Sie sind auf Seite 1von 73

Validation of a Smoke Detection Performance Prediction Methodology Volume 3.

Evaluation of Smoke Detector Performance

Prepared by:
James A. Milke and Frederick W. Mowrer University of Maryland Pravinray Gandhi Underwriters Laboratories Inc.

October 2008 Fire Protection Research Foundation

FOREWORD
This report presents the results of a Foundation project whose goal was to develop a validated engineering methodology to calculate and accurately predict the response time of spot-type and aspirated smoke detection systems exposed to incipient fires and growing fires. The report, divided into four volumes, describes the test methods, test results, computer simulations and analyses used for this project, which addresses the validation of a smoke detection performance prediction methodology.

The four volumes of this report include: Volume 1, which addresses the characterization of the heat and smoke release rates of eight incipient fire sources selected for this project; Volume 2, which addresses the large-scale room fire tests conducted as part of this project; Volume 3, which addresses evaluation of smoke detector performance in the large-scale room fire tests conducted as part of this project; Volume 4, which addresses comparisons of current FDS smoke detection prediction methodologies with actual smoke detector performance in the largescale room fire tests. The Research Foundation expresses gratitude to the project sponsors and technical panelists listed on the following page. The content, opinions and conclusions contained in this report are solely those of the authors.

Validation of a Smoke Detection Performance Prediction Methodology Research Project


Technical Panel
Shane Clary, Bay Alarm Company Kenneth Dungan, PLC Foundation Jay Ierardi, R.W. Sullivan Engineering Kevin McGrattan, National Institute of Standards and Technology Dan Nichols, NYS Code Enforcement and Administration Ali Rangwala, Worcester Polytechnic Institute Joseph Su, National Research Council of Canada

Principal Sponsors
Honeywell Life Safety National Electrical Manufacturers Association Siemens Building Technologies, Inc. SimplexGrinnell

Contributing Sponsors
Automatic Fire Alarm Association Bosch Security Systems Xtralis, Inc.

Validation of a Smoke Detection Performance Prediction Methodology Volume 3. Evaluation of Smoke Detector Performance

Prepared for: Kathleen Almand Fire Protection Research Foundation 1 Batterymarch Park Quincy, MA 02169

Prepared by: James A. Milke and Frederick W. Mowrer University of Maryland Pravinray Gandhi Underwriters Laboratories, Inc.

October 10, 2008

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. ii

Executive Summary This report, divided into four volumes, describes the test methods, test results, computer simulations and analyses used for this project, which addresses the validation of a smoke detection performance prediction methodology. This project was conducted jointly by the University of Maryland (UM) and Underwriters Laboratories, Inc., (UL) under the auspices of the Fire Protection Research Foundation (FPRF). The financial and technical support of the FPRF, the project sponsors and the project technical panel are gratefully acknowledged. The four volumes of this report include: Volume 1, which addresses the characterization of the heat and smoke release rates of eight incipient fire sources selected for this project; Volume 2, which addresses the large-scale room fire tests conducted as part of this project; Volume 3, which addresses evaluation of smoke detector performance in the large-scale room fire tests conducted as part of this project; Volume 4, which addresses comparisons of current FDS smoke detection prediction methodologies with actual smoke detector performance in the large-scale room fire tests. The overall objective of this project has been to evaluate the capabilities of the current release version (5.1.0) of the Fire Dynamics Simulator (FDS) to predict smoke detector activation in response to relatively low energy incipient fire sources. The project was subdivided into four tasks, consistent with the four volumes included in this report. The first task was to characterize the heat and smoke release rates of eight incipient fire sources selected for this project. The incipient fire sources are described in Table E1; the fire sources include four flaming fire sources and four smoldering/pyrolyzing fire sources. The heat and smoke release rates of these incipient fire sources were measured in the same IMO intermediate scale calorimeter that UL used previously as part of its FPRF-sponsored smoke characterization project [Fabian, et al., 2007]. Three replicate tests were conducted for each of the eight incipient fire sources to provide a measure of the repeatability of these tests. Volume 1 of this report provides descriptions of the incipient fire source fuels and ignition sources, the fire test apparatus and instrumentation used for this task, and the results of these tests. Volume 1 also addresses FDS simulations of these tests conducted in the IMO calorimeter as a means to evaluate the predictive capabilities of the FDS model on a preliminary basis. These FDS simulations were not originally planned, but have proven valuable in troubleshooting issues related to the simulation of fires involving these incipient sources. They provide an indication of the uncertainty in simulating the fire source terms in FDS. The second task was to perform large-scale room fire tests using the eight incipient fire sources characterized in Task 1. The large-scale room fire tests were conducted in two rooms at the UL facility in Northbrook, IL. The first set of large-scale tests was conducted under unventilated conditions in the standard room used to test smoke detectors for the UL 217/268 standards; this room measures 10.8 m (36 ft.) long by 6.6 m (22 ft.) wide by 3.0 m (10 ft.) tall. Three replicate

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. iii

tests were conducted with each of the eight incipient fire sources, for a total of 24 unventilated room fire tests. The second set of large-scale tests was conducted in a 7.2 m (24 ft) long by 7.2 m (24 ft) wide by 3.0 m (10 ft) high room constructed specifically for this project to represent a mechanically ventilated commercial space. This room was provided with mechanical injection ventilation and a ceiling return air plenum to represent a typical commercial type of installation. Three replicate tests were conducted with each of the eight incipient fire sources at nominal mechanical ventilation rates of 6 and 12 air changes per hour; two replicate tests were conducted with each of the incipient fire sources under unventilated conditions in this room. Thus, 64 fire tests were conducted in the ventilated room, for a total of 88 large-scale fire tests in the two rooms. A matrix showing the designations of the 88 large-scale tests is provided in Table E2. Table E1. Incipient fire sources Fuel source Shredded office paper Flexible PU foam / microfiber fabric Flexible PU foam / microfiber fabric Ponderosa pine Cotton linen fabric PVC wire Computer case Printed circuit board Ignition source Small flame (50 W) Small flame (50 W) Hotplate Hotplate Hotplate Electric overcurrent Small flame (UL 94) Small flame (ATIS T1.319) Fire type Flaming Flaming Smoldering/pyrolysis Smoldering/pyrolysis Smoldering/pyrolysis Smoldering/pyrolysis Flaming Flaming

Table E2. Matrix of large-scale room fire test designations Incipient fire source Unventilated room 1, 2, 3 4, 5, 6 Ventilated room 12 ach 49, 50, 51 52, 53, 54

Shredded office paper Flaming PU foam / microfiber fabric Smoldering PU foam / 7, 8, 9 31, 32, 33 55, 56, 57 microfiber fabric Ponderosa pine 10, 11, 12 34, 35, 36 58, 59, 60 Cotton linen fabric 13, 14, 15 37, 38, 39 61, 62, 63 PVC wire 16, 17, 18 40, 41, 42 64, 65, 66 Computer case 19, 20, 21 43, 44, 45 67, 68, 69 Printed circuit board 22, 23, 24 46, 47, 48 70, 71, 72 ach = nominal mechanical injection ventilation rate in air changes per hour

6 ach 25, 26, 27 28, 29, 30

0 ach 73, 74 75, 76 77, 78 79, 80 81, 82 83, 84 85, 86 87, 88

The large-scale rooms were instrumented with a number of thermocouples, velocity probes and light obscuration measurement devices to provide a basis for evaluating the current capability of FDS to predict fire-induced conditions throughout a room in response to incipient fire sources.

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. iv

The rooms were both equipped with a number of spot-type commercial smoke detectors from two manufacturers. The ventilated test room was also equipped with three aspirated smoke detection systems from one manufacturer. The response of these different smoke detection devices during these tests provides a basis for evaluating the current capability of FDS to predict smoke detector activation in response to incipient fire sources. Volume 2 of this report describes the details of the large-scale room fire tests and provides the instrumentation and detection data from these 88 fire tests. More than 1,200 graphs have been developed to illustrate the results of these 88 tests; these graphs are too voluminous to print, so they are provided on electronic media in Excel files associated with each test. This large-scale room fire test data set should prove useful for future smoke transport and smoke detection validation exercises as well as for this one. The third task was to evaluate smoke detector performance during the large-scale fire tests. For this task, the response of the spot-type and aspirated smoke detectors during the fire tests was evaluated and characterized. These results were then compared with methodologies available in the fire safety literature for predicting the activation of smoke detectors. Volume 3 of this report describes the details of these comparisons. One objective of this project has been to develop the means, based on experimental data, to estimate the response of smoke detectors using the simulated results of the smoke conditions computed by FDS. Smoke conditions estimated by FDS throughout the domain include temperature, velocity and mass fraction of smoke (which can be related to light obscuration or visibility). One of the relatively unique aspects of this study is an examination of the role of ventilation conditions in identifying surrogate measures to predict smoke detector response. Within the last 10 years, there have been five significant studies examining the response of smoke detectors. These studies, examined as part of this project, include: Kemano by the National Research Council of Canada Naval Research Laboratory and Hughes Associates tests for shipboard applications Home Smoke Alarm Project by NIST Smoke Characterization Project by Underwriters Laboratories for the Fire Protection Research Foundation Experiments program in this project. These experimental programs include a sufficiently wide variety of spaces, fuels and ventilation conditions to form a substantial basis for the development of robust, simple guidelines for estimating smoke detector response. Unfortunately, the smoke detector responses appear to be strongly dependent on the specific characteristics of the smoke and in some cases on the detector technology. Consequently, proposing a single set of guidelines for obscuration, temperature rise and velocity which can apply to a wide range of applications is difficult, other than suggesting guidelines which would be very conservative in some applications. For flaming fires, the obscuration level in tests without forced ventilation ranged from 1.4 to 10.7 %/ft for ionization detectors and from 2.7 to 12.9 %/ft for photoelectric detectors. Given the

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. v

noted range in the 80th percentile values of obscuration at the time of response, a guideline which reasonably captures much of the data for smoke detectors of either type of technology is 8 %/ft. In the case of flaming fires in ventilated rooms, the 80th percentile values of the obscuration levels differ appreciably for the two detection technologies. For flaming fires with ventilation, the 80th percentile values of the obscuration level for photoelectric smoke detectors were 4.3 to 4.9 %/ft. In contrast, the 80th percentile values of the obscuration level for ionization smoke detectors were 8.0 to 10.3 %/ft, although it is noted that the 10.3% is based on only two tests conducted at a forced ventilation rate of 12 ACH. As such, a possible guideline of obscuration levels for photoelectric detectors could be 5 %/ft for ventilation rates ranging from 6 to 12 ACH. For ionization detectors, the 8 %/ft value appears to be an appropriate guideline considering only the results from the tests with 6 ACH. With the limited number of tests conducted at 12 ACH where ionization detectors responded, a guideline to estimate their response cannot be suggested. For non-flaming fires without ventilation, the 80th percentile values of the obscuration levels ranged from 4.4 to 18.5 %/ft for ionization smoke detectors and 1.6 to 12.1 %/ft for photoelectric smoke detectors. The 80th percentile values of the obscuration levels for non-flaming fires with ventilation were all less than 1 %/ft in this study and approximately 5 %/ft for ionization detectors in the NRL study. Given the limited data in this area, a recommendation for establishing a guideline of only 1 %/ft is questionable, especially in light of the difference in results obtained from experiments conducted as part of this study and the NRL study. Until further data is obtained, a value in excess of 1 %/ft is recommended and should perhaps be as large as 2.5 %/ft. The temperature rise at the time of detection response for flaming fires with no forced ventilation is highly dependent on the detection technology. A temperature rise of approximately 5 K can be suggested as a reasonable conservative guideline for ionization detectors, though should be greater than 5 K, e.g. 15 K given the measurements obtained in the NRL and NIST tests. For non-flaming fires and all fires with forced ventilation a temperature rise of approximately 3 K appears to be a reasonable guideline to estimate smoke detector response of either technology. Because the velocities associated with the forced ventilation provided in the test room were appreciably greater than the ceiling jet velocity, a guideline based on velocity cannot be recommended for such cases. An appreciable variation of smoke conditions was noted at the time of response of smoke detectors in all of the experimental programs reviewed. While guidelines of obscuration level or temperature rise can be suggested, these are very approximate in nature and may involve appreciable errors. One reason for this error is the fact that light obscuration and temperature are not related to the operating mechanisms of current smoke detector technologies, i.e. light scattering and ionization. Volume 3 presents an outline of additional research which could be used to better correlate light obscuration with light scattering measurements. The fourth task of this project was to evaluate the capabilities of the current release version (5.1.0) of the Fire Dynamics Simulator (FDS) to predict smoke detector activation in the two

