Warehouse fires have long posed a unique challenge to the fire protection engineering community. The rack-storage configuration, while being practical, economical, and efficient, also produces a challenging scenario with high densities of flammable goods stored at great heights over a vast floor space.

The general approach taken to protect warehouse storage configurations has been that of suppression, where commodity classification is used to design the parameters of suppression necessary to contain or extinguish fires. In commodity classification, full-scale tests on standardized commodities with appropriate fire suppression systems have established acceptable criteria for the protection of stored goods.or extinguish fires. In commodity classification, full-scale tests on standardized commodities with appropriate fire suppression systems have established acceptable criteria for the protection of stored goods.1,2

Figure 1: Fire Development Over a Group A Plastic Commodity5

Figure 2: Three Stages of Burning of a Group A Plastic Commodity6

While implementation and continued development of these standards have greatly reduced the number of warehouse fires, from more than 4,700 a year in 1980 to just 1,200 in 2011, the value of direct property damage has not shown a similar decrease.3 Between 2007 and 2011, storage fires still cost $16 million per month on average.3

As storage facilities continue to grow larger and taller, the practicality of large-scale testing for all possible scenarios has become increasingly impractical. Some means of determining adequate protection from smaller-scale test results as well as relating known protection schemes to new, diverse commodities should be developed.

Unfortunately, the dimensional and material complexity of real world storage commodities is a formidable obstacle. A rigorous approach includes computational fluid dynamics checked against full scale experiments.

In the hopes of systematically reducing the prohibitive costs (actual and computational) associated with this approach, the industry has already established significant momentum in this direction, particularly at FM Global; however, a description of a mixed commodity to use within models has yet to be ascertained.

This study, funded by the SFPE Educational and Scientific Research Foundation, sought to develop a method to ascertain the flammability (including burning rate, flame spread rate, etc.) of a mixed warehouse commodity as a first step towards tackling this problem.

Group A Plastics Test

The classification scheme currently used in the U.S. places commodities into one of seven groups, Classes I–IV for general commodities or Groups A–C for plastic commodities.1,2 The Group A plastic commodity represents the greatest "benchmark commodity” fire hazard, consisting of crystallized polystyrene cups placed within a compartmentalized, corrugated cardboard box.

Although more challenging fire hazards exist, such as expanded meat trays, polyurethane foams, etc., the basis of current commodity classification approaches for plastics is based around this Group A plastic; therefore, it was chosen for this study.4

In testing, the commodity was insulated on all sides except for the front face and ignited at the base, in many ways simulating ignition during an early-stage, rack-storage test. Thermocouples, load cells, cameras, and heat flux gauges provided data that was used to assess flame spread and burning rates of the commodity over time. The mixed commodity was found to progress through three distinct stages of burning, indicated in Figure 1, due to its unique geometry and material distribution.5,6

After ignition of the front face of the commodity, flames spread upward along the front face of the box with little involvement of interior material. Therefore, only the properties of corrugated cardboard are necessary to describe the upward flame spread process, described in detail later.

As the front face of the box chars and falls off, it reveals the first inner layer of segregating cardboard and unexpanded polystyrene cups, indicated as Stage I in Figure 2. This first layer of cardboard also pyrolyzes and burns as it is exposed to flames and outside air, contributing to the burning rate; however, the polystyrene cups inside do not heat sufficiently to ignite, and only begin to soften and melt. The resulting heat-release rate in Stage I of burning increases from 0 to a peak of approximately 25 kW over approximately one minute, with flame heights reaching 1 m (twice the height of the commodity), contributing to rapid involvement of additional fuel above the ignited commodity.

Once the first layer of cardboard burns out, not enough heat has been absorbed by the polystyrene cups to ignite them, nor have flames penetrated the second mixed layer of cardboard and cups; therefore, the heat-release rate and flame heights decay. With only smoldering combustion remaining, the commodity transitions to Stage II, where, on average, low heat-release rates of 10 kW and flame heights of 0.5 m provide a probable opportunity for extinguishment before ignition of the plastic product.

