Sprinkler Protection for Cloud Ceilings Phase 2: Small Area Clouds

Transcription

1 Sprinkler Protection for Cloud Ceilings Phase 2: Small Area Clouds Final Report Prepared by: Dr. Jason Floyd Steve Strege Matt Benfer Hughes Associates, Inc. Baltimore, MD August 214 Fire Protection Research Foundation THE FIRE PROTECTION RESEARCH FOUNDATION ONE BATTERYMARCH PARK QUINCY, MASSACHUSETTS, U.S.A WEB:

2 Page ii

3 FOREWORD Cloud ceilings are ceiling panels that sit beneath the structural ceiling of a room or space and are seen increasingly in commercial and industrial buildings. Cloud panels range in area from discrete ceiling panels with large spaces in between, to close-to-full-room-area contiguous coverage with small gaps at the perimeter wall location. NFPA 13, Standard for the Installation of Sprinkler Systems, does not have definitive guidance on automatic sprinkler installation requirements for these ceilings and in some conditions requires sprinklers at both the structural ceiling and cloud ceiling panel elevations. Recent NFPA 13 change proposals were rejected based on a lack of validation of modeling results. The Fire Protection Research Foundation initiated this project to obtain an understanding of how cloud ceiling panels impact sprinkler actuation thresholds with an overall goal to provide the technical basis for sprinkler installation requirements. A Phase 1 study investigated the effectiveness of sprinklers on large area clouds. Phase 2 of this work, which is covered in this report, focused on developing guidance for sprinkler installation requirements for small area clouds by determining the maximum gap size between the wall and cloud edge at which ceiling sprinklers are not effective. The Research Foundation expresses gratitude to the report authors Dr. Jason Floyd, Steve Strege, and Matt Benfer, who are with Hughes Associates, Inc. located in Baltimore, MD. The Research Foundation appreciates the guidance provided by the Project Technical Panelists, the funding provided by the project sponsors, and all others that contributed to this research effort. The content, opinions and conclusions contained in this report are solely those of the authors. About the Fire Protection Research Foundation The Fire Protection Research Foundation plans, manages, and communicates research on a broad range of fire safety issues in collaboration with scientists and laboratories around the world. The Foundation is an affiliate of NFPA. About the National Fire Protection Association (NFPA) NFPA is a worldwide leader in fire, electrical, building, and life safety. The mission of the international nonprofit organization founded in 1896 is to reduce the worldwide burden of fire and other hazards on the quality of life by providing and advocating consensus codes and standards, research, training, and education. NFPA develops more than 3 codes and standards to minimize the possibility and effects of fire and other hazards. All NFPA codes and standards can be viewed at no cost at Keywords: automatic sprinkler systems, cloud ceilings, automatic sprinkler installation Page iii

4 Page iv

5 PROJECT TECHNICAL PANEL Jarrod Alston, Arup Melissa Avila, Tyco Fire Protection Products Bob Caputo, Fire and Life Safety America Dave Fuller, FM Global Dave Lowrey, City of Boulder Fire Rescue Jamie Lord, ATF Fire Research Laboratory Steven Scandaliato, SDG LLC Karl Wiegand, Global Fire Sprinkler Corporation Matt Klaus, NFPA Staff Liaison PROJECT SPONSORS American Fire Sprinkler Association National Fire Sprinkler Association The Reliable Automatic Sprinkler Company Viking Corporation Page v

6 Page vi

7 Sprinkler Protection for Cloud Ceilings, Prepared for Amanda Kimball National Fire Protection Research Foundation 1 Batterymarch Park Quincy, MA 2169 Prepared by Dr. Jason Floyd Steve Strege Matt Benfer Hughes Associates, Inc. 361 Commerce Dr., Suite 817 Baltimore, MD July 31, 214 FIRE SCIENCE & ENGINEERING

8 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE ii TABLE OF CONTENTS 1. BACKGROUND TASK 1: EXPERIMENTAL PROGRAM AND RESULTS Experimental Setup Cloud Array Instrumentation Test Matrix Experimental Procedure Experimental Results FDS Modeling of Experiments Grid Study Results of Experimental Simulations Task 1 Summary TASK 2: NUMERICAL MODELING OF CLOUD CEILING CONFIGURATIONS Methodology FDS Model Performance Criteria Analysis Approach Results and Analysis First Pass Results Second Pass Results Third Pass Results Fourth Pass Results Summary of Simulations and Development of Installation Guidance SUMMARY Summary of Task 1 and Task Limitations of Study REFERENCES... 2 APPENDIX A Experimental Average Temperature Data APPENDIX B Operated Sprinkler Heads for FDS Simulations B1 Corner Fires B1.1 8 ft Ceiling Height B ft Ceiling Height... 4 B1.3 2 ft Ceiling Height... 4 B2 Cross Fires... 9 B2.1 8 ft Ceiling Height... 9 B ft Ceiling Height... 6