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. vi

rooms described in Task 2 in response to the relatively low energy incipient fire sources characterized in Task 1. As part of this task, FDS simulations were performed of the 32 different room fire scenarios conducted as part of this project. The FDS simulated results were then compared with the experimental results. Volume 4 of this report describes the details of these simulations and comparisons. The baseline FDS simulations of the room tests were performed with a uniform grid size of 10 cm (4 in.). This resulted in a total number of 233,280 computational cells in both the unventilated enclosure domain, which had dimensions of 10.8 m (108 cells) by 7.2 m (72 cells) by 3.0 m (30 cells) high, as well as in the ventilated enclosure domain, which had dimensions of 7.2 m (72 cells) by 7.2 m (cells) by 4.5 m (45 cells) high. On a single-processor PC, it took a few hours to run the 5 to 10 minute simulations of the flaming fire sources to a few days to run the 80 to 90 minute simulations of the smoldering fire sources at this resolution. Doubling the grid resolution from 10 cm (4 in.) to 5 cm (2 in.) changes these run times from a few days to a number of weeks and consequently would be unreasonable for most applications. It is difficult to generalize about the comparisons of the FDS simulations of detector activation in the room tests with the actual room test detection data because of the wide range of results. In some cases, the simulated and actual smoke conditions at the detection stations were relatively close to one another and within the experimental scatter, while in other cases, the simulated smoke concentrations exceeded the measured smoke concentrations by relatively large margins. There are at least three potentially significant sources of uncertainty associated with FDS simulation of smoke detector performance in room fire scenarios: Uncertainties in the initial and boundary conditions specified for a scenario, including uncertainties in specification of the fire heat and smoke release rate histories and in specification of the mechanical ventilation; Uncertainties in the calculations performed by FDS to simulate heat and smoke transport; Uncertainties in the empirical models FDS currently uses to calculate smoke detector response and to predict smoke detector activation. Quantitative uncertainty analysis has not been performed as part of this project, but qualitatively it appears that the greatest uncertainties are associated with the first and third sources of uncertainty identified here. The eight incipient fire sources used for this project each exhibited a range of fire growth, heat release and smoke release rates that limited the reproducibility of the bench-scale and large-scale fire tests. It is unreasonable to expect the simulation of these fire scenarios to be any better than the scatter in the experiments being simulated. It is suspected that the treatment of mechanical ventilation represents another source of considerable uncertainty in the FDS simulations performed as part of this project. Real ventilation grilles and resulting airflows are more complicated than the simulated grilles and airflows in the ventilated enclosure. More work is needed to more fully explore this issue.

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. vii

Before this project was undertaken, the prediction of smoke production in FDS was based only on a user-specified constant soot yield tied to the heat release rate of a fire. During this project, at least three limitations with this approach to predicting smoke production were recognized: Only a single fire source could be specified, which did not allow separate specification of an ignition source and other fuels subsequently ignited; The smoke release rate could not vary independently of the heat release rate, so products with variable smoke yields could not be modeled properly; Smoldering and pyrolyzing smoke sources that produce substantial quantities of smoke but little heat could not be modeled properly. As a result of these limitations, the developers of the FDS model incorporated a new algorithm that permits the user to specify smoke release independently of heat release. This new feature was used to specify smoke release rates for this project. The primary findings of this project can be summarized as follows: The smoke release rates of eight different incipient fire sources, including four flaming sources, three smoldering sources and one overheated electrical wire, have been measured under well-characterized conditions in replicate bench-scale tests conducted in the IMO intermediate scale calorimeter at Underwriters Laboratories in Northbrook, IL. The primary smoke signature of interest in this project was the obscuration of visible light. Additional data was gathered during the bench-scale tests, including particle count density, mean particle diameter, carbon monoxide production and carbon dioxide production. This additional data may be of use in future investigations, but has not been analyzed for this project. Smoke obscuration was measured in the exhaust duct of the IMO intermediate scale calorimeter by projecting a white light beam across the diameter of the exhaust duct onto a photocell and measuring the change in voltage at the photocell caused by smoke particles in the light beam. Smoke release rates are characterized in units of m2/s, where the smoke release rate is calculated as the product of the smoke extinction coefficient, k (m-1), by the volumetric flow rate in the exhaust duct, V (m3/s):
S kV ln(I o / I ) V L

The total smoke release (TSR) is characterized in units of m2 and is calculated as the integral of the smoke release rate over the period of a test, i.e., the area under the smoke release rate curve:

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. viii

TSR
0

Sdt

The mass of smoke released during a test is characterized in units of gs and is calculated as the quotient of the total smoke release to the specific extinction coefficient, km, which was assumed to have a constant value of 8.7 m2/gs:
ms TSR km

The average smoke yield during a test is calculated as the quotient of the mass of smoke released to the fuel mass loss during a test:

ys

ms mf

When calculated in this way, the average smoke yields obtained for the eight incipient sources in the IMO apparatus are shown in Table E3 along with other data from the IMO tests. These data provide an indication of the variability in the replicate tests. When this project started, smoke production was calculated in FDS only in terms of constant smoke yield factors tied to the specified heat release rate through the mixture fraction model used by FDS. During this project, it became apparent that smoke yields for the eight incipient sources are not constant and that characterizing smoke production in terms of a constant smoke yield factor would not produce accurate smoke production or transport results in FDS for these incipient fire sources. During this project, the developers of FDS implemented a new method to specify smoke production independently of heat release. Called the species ID method, this method was used throughout this project to specify smoke production in FDS for both the IMO test simulations and the room fire simulations. The bench-scale tests conducted in the IMO apparatus were simulated in FDS as one means to validate the capabilities of FDS to model smoke production and transport. For these FDS simulations, a uniform grid size of 2.5 cm was used. These simulations of the IMO tests showed that the calculated smoke quantity transported past the smoke eye in the exhaust duct was similar to the quantity of smoke released from the fuel package, as shown in Table E4. The largest variation between output and input was 5.4%. Differences in the peak obscuration values and the times to reach these peaks between the IMO physical tests and FDS simulations are shown in Table E5. The simulated peak smoke release rate was within 17.3% of the specified peak smoke release rate for all fuels except the PVC insulated wire. The FDS simulated time to peak obscuration lagged the specified peak time by 4 to 33 seconds, with two exceptions. This lag time is most likely related to the transport lag between smoke release at the fuel source and measurement at the smoke eye in the exhaust duct. The IMO apparatus smoke test data was not corrected

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. ix

for transport lag; this suggests that the actual smoke release in the IMO tests occurred earlier than represented in the smoke release rate curves for these tests. For the FDS simulations of the IMO tests, one replicate test for each fire source was selected for simulation and comparison with the measured data from that test. For the FDS simulations of the room fire tests, the IMO test data was typically averaged for each fire source and this average data was used as input to the FDS simulations. The expected uncertainty in the FDS input data based on this approach has not yet been characterized.

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. x

Table E3. Summary of data obtained from tests conducted in IMO apparatus

Sample Description

Mode

Peak HRR

Peak SRR Smoke Yield

Total SR

Total HR

(kW) Shredded Paper-1 Shredded Paper-2 Shredded Paper-3 Shredded Paper PU Foam/Microfiber-1 PU Foam/Microfiber-2 PU Foam/Microfiber-3 PU Foam/Microfiber Circuit Board-1 Circuit Board-2 Circuit Board-3 Circuit Board Computer Case-1 Computer Case-2 Computer Case-3 Computer Case PU Foam/Microfiber-1 PU Foam/Microfiber-2 PU Foam/Microfiber-3 PU Foam/Microfiber Ponderosa Pine-1 Ponderosa Pine-2 Ponderosa Pine-3 Ponderosa Pine Cotton Linen Fabric-1 Cotton Linen Fabric-2 Cotton Linen Fabric-3 Cotton Linen Fabric PVC Insulated Wire-1 PVC Insulated Wire-2 PVC Insulated Wire-3 PVC Insulated Wire Flaming Flaming Flaming Average Flaming Flaming Flaming Average Flaming Flaming Flaming Average Flaming Flaming Flaming Average Smoldering Smoldering Smoldering Average Smoldering Smoldering Smoldering Average Smoldering Smoldering Smoldering Average Smoldering Smoldering Smoldering Average N/A N/A N/A N/A N/A N/A N/A N/A N/A 7.76 8.91 12.49 9.72 8.54 11.23 9.79 9.85 1.90 2.41 2.59 2.30 0.00 0.73 0.63 0.45 N/A N/A N/A

(m2/s) 1.350 0.990 2.000 1.447 0.432 0.512 0.513 0.486 0.534 0.491 0.587 0.537 0.119 0.245 0.292 0.219 0.066 0.073 0.040 0.059 0.161 0.1219 0.1458 0.143 0.084 0.118 0.086 0.096 0.072 0.155 0.094 0.107

(g/g) 0.094 0.083 0.097 0.091 0.094 0.096 0.095 0.095 0.215 0.221 0.319 0.252 0.785 0.967 0.878 0.877 0.085 0.089 0.073 0.082 0.141 0.142 0.140 0.141 0.254 0.240 0.168 0.221 0.237 0.258 0.256 0.250

(m) 110.5 105.8 117.4 111.2 74.2 84.2 80.6 79.7 22.0 20.3 24.7 22.3 9.2 19.1 19.6 15.9 39.9 43.8 36.2 39.9 182.7 182.3 183.0 182.7 43.1 40.9 29.9 38.0 2.1 2.4 2.5 2.3

(MJ) 0.388 0.612 0.589 0.530 1.896 2.240 1.974 2.037 0.826 0.924 1.120 0.957 0.000 0.129 0.078 0.069 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xi

Table E4. Variation in FDS modeling results of smoke measurement in IMO apparatus
Model Output to Input Fuel Source Flaming -0.8% Shredded Office Paper PU Foam with Micro-5.4% fiber Fabric 0.1% Printed Circuit Board Computer Case ABS 4.3% Plastic Smoldering PU Foam with Micro-3.1% fiber Fabric -1.6% Ponderosa Pine -0.9% Cotton Linen Fabric -1.6% PVC Insulated Wire Model Output to Test 4.8% -4.7% -1.6% 4.7%

-1.5% 1.9% 1.3% -5.2%

Table E5. Peak obscuration values and times in the IMO physical tests and FDS simulations.

The 88 room fire tests conducted as part of this project provide a wealth of data on the conditions resulting from the eight incipient fire sources and the response of spot, beam and aspirated detection systems to these conditions in both unventilated and mechanically ventilated enclosures. Only a fraction of this data has been analyzed in detail as part of this project, but all the data acquired during this project has been summarized in tabular

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xii

and chart form in Excel spreadsheet files and will be made available for future analysis. More than 1,200 data charts have been generated to illustrate the data from these tests. The responses of the two brands of photoelectric detectors used in this project were generally consistent with each other, but the levels of smoke obscuration reported by these detectors was not always consistent with the smoke obscuration levels measured at the adjacent detection stations. This may be due to the different methods used to measure smoke obscuration by the detectors, which use light reflection, and by the adjacent photocell assemblies, which use light obscuration. The levels of smoke obscuration reported by the spot detectors are based on correlations developed from testing in the UL smoke box using only a single smoke source, a smoldering cotton wick. This correlation has not been demonstrated for the smoke sources used in this project; this may account for at least some of the differences between the smoke obscuration levels reported by the spot detectors and those measured by the adjacent photocell assemblies. Based on analysis of the smoke detector data from the room fire tests in this project, the smoke obscuration at detection, represented in %/ft and based on the 80th percentile values, are shown in Table E6 for the different ventilation conditions, fire conditions and detector types. Table E6. Smoke obscuration at detection in room tests based on 80th percentile values.
Ventilated Unventilated 6 ACH Flaming Ionization Photoelectric Nonflaming Ionization Photoelectric 8 8 12 10 8 5 1? 1? 12 ACH Insuff. Data 5 Insuff. Data 1?

Based on analysis of the smoke detector data from the room fire tests in this project, the temperature rise at detection, represented in K and based on the 80th percentile values, are shown in Table E7 for the different ventilation conditions, fire conditions and detector types. Table E7. Temperature rise at detection in room tests based on 80th percentile values.

