Document Sample
rodricks Powered By Docstoc


Mr Brian S. Rodricks, BTech (Hons), MBA,
(Thorn Security Limited)

To be read at 1730 on Tuesday 14th November 2000
at the Institute of Marine Engineers, 80 Coleman Street, London EC2R 5BJ

NAME :                  Brian S. Rodricks

NATIONALITY :           British

QUALIFICATIONS :        BTech (Hons), MBA, AMIMarE

TITLE OF PAPER :        Fire Detection Systems for the Millennium

PRESENT POSITION :      General Manager - Marine, Offshore and International

PREVIOUS EXPERIENCE :   Brian Rodricks is a graduate Mechanical Engineer who
                        was employed as a senior sea engineer for over 10 years. During
                        that time he completed his Class One Certificate of Competency
                        (Motor). After completing an MBA in International Business at the
                        City University Business School he joined Thorn Security in 1989
                        as the Marketing Manager for their International Division. For the
                        last 7 years he has been managing the Marine and Offshore
                        Division of Thorn Security and for the last 3 years has also taken
                        responsibility for the Company’s International Division.

                              Page 2 of 28

It is nearly 10 years since a paper was read on marine fire detection systems . This paper will try and
address some of the new developments. Enhancements to the traditional fire detection principles and the
development of new technologies are reviewed.

It includes the use of Carbon Monoxide Fire Detection, Smoke and Flame Detection using Closed Circuit
Television and Optical Fibre Real Time Temperature Sensing. Detection systems have also been enhanced
using digital protocol knowledge based expert Systems and multi-sensor detectors.

The use of new technology is becoming more widespread as ship operators increasingly use risk
management based philosophies rather than relying on prescriptive SOLAS regulations to protect their

                                             Page 3 of 28

A Fire Detection System can be divided simply into four sections:-

1)      The sensing of the fire by detectors.
2)      The analysis of the information, using algorithms for analogue addressable systems.
3)      The annunciation of the alarm and subsequent actions via the Operator Interface.
4)      The communication between sensors and the Operator Interface.

Sensor Technology

a)      The High Performance Optical Detector

        Over the years, environmental responsibility has meant that most operators have shied away from
        fitting radio active ion-chamber detectors. In addition to this is the stringent regulation required in
        disposing, cleaning, storing and transporting these radio-active devices. There has therefore been a
        demand for an optical smoke detector that has “Middle of the Road” sensitivity to both hot, fast,
        “Clean” burning fires (domain of the ion-chamber detector) and cold, slow smouldering fires (domain
        of the optical detector).

        The High Performance Optical Detector meets this demand. The design is based on the experience
        and testing done which confirmed that flaming fires generate a significant rise in air temperature in
        the early stages of the fire together with a small increase in visible smoke. The High Performance
        Optical Detector senses both these characteristics. It then uses the rapid change in temperature to
        adjust the sensitivity of the Optical Detection chamber, to give a considerable improvement in the
        detectors smoke response characteristic. It is important to note that this improvement in sensitivity
        is achieved without an associated increase in the risk of nuisance alarms.            This is due to the
        patented design of the sensing chamber and the optics.


        The High Performance Optical Detector incorporates a rate-of-rise thermal detection function added
        to the light scattering optical detection principle; whereby the thermal element is used to control the
        sensitivity of the optical detector. Our fire test studies concluded it was possible that the rapid rise in
        ambient temperature associated with a flaming fire could be sensed and used to adjust the optical
        detection sensitivity without increasing the probability of nuisance alarms.

        To sense this temperature rise, two thermistors are arranged in a similar fashion to that found in a
        standard rate-of-rise heat detector (Figure 1). One thermistor is mounted so as to be exposed to the
        air while the second is shielded inside the detector’s body. If the temperature rises slowly then the
        thermistor temperature will be approximately equal and no adjustment to optical sensitivity occurs.

                                                Page 4 of 28
                           Figure 1: High Performance Optical Detector

If however the air temperature changes very rapidly, the exposed thermistor will heat more quickly
than the reference thermistor (heat shielded by the detector body) and a temperature difference will
be established. The electrical circuit senses that the exposed thermistor is hotter than the reference
thermistor and reduces the alarm threshold of the optical sensor accordingly.

If there is smoke present at a level above the reduced threshold then an alarm will be raised.
Otherwise the detector will remain in its enhanced sensitivity state, without giving an alarm until the
temperature stabilised.

For most day to day events, such as stopping and starting of heavy machinery, the temperature rise
is far too slow and too small to give any significant enhancement. So, for most of its life the detector
behaves as a normal optical detector.

The response of the various smoke detectors to the EN54 fire tests is shown in Table 1. These
tests are also used by the marine industry to evaluate smoke detectors. EN54 does not require that
each detector provides the same sensitivity to all fires. Rather, each detector is classified by its
sensitivity to a given fire type, A, B, C or N where A denotes the highest sensitivity level and N is
defined as unclassified or unsuitable. Under EN54 approval, a detector is deemed suitable for use
provided the test results achieve either an A, B or C classification only detectors classified as N are
unsuitable for the defined application.

