Research supporting the development of a video fire detector test standard Raman Chagger Principal Consultant, Fire Safety Group, BRE FIREX, 20 th June 2018 Part of the BRE Trust
Background • Video fire detectors use cameras to monitor a space and analyse signals to detect the presence of smoke or flame. • Complex algorithms and very different methods to identify the smoke or flame signature of a fire. • Currently no European standard for this technology. • BS 5839-1:2017 states video fire detectors : • … Video fire detectors can also be used as a means of verifying a fire. Where they are the sole means of detection, they need to be designed, and their suitability and performance verified, by a qualified specialist. • … performance of video fire detectors and reliance is placed on manufacturers’ recommendations. • … tend to be used as supplemental detection or for special • applications where other detection techniques are inappropriate or ineffective. • No plans for standard development at CEN level
Background – VFD work • In 2012 the Fire Industry Association (FIA) formed the Video Fire Detectors task group • Aim was to support the development of the ISO 7240-29 test standard • Need to develop for Video Flame Detectors (VFD) and Video Smoke Detectors (VSD): • benchmark tests • operational performance tests • March 2014 a Video Fire Detector FIA/BRE research group • All known VSD and VFD manufacturers were approached with requested to contribute to the research A D Holdings
Aims of the research group To develop: • reproducibility measurement methodology for Phase 1 • VFD • VSD • Operational performance methodology for • Phase 2 VFD • VSD - Limited to flames in the visible spectrum - Where possible to use EN 54 test methods - Existing standards were reviewed but did not provide sufficient guidance on how to perform these tests
Video Flame Detection- bench tests Modified EN 54-10 bench test methodology was used Two systems were tested and would not respond to the same test method. The following were developed: 1) a looped video of flaming Bunsen burner 2) a constant un-flickering flame
Video Flame Detection- bench test results System 1 (steady flame) - Reproducibility Specimen No. Response point (mm) D max & D min Ratio 1 D max : D mean Ratio 2 D mean : D min 1 2050 2 2200 D max 3 2100 1.11 1.16 4 1700 D min 5 2100 6 1700 Requirements of EN 54-10:2002: D max : D mean 1.15, D mean : D min 1.22 System 2 (flame video) - Reproducibility Specimen No. Response point (mm) D max & D min Ratio 1 D max : D mean Ratio 2 D mean : D min 1 2500 2 2300 3 2200 D min 4 2550 D max 1.09 1.06 5 2250 6 2250 7 2250 8 2400 Requirements of EN 54-10:2002: D max : D mean 1.15, D mean : D min 1.22 ☺ 1_4
Video Smoke Detection- bench tests The metric root mean squared error (RMSE) deviation was identified that enabled smoke change to be “measured” One pixel of four from white to black RMSE = 25% All four pixels from white to 10% black RMSE = 10% Thousand of pixels RMSE ~0.02%
Video Smoke Detection- bench tests - EN 54-7 fire tests were used to produce a video of smoke growth - External conditions fully controlled - Video of the fire was played back on a monitor display at a fixed distance from the camera.