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xiii

Ventilated Unventilated 6 ACH Flaming Ionization Photoelectric Nonflaming Ionization Photoelectric 5 5 3 3 3 3 3 3 12 ACH Insuff. Data 3 Insuff. Data 3

Based on analysis of the smoke detector data from the room fire tests in this project, substantial errors are indicated in using simplistic guidelines of obscuration and temperature rise based on 80th percentile values. The values reported in the previous tables both overestimate and underestimate response times in specific tests. These errors may be reduced through use of a dual parameter approach, e.g. obscuration and velocity in unventilated rooms: o Flaming fires, photoelectric detectors: 5.5-9.5 %/ft and 0.14-0.33 m/s o Non-flaming fires, photoelectric detectors: 1.5-2.5 %/ft and 0.03-0.07 m/s The near-ceiling velocity of ventilation in the ventilated room tests with 6 and 12 ACH exceeds the velocity of the ceiling jet from the incipient fires in these tests. The near-ceiling velocity field caused by the mechanical injection of air at 6 and 12 ACH has not been experimentally characterized. The responses of the aspirated systems in the 64 tests in the ventilated room have been summarized in Excel spreadsheets, but have not yet been analyzed. The data from the aspirated systems has not yet been synchronized with the other experimental data due to technical difficulties with the synchronization process. Baseline FDS simulations have been conducted for 32 different room fire scenarios involving the 8 incipient fire sources under 4 different conditions, including unventilated tests conducted in the UL 217/268 standard smoke room, unventilated tests conducted in the ventilated room constructed for this project, and ventilated tests conducted at 6 and 12 air changes per hour in this ventilated room. For the baseline FDS simulations, a 10 cm uniform grid was used, resulting in a total of 233,280 computational cells for both the unventilated and ventilated enclosures. For the baseline FDS simulations, the specified smoke release rate was based on measurements of smoke release rate in the IMO benchscale tests and was not corrected for transport lag. Additional FDS simulations have been conducted for a few scenarios using the multimesh feature of FDS to provide a higher level of resolution of 5 cm in the fire plume and ceiling jet regions of the two enclosures, but these simulations have not yet been compared with the experimental data or the baseline FDS simulations. These results and comparisons will be reported separately. Additional FDS simulations have also been conducted for the 16 mechanically ventilated scenarios using a different description for the ceiling vents than in the baseline calculations. These simulations use a uniform cell size of 10 cm, but they have not yet

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xiv

been compared with the experimental data or the baseline FDS simulations. These results and comparisons will be reported separately. The 32 FDS baseline simulations demonstrate a wide range of results in comparison with the related room fire tests so it is difficult to generalize about the current capability of FDS to predict smoke detector activation over the range of fuels and ventilation conditions evaluated in this project. In many of the 32 baseline FDS simulations, the predicted maximum level of smoke obscuration is higher than the measured level of smoke obscuration in the related room fire tests. This may be due to the relatively coarse resolution of 10 cm used for the baseline FDS simulations. In these simulations, it appears that the dynamics of plume entrainment is not fully captured, which would lead to higher concentrations of smoke in the FDS simulations. Another factor that may contribute to the higher predicted smoke obscuration levels is smoke deposition to room surfaces, which is not currently addressed in FDS. The levels of smoke obscuration measured by the photocell assemblies at the detection stations during the mechanically ventilated tests were low in comparison with the levels of smoke obscuration reported by the adjacent smoke detectors and in comparison with the levels of smoke obscuration predicted by the associated FDS simulations. The reason for this has not yet been determined. The mechanically ventilated tests conducted at 6 and 12 ACH demonstrated conditions different from those observed in the unventilated tests. In particular, smoke did not readily transport past the plane defined by the line between the two injection louvers at the center of the room. Instead, the smoke tended to stack up on the fire side of this plane, suggesting that the mechanical injection of air was acting as an air curtain. Qualitatively, this was observed in both the room fire tests as well as in the baseline FDS simulations. This also had the effect of delaying smoke detector response on the downstream side of the injection louvers. The impact of mechanical ventilation on smoke detector response warrants further investigation. Recommendations for further study include: Develop the relationship between light scattering and light obscuration for fuels of primary interest (UL 217 fuels, PU foam, etc.) as a means to resolve the differences in smoke obscuration levels reported by the smoke detectors and those measured by the adjacent photocell assemblies. Perform additional FDS simulations at higher resolutions to evaluate the effects on predicted smoke obscuration levels. Perform additional mechanically ventilated room tests to characterize the velocity field caused by the injection of air through representative air louvers. Establish methods to more accurately simulate the injection of air through representative air louvers in FDS. Further investigate the impact of mechanical ventilation on smoke detector response. In summary, this project has generated a wealth of new data on the fire-induced conditions in the room of origin resulting from a range of different incipient fire sources under both unventilated

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xv

and mechanically ventilated conditions. It has also generated a wealth of data on the response of both spot-type and aspirated smoke detection systems to these conditions. Thirty-two different room fire scenarios were conducted in replicate in 88 large-scale tests and each scenario was simulated using the current release version (5.1.0) of the Fire Dynamics Simulator to evaluate the current capabilities of FDS to predict smoke detector response and activation. In light of the large number of room fire tests conducted and FDS simulations performed, it has not been possible to perform a comprehensive analysis of the results. The data from these tests and FDS simulations demonstrate a range of results that warrants further analysis.

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xvi

Validation of a Smoke Detection Performance Prediction Methodology Volume 3. Evaluation of Smoke Detector Performance

Prepared for: Kathleen Almand Fire Protection Research Foundation 1 Batterymarch Park Quincy, MA 02169

Prepared by: James A. Milke and Frederick W. Mowrer University of Maryland Pravinray Gandhi Underwriters Laboratories, Inc.

October 10, 2008

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xvii

CONTENTS Page 1.0 Introduction 1.1 Smoke Detector Principles 2. Modeling Response of Smoke Detectors 2.1 Threshold Guidelines, NFPA 72 2.2 Zone Models 2.3 FDS 3. Experimental Data of Smoke Detector Response 3.1 Kemano 3.2 NRL 3.3 Home Smoke Alarm Project 3.4 Smoke Characterization Project 3.5 Experimental program in this project 4. Discussion 4.1 Trends in detector response 4.2 Variation in obscuration levels at detector response 4.3 Velocity 4.4 Dual parameter guidelines 5. Summary 6. References Appendix. Conditions at the time of response for Ventilated Room 1 1 4 4 5 5 6 7 8 11 17 24 34 34 38 39 40 42 42 45

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xviii

List of Figures Page Figure 1. Mean particle diameters for light obscuration of 0.5 %/ft 4 Figure 2. Cumulative distribution of optical density at the time of smoke detector response 8 Figure 3. Cumulative distribution of optical density at the time of smoke detector response, flaming fires 9 Figure 4. Cumulative distribution of optical density at the time of smoke detector response, non-flaming fires 10 Figure 5. Cumulative distribution of temperature rise at the time of smoke detector response, flaming fires 10 Figure 6. Cumulative distribution of velocity at detector response, flaming fires 11 Figure 7. Obscuration levels at detector response, flaming fires 12 Figure 8. Obscuration levels at detector response, non-flaming fires 13 Figure 9. Temperature rise at detector response, flaming fires 13 Figure 10. Temperature rise at detector response, non-flaming fires 14 Figure 11. Cumulative distribution of obscuration at detector response, flaming fires 15 Figure 12. Cumulative distribution of obscuration at detector response, non-flaming fires 15 Figure 13. Cumulative distribution of temperature rise at detector response, flaming fires 16 Figure 14. Cumulative distribution of temperature rise at detector response, non-flaming fires 16 Figure 15. Obscuration levels at detector response, flaming fires 17 Figure 16. Obscuration levels at detector response, non-flaming fires 18 Figure 17. Temperature rise at detector response, flaming fires 18 Figure 18. Temperature rise at detector response, non-flaming fires 19 Figure 19. Velocity at detector response, flaming fires 19 Figure 20. Velocity at detector response, non-flaming fires 20 Figure 21. Cumulative distribution of obscuration at detector response, flaming fires 21 Figure 22. Cumulative distribution of obscuration at detector response, non-flaming fires 21 Figure 23. Cumulative distribution of temperature rise at detector response, flaming fires 22 Figure 24. Cumulative distribution of temperature rise at detector response, non-flaming fires 22 Figure 25. Cumulative distribution of velocity at detector response, flaming fires 23 Figure 26. Cumulative distribution of velocity at detector response, non-flaming fires 23 Figure 27. Light obscuration measurement in shredded paper test 24 Figure 28. Cumulative distribution of obscuration at detector response, non-ventilated room 26 Figure 29. Cumulative distribution of temperature rise at detector response, non-ventilated room 26 Figure 30. Obscuration levels at detector response, ventilated room 27 Figure 31. Temperature rise at detector response, ventilated room 27 Figure 32. Cumulative distribution of obscuration at photoelectric detector response, flaming fires, ventilated room 29

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xix

Figure 33. Cumulative distribution of temperature rise at photoelectric detector response, flaming fires, ventilated room Figure 34. Cumulative distribution of obscuration at photoelectric detector response, non-flaming fires, ventilated room Figure 35. Cumulative distribution of temperature rise at photoelectric detector response, non-flaming fires, ventilated room Figure 36. Cumulative distribution of obscuration at ionization detector response, flaming fires, ventilated room Figure 37. Cumulative distribution of temperature rise at ionization detector response, flaming fires, ventilated room Figure 38. Cumulative distribution of obscuration at ionization detector response, non-flaming fires, ventilated room Figure 39. Cumulative distribution of temperature rise at ionization detector response, non-flaming fires, ventilated room Figure 40. 80th percentile obscuration level at detector response Figure 41. 80th percentile obscuration level at detector response in recent experimental programs Figure 42. Visibility to illuminated exit sign through smoke Figure 43. 80th percentile temperature rise at detector response Figure 44. 80th percentile obscuration level at detector response in recent experimental programs Figure 45. Near ceiling velocities in ventilated room tests Figure 46. Light obscuration and velocity at photoelectric detector response, unventilated room Figure 45. Near ceiling velocities in ventilated room tests Figure 46. Light obscuration and velocity at photoelectric detector response, unventilated room

30 30 31 32 32 33 33 34 35 36 37 37 40 40 41 41

Validation of a Smoke Detection Performance Prediction Methodology October 10, 2008

Volume 3 p. xx

List of Tables Page E1. Incipient fire sources E2. Matrix of large-scale room fire test designations E3. Summary of data obtained from tests conducted in IMO apparatus E4. Variation in FDS modeling results of smoke measurement in IMO apparatus E5. Peak obscuration values and times in the IMO physical tests and FDS simulations. E6. Smoke obscuration at detection in room tests based on 80th percentile values. E7. Temperature rise at detection in room tests based on 80th percentile values. 1. Obscuration levels for response of smoke detectors to flaming fires (%/ft) 2. Temperature rise for response of smoke detectors to flaming fires (K) 3. Detector response statistics for flaming and non-flaming fires, Home Smoke Alarm Project 4. 80th percentile values of parameters at detector response, Home Smoke Alarm Project 5. Detector response statistics for flaming and non-flaming fires, Smoke Characterization Project th 6. 80 percentile values of parameters at detector response, Smoke Characterization Project 7. Photoelectric detector response statistics for flaming and non-flaming fires 8. 80th percentile values of parameters at detector response 9. Detector response statistics for flaming and non-flaming fires, photoelectric detector SG 10. Detector response statistics for flaming and non-flaming fires, photoelectric detector SS 11. Detector response statistics for flaming and non-flaming fires, ionization detector SG 12. 80th percentile values of parameters at detector response, photoelectric detector SG 13. 80th percentile values of parameters at detector response, ionization detector SG iii iii ix x x xi xi 4 5 14 16 20 24 25 26 28 28 29 31 33