The purpose of the fire sensitivity classification is to give the user an indication of the suitability of a
detector type for a particular potential fire risk.

                                          Page 5 of 28
        Performance Results

        Suitability Test Results : High Performance Optical (HPO)

Table 1 Response to EN54: Part 9 test fires of ion-chamber,
Optical and high-performance optical smoke detectors

                                                                  ION   OPTICAL             HPO
          TF1        Open cellulosic fire (wood)                  A     N                   C
          TF2        Smouldering pyrolysis fire (wood)            C    A                    B
          TF3        Glowing smouldering fire (cotton)            B     A                   B
          TF4        Open plastics fire (polyurethane)            A     C                   B
          TF5        Liquid fire (n-heptane)                      A     C                   B
          TF6        Liquid fire (methylated spirits)             N     N                   N

        From the test results it can be seen that the High Performance Optical offers a significant
        performance improvement over standard optical detectors, with a much more uniform performance,
        across TF1 to TF5. This “middle of the road” response will help reduce the number of unwanted
        false alarms.

        Additional more realistic fire tests have been carried out for THORN at the Fire Research Station
        Laboratory Cardington.

        A typical example of one of the tests involved a paper sack full of mixed rubbish, typical of that
        emptied from waste paper baskets. The difference in performance between the ion-chamber and
        the High Performance Optical is very small whereas the normal optical detector only just reaches
        the alarm level sometime later. Similar results were found in other flaming fires, while in slowly
        developing fires, the high performance optical outperformed the ion-chamber detector with
        responses close to that of a normal optical detector.


        The high performance optical detector design is based upon proven low-cost optical scatter
        technology. The incorporation of the heat sensing technique allows the detector to be sensitive to
        both flaming and smouldering type fires.

        The high performance optical detector operations characteristics provide an inherent, further
        immunity from nuisance alarms in that the normal sensitivity of the detector is maintained at its
        lowest level. This is accomplished with complete confidence that any rapid change in ambient air
        temperature will begin to ‘sensitise’ the detector and allow for immediate detection of the fire.

                                                Page 6 of 28
     The use of this detector that contains no radio-active material, together with its systems design
     flexibility, now offers the ship operator a cost effective, stable, false alarm free alternative to the ion-
     chamber detector.

b)   Multi-Sensor Detectors

     Early use of multi-sensors were confined to detectors where shortcomings (increased false alarm
     rates) of one type of sensor (for example Optical/Ion) were supplemented with another detection
     principle (heat detection) to reduce the amount of false alarms. This in most cases was a “work
     around” without addressing the inherent problem in the detector design. Smoke is propagated by air
     currents while heat is radiated, the best position for a heat detector is normally not the best place for
     a smoke detector. This meant that location of the detector was always a compromise.

     Nowadays, with attitudes to risk management changing, operators use combined smoke and heat
     detectors where individual outputs are available for the smoke signal and heat signal to help them
     visualise the emergency scenarios and decide on the best course of action. The smoke detectors
     would follow the spread of smoke while the real time readout of the temperature at the same time
     would help the operator to identify the spread of “fire” and any hot spots. Control Panels are now
     available which can give you this information and carry out a trend analysis.

     The construction of the detector is similar to the high performance optical detector but the heat and
     smoke readings are now totally independent of each other. What you now have is a multi-sensor
     virtual device with various modes (normal, low or high sensitivity) configured in software, including
     the high performance optical option.

c)   Carbon Monoxide (CO) Fire Detectors

     Irrespective of how intelligent a smoke detector is it still needs the smoke to be introduced to the
     detector before it can be sensed and an alarm decision made. In order for this to happen two
     assumptions are made, firstly that the seat of the fire will be such that smoke will travel to the smoke
     detector. Secondly, that the combustion characteristics of the fire will be such that a detectable
     amount of smoke is produced in sufficient time for effective evacuation or other actions to take

     Detecting real fires with Carbon Monoxide Detectors

     Over 10-year’s research has shown that these assumptions are fundamentally flawed in many
     applications. When detectors are positioned, the location of the seat of any potential fire cannot be
     known. Also usually unknown, or not always taken into account, when decisions are made on the
     positioning of detectors is airflow in the area. Installation standards and codes of practice around
     the world help to limit the room for error. However, it is almost impossible to position a smoke

                                              Page 7 of 28
detector and guarantee the smoke from any potential fire will reach the detector quickly. This is
even more difficult if the protected area is large and open or the seat of the fire is in a hidden area
such as a linen locker or adjacent unprotected room.

Research into slow smouldering fires , typical of those started by discarded cigarette ends in soft
furnishings or smouldering sawdust and other organic materials shows that smoke may not be given
off for many minutes, even several hours in certain situations, after ignition (See Figure 2). During
this time the insidious carbon monoxide gas can build up to a level sufficiently high so that, on
awakening, sleeping persons are too disoriented to evacuate the area.