9 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 1 1. BACKGROUND Cloud ceilings are increasingly seen in commercial and industrial facilities. The ceilings consist of ceiling panels separated by gaps that are suspended beneath the structural ceiling. Designs for cloud ceilings can vary greatly in terms of the shape and size of the panels, the gaps between panels, and the spacing between the panels and the structural ceiling. The use of cloud ceilings presents challenges for sprinkler protection that are not definitively addressed in NFPA 13. These challenges result from 1) heat from the fire plume entering the gaps between the panels and rising to the structural ceiling which may prevent sprinklers below the clouds from activating and 2) that sprinklers above the clouds may have their spray distribution blocked by the clouds. As a result, in some conditions the code would require sprinklers both below the clouds and at the structural ceiling. A prior study [1] investigated the effectiveness of sprinklers on large area clouds. A large area cloud was defined as a cloud whose extents were large enough to require at least one sprinkler per cloud. A combination of full scale testing and CFD fire modeling was used to examine the effectiveness of sprinklers on large area clouds in order to determine conditions where only sprinklers on the undersides of clouds would be required. The study concluded that where the clouds are level and co-planar, sprinklers can be omitted on the structural ceiling if: The gap between a wall and any cloud is less than or equal to 1 inch of gap per foot of ceiling height, or The gap between any two adjacent clouds is less than or equal to 1 ¼ inch of gap per foot of ceiling height. The study also made a number of recommendations including the following recommendation If clouds are small enough (or have a large enough aspect ratio) that at least one sprinkler per cloud is not required based upon the listed sprinkler spacing, then a ceiling jet might encounter additional gaps between clouds. Depending upon the gap size and cloud size, the ceiling jet may not have the strength (e.g. velocity) to jump the gap in order to reach a sprinkler. Conditions under which only below cloud sprinklers would be allowed for small area clouds are likely to be much more limited than for large area clouds. A study of similar effort to this study is recommended. This report documents efforts to address the above recommendation. Specifically this project has two tasks: 1. Selected fire dynamics modeling of cloud ceiling configurations, exploring the impact of cloud and ceiling height, plenum height, gap distances, fire growth rates, and fire locations on sprinkler actuation time and temperatures at the cloud and structural ceiling levels. Configurations of cloud ceilings will include multiple clouds with a range of gap distances between clouds as well as between clouds and walls. 2. Recommendations for appropriate sprinkler installation criteria for cloud ceilings constructed with smaller clouds based on these results. To accomplish the tasks above a two task work plan was proposed for the project and accepted by the technical panel. Task 1 of the plan is a short experimental program using the Hughes movable ceiling apparatus. The primary goal of the experimental program is to collect data on fire plume interactions with small clouds in order to develop appropriate CFD model inputs for Task 2. Task 2 of the plan is to execute a matrix of simulations for the variables in Task 1 above and use those results to develop the recommendations for Task 2 above. The remainder of this report documents Task 1 and Task 2 of the work plan.

10 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 2 2. TASK 1: EXPERIMENTAL PROGRAM AND RESULTS 2.1. Experimental Setup Cloud Array A square array of nine 2 ft x 2 ft cloud panels was constructed with the panels arranged in a 3 x 3 grid. The panels were.2 inch thick gypsum drywall and had a 4 in. separation between panels. The panels were connected by a frame work of 2 by 4 and 2 by 2 dimensional lumber (i.e., studs). Pairs of 2 x 4 studs were placed on end approximately 16 inches apart; sets of three clouds were then centered and screwed to the pair of studs. The three sets of three clouds were then connected by three 2 x 2 studs placed at the approximate centers of the panels, perpendicular to the 2 x 4 studs. The panels were attached to the ceiling framework using four 2 x 2 studs placed near the corners of the array. These studs provided a rigidity and stability for the cloud array. The cloud array was centered beneath a 12 ft x 12 ft layer of.2 inch drywall that was attached to the existing structural ceiling. The bottom surface of the clouds were 18 inches below the drywall attached to the structural ceiling. A photograph of the cloud array is shown in Figure 1. There were no walls or baffles attached to the structural ceiling; therefore, this setup is equivalent to an unconfined ceiling with no layer buildup. Figure 1 Cloud array mounted on movable ceiling A 12 inch x 12 inch propane sand burner was used to provide the heat release rates desired for testing. The flowrate to this sand burner was controlled using an Alicat mass flow controller. A shroud was constructed for the burner to prevent ambient airflows from causing excessive lean of the fire plume. The shroud consisted of a square built from four pieces of drywall measuring 2 ft x 4 ft and placed on top of four standard bricks laid on end. The propane burner was placed on the ground and centered within the shroud. A photograph of the shroud is shown in Figure 2.

11 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Instrumentation Figure 2 Burner with shroud Eighteen thermocouples were mounted to the clouds and the moveable ceiling. The thermocouple (TC) locations are shown in Figure 3. At the centers of each cloud, type K,.32 diameter TCs were mounted with the beads 2 inches below both the cloud and the structural ceiling. Data was recorded at a rate of 1Hz using National Instruments cdaq hardware and LabView software. Figure 4 shows the cloud numbers and fire locations (X s) for referencing the data spreadsheets. 8 ft 2 ft Structural Ceiling Cloud TC (2 below both clouds and structural ceiling) Fire Location Figure 3 Plan view of ceiling plenum