FPRF incipient fire project final report Volume 3 1. Introduction

p. 1

An objective of this project is to develop the means to estimate the response of smoke detectors using the computed results of the smoke conditions by FDS. This objective is consistent with the growing interest to develop engineering methods for estimating the response of fire protection systems, either for applications in design or fire investigation. Methods have been developed to estimate the response of flame, thermal and smoke detectors. The method for estimating the response of thermal detectors has been included in NFPA 72 for many years and is based on an estimate of the temperature of the ceiling jet, originally developed by Alpert [1972]. The response of flame detectors is based on an estimate of the radiation heat transfer from a flame plume to a detector. A computational approach to provide this estimate has been more recently included in NFPA 72 [2007]. In both cases, the sensitivity of the heat or flame detectors to the respective signature needs to be determined via experiment. 1.1 Smoke Detector Principles The response of traditional spot smoke detectors is dependent on the characteristics of the smoke in the vicinity of the detector and the characteristics of the detector. Most of the current smoke detectors operate based on one of two types of detection technologies: photoelectric or ionization. Contemporary, photoelectric smoke detectors respond based on the scattering of light caused by smoke particles [Schifiliti, et al., 2002]. The response of an ionization smoke detector is based on a change in the local ionization field within the detection chamber caused by introduction of smoke particles within the chamber. Both types of alarms activate when a set threshold is reached. Two of the principal characteristics of the smoke affecting the response of the smoke detector technologies include the distribution of particle sizes and concentration of smoke particles. The role of the two characteristics depends on the sensing technology included in the detector. Relationships between light scattering and current change within an ionization chamber relative to the characteristics of smoke particles are provided in the literature [Schifiliti and Pucci, 1996]. While light scattering is dependent on the smoke particle characteristics (size distribution and concentration), it is also affected by the wavelength of the transmitted light, refractive index of the smoke and the angle between the light source and the light receiver. There are three regions of light scattering behavior described in the literature for single, spherical particles. The three regions depend on the ratio of the particle diameter (d) and wavelength of light, , as follows: Rayleigh: d/ < 0.1 Mie: 0.1 < d/ < 0.4 Bricard: d/ > 0.4 Considering the wavelength of light used in photoelectric detectors and the range of particle sizes produced in fires, Mie theory can be applied to appreciate the effect of smoke particle characteristics on photoelectric smoke detector responses. According to Mie theory, light

FPRF incipient fire project final report Volume 3

p. 2

scattering is linearly proportion to the number of smoke particles and the square of the diameter of the particles:
LS ni di
2

(1)

Performance of the ionization detector is dependent on the attachment coefficient of air-molecule ions to the soot particles, , where = 2 Ddm [Litton, et al., 2004] Thus, the MIC response is related to the product of particle count and diameter as shown in Eq. 2. MIC dini (2)

The conditions in a ceiling jet or smoke layer determined via engineering methods generally do not include smoke particle size and concentration. Instead, the methods include estimates of involve light obscuration, temperature and velocity. As indicated by Schifilti, et al., this provides an inherent difficulty in estimating the response of smoke detectors. The ability of smoke particles to obscure light is described by Beers Law, expressed in terms of the optical density per unit path length. According to Beers Law, the optical density (i.e. light obscuration) is linearly proportional to the concentration of smoke, as expressed in Eq. 3.
OD Cs

(3)

Where OD is the optical density, is the path length, and Cs is the smoke concentration. The smoke concentration, Cs, is proportional to the smoke number density as follows:
Cs ni di
3

(4)

Where ni, and di are the number count (density) and particle diameter for a small range of particle size i, referred to as a bin in the recent UL study [Fabian, et a., 2007]. A relationship between optical density per path length and the smoke number density count at a given time can be developed by combining Eq. 3 and Eq. 4:
D ni d i
3

(5)

Optical density and obscuration per foot can be related as expressed in Eq. 6:

OBS 100 1 10

(6)

Where OBS is obscuration per meter and D is the optical density (m-1). Unfortunately, a particular level of obscuration does not uniquely describe the characteristics of a particular smoke. This issue was evident in the recent Smoke Characterization Project conducted

FPRF incipient fire project final report Volume 3

p. 3

by UL collected highly detailed information of the smoke particle size, concentration, light obscuration, and other parameters relative to smokes produced by several different fuels [Fabian, etal., 2007]. The mean particle diameter measured for smoke from flaming and non-flaming fires that produced a light obscuration of 0.5%/ft in the Smoke Characterization Project are presented in Figure 1. For the series of experiments indicated in Figure 1, the mean particle diameter producing the same level of light obscuration ranges from 0.08 to 0 0.22 microns. As such, even though the level of light obscuration is the same for these smokes, their detectability by a light scattering detector or ionization detector would vary, given that these detection technologies are dependent on the square and first power of the particle diameter, respectively. Nonetheless, the purpose of this project is to identify relationships between smoke detector and smoke parameters which are included within current numerical models. Thus, being that optical density is computed by several numerical models, the relationship of smoke detector response to light obscuration is sought in this project. One of the relatively unique aspects of this study is an examination of the role of ventilation conditions in identifying surrogate measures to predict smoke detector response. Most of the previous studies in this area have been conducted in quiescent atmospheres.
0.25 Flaming Non-flaming 0.20

Mean Particle Diameter (micron)

0.15

0.10

0.05

0.00

en e

am

ne

cy an ur at e

ne

ak er

n/ Po ly

to n

Po l

ca rp

ap

ou gl as

/T ol ue

sp

ty r

de ro s

Po ly s

PU

ot to

ew

in

of fe

ne

am

yl on

am

in

fo

Po n

ep ta

fo am

Figure 1. Mean particle diameters for light obscuration of 0.5 %/ft [Fabian, et al., 2007]

2. Modeling Response of Smoke Detectors 2.1 Threshold Guidelines, NFPA 72 Annex B of NFPA 72 includes an engineering approach for estimating the response of smoke detectors to flaming fires. Three parameters identified in the Annex include temperature rise, obscuration (or optical density) and velocity. Given that neither of the two detection technologies respond to conditions represented by any of these three parameters, inherent errors

PU

PU

Po ly i

in

fo

PU

so

Br e

pi

fo

ot

ad

er

fir

et

FPRF incipient fire project final report Volume 3

p. 4

are to be expected when applying any of these parameters for estimating the response of smoke detectors [Schifiliti and Pucci, 1996]. Heskestad and Delichatsios [1977] suggested values of the optical density that coincided with smoke detector response based on their measurements. Their suggestions are incorporated into the optical densities noted in Annex B of NFPA 72. The optical densities noted in NFPA 72 are converted into obscuration levels using equation 4 and are included in Table 1. As indicated in Table 1, the obscuration level associated with detector response varies by the detection technology and fuel. The range in obscuration levels for smokes from the various fuels varies by a factor of 11 for photoelectric detectors and 180 for ionization detectors.

Table 1. Obscuration levels for response of smoke detectors to flaming fires (%/ft) Material Wood Crib Cotton fabric Polyurethane foam PVC Photoelectric 3.39 1.83 10.9 20.6 Ionization 1.14 0.12 10.9 20.6

Schifiliti and Pucci [1996] estimated the temperature rise necessary for detection to fires involving the fuels noted in Table 1 based on ratios of the optical density and temperature at detector response determined by Heskestad and Delichatsios. The resulting temperature rises suggested to estimate smoke detector response are included in Annex B of NFPA 72 and are reproduced here in Table 2. A default temperature rise of 13 K, presumably applicable to any fuel is sometimes suggested, though it is apparent that for fires involving wood cribs and cotton fabrics, such a value would provide an optimistic view of detector response. Geimans review of previous experiment programs indicated that a significant proportion of ionization smoke detectors responded at temperature rises much less than those indicated in Table 2.

Table 2. Temperature rise for response of smoke detectors to flaming fires (K) (NFPA 72) Material Wood Crib Cotton fabric Polyurethane foam PVC Photoelectric 41.7 27.8 7.2 7.2 Ionization 13.9 1.7 7.2 7.2

The critical velocity of the ceiling jet associated with the response of smoke detectors ranges from 0.13-0.15 m/s for flaming fires [Borzovski, 1989][Geiman, 2003]. The critical velocity is independent of the type of smoke detector.

FPRF incipient fire project final report Volume 3 2.2 Zone Models

p. 5

Some zone models include an algorithm to estimate the response of smoke detectors. CFAST is one zone model that includes such an algorithm [Jones, et al., 2005]. The approach used in CFAST is to treat a smoke detector as a thermal detector, where the sensitivity of the smoke detector is expressed as a Response Time Index and the operating temperature is expressed as a magnitude of temperature rise above ambient. Guidance on the temperature rise to select is currently available in Annex B of NFPA 72, though much of the purpose Geimans research was to assess the appropriateness of the guidance included in NFPA 72. An elementary heat transfer calculation is conducted (treating the smoke detector as a thermal detector) to determine when the smoke detector reaches the activation temperature. 2.3 FDS The algorithm within Fire Dynamics Simulator (FDS) [McGrattan, et al., 2008] is based on the mass fraction of smoke in the sensing chamber of the smoke detector and velocity within the ceiling jet in the vicinity of the smoke detector. The mass fraction is related to the obscuration per unit length as expressed in equation (7).

OBS 100 1 exp

Yc l

(7)

The FDS Technical Manual [McGrattan, 2007] suggests a value of 8700 m2/kg 1100 m2/kg at a wavelength of 633 nm for for most flaming fires. The FDS model relates mass fraction of smoke in the sensing chamber of the detector to the mass fraction of smoke in the ceiling jet outside the detector by one of two approaches. The simpler of the two approaches is that described by Heskestad [1975]. Heskestads approach is presented in the governing equation included as Eq. 8:

dYc dt

Ye t

Yc t Lu

(8)

where Ye and Yc are the mass fraction of smoke external to the detector and in the detectors sensing chamber, respectively, L is the characteristic length of the detector and u is the ceiling jet velocity at the location of the smoke detector. The ceiling jet velocity and mass fraction external to the detector are parameters included in the basic calculations performed by FDS. The characteristic length of the detector needs to be determined via experimentation. The characteristic length for the light scattering detectors determined in previous experimental efforts range from 2.6 to 15 m [Schifiliti, et al., 2002]. An alternative approach to estimate the response time of smoke detectors included within FDS is an approach suggested by Cleary [2000]. This approach requires a determination of four parameters (rather than Heskestads one) to address the entrance lag issues. Roby, et al. [2007] suggest that the Cleary model is a better predictor of smoke detector response for cases where the detector is exposed to ceiling jets with velocities less than 0.5 m/s.

FPRF incipient fire project final report Volume 3

p. 6

In Clearys approach, the entrance time is divided into two segments. The time segments are associated with the time required for the smoke to pass through the exterior housing and then baffles in order to fill the sensing chamber. The two time segments are approximated using Eq. 9 and 10 [Cleary, et al., 1999].

te tc

e c

u u

(9) (10)

The s and s need to be empirically determined, with suggested values for these four parameters for smoke detectors included in the FDS Users Guide [McGrattan, et al., 2008a]. With this approach, the governing equation for the change in the mass fraction of smoke in the sensing chamber is presented as Eq. 11.

dYc dt

Ye t

te tc

Yc t

(11)

3. Experimental Data of Smoke Detector Response Within the last 10 years, there have been five significant studies examining the response of smoke detectors. These studies are: Kemano [Su, at al., 2003] Naval Research Laboratory (NRL) [Gottuk, et al., 1999][Rose-Pehrsson, et al., 2000][Wong, et al., 2000] Home Smoke Alarm Project by NIST [Bukowski, et al., 2008] Smoke Characterization Project [Fabian, et al., 2007] Experiments program in this project. Observations and trends of the response of detectors included in each of the projects will be discussed in their own section. 3.1 Kemano The Kemano tests were conducted by the National Research Council of Canada (NRC). These tests were conducted in residences consisting of a one-story bungalow and a two-story singlefamily home. The one-story bungalow had approximate internal dimensions of 7.7 x 12.2 x 2.4 m. The internal dimensions for each story of the two-story single-family home were 6.9 x 8.9 x 2.4 m. A total of 12 tests, were conducted where detectors responded. All of the test scenarios included the fires starting as non-flaming fires and transitioning to flaming fires. In this section, particular tests are referred to as flaming or non-flaming fires based on the mode of combustion present when detectors activated.

FPRF incipient fire project final report Volume 3

p. 7

Both battery-powered ionization and photoelectric smoke alarms were used in the tests. Measurements for the tests included optical density and temperature. The optical density meters used a 940 nm, pulsed, near-IR LED light source with a photodiode, separated by a distance of 0.6 m. Temperature was measured with 26 AWG, Type K thermocouples. The cumulative probability distribution for the optical density at the time or response of a smoke detector is presented in Figure 2. Geiman, et al., observed that the optical density for flaming fires was very similar for both ionization and photoelectric detectors. The optical density associated with the ionization smoke detector response for flaming fires at the 80th percentile level was approximately 0.11 m-1, (7.4 %/ft) and 0.14 m-1 (8.4 %/ft) for photoelectric smoke detectors. The 80th percentile levels for the non-flaming fires were 0.19 m-1 (10.8 %/ft) for the photoelectric detectors and 0.21 m-1 (13.7 %/ft) for the ionization detectors. The 80th percentile levels for the two types of detectors were very similar for a particular type of fire.