When smoke is given off and has reached the detector it can frequently be too late to stop the rapid
spread of the fire – as has been seen in a number of high profile incidents in recent years.

        FIGURE 2 : Results from Fire 20 : Cloth Covered Fire retardant Polyurethane foam
        The fire is started 5.5 hours before flaming ignition.    The ceiling mounted CO fire
        detector operates 3.5 hours after ignition. The floor mounted CO fire detector operates
        4 hours after ignition. The first smoke detector alarms 4.5 hours after ignition by which
        time CO levels are in excess of 250ppm, severely affecting occupants ability to escape.

                                         Page 8 of 28
It is also well known that smoke escaping into corridors can cool and fall to the floor thus making them
impassable by the time the smoke reaches the detectors at the ceiling and generation and alarm condition
(Figure 3). Smoke can also be prevented from reaching the detectors by barriers of hot air building towards
the ceiling.

                 FIGURE 3: Measurements in escape corridor adjacent to a burning room (door

                 The smoke detector alarms 20 minutes after the CO fire detector when the
                 corridor is full of smoke

         Smoke Detector Test Fires

         Detector test fires, such as those used in EN54 part 9, provide a repeatable method of comparing
         the performance of smoke detectors. To this purpose, they are specifically designed to generate
         predictable levels and types of smoke that can be used in determining and classifying the
         performance of smoke detectors. They are not designed to generate and quantify other products of
         combustion and. Therefore, they do not, nor do they intend to represent real fires. Unlike a real fire,
         the seat of the fire, the position of the detector relative to the fire, and the air movement are known.
         In addition, some tests are designed to produce large quantities of smoke through pyrolysis prior to
         combustion taking place.

         These issues are not highlighted to invalidate the tests but more to illustrate that their function is to
         provide comparative tests of smoke detectors – something they do quite well, and to suggest that,
         where smoke is not the prime product of combustion being detected, alternatives to the test fires
         currently specified in the standards need to be found.

                                                 Page 9 of 28
As can be expected, CO fire detectors react well to some smoke detector tests and badly to others.
For example, a good response will be obtained in both EN54 TF2 and TF3 whilst TF4 and TF5 do
not produce sufficient CO gas, during the period of the test, to trigger an alarm. At the time of
writing, the performance of CO detectors in the Underwriters Laboratories (USA) test fires had not
yet been established, however, it has been found that CO detectors provide a response similar to
that of ionisation detectors in the fire test specified in the Australian standard. The ability of CO fire
detectors to pass smoke detector tests clearly depends on the nature of the test. This, by no means,
indicates that CO fire detectors are not a good indicator of real fires.

        FIGURE 4 : Ship Fire 3, Slow Smoulder, All Sensors Board 3

        During the period of the fire the smoke detectors gave no reaction

Real Fire Tests

In order to assess the capability of CO fire detectors to detect fires several years of fire tests were
carried out including a variety of fires in real situations using purpose built test rigs at the Moreton-in-
the-Marsh Fire Research Centre.

The first rig was designed to simulate a high roofed area such as an atrium or ship’s hold. Tests
carried out on this rig with a variety of fires showed that Co fire detectors located above the seat of
the fire reacted in a very similar time to smoke detectors. However, CO fire detectors mounted
further away reacted much quicker and were immune to obstructions such as when they are
recessed (Figure 4).

                                        Page 10 of 28
The second rig simulated a large open area such as a warehouse, manufacturing facility or other
similar large open area. Whilst all the different types of detectors mounted close to the seat of the
fire reacted in similar times, the CO detectors mounted at ceiling height react considerably faster and
were far less affected by thermal barriers.

A third rig simulated adjacent rooms. Waste paper bin fires showed CO fire detectors reacting
quicker than smoke detectors especially in rooms with no ventilation. Even CO detectors above the
ceiling tile reacted faster than some smoke detectors below the tiles (Figure 5).

        FIGURE 5: Relative rate of detection of a smouldering waste bin fire in an office
        The CO Fire Detector picks up the fire in half the time of the photo-electric smoke
        detector. Even CO fire detectors mounted behind the ceiling tiles reacted more quickly
        than the ionization detector.

CO Fire Detectors Resistance to Unwanted and False Alarms

There are many sources of false alarms for smoke detectors which CO fire detectors are inherently
immune to. Properly designed CO fire detectors are completely immune to false alarm sources such
as steam, dust, chemical aerosols, theatrical smoke and other airborne particles.

Even false alarm sources where carbon monoxide is produced are less prone to unwanted alarm as
the carbon monoxide distributes evenly through a space rather than concentrating in plumes to
provide a false alarm source as seen with smoke detectors. Whilst not completely immune, CO fire
detectors have been shown to be far more resilient than smoke detectors to false alarm sources
from smokers and overheating food.

