488

Minimum Water Flows for Fire Suppression with In-Rack Sprinklers at Increased Vertical Intervals

de Vries J.*, Meredith K. V., Jamison L. T.

FM Global, Research Division, Norwood, MA, USA *Corresponding author email: jaap.devries@fmglobal.com

ABSTRACT The use of in-rack automatic sprinklers (IRAS) can be a valuable tool when protecting high warehouse storage. Until now, IRAS applications were limited to small orifice sprinklers placed at vertical intervals less than 3.1 m. This study focusses on IRAS water flow tests and intermediate-scale fire suppression tests aimed to increase the vertical IRAS intervals to 9.1 m. This increase was accomplished using larger orifice (K-factor of 200 and 360 lpm/bar1/2) sprinklers, placed inside the longitudinal flue and at the face of the array, at flow rates ranging from 200 to 946 lpm. Both test sets were part of a comprehensive hybrid approach combining both physical testing and water film modeling to determine adequate protection. All intermediate-scale tests were conducted with 9.1 m high arrays. For the fire suppression tests, the heat release rate was measured via a 20 MW fire products collector. During the water flow tests, water flow rates reaching the bottom of the test array were measured using a set of 24 water collection bins, resulting in detailed spatially resolved mapping of both transient and steady state water flows from a variety of sprinkler locations, pressures, and flow rates. The water test results were used for model validation, while the intermediate-scale fire tests results led to the determination of minimum water flows required for both fire suppression and to prevent lateral fire spread. Adequate protection determined using these reduced-scale tests were finally validated via six large-scale (8 to 11 tier-high) fire suppression tests. This work showed that 9.1 m vertical intervals between IRAS can provide adequate protection. Furthermore, the per sprinkler water flow required can be determined using fewer expensive large-scale fire suppression tests when a combined modeling/reduced-scale test approach is taken.

KEYWORDS: In-rack sprinklers, fire suppression tests, water film modeling, large-orifice sprinklers.

INTRODUCTION AND OBJECTIVES

Warehouse operators explore strategies to cut costs through innovative warehouse designs. Building owners are expanding upwards rather than outwards [1]. This upward trend challenges the fire protection community. Increased fire loading causes hazardous fire scenarios and increased ceiling height delays sprinkler operation. Often, installing IRAS is the only viable fire protection option for high storage arrays. In-rack sprinklers activate when a fire is smaller, requiring less water for control [2]. However, testing new protection designs is challenging when the scale of the problem exceeds the size of the available test facility.

Since 1968, FM Global has conducted roughly 160 large-scale fire tests involving in-rack sprinklers. These tests resulted in both NFPAʼs [3, 4] and FM Globalʼs [5] current fire protection standards. Prior rack-storage test programs showed that IRAS could protect rack-storages of various heights [6]. Adequate protection designs required both IRAS and ceiling sprinklers operation. In addition, storage 9.1 m or higher should have face sprinklers installed. The test data only covered small K-factors sprinklers, (with flows less than 114 lpm). Currently, the greatest vertical separation distance allowed is two tier, or approximately 3.1 m. Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8), pp. 488-496 Edited by Chao J., Liu N. A., Molkov V., Sunderland P., Tamanini F. and Torero J. Published by USTC Press ISBN:978-7-312-04104-4 DOI:10.20285/c.sklfs.8thISFEH.049

Part II Fire

489

Increasing vertical separation distance reduces the total number of sprinklers required. This could reduce costs and the possibility of forklifts or shifting pallets to damaging sprinklers. Providing adequate IRAS protection with increased vertical separation requires larger flows and larger orifice sprinklers. This report focuses on intermediate-scale fire suppression and water flow tests. The water flow tests were used for CFD model validation. More details on the CFD model used are listed in Refs [7-10]. The fire suppression tests and modeled results combined determined the smallest per sprinkler water flows. A reduced number of full-scale tests validated these water flows. Test results discussed here affect guidelines [5] for FM Globalʼs clients.