12 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Figure 4 Instrumentation numbering for data acquisition and fire locations (blue x) Test Matrix Twelve tests were run as part of this test series. Each test shown in the test matrix (Table 1) was run in duplicate. Three fire locations, which are illustrated in Figure 3 and Figure 4, were used: centered below a cloud (cloud center), centered between two clouds (cloud-cloud-slot), and centered between four clouds (cloud-cloud-cross). The cloud ceiling was set to two different heights: 8 ft and 16 ft above the floor. For each test, two different fire sizes were used sequentially (see Section 3.). Table 1 Test Matrix Test Configuration Ceiling Height (m [ft]) Fire Size (kw) 1 Cloud Center 2.4 [8], 1 2 Cloud-Cloud-Cross 2.4 [8], 1 3 Cloud-Cloud-Slot 2.4 [8], 1 4 Cloud Center 4.9 [16] 1, 2 Cloud-Cloud-Cross 4.9 [16] 1, 2 6 Cloud-Cloud-Slot 4.9 [16] 1, Experimental Procedure Prior to testing, the ceiling was raised to the appropriate height. The bottom of the cloud ceiling was set to either 8 ft or 16 ft. The cloud ceiling was leveled using adjustment straps attached to the sides of the structural ceiling. The propane burner was placed in the appropriate location for the specific test. All ventilation and circulation fans in the lab were turned off to prevent air currents from causing excessive flame lean. The DAQ system was turned on for a period of 1 minute or more to ensure the system was operational and to capture background temperatures. The mass flow controller was set to a zero flow and valves for the propane system were opened. A lit handheld propane torch was positioned near the propane burner prior to ignition. The mass flow controller was set to the first output level and the burner was ignited. The burner was allowed to burn at the first level for a minimum of minutes. After minutes, the mass flow controller was set to the second output level for a minimum of minutes. The minute period ensured that the temperatures reached steady-state levels. After the second minute burn, the mass flow controller was secured; data was secured after the temperatures reached near ambient conditions. Overhead exhaust ventilation fans were operated until the space above the structural ceiling was clear of combustion products.

13 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 2.3. Experimental Results During some of the testing, the propane flame experienced some leaning as is illustrated in Figure. This caused the plume to shift somewhat from the intended location; however, the leaning was intermittent and varied from test to test. Figure Photograph showing a typical plume lean The ambient temperatures were between 19 and 2 C at the beginning of the tests. Table 2 lists the ambient temperatures for each test. These temperatures were average temperatures taken over all 18 thermocouples during the 6 second background data acquisition time. Test Table 2 Test Matrix Configuration Ceiling Height (ft) Ambient Temperature ( C) 1A Cloud Center B Cloud Center 8 2 2A Cloud-Cloud-Cross B Cloud-Cloud-Cross A Cloud-Cloud-Slot B Cloud-Cloud-Slot A Cloud Center B Cloud Center A Cloud-Cloud-Cross B Cloud-Cloud-Cross A Cloud-Cloud-Slot B Cloud-Cloud-Slot Steady-state average temperatures were calculated for each burner output level during each test. An interval of 2 seconds was selected from each steady state period and the temperature values were averaged across this period. An example of the steady state period selection is shown in Figure 6.

14 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 6 Figure 6 Example of Data Processing Methodology Average steady state temperatures for Test 1A is shown in Figure 7. Appendix A contains all the average test data. The fire location is marked with a blue X on each figure. In general, the tests were very repeatable. Differences in average steady state temperatures for the same cloud in the repeat tests were generally less than C for all clouds in tests 2, 4,, and 6. In test 3, the largest differences between average steady state temperatures were as high as 6 C. For test 1, at the kw burner output, the differences in average steady state temperatures for the same cloud were less than C, but the largest differences at the 1 kw burner output were up to 1 C. The largest difference of 1 C was located below the cloud (#7) directly located above the propane burner. The average temperatures at this location were the highest out of all of the tests at 131 C (test 1A) and 16 C (test 1B). For all the tests, the figures show a temperature rise over all clouds for all fire locations. However, for Tests 1 and 4 with the fire centered below a cloud, only a small rise in temperature is seen for the opposite corner cloud especially as compared to the structural ceiling cloud. Taken together these observations indicate that there is the potential to activate a sprinkler over multiple cloud gaps from the fire location, but the permissible gap size and number of gaps is not likely to be large.

15 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 7 kw Average, SS Temperatures 1 kw Average, SS Temperatures Below Structural Ceiling Below Structural Ceiling Below Cloud Below Cloud FDS Modeling of Experiments Figure 7 Test 1A results (x is fire location) FDS [2-6] was used to simulate each of the 6 tests after performing a grid study. FDS comparisons were made to the average of each pair of identical tests. The procedure in the FDS Validation Guide [] was followed in making the comparisons to determine model error and bias. The experimental error was taken by performing a propagation of error on the test data using the standard error of the two test average, the estimated error in the thermocouple measurement (expanded error of % []), and the manufacturer reported error for the mass flow controller ( kw 2.3 %, 1 kw 1.4 %, and 2 kw.9 %) adjusted to temperature []. Comparisons were made for each fire size separately for the cloud ceiling and the moveable ceiling locations as well as for each fire size for all locations combined for each test. Two sets of comparisons were made. The first set used the experimental data as collected. The second set attempted to account for plume lean of the fire by averaging symmetric locations. The symmetric locations are shown in Figure 8 below.