Extensive research has also shown that an alarm threshold can be selected for CO detectors that,
whilst providing a very early warning of a potential fire, leaves the detectors insensitive to normally

                                        Page 11 of 28
expected sources of ambient CO such as cookers, heaters, running diesel engines and general


The CO Detector uses an electrochemical cell to detect the build up of carbon monoxide generated
by fires. The cell operates by oxidising carbon monoxide on a platinum sensing electrode. On a
corresponding counter electrode the other half of the reaction takes place. The sensing cell is
represented diagrammatically in Figure 6.

         FIGURE 6: Representational Diagram of CO Sensing Cell

When this reaction takes place the potential across the cell tries to change and this causes a current
to flow within the circuit around the cell. The current is mirrored into a current to voltage conversion
circuit the resulting output is directly proportional to the carbon monoxide concentration.

The cell itself has a diffusion limiting component to ensure that all carbon monoxide in the area
proximate to the sensing electrode is continuously oxidised. This means that the rate of transport of
carbon monoxide to the cell is directly proportional to the external concentration and independent of

Main Applications for CO Fire Detectors

CO fire detectors are particularly well suited to accommodation areas where there is a risk of slow
smouldering fires causing death through the build up of CO, limiting occupants ability to evacuate.

The flexibility of detector positioning relative to the seat of the fire make CO fire detectors ideal in
large open and complex areas.        These include: cargo holds, large atria and foyers, theatres,
cinemas, escape corridors, areas with complex and uneven ceilings and hidden storage cupboards.

                                       Page 12 of 28
        The resistance of CO fire detectors to false alarms makes them well suited to cabins and laundry
        rooms where steam and water mist can cause problems.              In galleys and mess rooms where
        scorched toast and similar causes of false alarms are prevalent, the alarm threshold will only be
        reached once the toast is actually burning.

        Like optical smoke detectors, CO fire detectors are far more environmentally friendly than ionisation
        detectors and can be manufactured to allow most of the detector to be recycled.


        Extensive research over the last ten years have shown that CO fire detection is a good general-
        purpose fire detector. In particular they are far more tolerant than smoke detectors to the location of
        the start of the fire and are far superior to smoke detection in the early detection of smouldering fires
        in accommodation areas. CO fire detectors have also proved to be particularly resilient to many
        common smoke detector false alarm sources.

        Like other fire detection technologies, CO fire detection does have some limitations that must be
        taken into account when selecting suitable applications. Using a combination of Co fire detectors
        and heat/optical algorithmic smoke detectors, such as the HPO detectors, across the range of fire
        risks provides a comprehensive life and property protection solution.          Applied correctly these
        detectors enable rapid detection, elimination of false alarms, reduction in unwanted alarms and
        complete elimination of the need for radioactive ionisation detectors.

Operator Interface

Advances in LCD technology and processing power means that it is possible for the Operator Interface to
present the operator with a considerable amount of information. Some systems on the market offer a
number of features. Some of the features are discussed below.

Temperature and Smoke Density Information

This allows the operator to identify the seat of any “hot spots” and to track this over time. The smoke
obscuration % allows the operator to make a judgement on the amount of smoke in an area. Together they
could give him an idea of the intensity of the fire.

Risk Management Information

                                                 Page 13 of 28
This may allow the operator to pre-programme information on risks in the areas of the incident (Chemical
storage) while giving details on position of fire fighting appliances and closest exits. Pre-defined instructions
can also be stored linked to various areas (Figure 7).
                         FIGURE 7: Operator Display Module

Group Detection

Where there are a number of detectors in the same open space, the signals from all the various detectors
could be analysed via the knowledge based expert system to identify fires earlier while reducing false

Extensive Cause and Effect Programming

As part of a fire fighting strategy it may be required to shut down damper, activate fans, shut down other
machinery. This can all be done via the addressable loops but controlled through the Cause and Effect

Dirtiness Levels in Detector

The Control Panel will identify when a detector need to be replaced. While at any time the operator can
check on how dirty the detector is.

Self-Verification Functions

SOLAS required operators to regularly test the system. By including inbuilt testing mechanisms in the
detectors it is now possible to test devices from the sensing chamber all the way back to the control panel.
They allow daily/weekly tests to take place at the touch of a button or event better pre-programmed with a
print-out of any devices which have failed the test.

                                               Page 14 of 28
Remote Diagnostics

With the pressure on reducing manning levels, it is becoming obvious that it would be advantageous if fault-
finding could be carried out remotely. It is now possible for the manufacturer to access the fire detection
system remotely from a service centre, so they can assist the ship’s staff in identifying and resolving
problems and assisting with maintenance.

Isolation by Zone/Type of Device

For maintenance or while loading on car decks, it is possible to isolate only smoke sensor parts leaving the
heat part of the sensor and callpoints working irrespective of how they are wired in the field.