Video Smoke Detection- bench test results System 1- Reproducibility Specimen No. Response time (min:sec) RMSE (%) RMSE Max/Min Max:Mean Mean:Min 0.0278 1 6:36 0.0284 2 6:40 0.0278 3 6:36 1.045 1.031 0.0271 4 6:31 0.0292 MIN 5 6:44 0.0275 MAX 6 6:34 Requirements of EN 54-7:2001: Max:Mean ≤ 1.33 and Mean:Min ≤ 1.5 System 2- Reproducibility Specimen No. Response time (min:sec) RMSE (%) RMSE Max/ Min Max:Mean Mean:Min 1 7:46 0.0439 MIN 2 7:46 0.0439 3 7:50 0.0440 1.070 1.019 4 7:46 0.0439 5 7:46 0.0439 6 7:59 0.0479 MAX Requirements of EN 54-7:2001: Max:Mean ≤ 1.33 and Mean:Min ≤ 1.5 ☺ 2_4
Video Flame Detection- performance tests EN 54-10 fire test methodology was used as a basis - 2 test fires (n-heptane and methylated spirits) performed in trays at distances up to 100m - Time of response measured from start of fire
Video Flame Detection- performance tests results 100 90 Time to response (sec) 80 70 N-heptane Day 1 VFD 1 60 Methylated spirits Day 1 VFD 1 50 N-heptane Day 2 VFD 1 40 Methylated spirits Day 2 VFD 1 30 Upper limit 20 Lower limit 10 0 0 20 40 60 80 100 Separation distance between VFDs and fire (m) - Point-type flame detectors are required to alarm within 30 seconds of exposure to the fire. - Appears to be a repeatable response up to 30m. - Proposed to use methodology and requirements of EN 54-10 - 8 samples are tested at claimed maximum distance to respond before 30 seconds for both test fires. ☺ 3_4
Video Smoke Detection- performance tests (Trial 1) To use “Illuminated smoke” in a dark ambient environment Aim to: • Control the ambient environment • Produce smoke only with a background to contrast against • VSDs saturated by the light • Ambient levels too low
Video Smoke Detection- performance tests (Trial 2) • To perform “in principle” tests at the BRE fire test room • Used black and white tarpaulin screens to contrast the smoke from the fire • TF2 to TF5 from EN 54-7 were performed with reduced fuel loads • Videos cropped and analysed using the BRE VSD analysis tool.
Video Smoke Detection- performance tests (Trial 2) • RMSE growth profiles observed for the 4 test fires • Identified that light from the fire reflected on the tarpaulin screen contributed to the RMSE analysis for the TF4 and TF5 fires • Identified that illumination of the screen would have to be indirect
Video Smoke Detection- performance tests (Trial 3) • To perform tests on full scale at BRE Middlesbrough (at night!) • Remove direct flames and reflections from screen
Video Smoke Detection- performance tests (Trial 3) • Reflected light observed on the tarpaulin screens from the fire • Direct light from the fire
Video Smoke Detection- performance tests (Trial 4) • Use of “chimney cap” to prevent direct or indirect light from flames
Video Smoke Detection- performance tests (Trial 4)
Video Smoke Detection- performance tests (Phase 4) Whilst not obvious to the human eye the tarpaulin screen was moving due to air movement in the space… 0.030 0.025 0.020 RMSE (%) 0.015 Original data 10 second average 0.010 0.005 0.000 0 20 40 60 80 100 120 140 160 180 200 Time (sec)
Video Smoke Detection- performance tests (Phase 4) Smouldering fires on white Flaming fires on black Smouldering fires- lacked repeatability Flaming fires- TF4 fire much more repeatable.
Video Smoke Detection- performance tests • The sets of fire tests demonstrate, in principle, the viability of the test method. • Succeeded in: • Illuminating the screen without producing large reflections • Removing light from the fire from reaching VSDs • Removed reflections on the screen • Illuminating the smoke • VSDs responded to the smoke from the fires • Hurdles: • Need to develop a means for producing a stable background • Need to demonstrate repeatable RMSE profiles for TF2-5 fires 4_4 X More work required…
Video Smoke Detection- Phase 3 • Extension of research (additional phase 3) • Completed 15 June 2018 • Utilised a fixed installation of plasterboard sheets that was: • Painted white for flaming fires (4xTF4 and 4xTF5 fires) • Painted black for smouldering fires (4xTF2 and 4xTF3 fires) • Results from the fires are currently being analysed.
Conclusion • Research has successfully achieved test methods that can be utilised in a video fire test standard. • Benefits of collaboration • BRE and the FIA are now pursuing the development of a Loss Prevention standard and associated CoP. • Briefing paper is currently being prepared (Sept/Oct 2018). • Interested in further information? https://www.bre.co.uk/preference- centre/register.jsp https://www.fia.uk.com/newslettersubscr iption.html
Thanks A D Holdings
Recommend
More recommend
