Duration and energy of flash fire: Testing the standards

Over the past two decades, there have been some reasonable debates and some less reasonable marketing on the duration and energy of hydrocarbon flash fires, despite the fact that existing North American standards were quite clear on the subject. The National Fire Protection Association (NFPA) and Canadian General Standards Board (CGSB) both defined flash fire with identical technical language: the main factors being diffused fuel in air, an ignition source, a rapidly moving flame front and a consequent duration of three seconds or less. 

The NFPA 2112 standard requires manikin test duration of three seconds precisely because it is viewed as the practical upper limit of a flash fire. Groundbreaking research was recently conducted to answer the debate and vet the standards.

The key differentiator between a fire and a flash fire is the fuel. In a fire, the fuel is concentrated (i.e. pool fires, jet fires) and thus is not a significant limiting factor in duration; it will burn for minutes or hours or even days if not actively extinguished. Conversely, in a flash fire the fuel is diffused in air (i.e. gas leak, vapor cloud, combustible dust), meaning it will be consumed very quickly once ignited, as the flame front moves very rapidly from the ignition point to the source — and/or to the limit of the cloud — and goes out. Thus, the duration of heat levels sufficient to ignite flammable clothing or cause second-degree burns to exposed skin is very brief in any single location within the flash.

This short duration is what makes these events survivable without respiratory protection, and with a single layer of flame resistant (FR) clothing, as opposed to SCBAs and turnout gear worn by firefighters (FR clothing will not ignite and continue to burn, but single layer, breathable FR does not provide sufficient insulation against protracted fire exposures). 

While the science and standards seem clear, the sales and marketing of FR clothing, sometimes, do not.  Some companies merely report they pass the NFPA 2112 manikin test (less than 50 per cent total second and third degree body burn at a three-second test duration), while others report the exact percentage with which they pass. Very few spend the time and the money to conduct complete research, and publish graphs which fully characterize body burn from inception of burn through the fabric to, or beyond, failure (>50% burn). 

The manikin test required by NFPA 2112 utilizes the ASTM International F1930 standard test method; ASTM F1930 features a full-size manikin, wearing a standardized coverall, in a burn chamber with propane torches capable of fully engulfing it. The manikin has more than 100 thermocouples evenly distributed over its surface, to predict the extent, severity and location of body burn.

Further complicating matters, there are three ASTM F1930 manikin chambers in North America, two of which are independent university labs; the other is owned and operated by a company with commercial interest in FR clothing. Data can vary marginally from lab to lab, but should not vary significantly when testing is performed in compliance with the standards. Thus, decision makers are faced with performance data that can be presented as a “pass,” a number or a graph. They can even have different data on the same product from different labs. This environment has understandably caused reasonable confusion and disagreement about what is correct and what is relevant. It has also fostered significant leeway in the marketing of performance comparisons; some products prefer to show end-users a particular niche in the performance spectrum, because that is the only place they record an advantage.

The two primary points of contention have been duration and heat flux. The NFPA 2113 standard historically defined flash fire duration as “three seconds or less” predicated on the science of a flame front moving rapidly through a diffuse fuel. As noted earlier, NFPA 2112 accordingly set the pass/fail performance test at three seconds to characterize performance in a worst case scenario.

Heat flux measures the rate of heat energy transfer per unit area per unit time, and is typically expressed as calories/square centimeter second (kilowatts per square meter); it is important to understand that because heat flows, what matters is average heat flux over the course of a single event. Average heat flux of diffused hydrocarbons burning in air was known to be about 2 cal/cm2*sec (84kW/m2), so the standard selected propane fuel and a 2 cal/cm2*sec heat flux. However, when results of this standardized testing are less than favourable to the commercial interests of a fabric, data has been presented at longer or shorter durations, and arguments made about higher or lower heat flux.

Many things are theoretically possible, but standardized testing focuses on what is probable, Independent consensus standards organizations like NFPA and CGSB attempt to quantify and protect the greatest number of people from the most prevalent hazards, based on real world conditions and experience.

Given the frequency and scope of the debate, it was time to quantify the duration and heat flux of actual outdoor flash fires, and confirm whether the standards were on target.  

The first two challenges in initiating such testing would be finding or creating enough field-deployable sensors, and a facility capable of reliably, repeatedly and safely creating the flash fires. The University of Alberta is one of only two independent facilities in North America with an ASTM F1930 flash fire manikin lab. Professor Mark Ackerman was responsible for the flash fire manikin lab at the University of Alberta and developed portable versions of the same thermal sensors used in the Protective Clothing & Environmental Research Facility (PCERF) manikin to create 3-D models of wildfires. These sensors proved perfect for the research. 

With equipment capable of quantifying the answers in hand, what was still needed was an outdoor, full-scale fire field. After an exhaustive search, Texas Engineering Extension Service (TEEX), part of Texas A and M University, in College Station, Texas, was selected. 

TEEX’s Brayton Fire Training Field is the largest industrial fire training facility in the United States, with 279 acres containing dozens of rigs, pipelines, industrial plant structures, tankers, railcars and other testing props — all designed to intentionally create huge fires, allowing firefighters and other emergency personnel to train under real conditions. What Disney World is to children, TEEX is to those of us interested in fire science. The Brayton Fire Field is designed to train industrial firefighters, not conduct research, but TEEX personnel immediately realized the value of the work and agreed to participate.