16 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Grid Study Figure 8 Symmetric locations for plume lean correction (like colors) Test 1 was used for a grid study. Three meshing schemes were evaluated during the study. The first scheme was a uniform 2 inch mesh. The second scheme was a uniform 1 inch mesh extending 1 ft to the sides of the cloud array and 2/3 ft above and below the array with the remainder of the domain a uniform 2 inch mesh. The third meshing scheme replaced the finer 1 inch mesh with an. in mesh. The three meshing schemes respectively placed 2, 4, or 8 cells across the gap between clouds. Table 3 and Table 4 below shows the results of the grid study. As can be seen in the tables, for all grid sizes the errors are 17 % or less. This is the same as the 16 % error seen for ceiling jets in the FDS Validation Guide. When all the data for a test is grouped together, little difference is seen between the three meshing strategies. However, when the data is spilt into cloud and movable ceiling measurement locations, differences are apparent. Presented in this manner the uniform mesh has a higher error than the other two meshes. The 1 inch vs. inch mesh around the clouds show similar levels of error. From this it is concluded that a modeling goal should be to target 4 cells across the gap. An additional observation is that FDS is slightly under predicting the temperatures overall (the bias over all data is less than 1). Much of this is likely the plume lean being towards the open edge of the cloud array which can be seen in Test 2 and Test where the outer edge clouds next to the fire have a higher temperature than the inner clouds next to the fire. When data is separated based on location, the cloud bias is less than 1 and the movable ceiling bias is greater than 1. This is a desirable outcome as under predicting the cloud temperatures will result in conservative predictions of sprinkler operation for below cloud sprinklers.

17 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 9 Table 3 Grid Study Results Unadjusted Data Dataset All kw data All 1 kw data kw Cloud kw Movable 1 kw Cloud 1 kw Movable Uniform 2 inch mesh 1 inch mesh around clouds. in mesh around clouds error bias error bias error bias Table 4 Grid Study Results Symmetrically Averaged Data Dataset All kw data All 1 kw data kw Cloud kw Movable 1 kw Cloud 1 kw Movable Uniform 2 inch mesh 1 inch mesh around clouds. in mesh around clouds error bias error bias error bias Results of Experimental Simulations Based on the grid study, the 1 inch mesh around the clouds with a 2 inch mesh for the remainder of the domain was used as the meshing strategy in FDS to simulate all 6 tests. Figure 9 below shows scatterplots for the measured vs. the predicted data where the diagonal line represents perfect agreement. The plots show a good agreement between FDS and the measured data. The plots indicate somewhat more scatter for the movable ceiling than the cloud ceiling. The negative bias observed in the grid study for the cloud ceiling locations can also be seen in the plots (more data below the diagonal line for the clouds than above it).

18 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE All Data Cloud Data 12 1 Movable Ceiling Data Predicted Temperature ( C) Predicted Temperature ( C) Predicted Temperature ( C) Measured Temperature ( C) Measured Temperature ( C) Measured Temperature ( C) Figure 9 Scatterplots of predicted vs. measured data Table and Table 6 below show the results for all simulations for the unadjusted and averaged test data. With the exception of the unadjusted data for Test 2 at 1 kw, all the predictions fall within the ceiling jet error noted in the FDS Validation Guide. As with the grid study, a slight negative bias is seen for the cloud predictions and slight positive bias is seen for the movable ceiling predictions. The average biases are not large. The average cloud bias is.9 (under predict by %), and the average movable ceiling bias is 1.2 (over predict by 2 %). Table All Test Simulation Results Unadjusted Data Test Fire Size All Data Cloud Movable (kw) error bias error bias error bias

19 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 11 Table 6 All Test Simulation Results Symmetrically Averaged Data Test Fire Size All Data Cloud Movable (kw) error bias error bias error bias Task 1 Summary A set of six experiments using an array of nine small clouds was conducted to collect data for model development and validation. The experiments varied fire size, fire location, and ceiling height. The experimental results indicate that a portion of the energy from the fire can cross multiple cloud gaps. FDS was used to simulate all six of the experiments in two parts. The first part used Test 1 and performed a grid study. The grid study determined that four cells across the gaps results in a reasonable predictive accuracy of the below cloud and above cloud conditions. The second part used the grid determined in the first part to simulate the six tests. For all test variables it was determined that selected meshing strategy resulted in FDS predictions as accurate as the ceiling jet results in the validation guide. Additionally the bias in the predictions was appropriately conservative with the cloud ceilings slight under predicted and the structural ceiling slightly over predicted. 3. TASK 2: NUMERICAL MODELING OF CLOUD CEILING CONFIGURATIONS The goal of Task 2 was to simulate a range of cloud ceiling configurations in order to develop installation guidance and code recommendations. Simulations were performed using FDS Methodology FDS Model Geometry Modeling was based upon the geometry used in the first cloud ceiling study. This was a 3 ft by 3 ft room with an open doorway along one wall. A fixed ceiling plenum height of 2 ft was used in this study. The prior study varied plenum height; however, it determined that ceiling plenum height had little impact on the permissible cloud spacing. Geometry variables in this study were cloud size, the gap between clouds, and ceiling height. A schematic of the room geometry is shown in Figure 1.