As heat is continually absorbed by the polystyrene cups in Stage II, they eventually absorb sufficient heat to ignite, significantly increasing the heat-release rate of the overall commodity, with a peak of 40-50 kW and observed flame heights of 1-1.5 m, shown in Figure 1 and Figure2 as Stage III. This stage continues as layer after layer of cups is exposed to air, illustrated in Figure 2. The segregated nature of the commodity allows burning to progress in a relatively steady manner, involving cardboard and plastic as earlier-ignited layers burn out.

The segregated nature of the commodity, illustrated in Figure 2, aides not only in a controlled transition between stages for the commodity, but also in access to fuel, providing a somewhat averaged behavior within each of the three stages, pointing to a potential means of simplifying the analysis of the mixed burning of the commodity. In Stage I, for instance, combustion is likely to be described by the geometry and properties of cardboard alone, while in Stage III, it is the burning rate and properties of the plastics, now melted and dripping while burning, that control the burning rate.

Nondimensional Approach

One objective of this work was to develop an approach that was appropriate to measure small-scale fire behavior (at the scale of one or more commodity packages) up to behavior in large rack-storage tests. This significant challenge was not accomplished under this short-duration project; however, some advancement and probable concepts were presented.

The B-number, which appears as a boundary condition at the fuel surface in the classical Emmon’s solution for forced-flow flames over a condensed fuel surface,7 was suggested as a possible means to present the burning behavior of a commodity package and serve as a relatively flammable comparison tool. This dimensionless parameter is a ratio that compares a summation of the various impetuses (e.g., heat of combustion) for burning to a summation of the various resistances (e.g., heat of vaporization) to the process. Originally a purely thermodynamic quantity, its definition can be extended to encompass effects of different heat-transfer processes, including radiative transport.8,9

The unexpected finding of three stages with distinctive burning behaviors added complexity to this approach by necessitating averages of the B-number for each stage of burning (1.8, 1.4, and 1.9 for Stages I, II, and III are reported5). This in some ways simplified matters, as Stages I and II only include flaming combustion and later smoldering of corrugated cardboard, while Stage III is a mixed product of cardboard and polystyrene combustion.

For Stage III, some possible methods for determining the B-number of mixed materials were presented,5,10 but more fundamental research needs to continue in order to establish a firm methodology for utilizing such averaged approximations.



Upward Flame Spread Over Corrugated Cardboard

One parameter of significance for suppression applications in warehouses is the time to sprinkler activation, which largely depends on early-stage flame spread and heat release rates. Focusing on Stage I of the Group A plastic commodity tests, flame spread rates were shown to increase with time to the 3/2 power profile rather than traditional time squared observations.11

Based on experimental results, this behavior was hypothesized to be due to the unique properties of C-flute cardboard, which consists of a corrugated layer of paperboard glued between two flat sheets. As the outer layer burns, it delaminates from the corrugated surface and "curls” directly into the boundary layer, obstructing the flow of hot gasses and projecting the flame outwards and away from unburnt cardboard, shown in Figure 4.

This behavior is significant as the projected flames reduce heating rates above the solid fuel surface, into the preheating region, thereby slowing the development of flame spread and possibly delaying sprinkler activation times. This reduction occurs even though progression of the burning process into the interior of the commodity (including involvement of plastics) will proceed as usual.

The results of these tests have yielded alternative scalings that may be better applicable to some situations encountered in practice in warehouse fires.11 Understanding the time-dependent interaction of both the upward flame spread and in-depth fire growth processes also may be important for proper predictions of fire behavior in warehouses.

Impacts on Practical Warehouse Design

Ultimately, it will take years for the fruits of this labor to directly impact the design of fire protection systems, but some of the general insights should be useful in everyday designs. First, the ultimate flammability or fire hazard of stored commodities may not be as simple as a percentage classification of plastics and cellulosic materials.1,2 The increasing number of exceptions to standard commodity classification listed in NFPA 13 and FM Data Sheet 8-1 is particularly revealing, in that the list of stored items that do not fall under traditional commodity classification schemes is growing; therefore, current methodologies cannot capture all relevant behavior without full-scale test methods.1,2