TECHNICAL APPROACH AND TEST SETUP

Methodology

The following three commodities were tested in an intermediate-scale or “modular” setting inside open-frame double row racks: Uncartoned Unexpanded Plastic (UUP), Uncartoned Expanded Plastic (UEP), and Cartoned Unexpanded Plastic (CUP) [11, 12]. The UUP commodity consists of seven plastic pallets on top of a hardwood pallet with a total weight of 200 kg (of which 178 kg plastic). The four-way entry plastic pallets are made from high-density polyethylene. The UEP commodity consists of expanded polystyrene meat trays in plastic bags, stored on top of a hardwood pallet. The total weight of the commodity is 52.3 kg (of which 29.9 kg is expanded plastic). Finally, the CUP commodity consists of rigid, crystalline polystyrene cups packaged, face down, in a single-wall corrugated containerboard box. The cups are compartmentalized in five 25-cup layers, yielding 125 cups per box. Eight boxes are placed on top of a hardwood pallet. A pallet load weights 93 kg (of which 50.3 kg are plastic cups and 20.3 kg is containerboard).

Apart from the obvious cost savings when testing at a reduced scale, the 9.1 meter-high tests are representative of a single module (discrete subset) of a larger storage array, as shown in Fig. 1(a). Minimum water fluxes at the bottom of the array required for suppression and to prevent lateral fire spread were determined from the intermediate-scale fire test results. The IRAS water flows required for these fluxes were determined from water flow modeling. A reduced set of full-scale fire tests validated these flow rates. Fig. 1(b) shows this “hybrid” (combining modeling with reduced-scale testing) methodology pictorially.

(a) (b)

Figure 1. Module as part of a greater array (a); FM Globalʼs IRAS test strategy (b).

Cold flow water test setup

The effect of pressure, K-factor, and location within the rack on water flow distributions was unknown prior to this study. IRAS water flow distribution tests were conducted using sprinklers placed at different locations within the array. A water collection apparatus, consisting of 24 individual bins,

Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)

490

was used to obtain spatially resolved water fluxes at the bottom of the array, see Fig. 2(a). These cold-flow tests resulted in transient and steady state water flow distributions. The results also aided water flow model validation.

All tests were performed with 3 × 2, 6-tier high arrays with IRAS placed above the sixth tier in 12 possible positions, see Fig. 2(b). The distance between the top of the commodity and the IRAS deflector was 15 cm. Vertical spacing between the sprinklers and the water flow collection apparatus was 9.1 m (six tiers of standard rack storage). Both the longitudinal and transverse flue spaces were 15 cm, measured between adjacent wood pallets. The 12 sprinkler locations were numbered starting from the North-West corner of FM Globalʼs Large Burn Laboratory (LBL). During each test, the branch lines feeding the face sprinklers were placed 46 cm from the outside face of the commodity. These pipes ran parallel to the longitudinal flue and sprinkler locations were positioned every 1.2 m along the length of each pipe. Two sprinkler K factors (360 and 200 lpm/bar1/2) were used. The collected water flows from the bins were converted into flow rates via a linear fit through the steady state portion of the test, as shown graphically in Fig. 2(c).

(a) (b) (c)

Figure 2. Water collection test using UUP (a); Potential sprinkler locations above the 6th tier (b); Steady state water flows (slope of red lines) for all 24 bins (c).

Intermediate suppression test setup

Identical storage array configurations of five pallets wide, two pallets deep and six pallets high were used for all 14 intermediate-scale fire tests. The seventh tier placed above the sprinklers consisted of bare metal liners to maintain proper airflow and to replicate the commodity for proper commodity/droplet interaction. The outside array dimensions measured 5.9 m wide, 2.6 m deep, and 10.3 m high.