20 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Fire The first cloud study varied fire growth and used one of five fire locations. It determined that growth rate was not a significant factor in the permissible cloud spacing; therefore, this study used a medium growth rate. The first study also determined that the gap sizes were driven by two fire locations: fire in a corner and fire directly below the intersection of four clouds. Based upon that observation, this study only used the two limiting fires hereinafter referred to as corner-fire and cross-fire. Each simulation used a medium growth fire with a 1 kj/m 2 heat release rate based on a woodplastic mix representative of typical ordinary combustibles. This was represented as C 2.13H 8N 4 with a % soot yield and a 3.8 % CO yield that had a heat of combustion of 17 kj/kg. Prediction of sprinkler activation time is dependent upon reasonably predicting the fire plume temperatures as a function of time. If the fire in FDS was defined as just a single burner, then until it grows in size it would be a diffuse heat source over a large area with low plume temperatures. To prevent this and keep the fire plume representative of a growing, flaming fire, the fire source in FDS was implemented as a series of concentric rectangles, see Figure 11. The fire would grow by increasing in size over one rectangle until it reached 1 kj/m 2 and then having the next rectangle start burning. In this manner medium growth t 2 fire was created that grew in size over time. Cloud Panel Varies Varies Door 9.1 m (3 ft) Figure 1 Schematic of simulation geometry

21 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 13 Cross-fire Corner-fire Figure 11 Fire source implementation in FDS Sprinklers Sprinkler activation was determined by FDS using its built-in sprinkler response model the RTI equation. All sprinklers were modeled as quick response (RTI = (m s) 1/2 ) with an activation temperature of 73.9 C (16 F). Sprinklers were positioned cm (2 inches) below the clouds. Sprinklers were located at cloud centers, centered in the gap at cloud corners, and centered in the gap at the cloud edges as shown in Figure 12 (also shown for reference are the fire locations). Structural Ceiling Cloud Sprinkler Fire Location Figure 12 Modeled sprinkler locations plus fire locations for a 3x3 cloud array

22 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Material Properties The walls, clouds, and structural ceiling were given the properties of 3/8 gypsum wallboard. In general one would expect these surfaces to be some form of insulating (i.e. low thermal conductivity) material and gypsum is a common interior finish. The floor of the room was given the properties of 1 cm (6 in) of concrete. The floor plays little role in the overall heat balance of the room since a configuration would be considered a failure if the hot layer reached the floor prior to sprinkler activation Performance Criteria Sprinkler requirements for cloud ceilings are in NFPA 13 [7]. The purpose of NFPA 13 is to provide to provide a reasonable degree of protection for life and property from fire through standardization of design, installation, and testing requirements for sprinkler systems, including private fire service mains, based on sound engineering principles, test data, and field experience. The goal of this project was to determine configurations where the sprinklers would not be needed (or effective) on the structural ceiling when a cloud ceiling is present. It is obvious, and borne out by prior results, that a porous ceiling will result in increased time to sprinkler activation. Therefore, determining if a cloud configuration would require sprinklers both above and below the clouds means determining at what point the delay in activation prevents a reasonable degree of protection for life and property. Since the listing standards (e.g. UL 199 [8]) for automatic sprinklers do not contain a pre-actuation temperature requirement for the compartment gas or structure, a metric was needed to evaluate the model results. This project decided to apply a similar metric as was done for the FPRF residential sprinkler on sloped ceiling project [9]. The objective of the criteria was define a performance level that should ensure that life and property would be protected in accordance with the purpose of NFPA 13. The criteria were: 1. Below cloud sprinklers must activate due to the fire plume (e.g. ceiling jet) and not due to the development of a hot layer [9]. 2. The temperature at 1.6 m (63 in) above the floor cannot exceed 93 C (2 F) away from the fire and cannot exceed 4 C (13 F) for over two minutes [8]. 3. The temperature below either the structural ceiling or the drop ceiling cannot exceed 31 C (6 F) at a distance of % of a standard flat ceiling sprinkler spacing [8]. 4. The backside temperature of the structural and cloud ceilings must remain below 2 C (392 F) [8]. Simulations were performed until one of the above criteria was met. The extent of sprinkler operation that had occurred prior to that time was then used in the development of spacing requirements. The rationale for the criteria are discussed below Plume vs. Layer Sprinkler Operation If the fire is able to grow large enough, at some point it will fill the plenum space above the clouds with hot gases. At that point in time the layer will drop below the clouds. Since the layer will be relatively uniform in temperature, this has the potential for near simultaneous operation of a large number of sprinklers. This condition could lead to reduced effectiveness over the sprinkler coverage area. This condition should be avoided. Rather it is desired that sprinkler operation result from the fire plume and ceiling jet where a small number of sprinklers closest to the fire operate. This is illustrated in Figure 13 below. These figures are for a simple 2x2 cloud array with a corner fire (upper left corner) and show temperature just below the clouds at the time of the first sprinkler

23 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 1 operation. On the left it is seen that the highest temperatures are on the cloud immediate over the fire. In this case the first sprinkler operation is the sprinkler closest to the fire, a desired outcome. On the right is the same fire with a larger gap between the clouds. At the time of the first sprinkler operation, the highest below cloud temperatures are in the opposite corner from the fire and at all the gaps between the clouds. In this case the first operation is away from the fire and results from the layer banking down below the clouds. This is an undesirable outcome. Figure 13 Below cloud temperatures for plume (left) vs. layer activation (right), sprinkler indicated by blue circles Head Height Temperatures One of the goals of NFPA 13 is life safety. In the context of a fire that, in part, means ensuring conditions remain tenable for occupants to safely egress. A primary hazard to persons egressing the room with the fire is the thermal environment. In this case there is the risk of exterior burns and pulmonary injury from the inhalation of hot gasses [1]. Sprinkler operation will result in the creation of large amounts of water vapor which can condense in the lungs releasing the sensible enthalpy of the water. Keeping any extended exposure below 6 C will greatly avoid the risk. Keeping the instantaneous exposure below 1 C will avoid exterior skin injury Ceiling and Cloud Temperatures The last two criteria address hazards to the structure and other combustible materials in the room. Preventing a large area of a layer from exceeding 31 C (6 F), a radiative flux of < 7 kw/m 2, would be expected to prevent the radiative ignition of most combustibles remote from the fire location. If this avoided, then there is a greatly reduced risk of flashover which would threaten both the structure and its occupants. Preventive high backside temperatures addresses the risk of fire spread and structural failure. Low backside temperatures reduce the risk that combustibles in contact with the backside of the ceiling will ignite. Low backside temperatures should also ensure that the structure and fasteners holding the ceiling to the structure should remain intact Analysis Approach Simulations were performed using the basic principles of spiral development. That is, an initial set of simulations was performed in a first cut attempt to bound the end result. Those simulations were analyzed, and the results used to inform the creation of the next set of simulations. The prior study determined when there is one sprinkler per cloud that a cloud-to-cloud gap spacing of 1 inch or less per foot of ceiling height would result in adequate sprinkler performance. With smaller

