A system capable of detecting a fire within seconds – in some cases even before smoke is visible – could prevent millions of pounds worth of damage anywhere from petrochemical plants to shopping centres.
Using the system, blazes such as the fire in the Channel Tunnel last November could have been caught before the burning train entered the tunnel.
VSD-8 was jointly designed and developed by Magnox Electric Technology and security company Intelsec Security Systems to deal with problems such as fires caused by oil leaking on to pipe lagging in power station turbine halls. This was the cause of a conflagration at Aberthaw power station last autumn.
The system uses video motion detection software in conjunction with closed circuit TV cameras and can be retrofitted to existing CCTV systems at low cost, say the developers. Detection over long distances is possible using suitable lenses.
VSD-8 was developed in a two-year research programme. The motion detection software runs on a PC, which can accept the input from up to eight cameras.
The field of view of each camera can be divided into 10 `windows’ and the sensitivity of each can be adjusted independently.
A unique feature is that the system can differentiate between smoke, steam and people crossing the field of view.
Magnox strategy engineer Andy Brown said: `Fires tend to grow exponentially, so it is important to catch them at the earliest possible stage.’ In a turbine hall or any large area, it is difficult to identify the early stages of a potential fire. Until now detection has relied on operators noticing the early signs of a fire on one of many CCTV monitors, said Brown.
Early tests at the Building Research Establishment showed that Intelsec’s existing motion detection system could detect the appearance of smoke in just over a minute from ignition. But it could not detect white smoke against a light background or fires which produced little smoke, such as propane.
And to prevent false alarms caused by people moving across the field of view, a delay was introduced into the system, requiring a continuous alarm for 30 seconds before the operator was alerted.
To improve performance and solve the white smoke problem, Intelsec developed filter techniques to differentiate smoke from other visible objects.
Using a pixel by pixel analysis of successive video frames, the system can recognise characteristics of smoke – for example, said Brown, it always moves upwards, and a wisp of smoke has a fuzzy edge, whereas people have a sharp edge.
This enabled the 30-second delay to be taken out. Instead, the system issues a pre-alarm for the first 30 seconds while it verifies that it is seeing smoke. It then triggers a full alarm via a flashing light or siren and by showing the alarm area, and the window in which it was detected, on the operator’s monitor.
Trials were carried out in real turbine halls using smoke generators and a hotplate to ignite oil and simulate an oil on logging fire. The system appeared to generate false alarms but in fact it was being triggered by uncombusted vapour rising from the hotplate before it ignited – in other words, detecting a potential fire before it started.
Brown describes the ability to do this using inexpensive off-the-shelf equipment as `a major breakthrough in fire detection’. Attempts to fool the system by placing people or equipment in the target area, and even generating steam in front of the target area to mask the vapour or smoke, failed.
Magnox is installing the system at its Wylfa power station and British Energy is to use it at three of its own plants. Eurotunnel is also apparently interested. `Because the Channel Tunnel uses electric trains, so there are no diesel fumes, it could use the system to monitor its marshalling yards,’ says Brown.
The Shuttle involved in the Channel Tunnel fire is known to have been in flames before it entered the tunnel.