Figure 2. Cumulative distribution of optical density at the time of smoke detector response [Geiman, et al.,2006]

The most significant difference in the optical density at response was observed for the nonflaming fires, at least for the lesser percentiles. The difference in optical density observed at the time of photoelectric and ionization smoke detector response for 80 percent of the cases moderated. For photoelectric detectors, the 80th percentile value of the optical density was approximately 0.20 m-1 (13.1 %/ft) and for ionization detectors was approximately 0.21 m-1 (13.7 %/ft), only a 5% difference. The temperature rise observed at the time of smoke detector response was in the range of 1-3 K and was dismissed as being too small to be meaningful.

FPRF incipient fire project final report Volume 3 3.2 NRL

p. 8

The NRL tests were conducted by Hughes Associates, Inc. and NRL. Some of these tests were conducted in a small compartment (6.1 x 3.6 x 3.0 m) while others were conducted in a mediumsize test compartment (5.9 x 8.1 x 3.0 m). Two ventilation scenarios were used in these tests, one including no forced airflow and the other providing an airflow of 0.22 m3/s in the small compartment and 0.45 m3/s in the medium compartment which represented 12 air changes per hour in the respective compartments. Forty-one flaming and non-flaming fires were conducted as part of this test series. Two models each of ionization and photoelectric smoke detectors were included in the Navy tests The optical density was measured with an optical density meter, which utilized an 880 nm infrared (IR) light emitting diode (LED) and receptor arrangement over a 1.0 m path length. Temperature was measured with Inconel-sheathed, type K thermocouples. A sonic anemometer was used to measure the gas velocity in two orthogonal, horizontal directions. The cumulative probability distribution for the optical density at the time or response of a smoke detector is presented in Figure 3. The data included in Figure 3 is for tests with and without ventilation. In these tests, the 80th percentile of the ionization detector response is at a lesser optical density (0.02 m-1 or 1.4 %/ft) than the photoelectric detector (0.02 m-1 or 2.8 %/ft). The smoke characteristics at response were observed to be largely unaffected by the presence of ventilation.
Cumulative Percentage of Detector Alarms

Navy Flaming Fires

80

Ion Photo
50

20

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Optical Density at Alarm (m-1)

Figure 3. Cumulative distribution of optical density at the time of smoke detector response, flaming fires [Geiman, et al., 2004]

The cumulative probability distribution of the optical density associated with ionization smoke detector response for non-flaming fires is presented in Figure 4. The 80th percentile of the optical density in the ventilated tests was approximately 0.07 m-1 (4.8 %/ft) as compared to 0.19

FPRF incipient fire project final report Volume 3

p. 9

m-1 (12.5 %/ft) for the unventilated tests. The difference in the optical density in the ventilated and unventilated tests at the time of response for the photoelectric detector was similar to that observed for the ionization detector.
0 Air Changes / Hour 12 Air Changes / Hour

Cumulative Percentage of Detector Alarms

80

50

20

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Optical Density at Alarm (m-1)

Smoldering Fires, Model 2 Ionization Detectors Figure 4. Cumulative distribution of optical density at the time of smoke detector response, non-flaming fires [Geiman, et al., 2004]

The temperature rise for flaming fires at the time of response is presented in Figure 5. The 80 th percentile value is approximately 3 K for the ionization detectors and 16 K for the photoelectric detectors. The temperature rise observed at the time of response for the non-flaming fires was negligible. The mean ceiling jet velocity at the time of detector response for flaming fires was 0.13 m/s. These experiments are part of the basis for the velocities indicated in NFPA 72. Geiman, et al., indicate that detectors did respond at velocities as small as 0.05 m/s, well below the critical velocity noted in NFPA 72, refuting the notion that the cited range of 0.13 to 0.15 m/s is a minimum threshold velocity below which detector response is not expected. The velocity at the time of detector response for non-flaming fires was negligible in most tests.

FPRF incipient fire project final report Volume 3

p. 10

Figure 5. Cumulative distribution of temperature rise at the time of smoke detector response, flaming fires [Geiman, et al., 2006]

Figure 6. Cumulative distribution of velocity at detector response, flaming fires [Geiman, et al., 2006]

FPRF incipient fire project final report Volume 3 3.3 Home Smoke Alarm Project

p. 11

The full-scale experimental program that was included as part of the Home Smoke Alarm Project was conducted by the National Institute of Standards and Technology (NIST). These tests were conducted in two spaces, a one-story manufactured home and a two story house. Details of the spaces in which the tests were conducted are included in Bukowski, et al. [2008]. The overall dimensions of the manufactured home were 10.12 m x 4.17 m. The ceiling height inside the home was sloped, ranging from 2.1 m at the exterior walls to 2.4 m in the center. Tests were conducted in a bedroom, living room and kitchen. No forced ventilation was provided during these tests.

The footprint of the two-story home had dimensions of 6.19 m x 9.65 m. The ceiling height on both levels is unspecified in the NIST report, though is on the order of 2.4 m. Again, no forced ventilation was provided during the tests. A total of 32 fire tests were conducted where detectors responded. The test scenarios included both non-flaming and flaming fires and involved a variety of fuel items. Both ionization and photoelectric smoke alarms were used in the tests. For the purposes of the analysis included in this document, only the response of the closest array of detectors to the fire was considered. Measurements for the tests included optical density and temperature. The optical density meters used a low-cost laser pointer consisting of Class II laser diode with a wavelength range of 630 nm to 680 nm (the separation distance between the light source and photocell is not specified in the NIST report). Temperature was measured with Type K thermocouples. The obscuration levels at which the ionization and photoelectric detectors responded is presented in Figure 7 for flaming fires and Figure 8 for non-flaming fires. As is evident in Figure 7, the obscuration level at the time of response varies appreciably. The obscuration level at the time of response for the non-flaming fires indicated in Figure 8 is greater than that for the flaming tests and is even more varied. The mean and standard deviation of the obscuration level for the flaming and non-flaming fires is indicated in Table 3.

FPRF incipient fire project final report Volume 3

p. 12

14 12

Obscuration (%/ft)

10 8 6 4 2 0 0 5 10 Test # 15 20 ion1 ion2 photo

Figure 7. Obscuration levels at detector response, flaming fires

The temperature rise at which the ionization and photoelectric detectors responded is presented in Figure 9 for flaming fires and Figure 10 for non-flaming fires. The mean and standard deviation of the temperature rise for the flaming and non-flaming fires is indicated in Table 3. For many of the non-flaming tests, the temperature rise indicated is very small.
30 25

Obscuration (%/ft)

20 ion1 15 10 5 0 0 2 4 6 Test # 8 10 12 ion2 photo

Figure 8. Obscuration levels at detector response, non-flaming fires

FPRF incipient fire project final report Volume 3

p. 13

40 35

Temperature Rise (K)

30 25 20 15 10 5 0 0 5 10 Test # 15 20 ion1 ion2 photo

Figure 9. Temperature rise at detector response, flaming fires


40 35

Temperature Rise (K)

30 25 20 15 10 5 0 0 2 4 6 Test # 8 10 12 ion1 ion2 photo

Figure 10. Temperature rise at detector response, non-flaming fires Table 3. Detector response statistics for flaming and non-flaming fires, home smoke alarm project Ionization Flaming 3.04 3.78 5.99 4.09 Photoelectric Flaming 5.73 3.08 15.5 10.5

Parameter Obscuration (%/ft) Mean Standard Deviation Temperature Rise (K) Mean Standard Deviation

Non-flaming 14.1 8.24 7.49 9.72

Non-flaming 8.13 8.83 0.90 0.68

FPRF incipient fire project final report Volume 3

p. 14

Cumulative distributions of the obscuration level and temperature rise at the time of response are provided in Figures 11 to 14. The 80th percentile values of the smoke conditions when the detectors responded in the various tests are presented in Table 4. The 80th percentile values follow the general trends observed in previous studies. In flaming fires, ionization detectors responded when the obscuration level and temperature rise were less than those associated with the response of photoelectric detectors. In most tests in this experimental program this was accomplished by the ionization detectors responding prior to the photoelectric detectors in the flaming tests. Conversely, in the tests with the non-flaming fires, the photoelectric detectors responded when the obscuration level and temperature rise were less than those values when the ionization detectors responded. Again, in most tests the ionization detectors responded after the photoelectric detectors, and in some cases appeared to operate just as the temperature began to increase significantly, perhaps as the fire was transitioning to flaming combustion.
100

Cumulative Percentage of Detector Alarms

80

60

40 ion 20 photo

0 0 2 4 6 8 10 12 14 Obscuration (%/ft)

Figure 11. Cumulative distribution of obscuration at detector response, flaming fires

FPRF incipient fire project final report Volume 3

p. 15

100

Cumulative Percentage of Detector Alarms

80

60

40 ion 20 photo

0 0 5 10 15 Obscuration (%/ft) 20 25 30

Figure 12. Cumulative distribution of obscuration at detector response, non-flaming fires

100

Cumulative Percentage of Detector Alarms

80

60

40 ion 20 photo

0 0 5 10 15 20 25 30 35 40 Temperature Rise (K)

Figure 13. Cumulative distribution of temperature rise at detector response, flaming fires

FPRF incipient fire project final report Volume 3

p. 16

100

Cumulative Percentage of Detector Alarms

80

60

40

ion photo

20

0 0 5 10 15 20 25 30 35 Temperature Rise (K)

Figure 14. Cumulative distribution of temperature rise at detector response, non-flaming fires Table 4. 80th percentile values of parameters at detector response, Home Smoke Alarm Project Flaming Ionization 5.55 7.70 Non-flaming Ionization 18.5 14

Parameter Obscuration (%/ft) Temperature Rise (K)

Photoelectric 8.06 23.4

Photoelectric 12.1 1.4

3.4 Smoke Characterization Project The experiments in the Smoke Characterization Project were conducted by Underwriters Laboratories (UL). These tests were conducted in ULs Fire Test Room with dimensions of 11.0 6.7 3.1 m. No airflow was provided during the tests. Detectors responded in thirty-three tests to flaming and non-flaming fire sources. Ionization and photoelectric smoke detectors were included in the tests. Obscuration measurements were taken using the same white light source and photocell assembly used in standard UL 217 tests. The lamp and photocell were spaced 1.52 m apart. Temperature was measured with Inconelsheathed, Type K thermocouples located 5.4 m from the fire source and 0.15 m below the ceiling. A sonic anemometer 5.4 m from the fire source and 0.1 m below the ceiling was used to measure the gas velocity in the radial and orthogonal, horizontal directions. The obscuration levels at which the ionization and photoelectric detectors responded is presented in Figure 15 for flaming fires and Figure 16 for non-flaming fires. In this experimental program, the range of obscuration levels at the time of response is comparable among both sets of tests, ranging from less than 1 %/ft to almost 30 %/ft. The mean and standard deviation of the obscuration level for the flaming and non-flaming fires is indicated in Table 5.

FPRF incipient fire project final report Volume 3

p. 17

30.0 25.0

OBS (%/ft)

20.0 15.0 10.0 5.0 0.0 0 5 10 Test # 15 20 Ionization Photoelectric

Figure 15. Obscuration levels at detector response, flaming fires

30.0 25.0 20.0 15.0 10.0 5.0 0.0 0 5 10 Test # 15 20 Ionization Photoelectric

Figure 16. Obscuration levels at detector response, non-flaming fires The temperature rise at which the ionization and photoelectric detectors responded is presented in Figure 17 for flaming fires and Figure 18 for non-flaming fires. The mean and standard deviation of the temperature rise for the flaming and non-flaming fires is indicated in Table 5.

OBS (%/ft)

FPRF incipient fire project final report Volume 3

p. 18

12 10
Temperature Rise (K)

8 6 4 2 0 0 5 10 Test # 15 20 Ionization Photoelectric

Figure 17. Temperature rise at detector response, flaming fires

2.5

Temperature Rise (K)

2.0

1.5 Ionization 1.0 Photoelectric

0.5

0.0 0 5 10 Test # 15 20

Figure 18. Temperature rise at detector response, non-flaming fires

The velocity at which the ionization and photoelectric detectors responded is presented in Figure 19 for flaming fires and Figure 20 for non-flaming fires. In the case of the non-flaming fires, all of the velocities are less than the critical velocity of 0.13 m/s noted in NFPA 72. The mean and standard deviation of the velocity for the flaming and non-flaming fires is indicated in Table 5.