Figure 3: (left) Front video footage during a representative test. The blue contour across the width indicates the measured height of the pyrolysis region. (right) Image taken from the side of a sample during a representative test. Curling of the front layer of cardboard is visible in both images, but the extent of three-dimensional effects is more clearly seen in the side image.5

Smaller test methods here are shown to capture some of the complex behavior of stored commodities that, with future incorporation of suppression system performance, may be one piece of future system designs. Increasing progress in numerically simulating warehouse fires may help in this regard, but a method for simulating the in-depth combustion of mixed materials must be firmly developed. The ability to extract nondimensional burning behavior from a single warehouse commodity also is one approach for developing a useful comparison between actual stored commodities and standard commodities used in full-scale tests, possibly limiting the number of large-scale tests in the future.

The focus should not be restricted to suppression systems alone because a closer look at individual commodities may be worth considering. For instance, if new packaging could be developed that significantly delays in-depth combustion while still allowing flames to quickly spread upward, triggering sprinkler activation, the large heat-release rates of Stage III may be prevented and the size of necessary extinguishment systems reduced. Similarly, different types of cardboard may be designed that speed or slow upward flame spread.

In essence, by looking at the constituent pieces of a warehouse fire, it may be possible to not only design a suppression system for a fire hazard, but also to modify the fire hazard to match a suppression system in the future. These approaches would require strict control of stored commodities; there are many occupancies where this is possible. Full-scale testing also would be necessary to finally validate these concepts; however, with further modeling and understanding, there is room for revolution in the ways storage occupancies are protected.


This work was supported in part by the SFPE Scientific and Educational Research Foundation and AON Fire Protection Engineering, Inc. Group A plastic commodities were donated by Tyco International, Ltd. The work of Prof. Forman A. Williams, Jonathan Perricone, P.E., Prof. Ali S. Rangwala, Dr. Kristopher Overholt, Todd Hetrick, and others throughout the duration of this project are gratefully acknowledged. Experiments were performed at both the University of California, San Diego, and Worcester Polytechnic Institute.

Michael Gollner is with the University of Maryland, College Park.


  1. NFPA 13, Standard for the Installation of Sprinkler Systems, National Fire Protection Association, Quincy, MA, 2010.
  2. Property Loss Prevention Data Sheets 8-1, Commodity Classification, FM Global, Norwood, MA, 2004.
  3. Karter, M., "Fire Loss in the United States During 2012,” National Fire Protection Association, Quincy, MA, 2013.
  4. Palenske, G., "NFPA 13 Sprinkler System Design Density Curves – Where Did They Come From?," Fire Protection Engineering, Second Quarter, 2012.
  5. Gollner, M., Overholt, K., Williams, F., Rangwala, A. and Perricone, J., "Warehouse commodity classification from fundamental principles. Part I: commodity and burning rates,” Fire Safety Journal, Volume 46, Issue 6, August 2011, pp 305-316.
  6. Gollner, M., "A Fundamental Approach to Storage Commodity Classification,” Master’s Thesis, University of California, San Diego, 2010.
  7. Emmons, H. "The Film Combustion of Liquid Fuel,” ZAMM – Journal of Applied Mathematics and Mechanics, 36 (1956), pp. 60-71.
  8. Torero, J., Vietoris, T., Legros, G. and Joulain, P. "Estimation of a Total Mass Transfer Number from the Standoff Distance of a Spreading Flame,” Combustion Science and Technology. 174 (11) (2002) 187-203.
  9. Jiang, F., Qi, H., de Ris, J. and Khan, M. "Radiation Enhanced B-Number,” Combustion and Flame, Volume 160, Issue 8, 2013, pp. 1510 -1518.
  10. Overholt, K., Gollner, M., Williams, F., Rangwala, A. and Perricone, J., "Warehouse Commodity Classification From Fundamental Principles. Part II: Flame Height Prediction.” Fire Safety Journal, Volume 46, Issue 6, 2011, pp. 317-329.
  11. Gollner, M., Williams, F., and Rangwala, A. "Upward Flame Spread Over Corrugated Cardboard.” Combustion and Flame, 158. 7 (2011):1401-1412.