Until recently, the causes of unwanted alarms from optical flame detectors have been poorly understood. The remoteness of the sites makes communications difficult. Descriptions of problems are often received second or third hand, so accuracy can be rather limited. Flare stacks have been a common cause of unwanted alarms because detectors can respond to reflections in standing water or on metal surfaces. ppoor communications means that these false alarms were reported to be caused by sunlight or even by people in the area wearing orange overalls. Poor communication and understanding have inevitably delayed finding good solutions.
BP first encountered the reflection problem in the early 1980's on the Top Deck gas processing areas that had a flare some 50m directly above them. We tried to resolve this problem using a UV background radiation subtraction system. The UV radiation from sources outside the fire area was continuously measured and a proportion of this signal used to offset the zero on the fire detectors. This compensated for reflected flare radiation in the fire area. The subtraction had to happen in "real-time" since the size of the flare could vary considerably, particularly during a process upset. In principle these systems were a good idea but in practice have been prone to errors and unwanted alarms. The main problems have been calibration drift of the UV sensors caused by their gradual deterioration, and uneven contamination of the optics from one sensor to another.
In the mid 1980's BP needed fire detection on another North Sea site that had top deck process areas but this one had two flares horizontally adjacent to the plant. Either, or both of these flares could be burning at any time so that there were a large and variable number of reflections of the flares in the plant.
To a conventional optical detector, the process plant looked like one big fire. Clearly we could not design a reliable optical fire detection system for this environment. Our earlier experience told us that UV subtraction systems would not be suitable for this application, nor were there any other optical detection systems available that could cope with such conditions. Even though flame detection technology was shifting from UV to Infrared types, the flare problem remained. Despite the efforts of several vendors, it proved impossible to produce detectors that could respond to reasonably small fires and also be immune to incident flare radiation. Essentially, a reflection of a flame still looks like a flame regardless of the frequency at which it is viewed.
In the early 1990's research was conducted on the management of major incidents offshore. One of the findings was that F&G detection systems were not giving a sufficiently high quality of information about the hazards. This can lead to the wrong actions being taken. In response, BP funded research to establish if CCTV could be developed as an automatic fire detection system. The rational was that a live picture of the detected incident should give far better information than a combination of audible alarms and flashing lights. The outcome of the study was that although such systems could be built, the technology was not yet sufficiently advanced for commercial development or for reliable operation.
For this site we were forced to resort to heat detection in the form of pressurized plastic tube. These tubes were mounted close to the process vessels so that any fire threatening the vessels would burst the tube. The loss of pressure was signaled to the control room as a fire alarm. It was our practice at that time to investigate alarms before taking action (partly due to the high unwanted alarm rate). We were concerned that the low sensitivity of the detection would mean that fires would be too big to control by the time we had confirmed that the alarm was genuine. To speed up the response time we installed Infra Red (IR) sensitive cameras covering the process areas so that alarms could be visually "confirmed" in the control room with the minimum of delay. This was the first time BP used Closed Circuit Television (CCTV) for fire detection purposes.
In the mid 1990's BP began to use more open designs for its sites, including Floating Production Storage and Off-loading (FPSO) vessels. To cope with the fire detection needs of such sites we developed the pressurised plastic tube detector to make it more reliable, more sensitive and faster. The design included a number of valves and pressure transmitters to optimise system response. We can now reliably control automatic deluge, either directly or via the Fire & Gas Panel with this technology. We can respond to kilowatt rather than megawatt sizes of fireS and within two minutes rather than ten. However, we continued to look for better ways to detect fires in the difficult environment of an open process area illuminated by an adjacent and variable flare hanging over the plant.
When UK oil exploration and production activities spread to the Atlantic waters west of the Shetland Islands the use of FPSOs became important. The compact, single-level nature of the process plant and the need for rapid and reliable detection required better performance than we could reasonably achieve, even with the advanced tube designs. Conventional optical detectors had also advanced a great deal, but still had no answer for the problems with FPSOs.
An additional problem for this type of site is that there is often insufficient firewater capacity to deluge all process fire areas simultaneously. The hugely increased flare of an emergency shutdown would probably trigger all conventional optical detectors and thus all the process area deluges. The one deluge that we were perhaps relying on to control the incident would become ineffective. It was time to look again at the CCTV concept.
The aim of the fire detection system was that it would alarm the presence of fires and prompt actions in time to prevent escalation. In design we used a target fire size of 50 kW Radiant Heat Output and a response time of between 2 and 3 minutes depending on the sensitivity of the protected equipment to fire. We have demonstrated by test that these systems can achieve this level of performance. However such a system was still not as sensitive or fast acting as an optical flame detection system and therefore was not in line with good industry practice or the ALARP concept. ALARP stands for As Low As Reasonably Practicable and is stated in current legislation covering our industry. In our case, the intent of ALARP is to use the best available tools and practice to minimize the threat to people and equipment from fires.
This led us to review the need for better fire detection in the process areas of the vessel. We visited with the suppliers of conventional flame detectors. Most did not understand, or admit to understand, the problem and none were able to provide a credible solution to our problem of rejecting reflections of fires. We then returned to the company who did the CCTV research work for us in the early 1990's. This company managed to convince us that the components for a CCTV fire detection system were now cheap enough to allow them to develop a system. This based on the company's track record in developing fire detection algorithms gave us the confidence we needed to award a fixed price contract for a system.
The system comprises of 3 basic components, the camera/detection element, the control panel and the display computers all connected by an RS485 network. The clever part is all performed by the camera/detection element. This unit contains, in a small hazardous area approved enclosure, the camera and computing facilities to determine whether or not the unit is seeing a fire. The device is programmed with a range of algorithms to determine whether the changes within its field of view are a fire or not, and if it is a fire the view of the camera is automatically displayed at the control point. The system has been designed to detect fires of 10kW RHO in less than 10 seconds. Due to getting early alarms at the control point, we decided that control action initiation would be manual. This is not to say that automatic control action was unreliable, but that we did not see the need for it.
Since the system was installed last year there have been some minor problems that have since been overcome. The development of the system was based on our understanding of an operating site and had not considered the construction phase. The system had been designed to ignore sources such as welding but we had not anticipated a strange tarpaulin that our constructor used that was almost a perfect band pass filter for the wavelength that the camera is particularly sensitive to. Therefore on bright sunny days, combined with particular types of tarpaulin movement, the system would false alarm. However once we understood this, the system was reprogrammed, and retested to ensure it had maintained its sensitivity to fire, and no longer responds to this non fire source. A neat feature of the system was the non-disruptive nature of reprogramming. The new software was loaded at the control panel and downloaded to each camera.
The FPSO has been producing oil since July 1998 with a flare burning in various sizes for most of this time. To date we have had no false alarms from any of the cameras. In the eight months prior to oil production we regularly exposed each camera to a range of fires. In addition to that we commissioned the building of a test facility at the vendor's works. This facility contains a small CCTV fire detection system covering a mock process area. This facility is being used for ongoing tests and development. Due to the ease of upgrade, our system will be maintained to latest standard as the system is developed.
This brings us to the future. Clearly with no other alternatives available we will continue to use such a system in areas where reflected flare radiation is a problem. However the main benefit of the system is the live image of the area of alarm. At last we have a system that provides clear unambiguous information on what it has detected. With this system no longer will offshore operators question whether the cause of alarm is false or a real fire.