Fire Detection Algorithms

The term “algorithms” has its origin in the Greek language and means “calculation process, running
according to a specific pattern”. In the broadest sense, every software program can also be said to involve
an algorithm. Here also, a process is performed according to a fixed, pre-determined pattern.

A new method used by the author’s company uses a multi-parameter, multi-criteria algorithm incorporating
fuzzy logic. A diagram of the evaluation method is given below.

            Signal preprocessing                         Signal evaluation

                   Fluctuation           Fuzzification

                   Gradient              Fuzzification
   Meas. Sig.                                                                                      1. Idle
                                                              Expert             Defuzzification
                                                                                                   2. Fault
                                                                                                   3. Alarm
                   Fire duration         Fuzzification

                   Quiesc.               Fuzzification

                          FIGURE 8: Algorithm using Fuzzy Logic

Fluctuation, gradient and fire duration characteristics are derived from the measurement over time with the
aid of specially developed algorithms.

Studies of large data samples of both fire and non fire situations have revealed that a marked fluctuation
normally indicates a fault situation and rarely indicates fire. The gradient concept provides information on
both the graduated duration of a signal rise. It was evident that extremely sharp signal rises can indicate a

                                                  Page 15 of 28
false alarm. Fire duration indicates how long a situation has been identified as an emergency.          Fires
normally develop relatively slowly, rather than in an explosive manner. Important inferences can be made on
the basis of the information which indicates how long a situation has corresponded to a fire.

The quiescent values can be used to obtain useful information from the signal pattern. A typical use would
be, for example, to indicate the degree of contamination of an individual detector.

                                                                              4 & 5
The evaluation process is now implemented on the basis of fuzzy logic             . Generally speaking, three
procedural steps can be distinguished in fuzzy logic:

1.      Fuzzification means that a “defined” value is converted into a “soft”, “fuzzy” value. The fuzzy value
        is not a unit and can therefore be processed in the knowledge base.

2.      In the knowledge base, the “soft” or fuzzified values are processed according to defined rules and
        an output variable is calculated from them.

3.      The defuzzification process converts the still fuzzy result of the output variable into a defined,
        technically usable parameter, which in fire detection terms is quiescent, fault, alarm.

For the system to work effectively, the knowledge base should be representative of fire and false alarm
situations. The knowledge base used consists of thousands of fire and false alarm situations and research
experiments collected over a period of 80 years.

The use of fuzzy logic in simplistic terms means that alarm thresholds can move quickly both upward and
downward resulting in earlier detection with less false alarms.

Loop Communication Protocol

Over the years we have all noticed that the way equipment can be affected by external electro-magnetic
radiation and radio frequency interference.        This is especially true on vessels with various types of
electrical/electronic machinery and controls. The use of a true digital protocol rather than an analogue
protocol helps to eliminate the effect of this interference on loop communication.         The communication
method is Frequency Shift Keying (FSK) with sinusodial waveform. The main differentiation between a logic
0 and a logic 1 is determined by the width of the sine wave. There are two important aspects to the signal,
these are the point where the frequency changes occur and the period.

                                               Page 16 of 28
                             ½ period of low frequency

                                                         Start bit
                                                                                   8 Data bits   Stop

Frequency changes
occur at ‘troughs’

                     Beginning of start

                     bit        FIGURE 9: Digital Protocol

The diagram in Figure 9 shows the general structure of one such digital protocol signal and its discriminated

The use of two frequencies of around 6K and around 3K means that it is much easier to eliminate both high
frequency and low frequency interference using intelligent discrimination circuits incorporating noise
reduction filters.

Signal edges do not get distorted due to cable capacitance/inductance.                           More common square waves
become rounded at the edges.

The rounded shape of the signal results in negligible radiated emissions. Square wave signals produce
much higher emissions due to the sharp rise and fall times of the signal edges.

All this means that a wide range of cable types including unscreened cable can be used.

Distributed Temperature Sensing using Fibre Optics

Conventional temperature sensing, using discrete sensors such as thermocouples or platinum resistance
thermometers, provides data at a single point, which may be interpreted as an average reading over a
localised area.

A more elegant solution is to use a distributed temperature sensor which is intrinsically multiplexed, allowing
many hundreds or even thousands of points to be monitored with a single sensor.

Optical Time Domain Reflectometry

                                                                                 Page 17 of 28
A distributed temperature sensing solution is now available based on the principles of Optical Time Domain
Reflectometry (OTDR).


A laser source launches a pulse of light into an optical fibre. As the pulse travels down the fibre, energy is
lost through scattering. A fraction of the scattered signal is retained within the fibre. A portion of this is
directed back along the fibre towards the laser source – this signal is called backscatter.

The backscatter signal is split off by a directional coupler, which is then optically filtered and presented to a

This technique has been traditionally used within the telecommunications industry to check the integrity of
the optical fibres.