24 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 16 clouds, there is the potential for there to be clouds without sprinklers. Every gap that a ceiling jet has to cross to reach a sprinkler will result in additional heat loss from the ceiling jet into the plenum space. It was anticipated that the prior study results would not result in adequate performance. Therefore, the first modeling pass consisted of cloud ceiling heights of 8 ft, 14 ft, and 2 ft; with 3x3, 6x6, and 9x9 cloud arrays; and gap sizes of. or 1 inch per foot of ceiling height. For each simulation the time when the first criteria in Section is exceeded is used to end the simulation. The sprinkler operations at that time are used to determine a permissible sprinkler spacing. The 3x3 cloud, for a 4 inch gap for an 8 foot ceiling is used below to demonstrate the analysis approach. Figure 14 shows the evaluation of the four criteria at the point in time of the first failure. In the upper left image is the backside temperature of the clouds (the fire is the white square in the upper left of the image). As can be seen there are no temperature in excess of 2 C over the cloud area. In the upper right is the head level temperature. There are no temperatures that exceed 93 C and while there are temperatures that exceed 4 C; they have not done so for over 2 minutes. The bottom right image is the below cloud gas temperatures. From this it can be seen that there is no hazardous hot layer forming and that the layer has not yet dropped below the clouds as the fire plume is clearly the source of the highest temperatures. In the bottom left is the below ceiling gas temperatures. Here it can be seen that the simulation is showing a large region which exceeds 31 C. This violates the third criteria. Figure 14 Evaluation of Section Criteria. The sprinkler operations at 24 s were then evaluated. The heads that operated were examined and the maximum and minimum radius of the region of operation was determined. The radius was measured from the center of the fire. The average radius was taken as the permissible sprinkler spacing. This is demonstrated in Figure 1 below. In the figure the average radius is 1 ft. This can be

25 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 17 expressed as a uniform sprinkler spacing by assuming it represents the diagonal of a square. This would yield a coverage area of 449 ft 2, Figure 16. Note that this is larger than the maximum coverage area of 22 ft 2 allowed in NFPA 13 for standard sprinklers. Therefore, if standard sprinklers were being used, the coverage area would have to be reduced to that given by the maximum sprinkler spacing as specified by the manufacturer. 3 Corner Fire -3x3 Cloud -4 in. gap, 8 ft. ceiling at 24 seconds R(max) Criteria 1: 288s Criteria 2: 282s Criteria 3: 24s Criteria 4: 27s Closed Heads Open Heads R(min) = 14.2 ft R(max)= 1.8 ft 1 R(min) Figure 1 Evaluation of Sprinkler Spacing ft 1 ft 21.2 ft 449 ft 2 Figure 16 Evaluation of Sprinkler Coverage Area.

26 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Results and Analysis First Pass Results The first set of simulations was a pass through the permutations of. in. of gap/ft ceiling height and 1. in. of gap/ ft of ceiling height for 3x3, 6x6, and 9x9 clouds with ceiling heights of 8, 14, and 2 ft. All permutations were run for the corner fire location while the cross-fire location was limited to. in. of gap/ft ceiling height for the 8 ft and 14 ft ceiling heights. These are respectively shown in Table 7 and Table 8 below. Note that plots of all operated heads for all scenarios are shown in Appendix A. Table 7 First Pass Results Corner Fire Cloud Array 3x3 6x6 9x9 Gap Ceiling Ratio R min R max Coverage (in.) (ft) (in./ft) (ft) (ft) (ft 2 ) Table 8 First Pass Results Cross Fire Cloud Array Gap (in.) Ceiling (ft) Ratio (in./ft) R min (ft) R max (ft) Coverage (ft 2 ) 3x x x The following observations are noted: The corner fire results are more limiting than the cross fire results. This echoes conclusions from the first study.