FPRF incipient fire project final report Volume 3

p. 19

0.30 0.25

Velocity (m/s)

0.20 0.15 0.10 0.05 0.00 0 5 10 Test # 15 20 Ionization Photoelectric

Figure 19. Velocity at detector response, flaming fires

0.14 0.12 0.10

Velocity (m/s)

0.08 0.06 0.04 0.02 0.00 0 5 10 Test # 15 20

Ionization Photoelectric

Figure 20. Velocity at detector response, non-flaming fires

FPRF incipient fire project final report Volume 3

p. 20

Table 5. Detector response statistics for flaming and non-flaming fires, Smoke Characterization Project Ionization Flaming 5.80 6.88 2.24 1.17 0.17 0.05 Photoelectric Flaming 4.74 2.56 4.30 2.29 0.18 0.04

Parameter Obscuration (%/ft) Mean Standard Deviation Temperature Rise (K) Mean Standard Deviation Velocity (m/s) Mean Standard Deviation

Non-flaming 4.52 3.78 1.15 0.17 0.08 0.03

Non-flaming 7.60 6.09 1.22 0.43 0.06 0.03

Cumulative distributions of the obscuration level, temperature rise and velocity at the time of detector response are provided in Figures 21 to 26.
100
Cumulative Percentage of Detector Alarms

80

60

40
Ionization Photoelectric

20

0 0 5 10 15 20 25 30 Obscuration (%/ft)

Figure 21. Cumulative distribution of obscuration at detector response, flaming fires

FPRF incipient fire project final report Volume 3

p. 21

100

Cumulative Probability of Detector Alarms

80

60

40

Ionization Photoelectric

20

0 0 5 10 15 Obscuration (%/ft) 20 25 30

Figure 22. Cumulative distribution of obscuration at detector response, non-flaming fires


100

Cumulative Percentage of Detector Alarms

80

60

40

Ionization Photoelectric

20

0 0 2 4 6 8 10 12 Temperature Rise (K)

Figure 23. Cumulative distribution of temperature rise at detector response, flaming fires

FPRF incipient fire project final report Volume 3

p. 22

100

Cumulative Probability of Detector Alarms

80

60

40 Ionization 20 Photoelectric

0 0.0 0.5 1.0 1.5 2.0 2.5 Temperature Rise (K)

Figure 24. Cumulative distribution of temperature rise at detector response, non-flaming fires The temperature rise at the time of response is less for the ionization detectors than the photoelectric detectors for the flaming fires. For the non-flaming fires, the temperature rise is less for the 70th percentile level and greater values for the ionization detectors for the nonflaming fires. The velocities at the time of response of the ionization detectors are less than that for the photoelectric detectors for flaming fires. The converse is true for the non-flaming fires.
100

Cumulative Percentage of Detector Alarms

80

60

40 Ionization Photoelectric 0 0.00

20

0.05

0.10

0.15 Velocity (m/s)

0.20

0.25

0.30

Figure 25. Cumulative distribution of velocity at detector response, flaming fires

FPRF incipient fire project final report Volume 3

p. 23

120

Cumulative Probability of Detector Alarms

100 80 60 40 Ionization 20 0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Velocity (m/s) Photoelectric

Figure 26. Cumulative distribution of velocity at detector response, non-flaming fires The 80th percentile values of the smoke conditions when the detectors responded in the tests in the Smoke Characterization Project are presented in Table 6. One of the interesting aspects of the results presented in Table 6 is the greater obscuration level noted at the 80th percentile value for the ionization detectors for flaming fires and for the photoelectric detectors for non-flaming fires. This trend is contrary to that found in previous studies that have shown ionization detectors to be generally more sensitive to flaming fires and photoelectric detectors for nonflaming fires. Consequently, the expectation was that the obscuration level at activation for ionization detectors should be less than that for photoelectric detectors for flaming fires, and the converse for non-flaming fires. Table 6. 80th percentile values of parameters at detector response, Smoke Characterization Project Flaming Ionization 8.57 3.15 0.20 Non-flaming Ionization 6.70 1.25 0.11

Parameter Obscuration (%/ft) Temperature Rise (K) Velocity (m/s)

Photoelectric 7.15 5.65 0.22

Photoelectric 7.05 1.55 0.09

The seemingly contrary results are attributed to the transient nature of some of the tests included in this experimental program. The shredded paper tests are the more notable tests with a significant transient nature. The obscuration measurements over the duration of one of the shredded paper tests are presented in Figure 27. In this test, the obscuration increases significantly in the early portion of the test prior to the test transitioning to flaming combustion. The ionization detector responds at 87 s when the obscuration is 18 %/ft, during the first peak in obscuration indicated in the figure. The photoelectric detector responds at 134 s, when the obscuration is 4.3 %/ft, i.e. after the second peak in the obscuration indicated in the figure.

FPRF incipient fire project final report Volume 3

p. 24

25

20

Obscuration (%/ft)

15

10

0 0 60 120 180 240 Time (s) 300 360 420 480

Figure 27. Light obscuration measurement in shredded paper test

3.5 Experimental program in this project The experimental protocol was described in Volume II. The characteristics of the environment in the vicinity of the detectors at the time of response are described in terms of the obscuration and temperature rise. Spot photoelectric smoke detectors from two manufacturers (SG and SS) were included in the room. The SS detector included an algorithm for detection, while the SG detector did not. Ionization detectors were included from one manufacturer (SG). The response of the detectors was collected on proprietary systems provided by each of the detection system manufacturers (SS and SG). The SG photoelectric detector was judged to operate when it reported an obscuration level of 2.5 %/ft. The response of the SS detector was established by the algorithm. The SG ionization detector operated at an obscuration level of 1.5 %/ft. Unventilated Room As with the Smoke Characterization Project, conditions in the vicinity of the detectors at the time of response were highly varied, in part due to the transient nature of some of the samples. Further, the response of multiple detectors of the same type at different locations was not always consistent, possibly because of the highly dependent nature of the smoke conditions on location. The average and standard deviation of the conditions at the time of response are provided in Table 7.

FPRF incipient fire project final report Volume 3 Table 7. Photoelectric detector response statistics for flaming and non-flaming fires Parameter Obscuration (%/ft) Mean Standard Deviation Temperature Rise (K) Mean Standard Deviation Flaming 8.80 6.82 1.00 1.13 Non-flaming 2.47 1.96 0.64 0.50

p. 25

The mean obscuration level of 2.47 %/ft is less than the nominal setting of the detectors. In some tests, counter-intuitive measurements were noted where the detectors recorded an obscuration of 2.5 %/ft prior to the light obscuration equipment outside the detector. The cumulative probability distributions of the characteristics of the environment at the time of response of the detectors for each of the tests are presented Figures 28 and 29. The 80th percentile values are presented in Table 8. The step-like nature of the temperature data is attributed to the temperature being reported in 1 K increments by the data acquisition system used.

100

Cumulative Percentage of Detector Alarms

80

60

40 flaming non-flaming 0 0 5 10 15 20 25 30 35 Obscuration (%/ft)

20

Figure 28. Cumulative distribution of obscuration at detector response, non-ventilated room

FPRF incipient fire project final report Volume 3

p. 26

100

Cumulative Percentage of Detector Alarms

80

60

40 flaming non-flaming 0 0.0 1.0 2.0 Temperature Rise (K) 3.0 4.0

20

Figure 29. Cumulative distribution of temperature rise at detector response, nonventilated room Table 8. 80th percentile values of parameters at detector response Parameter Obscuration (%/ft) Temperature Rise (K) Flaming 8.37 1.60 Non-flaming 2.97 1.00

FPRF incipient fire project final report Volume 3 Ventilated room

p. 27

The conditions at the time of detector response for the ventilated room are characterized in terms of obscuration and temperature measurements in the vicinity of the respective detectors. As an example of the obscuration and temperature rise observed for the tests, Figure 30 and 31 present the conditions at the time of response of the photoelectric detectors for the flaming fires in the ventilated room where no ventilation was provided.

25 SG Station 1 SG Station 2 15 SG Station 3 SG Station 4 10 SS Station 1 SS Station 2 SS Station 3 5 SS Station 4

20

Obscuration (%/ft)

0 73 74 75 76 85 86 87 88 Test #

Figure 30. Obscuration levels at detector response, ventilated room

6 5 SG Station 1 SG Station 2 4 3 2 1 0 73 74 75 76 85 86 87 88 Test # SG Station 3 SG Station 4 SS Station 1 SS Station 2 SS Station 3 SS Station 4

Figure 31. Temperature rise at detector response, ventilated room

Temperature Rise (K)

FPRF incipient fire project final report Volume 3

p. 28

As noted in the NIST study, the range of obscuration levels at the time of response indicated in Figure 30 is significant. The temperature rise being indicated in Figure 31 in 1 K increments is because the data was reported with that level of precision. The response times of the detectors did show a distinct bias based on detector location, with the detectors at stations 1 and 2 responding more quickly than the detectors at station 3 and 4. This is attributed primarily to stations 1 and 2 being closer to the fire location than stations 3 and 4. However, the obscuration and temperature levels do not indicate as clear a trend because of the transient nature of many of these tests. The average and standard deviation of the conditions at the time of response of the three types of detectors are included in Tables 9-11. Table 9. Detector response statistics for flaming and non-flaming fires, photoelectric detector SG No Ventilation Flaming Nonflaming 9.52 6.29 1.31 1.53 1.32 0.65 1.48 0.77 6 ACH Flaming 12 ACH Flaming

Parameter Obscuration (%/ft) Mean Standard Deviation Temperature Rise (K) Mean Standard Deviation

Nonflaming 0.28 0.25 1.33 0.66

Nonflaming 0.42 0.53 1.44 1.03

2.55 2.60 1.90 2.05

1.90 2.47 1.53 2.09

Table 10. Detector response statistics for flaming and non-flaming fires, photoelectric detector SS No Ventilation Flaming Nonflaming 6.88 5.01 1.09 1.22 0.94 0.69 1.25 0.79 6 ACH Flaming 12 ACH Flaming

Parameter Obscuration (%/ft) Mean Standard Deviation Temperature Rise (K) Mean Standard Deviation

Nonflaming 0.30 0.28 1.24 0.72

Nonflaming 0.38 0.42 1.25 0.80

2.72 3.55 1.27 1.26

1.28 1.40 1.10 1.21

The cumulative probability distributions of the obscuration level and temperature rise at the time of response of the SG photoelectric detector for the three airflows are presented in Figures 32 to 35. The 80th percentile values are included in Table 12.

FPRF incipient fire project final report Volume 3

p. 29

Table 11. Detector response statistics for flaming and non-flaming fires, ionization detector SG No Ventilation Flaming Nonflaming 7.18 4.32 0.75 0.68 2.33 2.31 1.33 0.52 6 ACH Flaming 12 ACH Flaming

Parameter Obscuration (%/ft) Mean Standard Deviation Temperature Rise (K) Mean Standard Deviation

Nonflaming 0.51 0.88 0.67 0.58

Nonflaming NR NR NR NR

4.18 3.89 0.75 0.46

10.1 1.89 2.00 0.00

NR = no ionization detectors responded for the non-flaming fires with 12 ACH.

100

Cumulative Percentage of Detector Alarms

80

60

40

0 ACH 6 ACH 12 ACH

20

0 0 5 10 15 20 25 Obscuration (%/ft)

Figure 32. Cumulative distribution of obscuration at photoelectric detector response, flaming fires, ventilated room

As in the research by Geiman, et al., the obscuration level at the time of detector response is less for the tests where ventilation was provided than in the cases where ventilation was not provided. This trend was observed in both the flaming and non-flaming tests. A similar trend in temperature rise was not evident.

FPRF incipient fire project final report Volume 3

p. 30

100

Cumulative Percentage of Detector Alarms

80

60 0 ACH 6 ACH 12 ACH 20

40

0 0 1 2 3 4 5 6 7 Temperature Rise (K)

Figure 33. Cumulative distribution of temperature rise at photoelectric detector response, flaming fires, ventilated room
100

Cumulative Percentage of Detector Alarms

80

60

0 ACH 6 ACH 12 ACH

40

20

0 0 0.5 1 1.5 2 2.5 3 Obscuration (%/ft)

Figure 34. Cumulative distribution of obscuration at photoelectric detector response, nonflaming fires, ventilated room

FPRF incipient fire project final report Volume 3

p. 31

100

Cumulative Percentage of Detector Alarms

80

60

40 0 ACH 20 6 ACH 12 ACH

0 0 1 2 3 4 5 Temperature Rise (K)

Figure 35. Cumulative distribution of temperature rise at photoelectric detector response, non-flaming fires, ventilated room Table 12. 80th percentile values of parameters at detector response, photoelectric detector SG No Ventilation Flaming Nonflaming 12.9 1.58 3.00 2.00 6 ACH Flaming 4.28 4.00 12 ACH Flaming 4.92 3.00

Parameter Obscuration (%/ft) Temperature Rise (K)

Nonflaming 0.47 2.00

Nonflaming 0.95 2.00

The cumulative probability distributions of the obscuration level and temperature rise at the time of response of the SG photoelectric detector for the three airflows are presented in Figures 36 to 39. The 80th percentile values are included in Table 12. As in the case of the photoelectric detectors, the obscuration levels at the time of response of the ionization detectors for the cases with 6 ACH of forced ventilation are less than those for the cases without. The tests with 12 ACH only included two tests where the ionization detectors responded (both to flaming fires) and thus are too small a sample to draw any significant conclusions from.