                         FIGURE 10: Principle of Backscatter Measurement

Scattering is due to:-

                                 Variations in density and composition of the medium (Rayleigh scattering)
                                 Bulk vibrations (Brillouin scattering)
                                 Molecular vibrations (Raman scattering)

                                               Page 18 of 28
                         FIGURE 11: Backscatter Spectrum

Rayleigh Scattering

Rayleigh scattering, although the strongest component of the scattered light spectrum, is only very weakly
sensitive to temperature and is therefore not able to be used for temperature sensing.

Brillouin Scattering

Brillouin scattering is temperature sensitive and produces a relatively strong signal.       Unfortunately, the
Brillouin and Rayleigh signals are relatively close to each other in the frequency spectrum, so that detecting
the Brillouin component is very difficult and requires special sources and filters. This too, is therefore not
suitable for temperature sensing.

Raman Scattering

Like the Brillouin scattering, Raman scattering is also temperature sensitive and of sufficient intensity to be
suitable for temperature sensing. The signal is split into two ‘bands’ (Stokes and Anti-Stokes) displaced,
roughly symmetrically, about the incident wavelength.

The displacement of the Raman bands is sufficiently large to enable the signals to be separated by means of
a filter from the other backscatter components and detected with an Avalanche Photodiode Detector (APD).
The band of longer wavelengths, (Stokes), is only weakly temperature sensitive but the band of shorter
wavelengths, (Anti-Stokes), exhibits a distinct sensitivity to temperature.

Signal Types

Stokes Band – The Stokes wavelengths are used to define the non-temperature sensitive (NTS) signal. This
is also known as the reference signal. It is used to analyse the integrity of the fibre. This includes measuring
the fibre losses and detecting breaks in the fibre.

Anti-Stokes Band – The Anti-Stokes band forms the temperature sensitive (TTS) signal. It is the main
component used in the temperature computation.           Knowing the speed of propagation at the various

                                                Page 19 of 28
wavelengths, it is possible to plot each of the signals on a graph showing backscatter power versus

                          FIGURE 12: Anti-Stokes Backscatter Power v Temperature Graph

The Optical Fibre Sensor

Optical fibre itself offers several important advantages as a sensing medium. The signals are immune to
electromagnetic interference thereby ensuring integrity of readings from electrically noisy areas, for example
around power cables and transformers. As no electric current is used in the sensing fibre and the fibre is a
relatively inert and dielectric (non-conducting) medium, it is safe technology to use in Zone 2 hazardous

The sensor element within the system is a communications grade optical fibre of the 62.5/125 graded index
multimode type.

The temperature range is predominantly a function of the coating used to protect the optical fibre as the fibre
itself is well behaved over a temperature range from below –50°C to approximately 300°C. Coatings have
been tested down to –190°C (acrylate) and up to 460°C (metallic).

The System measurement accuracy suffers below –50°C due to non-linear effects in the opto-electronics
within the system. Also at the lower cryogenic temperatures, care must be taken that the coating does not
exert mechanical stress on the fibre, thereby affecting its durability or causing micro bending, which would
increase its loss.

The factor affecting the distance at which the system can make a temperature measurement is the
attenuation of the optical fibre sensor (ie. rate at which power is lost).

                                                 Page 20 of 28
The System can continue to operate in the event of a fibre break by exploiting the signal processing
techniques outlined. Provided that the “break-detect” option has been selected and the system has been set
up in a loop configuration, the System will automatically invoke single ended processing from both ends of
the loop when the fibre break is detected. The System can reconstitute the temperature profile of the entire
fibre length regardless of the position of the break.        Depending on the nature of the break a few
measurement points in the immediate vicinity of the break may be lost. In the case of multiple breaks, the
length accessible to the System will continue to be measured.

Temperature resolution overcomes the uncertainty in the temperature information resulting from inherent
noise in the opto-electronics. Thus the temperature measured at a given point in the fibre may vary between
successive measurements.

The accuracy of the measurement is determined by a range of factors: these factors include the linearity and
reproducibility of the electronics, the robustness of the signal processing, and the calibration of the system to
the fibre to be measured.

The overall result is a complete system that can be programmed to operate as a fixed temperature or rate-
of-rise temperature detector, with accurate temperature information from many discrete points along the
entire fibre length.

A complete temperature profile of the fibre sensor cable is thus obtained. The System configuration is as in
Figure 13. This temperature profile information is typically available from a RS232 port on the terminal
equipment. The temperature profile information can be viewed via special PC based applications. Once the
information is input to a PC, many other features are immediately available. These include remote
diagnostics and system interrogation via a modem connection. The information can also be integrated into a
fire detection graphics package.

FIGURE 13: System Components

                                               Page 21 of 28
The systems can be programmed into multiple fire detection zones. These fire detection zones are of
programmable length and position along the fibre sensor. Each fire detection zone can have multiple
(programmable) fire detection trip levels. The temperature trip levels can be set as a function of fixed and/or
rate-of-rise in temperature. The systems can thus produce early warning (pre-alarm) signals, and alert
attention to a potential fire alarm condition prior to an executive action being taken.