27 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 19 For all the cases, the. in. of gap/ft of ceiling height results in a sprinkler spacing that is larger than the 22 ft 2 allowed in the code for a standard sprinkler. This indicates that the permissible gap size can be larger. At 1. in. of gap/ft of ceiling height the coverage area drops significantly in most cases from >22 ft 2 to 6 to 1 ft 2. This indicates that a larger gap should not be used. The coverage area decreases as the cloud size decreases Second Pass Results Since there was a significant drop in coverage area between gaps of. in./ft of ceiling height and 1. in./ft of ceiling height the second pass targeted gaps between those size. A gap ratio of.7 was used for the 8 ft ceiling. To maintain the current FDS gridding for the 14 and 2 ft ceiling heights (where the cloud grid was coarser than for the 8 ft ceiling heights) the gap ratios were.71 and.86. Results are shown in Table 9 for corner fires and in Table 1 for cross fires. Table 9 Second Pass Results Corner Fire Cloud Array 3x3 6x6 9x9 Gap (in.) Ceiling (ft) Ratio (in./ft) R min (ft) R max (ft) Coverage (ft 2 ) Table 1 Second Pass Results Cross Fire Cloud Array 6x6 9x9 Gap (in.) Ceiling (ft) Ratio (in./ft) R min (ft) R max (ft) Coverage (ft 2 ) The following observations are noted: As previously seen, corner fire results are more limiting than the cross fire results. There is an increasing drop in coverage area as the gap is increased from. in./ft of ceiling height. Gaps on the order of.7 in./ft of ceiling height still provide coverage areas of 1 ft 2 or larger.

28 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Third Pass Results For the third pass the cloud arrays were increased to 12x12 and 1x1 in order to evaluate the effects of smaller clouds. Only corner fire cases were run for these simulations. Results are shown in Table 11. Table 11 Third Pass Results Corner Fire Cloud Array 12x12 1x1 Gap Ceiling Ratio R min R max Coverage (in.) (ft) (in./ft) (ft) (ft) (ft 2 ) The following observations are made: Coverage areas have decreased from the larger cloud sizes. No 22 ft 2 or larger coverage areas were seen for the. in./ft of ceiling height cloud spacing. Coverage decreases with decreasing cloud size with the 1x1 clouds having a smaller coverage area than the 12x12 clouds Fourth Pass Results For the fourth pass two simulations were run with a 12x6 cloud array to examine the impact of having a non-square cloud. Only corner fires were run. Results are shown in Table 12. Table 12 Fourth Pass Results Corner Fire Cloud Array 12x6 Gap (in.) Ceiling (ft) Ratio (in./ft) R min (ft) R max (ft) Coverage (ft 2 ) It can be seen that the coverage areas for the 12x6 clouds are close to the coverage areas for the equivalent 12x12 case. This indicates that when determining how to apply spacing rules, that the smaller dimension of the cloud should be used Summary of Simulations and Development of Installation Guidance Plots of coverage area vs. gap ratio for corner fires are shown by cloud size in Figure 17. Coverage area is clearly seen to decrease with both decreasing cloud size and with increasing gap size. Figure 18 plots all ceiling heights on a single plot. In this plot there is a clear pattern that coverage area decrease with the size of the cloud array. In other words, coverage area decreases with the fraction of area that the ceiling is open.

29 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 21 8 ft Ceiling 14 ft Ceiling 9 7 Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x1 Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x Gap to Ceiling Ratio (in/ft) Gap to Ceiling Ratio (in/ft) 2 ft Ceiling Coverage Area (ft 2 ) Gap to Ceiling Ratio (in/ft) 3x3 6x6 9x9 12x12 Figure 17 Sprinkler Coverage Area by Height, Gap Ratio, and Cloud Array for Corner Fires Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x1 Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x Gap to Ceiling Ratio (in/ft) Gap to Ceiling Ratio (in/ft) Figure 18 Sprinkler Coverage Area by Gap Ratio and Cloud Array for Corner Fires. Figure 19 below presents the same data as the prior two figures in a slightly different manner. For each cloud configuration the total area of the gaps was computed and then normalized by the total ceiling area. This results in the open area fraction of the ceiling. While the data in this figure does not show a clear pattern with the cloud array, it does indicate if the gap area fraction is less than 2 % that a 22 ft 2 or larger coverage area can be used. However, immediately after crossing that threshold coverage areas for some configurations drop to -6 ft 2. Closer examination of the reveals that the points in the lower left are primarily from the 8 ft ceiling height and that ceiling height increases towards the upper right of the point in the plot.

30 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x1 Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x Gap Area Fraction Gap Area Fraction Figure 19 Sprinkler Coverage Area by Gap Fraction to Height Ratio and Cloud Array for Corner Fires Figure 2 and Figure 21 take the data from Figure 19 and normalizes the independent axis by the ceiling height. Figure 2 shows this data by ceiling height and Figure 21 shows all the data on one plot. For the 14 ft ceiling height, which has the most simulations, the data collapses to a clear trend. The same basic trend is seen on the other two ceiling heights. Coverage Area (ft 2 ) ft Ceiling Gap Area Fraction to Ceiling Height (1/ft) x3 6x6 9x9 12x12 1x1 Coverage Area (ft 2 ) 2 ft Ceiling ft Ceiling Gap Area Fraction to Ceiling Height (1/ft) 3x3 6x6 9x9 12x12 1x1 Coverage Area (ft 2 ) x3 6x6 9x9 12x Gap Area Fraction to Ceiling Height (1/ft) Figure 2 Sprinkler Coverage Area by Gap Fraction to Height Ratio, Ceiling Height, and Cloud Array for Corner Fires.

