National Institute of Standards and Technology (NIST).

Because the new device uses fiber optics to detect radiation from a fire, optical fibers could easily be run from a central detection unit to each room in a multistory building, Gore says, providing "blanket" coverage.

Also, the Purdue device would benefit those people who periodically turn off their smoke detectors to avoid a false alarm. The Purdue engineers say their device would cut down on false alarms because it doesn't respond to common household occurrences, such as a hot plate or an overcooked hamburger.

The prototype detector is bulky, but Sivathanu has established a small business, En'Urga Inc., at Purdue's Business and Technology Center to develop the detector for industrial and commercial use within the next three to five years. Sivathanu presented information on the new detector Aug. 15 at the Second International Conference on Fire Research and Engineering in Gaithersburg, Md.

The device, which detects the characteristic light given off by an uncontrolled flame, could be hooked into a telephone or personal computer to automatically notify the fire department and give off audio safety instructions when a fire starts, Gore says.

"If the device were connected to a personal computer, as we have it in the lab, the computer can be programmed to repeat safety instructions to people in the house, such as 'Stay close to the floor,' or 'Crawl to the nearest door,'" Gore says. "This can be very helpful in an emergency, when people may be panicking, but it is especially important for small children who are frightened by scary, loud alarms. It could be very calming to hear a parent's voice, even if it is coming from the computer.

"I believe that in the next few years, the home PC will control the home security and safety systems and that a fiber optic communication network will carry such signals in addition to voice and data. This detector could be easily integrated into such a communications network."

The wavelengths of light that the detector picks up are in the near infrared -- in other words, heat. "The idea of using heat to detect fires is not new, but we have applied a unique discrimination algorithm to the process to eliminate false alarms," says Gore, whose previous research on different types of flames helped NIST researchers in their analysis of oil well fires in Kuwait.

Over the past two years, as they were developing the detector, Sivathanu, Gore and their graduate students examined the near-infrared radiation given off by several "standard" types of fires. These type of fires cover a wide range of combustible materials.

"We have analyzed flames based on the way the intensity of the light fluctuates as they burn," Gore explains. "For example, we may see a peak in intensity every tenth of a second, and then the pattern is repeated. That type of fluctuation frequency is characteristic of an uncontrolled fire."

Once a flame has been detected, a sophisticated computer program analyzes the fluctuations in its near-infrared intensity -- its so-called infrared signature -- and determines whether to sound an alarm. While infrared signature analysis has been used by the military in a variety of ways, using it for fire detection is a new application.

"We have 'taught' the detector not to respond to common household flames such as candles, gas stoves and cigarette lighters," Sivathanu says. "It also does not respond to fluctuations from hot plates, solar radiation or fluorescent light, which are different than those from uncontrolled flame.

"We also installed a corrective filter when we learned that the detector went off when someone waved their hand quickly in front of a hot plate placed in direct view of the detector. We still have a problem with the alarm going off from a fireplace, which is an uncontrolled flame, but we are developing ways that may make the device 'blind' to certain spots in the room. Conventional smoke detectors also go off from smoke from fireplaces."

Another bug to be worked out of the new detector is its difficulty in picking up smoldering fires, an area where smoke detectors also are slow to respond because of the time it takes for the smoke to reach the detector.

"The intensities obtained for smoldering fires are too low for our detector to successfully discriminate them from background noise," Gore says. "However, new, more sensitive infrared technologies are now available that might boost this capability in our detector."

Sivathanu and Gore have received a new, three-year grant from NIST for research that will focus on detecting the far-infrared radiation emitted from a fire, the type of radiation associated with smoldering.

Sources: Jay Gore, (765) 494-1452; e-mail, gore@ecn.purdue.edu

Yudaya Sivathanu, (765) 494-9364; e-mail, sivathan@ecn.purdue.edu

Writer: Amanda Siegfried, (765) 494-4709; e-mail, amanda_siegfried@uns.purdue.edu

Purdue News Service: (765) 494-2096; e-mail, purduenews@uns.purdue.edu

ABSTRACT

Experimental and numerical evaluation of a near-infrared fire detector

Ying-Jie Zhu, Andrew Lloyd, Yudaya Sivathanu, Jay Gore - Thermal Sciences and Propulsion Center, School of Mechanical Engineering, Purdue University

Near-infrared fire detectors work on the principle of detecting fires based on a statistical analysis of the apparent source temperature of fires. The apparent source temperatures are estimated from the radiation intensity incident on the fire detector at two near-infrared wavelengths. However in some instances, the fires are not in the direct view of the detector, and most of the radiation which is incident on the detector reaches it after multiple reflections from the walls of the building. An experimental and numerical study of the effects of these reflections on the temperatures inferred by a near-infrared fire detector are presented. The experimental evaluation was conducted using three open and two smoldering fires. The results show that the near-infrared fire detector is capable of discriminating open fires from reflected radiation. However, for smoldering fires, the intensities obtained from reflected radiation are too low to be successfully discriminated from background noise. Numerical evaluation of the performance of the near-infrared fire detector in cylindrical and rectangular enclosures were conducted utilizing a photon tracing algorithm in conjunction with the discrete probability function method. The numerical evaluation confirms that the detector can successfully detect fires from reflected radiation if its sensitivity is sufficiently high.