The optical fibre temperature sensing system has wide ranging applications especially where:-

a)      Maintenance is very difficult and a “fit and forget” solution is required
        •   Drilling Areas on platforms and drill rigs
        •   Cable Tunnels and Trays
        •   Cargo Holds
        •   Conveyor belts on self-unloading vessels

b)      Small changes in temperature need to be detected
        •   Pipe leakages
        •   Overheating of sensitive equipment
        •   Magazine Areas on Warships

c)      The above needs to be carried out in a hazardous environment.

Closed Circuit Smoke and Flame Detection System

As with most point fire detectors, location of the detector is important. In areas where we have large open
spaces a mixture of smoke and flame detectors are used. In cathedral type machinery spaces, atria and
open spaces on fighting ships, it is important that an alarm is raised as quickly as possible preferably the
moment smoke is produced. Closed Circuit Smoke and Flame Detection have this advantage. The system
does not require the operator to be constantly watching the monitors as it annunciates an alarm

The system functions by comparing one frame with the next, so that any change can be evaluated.
Compound Obscuration evaluates the total attenuation of light from the camera to the furthest point in the
field of view. The algorithm is able to de-couple smoke quantity from smoke density i.e. large clouds of thin
smoke can be identified as well as small areas of dense smoke.

The 64,000 pixels that make up the full screen are evaluated every second.                Any change in the light
attenuation will cause the system to alarm depending on the thresholds set.

                                                Page 22 of 28
A schematic of a system is shown in Figure 14.


                                             Keyboard                     Mouse

                                                 Video Fire Detection Hub                 Alarm

                                                                        Camera           CCTV
                                                                       Multiplexer       Monitor

                            Risk Area             Equipment Area                     Response Area

                        FIGURE 14: Schematic Video Smoke and Flame
The system uses standard CCTV Cameras and takes a feed off the cameras before the multiplexer. The
signals are evaluated using the algorithms in the Video Fire Detection Hub and are presented at the Alarm
Annunciator. The detection system can then be configured to alarm if a user selectable number of pixels are
obscured to a defined density of smoke. Because the system uses standard CCTV equipment it is eminently
suitable for being added to existing CCTV systems onboard ships. As all the evaluation is done in software
and is programmable, the system can also be used to detect visible oil mist, high pressure oil leakage from
pipes, and steam leaks the moment they occur. When used in exposed areas, there is a danger that false
alarms could be generated because of fog and this will need to be addressed.

Flame Detection

On offshore installations and onboard ships there has been a tendency to move away from Ultra Violet (UV)
flame detectors and use triple wave-band infra red flame detectors instead. The reason for this is more
easily explained by Figure 15.

                                             Page 23 of 28
                         FIGURE 15: Comparison of Ultra Violet and Infra Red Flame Detectors

UV flame detectors are very sensitive to arc-welding, electrical arcs, x-rays and lighting. Although it is
possible to eliminate false alarms from lighting and electrical arcs by the inclusion of time delay processing
the elimination of false alarms from arc welding and x-rays is much more difficult to achieve. The detectors
sensitivity to these false alarm sources can be a significant problem. There are external influences, whose
presence can have a detrimental effect on the ability of the detector to see flame radiation. The main
inhibitors of UV propagation are oil mists or films, heavy smoke or hydro carbon vapour and water films.
These phenomenon are present in machinery spaces and on offshore platforms and can significantly reduce
the intensity of the UV signal if present in the flame detection path.

The shortcoming of UV detectors for offshore and machinery space applications has resulted in operators
preferring the Triple Wavelength Infra Red Flame Detectors.

The use of Triple Wavelength Infra-Red Detection principles has meant that the main shortcoming of Infra
Red Flame Detectors, namely response to solar radiation and black body radiation had been overcome.

Triple Wavelength Infra Red Flame Detectors

When organic material burns, (refer to Figure 16) large amounts of hot Carbon Dioxide are produced and
thus the burning process emits infra red radiation at 4.3 microns. On the other hand atmospheric adsorption
provides high IR absorption at 4.3 microns. This atmospheric adsorption band significantly reduces solar
radiation reaching the Earth’s surfaces at 4.3 microns. It also has the effect of limiting the effective range of
an Infra red detector, since the atmospheric absorption will affect the radiation emitted from a carbonaceous

                                                Page 24 of 28
Some infra red detectors are tuned to respond to the H2O emission band by using a single infra red channel
tuned to 2.9 microns.    Doing so can enable these sensors to detect fire from non-hydrocarbon based

                         FIGURE 16: Typical Hydrocarbon Fire Spectrum

Infra red detectors generally incorporate lens/filter arrangements, with a band pass filter operating at 4.3
microns making the IR detector blind to solar radiation.       More sophisticated filtering techniques and/or
selected electronic signal processing is used in the design of the latest, multi-channel infra red flame
detectors to reduce false alarms further.