FPRF incipient fire project final report Volume 3

p. 32

100

Cumulative Percentage of Detector Alarms

80

60

0 ACH 6 ACH 12 ACH

40

20

0 0 3 6 9 12 15 Obscuration (%/ft)

Figure 36. Cumulative distribution of obscuration at ionization detector response, flaming fires, ventilated room
100

Cumulative Percentage of Detector Alarms

80

60

0 ACH 6 ACH 12 ACH

40

20

0 0.0 0.5 1.0 1.5 2.0 2.5 Temperature Rise (K)

Figure 37. Cumulative distribution of temperature rise at ionization detector response, flaming fires, ventilated room

FPRF incipient fire project final report Volume 3

p. 33

100

Cumulative Percentage of Detector Alarms

80

60 0 ACH 40 6 ACH

20

0 0 1 2 3 Obscuration (%/ft) 4 5 6

Figure 38. Cumulative distribution of obscuration at ionization detector response, nonflaming fires, ventilated room
100

Cumulative Percentage of Detector Alarms

80

60

40

0 ACH 6 ACH

20

0 0.0 0.5 1.0 1.5 2.0 2.5 Temperature Rise (K)

Figure 39. Cumulative distribution of temperature rise at ionization detector response, non-flaming fires, ventilated room Table 13. 80th percentile values of parameters at detector response, ionization detector SG No Ventilation Flaming Nonflaming 10.7 4.35 1.0 1.0 6 ACH Flaming 7.96 1.0 12 ACH Flaming 10.3 2.0

Parameter Obscuration (%/ft) Temperature Rise (K)

Nonflaming 0.61 2.0

Nonflaming NR NR

NR = no ionization detectors responded for the non-flaming fires with 12 ACH.

FPRF incipient fire project final report Volume 3 4. Discussion

p. 34

The series of experimental programs include a sufficiently wide variety of spaces, fuels and ventilation conditions to form a substantial basis for the development of robust, simple guidelines for estimating smoke detector response. Unfortunately, the smoke detector responses appear to be strongly dependent on the specific characteristics of the smoke and in some cases on the detector technology. As such, proposing a single set of guidelines for obscuration, temperature rise and velocity which can apply to a wide range of applications is very difficult, other than suggesting guidelines which would be very conservative in some applications. 4.1 Trends in detector response The 80th percentile obscuration levels measured at the time of smoke detector response in all of the previous experimental programs described in section 4 are illustrated in Figure 40. The variation of the obscuration levels in the experimental programs is provided in Figure 41 to explore any lab-related differences in the 80th percentile values of the obscuration at the time of smoke detector response. The compilation of experimental programs reviewed in this report used a variety of fuels and were conducted in spaces with different dimensions. Nonetheless, the 80th percentile values for the obscuration level at smoke detector response were relatively similar among most of the experimental programs, with the principal exception being that associated with the response of ionization detectors in non-flaming tests conducted by NIST.
20 18 16 Kemano NRL NIST UL UMD-UL(1) UMD-UL(2)

Obscuration (%/ft)

14 12 10 8 6 4 2 0 Ion-FL Photo- Ion-NF Photo- Ion-FL- Ion-FL- Photo- Photo- IonFL NF 6 ACH 12 FL-6 FL-12 NF-6 ACH ACH ACH ACH

Ion- Photo- PhotoNF-12 NF-6 NF-12 ACH ACH ACH

Figure 40. 80th percentile obscuration level at detector response

FPRF incipient fire project final report Volume 3

p. 35

20 18 16 Ion-FL Photo-FL Ion-NF Photo-NF

Obscuration (%/ft)

14 12 10 8 6 4 2 0 Kemano NRL NIST UL

UMD-UL(1)

UMD-UL(2)

Figure 41. 80th percentile obscuration level at detector response in recent experimental programs

For flaming fires, the obscuration level in tests without forced ventilation ranged from 1.4 to 10.7 %/ft for ionization detectors and 2.7 to 12.9 %/ft for photoelectric detectors. Given the noted range in the 80th percentile values of obscuration, a guideline for flaming fires which reasonably captures much of the data for smoke detectors of either type of technology is 8 %/ft. In the case of flaming fires in ventilated rooms, the 80th percentile values of the obscuration levels differs appreciably for the two detection technologies. For flaming fires with ventilation, the 80th percentile values of the obscuration level for photoelectric smoke detectors were 4.3 to 4.9 %/ft. In contrast, the 80th percentile values of the obscuration level for ionization smoke detectors were 8.0 to 10.3 %/ft (though the 10.3% is based on only two tests being conducted at a forced ventilation rate of 12 ACH). As such, a possible guideline of obscuration levels for photoelectric detectors could be 5 %/ft for ventilation rates ranging from 6 to 12 ACH. For ionization detectors, the 8 %/ft value appears appropriate as a guideline considering only the results from the tests with 6 ACH, given the limited data at the higher ventilation rate. For non-flaming fires without ventilation, the 80th percentile values of the obscuration levels ranged from 4.4 to 18.5 %/ft for ionization smoke detectors and 1.6 to 12.1 %/ft for photoelectric smoke detectors. Excluding the NIST study, suggested guidelines for obscuration level could be 12 %/ft for ionization detectors and 10 %/ft for photoelectric detectors. The 80th percentile values of the obscuration levels for non-flaming fires with ventilation were all less than 1 %/ft in this study and almost 5 %/ft for ionization detectors in the NRL study. Given the limited data in this area, a recommendation for establishing a guideline of only 1 %/ft is questionable, especially given the difference in results from this study and the NRL study. Until further data is obtained, a value in excess of 1 %/ft is recommended and should perhaps be as large as 2.5 %/ft.

FPRF incipient fire project final report Volume 3

p. 36

As a benchmark to appreciate the smoke density associated with the various obscuration levels, the relationship of obscuration level to visibility is presented in Eq. 12.
V 0.43 K log 1 OBS

(12)
100

Where V is the visibility in ft, OBS is the obscuration level in %/ft and K is a constant related to the object being viewed, e.g. has a suggested value of approximately 4 for illuminated exit signs [Jin, 2002]. The relationship of obscuration level and visibility is illustrated in Figure 42.
200 175 150

Visibility (ft)

125 100 75 50 25 0 0 4 8 12 16 20 Obscuration (%/ft)

Figure 42. Visibility to illuminated exit sign through smoke The 80th percentile values of the temperature rise measured at the time of smoke detector response on the experimental programs described in the previous section are illustrated in Figure 43. The variation of the 80th percentile temperature rise obtained in the experimental programs is provided in Figure 44. In general, the data included from the UL smoke characterization project and the current projects have the smallest values of the 80th percentile values of temperature rise. The UL and current project data indicates that smoke detector response occurs when the temperature rise is 5 K or less. The data from the current project suggests that smoke detectors respond when the temperature rise is very modest (on the order of 1 to 2 K).

FPRF incipient fire project final report Volume 3

p. 37

25 NRL

Temperature Rise (K)

20

NIST UL UMD-UL(1) UMD-UL(2)

15 10

0 Ion-FL Photo- Ion-NF Photo- IonIon- Photo- Photo- Ion- Photo- PhotoFL NF FL-6 FL-12 FL-6 FL-12 NF-6 FL-6 FL-12 ACH ACH ACH ACH ACH ACH ACH

Figure 43. 80th percentile temperature rise at detector response

The temperature rise at the time of detection response for flaming fires with no forced ventilation is highly dependent on the detection technology. A temperature rise of approximately 5 K can be suggested as a reasonable conservative guideline for ionization detectors. However, based on the measurements obtained in the NRL and NIST tests, a temperature rise guideline greater than 5 K, e.g. 15 K, could be justified. For non-flaming fires and all fires with forced ventilation a temperature rise of approximately 3 K appears to be a reasonable guideline to estimate smoke detector response of either technology.
25 Ion-FL 20 Photo-FL Ion-NF Photo-NF 15

Temperature Rise (K)

10

0 NRL NIST UL UMD-UL(1) UMD-UL(2)

Figure 44. 80th percentile obscuration level at detector response in recent experimental programs

FPRF incipient fire project final report Volume 3 4.2 Variation in obscuration levels at detector response

p. 38

Estimates of the time for smoke detector response using the guidelines suggested in the previous section may yield substantial errors. In most cases, the error would be conservative. However, some of the estimates would not be conservative being based on 80th percentile values (in one test, an obscuration level of 31 %/ft was measured at the time of detector response). In addition, 80th percentile values determined from some experimental programs (e.g. NIST) were appreciably greater than the values noted in other programs. Another issue affecting the accuracy of estimates from the guidelines suggested in the previous relationship is the tenuous relationship between light scattering and light obscuration. The detector sensitivity was set at 1.5 %/ft for ionization detectors and 2.5 %/ft for photoelectric detectors, e.g. the obscuration levels for smoke inside the sensing chamber associated with the response of the detectors. The smoke detector sensitivities are established through tests conducted in the UL 217 Sensitivity Test Smoke Box with a single smoke source which produced a light gray smoke. From this test, a relationship is developed by manufacturers to relate light obscuration to light scattering (light scattering is measured implicitly by noting the detector output signal. With the detector sensitivities set at a particular obscuration, the expectation is that the detectors should not alarm until the obscuration outside the detector reaches at least that level. However, the mean obscuration level for the photoelectric detectors at the time of alarm was less than 2.5 %/ft in all test conditions involving non-flaming fires (see Tables 7, 9 and 10), being as low as 0.28 %/ft for the tests with 6 ACH of forced ventilation. The mean obscuration level at the time of response for ionization detectors in the case with 6 ACH of forced ventilation was less than the 1.5 %/ft sensitivity setting (see Table 11). In the cases without forced ventilation, the obscuration level at the time of detector response noted in the experiments with flaming fires conducted in this study were less than 1 %/ft. These results indicate that the light obscuration measurements collected in this experimental program are not a good predictor of smoke detector performance. Measurements from the air sampling detectors may be able to provide further insight into this issue. Looking forward, there are several possible reasons which can be suggested for this apparent discrepancy in the measurements noted above. Smoke detector sensitivity analysis Light scattering or ionization chamber responses are not directly related to light obscuration, as described in section 1.1 of this report. The smoke detector sensitivity of 1.5 %/ft for ionization smoke detectors and 2.5 %/ft for photoelectric smoke detectors are established through tests conducted in the UL 217 Sensitivity Test Smoke Box with a single smoke source which produces a light gray smoke. From this test, a relationship is developed by manufacturers to relate light obscuration to light scattering (light scattering is measured implicitly by noting the detector output signal). However, this relationship is only relevant to a particular smoke with a specific characteristics relating to color, particle size distribution, etc., as well as the wavelength

FPRF incipient fire project final report Volume 3

p. 39

of light used to make the measurement. As the detector is exposed to smokes with different characteristics, that relationship breaks down. This situation could be improved by conducting tests in the UL 217 Sensitivity Test Smoke Box which include a variety of smokes to obtain a relationship between light scattering to light obscuration for a greater range of applications. From tests of this nature, multiple relationships for light obscuration and light scattering could be generated to correspond to a variety of smokes. Guidelines could perhaps be developed which yield maximum, minimum and nominal values of light obscuration at the time of response for the various smokes. Ambient light In the tests conducted in the ventilated room, fluorescent lights were left on during the tests. The photocell used for the light obscuration measurements was not shielded from ambient light. As such, the photocell could have received additional scattered light from smoke particles, thereby yielding a lesser smoke obscuration measurement. If subsequent experiments are conducted with ambient light, the photocell should be shielded as much as possible from ambient light. 4.3 Velocity The use of ceiling jet velocity as a guideline to indicate smoke detector response has been described in previous efforts and in NFPA 72. The radial velocity measurements in the ventilated room for the shredded paper tests for the three ventilation conditions (none from test 25, 6 ACH from test 49 and 12 ACH from test 74) are presented in Figure 45. The velocities included in the figure are a 10 point moving average over a 1.0 sec period. The graphs begin one minute prior to the start of the tests to indicate the airflows associated solely with the forced ventilation. For the test without forced ventilation, the change in velocity at approximately 60 sec after the beginning of the test in the room is apparent with a peak near-ceiling velocity of approximately 0.3 m/s being recorded. However, similar changes in velocity are not evident in the two tests with forced ventilation. The velocities associated with the forced ventilation are approximately 0.6 and 1.4 m/s, which are appreciably greater than that associated with the near-ceiling smoke flow. Being the data depicted in Figure 45 is from one of the flaming tests, any ceiling jet velocity for the tests with a non-flaming source would be expected to be appreciably smaller. Consequently, the use of near-ceiling velocities as a means of indicating smoke detector response in rooms with forced ventilation cannot be recommended.