All intermediate suppression tests were performed underneath a 20 MW fire products collector (FPC) inside FM Globalʼs Large-burn Laboratory (LBL). The 20 MW FPC consists of a 10.7 m diameter cone positioned 11.3 m above the lab floor. A single gas-sampling probe extracts gas from the duct. The gas analysis is capable of providing real time chemical heat release rate (HRR) information based on the total gas mass flow rate and measured concentrations, (using a Rosemount NGA 2000 gas analyzer), of CO and CO2. Additional instrumentation consisted of 12 embedded K-type thermocouples installed at each sprinkler location, embedded cameras, and a long wave infrared thermography (LWIR) via a FLIR® SC655 camera. For all tests, the duct flow was set to 94 m3/s.

During each test, in-rack sprinklers were installed above the sixth tier at 9.1 m elevation; their deflector was 15 cm above the top of the commodity. Both 200 lpm/bar1/2 and a 360 lpm/bar1/2 K-factor sprinklers were used with flow rates ranging from 314 lpm to 946 lpm. These sprinklers are identical to those used in the water transport tests. Both pendent sprinklers had quick response (QR) soldered thermal elements with a 74 °C temperature rating.

Loc. 3 458 lpm

K200 lpm/bar1/2

Ignition

Part II Fire

491

Twelve in-rack sprinkler positions were available for each test. These sprinklers were numbered from 1 to 12 starting in the North-West corner of the array, similar to the water flow tests as shown in Fig. 2(b). Each flue sprinkler was placed in the intersection of the longitudinal and transverse flue. The face sprinklers were centered inside the transverse flues and placed 45 cm from the face of the array. When IRAS are closely spaced together, sprinkler skipping was found to be an issue. Sprinkler skipping occurs when a sprinkler is wetted and cooled by an adjacent active sprinkler, hindering activation [13]. During some of the intermediate-scale tests, selected sprinklers (# 6, 7, 8, 10 and 12) were plugged to simulate this skipping phenomenon.

A target array, consisting of fiberglass panels, was placed at 1.2 m from the main array. Four, Schmidt-Boelter-type, 100 kW/m2-rated heat flux gauges (HFG) were installed on the target panels at 2.2 m vertical intervals from the laboratory floor. The target array acted to maintain airflow to the face of the array, where ignition was located. The HFG data can be used to estimate if radiant ignition of the target array was likely. Figs. 3(a), (b), and (c) shows an example of an intermediate-scale test with UUP, UEP, and CUP, respectively and the inert target panel.

(a) (b) (c)

Figure 3. Intermediate scale test with UUP (a); UEP (b); and CUP (c).

For all tests, two FM Global standard half igniters were used, each half igniter is a 76 mm diameter, and 76 mm long cylinder of rolled cellucotton soaked in 118 ml of gasoline and sealed in a plastic bag. The ignition location was centered at the face of the main array facing the fiberglass target. Face ignition should be consisted with a reasonable worst-case ignition scenario since sprinkler activation can be significantly delayed.

RESULTS

Water flow distribution test results

Transient water flow profiles differ between “flue” bins and “pallet” bins, where flue bins refer to the water collection bins that are located in the flue spaces and the “pallet” bins refer to the water collection bins that are located directly below each pallet at the bottom of the array. The time delays were greater for water fluxes measured below the pallets, due to the cascading nature of the water film through each commodity. The flue water fluxes showed faster transport times, since water can reach the water collection bins without much obstruction.

When studying the time histories of the transient water flow tests, both the average “flue” flux and the “pallet” flux could be described as a first order response to a step input function. Therefore, the total water flow can simply be estimated as the sum of the “flue” and “pallet” fluxes so that

Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)

492

0,F 0,P

F Ptot ss,F ss,P= (1 e ) (1 e ).

t t t t

q q qτ τ

− −− −

′′ ′′ ′′− + − (1)

In Eq. (1), totq′′ is the collected water flux in mm/min, ssq ′′ is the steady state water flux value, τ is the time constant in seconds, t and to correspond to the time and the time delay (lag), respectively. The subscripts “F” and “P” refer to “flue” and “pallet” water fluxes, respectively. Fig. 4(a) shows an example of a transient water flux profile. Good agreement can be seen between the measurement and the flux described by Eq. (1). All measured water flux data were compared against that computed by the water film model. As an example, Fig. 4(b) shows a scatter plot for the UUP commodity. The abscissa represents the experimental results and the ordinate the simulated results. A black line with a slope of one represents perfect agreement between the model and the experiment. The numerical simulation over predicts the measurements by 11%. Overall, the modeled results show good agreement with the data and generally fall within the experimental uncertainty, which was found to be 14%, 15%, and 24% for CUP, UEP, and UUP, respectively.