The detectors monitor the infra red spectrum at three chosen frequencies. One sensor monitors the CO2
emission bands at 4.3µm. The other two frequencies are used to monitor the background infra red level.
They are normally chosen at frequencies on either side of the CO2 emission band. The main objective of
using the two background frequencies on either side of the emission band is to allow the detector to more
accurately predict the amount of black body radiation present in the field of view.

The detector can account for the differences in hot and cold blackbody radiation present – a function that
cannot be accurately predicted by dual channel IR detectors. The detector can detect fires in the presence
of black body radiation. This can vary significantly depending on the design of the detector. In particular
some detectors may be less sensitive to genuine fire conditions than others particularly in the presence of
black body radiation from a cold black body. Using signal-processing techniques, the three signals are
correlated and the device decides if a true alarm condition is present. Typical parameters used in these
detectors are:

        1.       Ratio of the reference sensors to the CO2 emission band sensor.

                                               Page 25 of 28
        2.      Correlation between the sensors.
        3.      The relative amplitude of received signal from each sensor.
        4.      The flicker frequency of each sensor.

Elimination of Nuisance Alarms from Modulated Black Body Sources

The design incorporates an optical filter which enables a single electronic infra red sensor to measure the
radiated energy present in two separate wavebands placed on either side of the flame detection waveband,
at 3.8 µm and 4.8 µm respectively (See Figure 14). The signal obtained from these ‘guard’ channels is
cross-correlated with the signal from the flame detection channel to provide an accurate prediction of the
non-flame energy present in the flame detection waveband.              This prediction is independent of the
temperature of the radiation source, allowing the S200 Plus to provide blackbody rejection over a wide range
of source temperatures.

                          FIGURE 17: Blackbody Rejection

Figure 17 shows the amount of energy given by a ‘hot’ object [blackbody] as viewed in the electromagnetic
spectrum. This curve has a peak, which moves further to the left with higher temperature objects. The
amount of energy seen between 3.8µm and 4.8µm can be approximated to a linear function. Thus, a
measurement of the energy at these two wavelengths will provide sufficient information to calculate the level
of blackbody radiation at the intermediate flame detection frequency of 4.3µm. The energy due to the
emission from hot carbon dioxide given by a flame, is superimposed on that from any blackbody in the
detector field of view without adding any significant emissions at 3.8µm or 4.8µm. This enables proper
segregation between non-flame signals and flame signals. As a large fire will possibly produce a large
amount of black smoke which will behave like a blackbody and may weaken the carbon dioxide peak, signals
greater than a pre-determined upper limit will be classed as a fire.

                                               Page 26 of 28
The use of an optical processing technique as opposed to the use of two separate electronic sensors
improves the overall reliability of the detector by reducing the number of components and eliminating the
need for complex calibration procedures during manufacture.

Detection of Flame in the presence of Blackbody Radiation

The ability of the detector to determine accurately the amount of non-flame radiation received at any one
time by the flame detection channel allows a variable alarm threshold to be determined (See Figure 18).
This threshold is positioned so as to minimise the possibility of a false alarm due to the presence of
modulated blackbody sources of different temperature and intensity.

                        FIGURE 18: Alarm/Blackbody Signals


The technologies and principles discussed allow the operator more flexibility in making decisions on how to
protect his Vessel.   They are by no means the only developments taking place, but I believe the
development discussed will play a major part in protecting Vessels in the future. They allow the operator to
design fire detection systems based on risk management scenarios which are more in keeping with
protecting the sophisticated vessels of tomorrow.


        P.C. Cooke and D.J. Stone: “Analogue Addressable Fire Detection Systems in
        Marine Applications”, Read on 19 March 1991 – Institute of Marine Engineers, London.

        HARWOOD, J.A. MOSELEY, P.T. PEAT, R and REYNOLDS, C.A.: “The Use of

                                              Page 27 of 28
        Low Power Carbon Monoxide Sensors to Provide Early Warning of Fire”, A Fire Journal

        KIRK, HUNTER, BAEK, LESTER and PERRY,: “Environmental Tobacco Smoke in
        Indoor Air”, Ambient Air Quality Conference Proceedings, (London 1988).

        HEKESTAD, G and NEWMAN, J.S.: “Fire Detection Using Cross-correlations of Sensor Signals”,
        Fire Safety Journal, 18, P335-374, (1992).

        Hans-Jηrgen Zimmermann: Fuzzy Technologien [Fuzzy Technologies]. VDI-Verlag,
        Dηsseldorf 1993.

        Jrg Kahlert, Hubert Frank: Fuzzy-Logik und Fuzzy-Control [Fuzzy Logic and Fuzzy
        Control]. Vieweg-Verlag, Braunschweig/Wiesbaden, 1994.


The author would like to acknowledge the contribution to the paper by his colleagues in the Research and
Development Department of Thorn Security Limited. They were instrumental in providing the information
and comment to allow this paper to be written.

                                             Page 28 of 28

Shared By:
Tags: technical