31 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x Gap Area Fraction to Ceiling Height (1/ft) Figure 21 Sprinkler Coverage Area by Gap Fraction to Height Ratio and Cloud Array for Corner Fires When all the ceiling heights are plotted together, Figure 21, there appears to be a hyperbolic section defined by the points. Figure 22 shows a best fit hyperbolic section through the data points. The function is limited to a lower bound of 36 ft 2 the minimum coverage area for a standard sprinkler per NFPA Area =.76xRatio -2 Coverage Area (ft 2 ) x3 6x6 9x9 12x12 1x Gap Area Fraction to Ceiling Height (1/ft) Figure 22 Figure 21 Fit with a Power Function to the Lower Edge of the Data Based on the previous figure a simple rule can be given: The coverage area, A, is defined by the gap area fraction to ceiling height ratio, R G, according the following formula. The computed coverage area should be limited by any NFPA 13 restrictions (e.g. limited a maximum of 22 ft 2 for a standard sprinkler or a manufacturer s listed spacing for an extended coverage sprinkler). A computed coverage area less than permitted by NFPA 13 would indicate the need for sprinklers above the clouds =.76 For reasons of aesthetics, it may be desirable to have a uniform array of sprinklers to match the uniform array of clouds. A sprinkler spacing rule which counts clouds may be useful. Figure 23 below plots the coverage from Figure 21 in terms of the number of clouds that could be skipped when installing

32 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 24 sprinklers. The figure is summarized as a tabular rule in Table 13. Note this table is applicable only far gap to ceiling height ratios of 1 in./ft or less. 4 3 Cloud Skipping 2 1 3x3 6x6 9x9 12x12 1x Gap Area Fraction to Ceiling Height (1/ft) Figure 23 Cloud Skipping by Gap Fraction to Height Ratio and Cloud Array for Corner Fires Table 13 Sprinkler Spacing Rule Table for Cloud Skipping Cloud Size 1 (ft) Gap Area Fraction to Ceiling Height (1/ft) Each Cloud 2 Every Other Every Third Cloud Cloud over 1 ft Up to 1 inch/ft ft to 1 ft >.2 <.2 3 ft 4 in. to ft >.4.3 gap.4 <.3 2 ft 6 in. to 3 ft 4 in. >..3 gap. <.3 under 2 ft 6 in. >.6.4 gap.6 <.4 1. Dimension is gap center to gap center based on smaller cloud dimension 2. 1 inch of gap/ft of ceiling height was the limit established for large clouds in [1]. 4. SUMMARY 4.1. Summary of Task 1 and Task 2 A two-part study was conducted to determine conditions under which sprinklers could be placed only below clouds for ceilings where the cloud size is less than the listed sprinkler spacing. The first part of the study was experimental, and the second part of the study was numerical. The experimental study was conducted to develop and validate a modeling approach for cloud ceiling sprinklers using FDS. Based upon the experimental results, it was determined that at least 4 grid cells are required across a gap to resolve the appropriate partitioning of plume flow through a gap vs. plume flow across the bottom of a cloud. Modeling of all experiments using the selected gridding approach resulted in FDS predictions whose modeling uncertainty matched those for ceiling jets in the FDS validation guide. Additionally, the modeling results showed a positive bias for the structural ceiling (over predicted temperatures) and a negative bias for the cloud ceiling (under predicted temperature). This was a conservative result for the purpose of this study as it decreased the chance of a below cloud sprinkler operating. The numerical study consisted of 44 simulations. The simulations varied ceiling height, cloud size, gap size, and fire location. The simulations were done in a series of sets of simulations. The first set used

33 Sprinkler Protection for Cloud Ceilings, 1JEF19. PAGE 2 the gap size recommendation from a previous study on large area clouds and one-half that gap size. The second set refined the gap size to an intermediate value. The third set looked at smaller cloud sizes. Finally, the fourth set examined the impact of non-square (rectangular) clouds. The numerical study showed that there is a complicated relationship between cloud size, gap size, and ceiling height. This relationship was best characterized by the ratio of the gap area fraction (area of gaps over the entire area of the ceiling) to the height of the cloud ceiling. Two simple rule sets were developed. The first simply looked at coverage area as a function of the gap area fraction to height ratio. The second expressed coverage area in terms of the number of clouds that could be skipped assuming a uniform array of sprinklers Limitations of Study The conclusions of this study are limited to following: Uniform gap sizes. Note that extrapolation to large gap sizes should be possible if area fractions are computed assuming all gaps are the largest size. Uniform cloud arrays. Note that extrapolation to non-uniform arrays (tessellations of multiple cloud sizes) should be possible if area fractions are computed assuming a uniform tessellation of the smallest cloud. Flat, level clouds all mounted at the same elevation. Ceiling heights of 8 ft to 2 ft; however, simple scaling laws would suggest the conclusions would be applicable to larger heights. Cloud sizes greater than 1.1 ft (The smallest cloud tested was for a 1x1 array with 1 inch gap and 2 ft cloud height).. REFERENCES 1. Floyd, J. and Dinaburg, J., Sprinkler Protection for Cloud Ceilings, Fire Protection Research Foundation, Quincy, MA, McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., Overholt, K., "Fire Dynamics Simulator User's Guide," NIST SP 119, National Institute of Standards and Technology, Gaithersburg, MD, McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., Overholt, K., "Fire Dynamics Simulator Technical Reference Guide Volume 1: Mathematical Model," NIST SP 118, National Institute of Standards and Technology, Gaithersburg, MD, McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., Overholt, K., "Fire Dynamics Simulator Technical Reference Guide Volume 2: Verification," NIST SP 118, National Institute of Standards and Technology, Gaithersburg, MD, McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., Overholt, K., "Fire Dynamics Simulator Technical Reference Guide Volume 3: Validation," NIST SP 118, National Institute of Standards and Technology, Gaithersburg, MD, 213.