(a) (b)

Figure 4. Transient water flow measurements against Eq. (1) (a); Comparison of simulated water flux vs. measured water flux for UUP (b).

The water flux can be spatially quantified at the bottom of the array after the water has been cascading through six tiers of commodity. Figs. 5(a) and (b) show the expected water flux XXXm′′& of commodity “XXX”, normalized by the sprinkler water flow, spkrm& , as a function of the radial distance away from that sprinkler (in the horizontal plane), for CUP and UUP, respectively.

Figs. 5(a) and (b) show that the measured water flux goes down to virtually zero when measured more that 2 meters away from the sprinkler. The water flux measured follows an exponential decay function of the form, e D BA −⋅ , where D is the radial horizontal distance away from the sprinkler (in m), and A and B are fitting parameters. The values for A and B for UEP, UUP, and CUP are listed in Table 1.

Table 1. Fitting parameter for exponential decay functions.

Parameter UUP UEP CUP

A (m-2) 0.0063 0.0044 0.0040

B (m) 0.42 0.69 0.56

The water flow test and modeling results linked sprinkler water flows to actual water fluxes at specific locations along the bottom of the test array. The water flux required for suppression was determined via a set of intermediate-scale fire tests.

Part II Fire

493

(a) (b)

Figure 5. Expected water flux versus horizontal distance away from the sprinkler measured for CUP (a) and UUP (b).

Intermediate scale fires test results

An illustration of the difference in fire development between the three commodities is shown in Figs. 6(a), (b), and (c) for UUP, UEP, and CUP, respectively. Fig. 6(a) shows that a single 200 lpm/bar1/2 sprinkler at 458 lpm is nearly capable of suppressing a UUP “module”. The fire lingered, and eventually grew again. Two subsequent sprinkler activations (5 and 9) suppressed the fire. Fire growth was much faster with the UEP commodity, quickly activating two sprinklers (11 and 5 @ 530 lpm) and suppressing the fire, as shown in Fig. 6(b). The CUP commodity has the benefit of fast vertical fire transport, quickly activating an IRAS. The cardboard boxes then assist with the water film transport down to the fire. Fig. 6(c) shows that a single 200 lpm/bar1/2-sprinkler at 375 lpm can fully suppress the fire.

(a) (b) (c)

Figure 6. HRR profiles for a UUP test (a), UEP test (b), and CUP test (c).

Figs. 7(a)-(c) show the fire at its peak HRR corresponding to Figs. 6(a)-(c). Face ignition can lead to difficulties when relying on IRAS installed in the longitudinal flue only. A single test with horizontal barriers and longitudinal flue IRAS only was unsuccessful. Intermediate-scale tests showed that face ignition presents the worst-case scenario. With face ignition, the initial fire is the furthest radial distance away from an IRAS, which delays the activation of IRAS. All successful tests required sprinklers installed at the face of the array.

Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)

494

(a) (b) (c)

Figure 7. Video frame taken at peak HRR for a UUP test (a), UEP test (b), and CUP test (c).

Testing with UEP showed that a single K360 lpm/bar1/2 sprinkler (pos. 11) flowing at 522 lpm is unable to suppress the fire; while two sprinklers (10 and 11) combining to 1044 lpm suppressed the fire completely. The critical amount of water should therefore be in between these two values. In general, fires involving uncartoned unexpanded plastics develop fast and grow rapidly in both the vertical and horizontal direction. The average HRR at first sprinkler activation for all UEP tests was 1073 kW. The average total heat released was 820 MJ.