FPRF incipient fire project final report Volume 3

p. 40

1.6 1.4 1.2

Velocity (m/s)

1 0.8 0.6 0.4 0.2 0 -60 0 60 120 180 240 Time (sec) 300 360 420 480 540

0 ACH 6 ACH 12 ACH

Figure 45. Near ceiling velocities in ventilated room tests

4.4 Dual parameter guidelines While the use of light obscuration as a means of estimating smoke detector response has limitations, the limitations may be moderated by examining light obscuration together with another parameter. This strategy was pursued for the tests conducted in the non-ventilated room. The pairs of parameters (light obscuration-velocity, light obscuration-temperature rise and temperature rise-velocity) at the time of response are depicted in Figures 46-48.
10 9 8

Obscuration (%/ft)

7 6 5 4 3 2 1 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 Velocity (m/s) flaming non-flaming

Figure 46. Light obscuration and velocity at photoelectric detector response, unventilated room

FPRF incipient fire project final report Volume 3

p. 41

10 9 8

Obscuration (%/ft)

7 6 5 4 3 2 1 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Temperature Rise (K) flaming non-flaming

Figure 47. Light obscuration and temperature rise at photoelectric detector response, unventilated room

0.35 0.3 0.25

Velocity (m/s)

0.2 0.15 0.1 flaming 0.05 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Temperature Rise (K) non-flaming

Figure 48. Temperature rise and velocity at photoelectric detector response, unventilated room

The parameter pair of obscuration and velocity results in the best cluster of data for both flaming and non-flaming fires. For non-flaming fires, detectors responded within a region with obscuration levels between 1.5 and 2.5 %/ft and velocities between 0.03 and 0.07 m/s. For flaming fires, the region for response is bounded by obscuration levels from 5.5 to 9.5 %/ft and

FPRF incipient fire project final report Volume 3

p. 42

velocities between 0.14 and 0.33 m/s (though the test with the 9.5 %/ft obscuration is appreciably greater than the obscuration measured at detection for the other tests and 0.33 m/s is an appreciably greater velocity than in the other tests). Thus, guidelines for the response of the photoelectric detectors that relates to all tests in the non-ventilated room could include a light obscuration of 2.5 %/ft and velocity of 0.07 m/s for non-flaming fires. For flaming fires, the guideline with both parameters could consist of 8.0 %/ft and 0.25 m/s.

5. Summary An appreciable variation of smoke conditions have been noted at the time of response of smoke detectors for experiments conducted as part of this study, as well as in other recent efforts. While guidelines can be suggested of obscuration level or temperature rise, these are very approximate in nature and may involve appreciable errors, which may not be conservative in all cases. Much of the reason for the errors is based on the fact that light obscuration and temperature are not currently related to the operating mechanisms of current smoke detector technologies, i.e. light scattering and ionization. Improved relationships of light scattering and light obscuration could be compiled through additional sensitivity experiments which expose to detectors to several smokes, rather than just a single gray smoke. In rooms with forced ventilation of 6 to 12 ACH, the air velocities associated with the ventilation are significantly greater than the ceiling jet velocities associated with small fires. As such, velocity is only relevant as a guideline in rooms without forced ventilation.

6. References Alpert, R., 1972, Ceiling Jets, Fire Technology, 8, 181. Borzovski, E., 1989, A Preliminary Approach to Siting Smoke Detectors Based on Design Fire Size and Detector Aerosol Lag Time, MS Thesis, Worcester Polytechnic Institute, Worcester, MA. Richard W. Bukowski, Richard D. Peacock, Jason D. Averill, Thomas G. Cleary, Nelson P. Bryner, William D. Walton, Paul A. Reneke and Erica D. Kuligowski, 2008, Performance of Home Smoke Alarms Analysis of the Response of Several Available Technologies in Residential Fire Settings, NIST Technical Note 1455-1, National Institute of Standards and Technology, Gaithersburg, MD. Thomas Cleary, Artur Chernovsky, William Grosshandler and Melissa Anderson, 2000, Particular Entry Lag in Spot-Type Smoke Detectors, Proceedings of the 6th Symposium, International Association of Fire Safety Science, Boston, MA. CFAST Consolidated Model of Fire Growth and Smoke Transport (Version 6) Technical Reference Guide, NIST Special Publication 1026, National Institute of Standards and Technology, Gaithersburg, MD.

FPRF incipient fire project final report Volume 3

p. 43

Geiman, Justin A., 2003, Evaluation of Smoke Detector Response Estimation Methods, MS Thesis, University of Maryland, College Park, MD. Geiman, J.A., Gottuk, D.T., and Milke, J.A., 2004, Evaluation of Smoke Detector Response Estimation Methods, NFPRF Fire Detection and Suppression Symposium, Orlando, FL. Geiman, J.A., Gottuk, D.T., and Milke, J.A., 2006, Evaluation of Smoke Detector Response Estimation Methods: Optical Density, Temperature Rise and Velocity at Alarm, J. of Fire Protection Engineering, 16, 4. D.T. Gottuk, S.A. Hill, C.F. Schemel, B.D. Strehlen, S.L. Rose-Pehrsson, R.E. Shaffer, P.A. Tatem and F.W. Williams, Identification of Fire Signatures for Shipboard Multi-criteria Fire Detection Systems, Naval Research Laboratory, Memorandum Report 6180-99-8386, Washington, DC. Heskestad, Gunnar, 1975, Generalized Characteristics of Smoke Entry and Response for Products of Combustion Detectors, Proceedings, 7th International Conference on Problems of Automatic Fire Detection, Rheinisch-Westaflischen Tehnischen Hochschule, Aachen, Germany. Heskestad, Gunnar and Delichatsios, Michael A., 1977 Environments of Fire Detectors Phase 1: Effect of Fire Size, Ceiling Height and Material, Measurements Vol I (NBS-GCR-7786), Analysis Vol II (NBS-GCR-77-95). National Technical Information Service (NTIS), Springfield, VA. Jin, Tadahisa, 2002, Visibility and Human Behavior in Fire Smoke, SFPE Handbook of Fire Protection Engineering, 3rd Ed., P.J. DiNenno, Editor, NFPA, Quincy, MA. Walter W. Jones, Richard D. Peacock, Glenn P. Forney and Paul A. Reneke, 2005, Fabian, T. Z., Gandhi, P.D., Patty, P. E., Chapin, J.T., 2007, Smoke Characterization Project: Final Report, Underwriters Laboratories Inc., Northbrook, IL. Kevin McGrattan, Simo Hostikka, Jason Floyd, Howard Baum, Ronald Rehm, William Mell and Randall McDermott, 2008, Fire Dynamics Simulator (Version 5) Technical Reference Guide, NIST Special Publication 1018-5, National Institute of Standards and Technology, Gaithersburg, MD. Kevin McGrattan, Bryan Klein, Simo Hostikka and Jason Floyd, 2008a, Fire Dynamics Simulator (Version 5) Users Guide, NIST Special Publication 1019-5, National Institute of Standards and Technology, Gaithersburg, MD. C.D. Litton, K.R. Smith, R. Edwards and T. Allen, 2004, Combined Optical and Ionization Techniques for Inexpensive Characterization of Micrometer and Submicrometer Aerosols, J. Sci. and Tech., 38 1054.

FPRF incipient fire project final report Volume 3

p. 44

NFPA 72, National Fire Alarm Code, 2007, National Fire Protection Association, Quincy, MA. R.J. Roby, S.J. Olenick, W. Zhang, D.J. Carpenter, M.S. Klassen, and J.L. Torero. A Smoke Detector Algorithm for Large Eddy Simulation Modeling. NIST GCR 07-911, National Institute of Standards and Technology, Gaithersburg, Maryland, 2007 S.L. Rose-Pehrsson, R.E. Shaffer, S. Hart, F.W. Williams, D.T. Gottuk, B.D. Strehlen and S.A. Hill, Multi-Criteria Detection Systems Using a Probabilistic Neural Network, Sensors and Actuators, B 69, 325-335. Schifiliti, Robert P., Meacham, Brian J., and Custer, Richard L.P., 2002, Design of Detection Systems, SFPE Handbook of Fire Protection Engineering, 3rd Ed., National Fire Protection Association, Quincy MA. Schifiliti Robert P. and Pucci, William E., 1996, Fire Detection Modeling, State of the Art, Fire Detection Institute. J.Z. Su, G.P. Crampton, D.W, Carpenter, C. McCartney, C. P.and Leroux, 2003, Kemano Fire Studies - Part 1: Response of Residential Smoke Detectors, Research Report, Institute for Research in Construction, National Research Council Canada, 108. J.T., Wong,, D.T. Gottuk, S.L. Rose-Pehrsson, R.E. Shaffer, S. Hart, P.A. Tatem and F.W. Williams, 2000, Results of Multi-Criteria Fire Detection System Tests, Naval Research Laboratory, Washington, DC.

FPRF incipient fire project final report Volume 3 Appendix. Conditions at the time of response for Ventilated Room

p. 45

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0 5 10 15 20 25 Obscuration (%/ft)

Figure A-1. Cumulative distribution of obscuration at detector response, flaming fires, ventilated room, no ventilation

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0 1 2 3 4 5 Temperature Rise (K)

Figure A-2. Cumulative distribution of temperature rise at detector response, flaming fires, ventilated room, no ventilation

FPRF incipient fire project final report Volume 3

p. 46

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 Obscuration (%/ft)

Figure A-3. Cumulative distribution of obscuration at detector response, non-flaming fires, ventilated room, no ventilation

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Temperature Rise (K)

Figure A-4. Cumulative distribution of temperature rise at detector response, non- flaming fires, ventilated room, no ventilation

FPRF incipient fire project final report Volume 3

p. 47

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0 2 4 6 8 10 12 14 Obscuration (%/ft)

Figure A-5. Cumulative distribution of obscuration at detector response, flaming fires, ventilated room, 6 air changes per hour

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0 1 2 3 4 5 6 Temperature Rise (K)

Figure A-6. Cumulative distribution of temperature rise at detector response, flaming fires, ventilated room, 6 air changes per hour

FPRF incipient fire project final report Volume 3

p. 48

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0.0 0.5 1.0 Obscuration (%/ft) 1.5 2.0

Figure A-7. Cumulative distribution of obscuration at detector response, non-flaming fires, ventilated room, 6 air changes per hour

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0.0 0.5 1.0 1.5 2.0 2.5 Temperature Rise (K)

Figure A-8. Cumulative distribution of temperature rise at detector response, non-flaming fires, ventilated room, 6 air changes per hour

FPRF incipient fire project final report Volume 3

p. 49

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0 2 4 6 8 10 12 Obscuration (%/ft)

Figure A-9. Cumulative distribution of obscuration at detector response, flaming fires, ventilated room, 12 air changes per hour

100

Cumulative Percentage of Detector Alarms

80

60

SG-Photo SS-Photo SG-Ion

40

20

0 0 1 2 3 4 5 6 Temperature Rise (K)

Figure A-10. Cumulative distribution of temperature rise at detector response, flaming fires, ventilated room, 12 air changes per hour

FPRF incipient fire project final report Volume 3

p. 50

100

Cumulative Percentage of Detector Alarms

80

60 SG-Photo 40 SS-Photo

20

0 0.00

0.25

0.50

0.75

1.00

1.25

1.50

Obscuration (%/ft)

Figure A-11. Cumulative distribution of obscuration at detector response, non-flaming fires, ventilated room, 12 air changes per hour

100

Cumulative Percentage of Detector Alarms

80

60 SG-Photo 40 SS-Photo

20

0 0 1 2 Temperature Rise (K) 3 4

Figure A-12. Cumulative distribution of temperature rise at detector response, non-flaming fires, ventilated room, 12 air changes per hour

Das könnte Ihnen auch gefallen