UEP showed fast lateral fire spread. Fig. 8 shows the result of an under protected (single sprinkler at 522 lpm) UEP fire 2 minutes after ignition. Clearly, the outer edges of the arrays are involved in the fire, resulting in a test failure.

Figure 8. Failed UEP test, a single IRAS at 522 lpm was not able to control the fire.

Uncartoned unexpanded plastics show a much slower initial fire growth. However, by the time the first sprinkler activates, the fire can be deeply embedded within the array. For this reason, the fire size was bigger at first sprinkler activation compared to UEP and CUP. Not including the test conducted without face sprinklers, the average HRR at first sprinkler activation was 2208 kW for the UUP case. The average total heat released of all UUP tests was 1395 MJ.

CUP has the advantage of the cardboard box aiding rapid fire growth in the vertical direction. The HRR at first sprinkler activation was the lowest among all three commodities, 885 kW. After first

Part II Fire

495

sprinkler activation, the cubic cardboard surfaces help transport the water more effectively compared to UEP and UUP, suppressing the fire quickly. The average total heat released for all CUP tests was only 205 MJ. This is almost 7 times less than UUP and 4 times less than UEP.

The intermediate-scale tests results showed that large-orifice sprinklers at 9.1 m vertical separation can suppressed all three commodities tested. No ceiling sprinklers were used to aid in the suppression process. IRAS sprinklers alone are capable of suppression.

DISCUSSION

The intermediate-scale fire suppression test results were used in conjunction with water film modeling to determine the minimum water flux, on a per pallet load basis, required for adequate protection. All sprinkler activation scenarios observed during the fire suppression tests were modeled. The suppression tests were evaluated on a per-pallet basis for both suppression performance and the prevention of lateral fire spread, i.e., to guarantee protection adequacy for the module tested, the fire would not be allowed to reach the outside boundaries of the test array. The amount of water required to suppress a burning pallet load and the amount of water required to prevent fire spread in the horizontal direction were obtained from the intermediate-scale fire suppression tests. It should be noted that these minimum water requirements are strongly correlated to the fire size at time of IRAS activation, which is dependent on the commodity, IRAS attributes and location in the rack. For instance the average HRR at first sprinkler activation was 1073 and 886 kW for UEP and CUP, respectively. Both commodities showed rapid vertical fire growth with relatively fast first sprinkler activation. The UUP commodity showed much slower initial fire growth and the fire was more established before first sprinkler activation. The average HRR at first sprinkler activation for UUP was 2876 kW.

Sprinkler skipping was prevalent during the intermediate-scale suppression tests. Therefore, sprinkler skipping between neighboring face sprinklers and skipping of all longitudinal sprinklers was assumed during the simulations. Water flow simulations were performed assuming this “skipped” sprinkler activation pattern using monotonically increasing values of sprinkler flow rate until the minimum water flux was achieved. This analysis resulted in a recommended per sprinkler water flow of 455, 455, and 250 lpm for UUP, UEP, and CUP, respectively. The minimum K-factors required for these flows are specified in FM Global Standards [5].

The hybrid approach described above gave reasonable estimates of the minimum per sprinkler flow required for each of the tested commodity. However, this approach, in its limited scale cannot assess the interaction of combustible adjacent commodity nor the interaction between IRAS and ceiling sprinklers. Therefore, the flow requirements determined via intermediate scale testing, with an added safety factor, were verified via 8 to 11 tier-high, large-scale fire-suppression tests below an 18.3 m high ceiling. More details on these tests are presented elsewhere [6]. Determination of a successful test was based on fire spread, the total number of IRAS activations and ceiling level gas temperatures. Protection was found to be adequate for all commodities tested. The flow rate was accurately predicted by combining the computational model with the intermediate-scale suppression test results. The results from this program show a powerful example of the efficiency and effectiveness of the innovative coupled experimental and modeling strategy.

CONCLUSION

This paper focussed on intermediate-scale water flow and fire suppression tests, which were both part of a more comprehensive hybrid test approach aimed at increasing vertical separation between automatic in-rack sprinklers (IRAS). Measured water flow data agreed well with sprinkler flow modeling. Sprinklers with K-factors smaller than 200 lpm/bar1/2 should be avoided. Water modeling

Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)

496

and cold flow water distribution tests demonstrated that longitudinal flue sprinklers are not capable of delivering enough water to the face of the array. Therefore, face sprinklers are required in all cases. Transient water flow tests showed that suppression water reaching the bottom tier behaves as a first order response function. Flow through the flue spaces of the open-frame double row rack storage reaches steady state values faster than the flows that cascades over and through the commodity itself. Both can be modeled as a separate step response function.

A set of intermediate-scale fire suppression tests were conducted to determine the minimum per sprinkler water flows required for adequate IRAS protection. These tests showed that large-orifice IRAS can protect rack stored commodities at 9.1 m vertical increments. A limited set of large-scale (8 to 11 tier-high) fire suppression tests led to a modification to FM Globalʼs Property Loss Prevention Data Sheet 8-9, Storage of Class 1, 2, 3, 4, and Plastic Commodities [5]. The number of expensive, large-scale tests required for these modifications was greatly reduced using a hybrid test approach; combining water flow modeling with intermediate scale fire suppression test results. Furthermore, the increase in vertical separation distance between IRAS levels greatly reduces sprinkler installation costs.

REFERENCES 1. Russel, E. Warehouse Designs are Getting a Makeover (2008, October 15). Retrieved from

http://www.themhedajournal.org/2008/10/15/warehouse-designs-are-getting-a-makeover/(accessed December 1, 2015).

2. Wolin, S. The Return of the In-Rack Sprinkler, Fire Protection Engineering, 3rd Quarter, 2015. 3. National Fire Protection Association Standard 231C (NFPA 231C). Standard for Rack Storage of Materials,

Quincy, MA, 1998. 4. National Fire Protection Association Standard 13 (NFPA 13). Standard for the Installation of Sprinkler

Systems, Quincy, MA, 2016. 5. FM Global Property Loss Prevention Data Sheet 8-9. Storage of Class 1, 2, 3, 4, and Plastic Commodities,

2015. 6. Jamison, K. L., De Vries, J., Baker, W., and Meredith, K. V. In-Rack Sprinkler Fundamentals and the Future

of Protection, Presentation slides, NFPA Conference and Expo, June 22-25, Chicago, IL, 2015. 7. Meredith, K. V., Jamison, K., De Vries, J., Zhou, X. Y., and Wang, Y. Simulating the Interaction of Spray

and Surface Film Flow with Complex Geometries in Fire Suppression Applications, Submitted to the 9th International Conference on Multiphase Flow, May 22-27, Firenze, Italy, 2016.

8. Meredith, K. V., De Vries, J., Wang, Y., and Xin, Y. B. A Comprehensive Model for Simulating the Interaction of Water with Solid Surfaces in Fire Suppression Environments, Proceedings of the Combustion Institute, 34(2): 2719-2726, 2013.

9. Meredith, K. V., Heather, A., De Vries, J., and Xin, Y. A Numerical Model for Partially-Wetted Flow on Thin Liquid Films, Computational Methods in Multiphase Flow, 70: 239-250, 2011.

10. Meredith, K. V., Xin, Y. B., and De Vries, J. A Numerical Model for Simulation of Thin Film Water Transport over Solid Fuel Surfaces, In: Spearpoint, M. (Ed.), Fire Safety Science–Proceedings of the Tenth International Symposium, 10: 415-428, 2011.

11. Approvals, F. M. Approval Standard for Automatic Control Mode Sprinklers for Fire Protection, Class Number 2000, March 2006.

12. Approvals, F. M. Approval Standard of Suppression Mode [Early Suppression Fast Response (ESFR)] Automatic Sprinklers, Class Number 2008, October 2006.

13. Croce, P. A., Hill, J. P., and Xin, Y. B. An Investigation of the Causative Mechanism of Sprinkler Skipping, Journal of Fire Protection Engineering, 15(2): 107-136, 2005.