The experiment

The ideal experimental design would feature a large, open outdoor area with a centrally located pipe to release hydrocarbon vapor; 360 degrees of unimpeded space to allow natural vapor cloud movement in all-wind conditions; externally operable ignition sources to create the flash; mounting surfaces adaptable to thermocouples and data loggers, good sightlines for HD cameras; and independent university labs and personnel. During a scouting trip to the TEEX Fire Field, Prop 66 proved to be nearly perfect and was selected for the experiments. 

The centre of the prop features a large diameter vertical pipe which releases propane, two rings of piping 10 and 25 feet from the propane release point, and an outer ring of torches 40 feet away. The experimental design focused on three concentric rings around the fuel source pipe: an inner ring of double sensors at 10 feet facing both out toward an oncoming flash and in toward the fuel leak, a second ring of single sensors at 25 feet facing out toward an oncoming flash, and the outermost ring of torches which would initiate combustion of the hydrocarbon vapor cloud (see diagram and photo).  The sensors were placed in rings to allow for changes in prevailing wind speed and direction, and remained stationary for the duration of the experiments.  There were also three cube arrays, each of which has five sensors, one per cube face, on a cube six-inch square; the sixth side of the cube houses an adjustable stand to deploy the array. These cubes are mobile, and were placed downwind to ensure maximum exposure to each flash. High-definition cameras were positioned perpendicular to wind direction to best capture movement of the flame front, and were adjusted as conditions dictated. 

Heat flux is fluid. Fires move and swirl and heat ebbs and flows, so each sensor was placed at upper-torso height of an average adult to optimize data capture in the area most relevant to a worker caught in a flash fire. Thirty-one sensors were deployed in each flash, and more than 60 flashes were created over several days. Each sensor measured heat every tenth of a second, which was recorded by a dedicated data logger for each unit. Ackerman then uploaded the data into a computer, which plotted precise flash duration as well as peak and average heat flux. 

The intentional release of huge quantities of propane outdoors can be a dangerous and intimidating process, especially with timing and location of ignition entirely determined by prevailing environmental conditions. TEEX personnel were in control of decisions and procedures at all times, and did an outstanding job of both successfully creating real flash fires, and keeping everyone safe.  

The research had three main goals: quantify duration, average heat flux and confirm that flash fires are a rapidly moving flame front. The concept of a moving flame front is central to understanding the brief duration of a flash fire. The length of time flame is visible overall when a flash traverses a significant distance is longer, and often much longer than the duration at any single location within the flash path (where a worker is standing). 

When a flash fire is initiated by an ignition source, the flame front will propagate from that source until it reaches the limits of the vapor and/or the source of the vapor. As the flame front moves, it quickly consumes all the diffuse fuel behind it. The net effect is that by the time further reaches of the cloud are on fire, the original area is extinguished. Thus, the fire moves like a wave through the cloud. And like a wave at the beach, the total duration is longer than the amount of time it affects a single person standing still in the surf. 

High-definition video cameras and high-speed still cameras are a good start to showing this phenomenon, but they only record the visible light energy of the fire itself. While visuals are nice to have and easy to understand, it is data that really counts. The cube arrays are outstanding in this capacity. While a camera is certainly capable of proving a flame front moves, it does not capture non-visible energy such as radiant heat, that can bracket the flame front and pose a hazard to people. The sensors do, and multiple sensors in the same location — one facing the flash and one facing away, in the “shadow” of the unit — were able to quantify the directional nature of flash fire.

Normally, scientific experimentation strives to control or limit all variables other than the ones being examined. This is necessary for repeatability of the results, which is central to the scientific process.  However, this research was specifically intended to replicate real-world flashes as closely as possible, so some variability in conditions was both inevitable and desirable.  

An experiment can compensate for higher variability with a higher volume of tests. Normal flash fire laboratory protocol for University of Alberta and other laboratories is to run three exposures and average the results; due to the real world design of this research and the variability of the weather conditions, Ackerman conducted in excess of 60 replications.

Environmental conditions fluctuated over a fairly wide range of temperature, wind speed, wind direction and humidity, just as they do in unplanned flash fires. Wind speed in particular is challenging. Little or no wind allowed more propane to be released prior to ignition, causing larger flash fires, while too much wind or too much rapid variation in direction made it more difficult to get the right concentration in the right place to ignite. However, once ignition occurred neither had any significant effect on duration.

Once all the data was uploaded and assimilated, the results were presented as graphs for each individual sensor in each individual flash fire. With so many thermocouples in so many locations through so many flash fires across so many environmental conditions, the data is compelling. No single sensor recorded a flash fire duration of three or more seconds. This was true regardless of position in the flash path, wind speed or direction, or amount of propane released. The vast majority of exposures were two to 2.25 seconds.


Evaluating the results

There are two ways to evaluate the duration results. The most conservative way, which yields the longest duration, is to count all the time the sensor records heat above ambient conditions, which was the protocol for this work. A second alternative is to approach the data from a perspective closer to the lab flash fire and arc flash analysis, which are predicated on avoiding or minimizing second-degree burn and worse. This can be done by looking at the total time each exposure spends above 1.2 cal/cm2*sec (50kW/m2), which is the threshold for second-degree burn. This will typically yield a slightly lower duration, and represents the amount of time that flash was directly hazardous to exposed human skin. We chose to examine and present the data from the more conservative perspective, but the two approaches generally yield data which differ by only fractions of a second.

Heat flux averages are discernible within a single location in a single exposure, across multiple locations in a single exposure, and across all exposures. The individual peak data showed somewhat greater variability than duration data, but average heat flux within an exposure and across all exposures were very consistent. Heat flux averaged 2 cal/cm2*sec (84kW/m2).

The pictures and video clearly showed that a flash fire is a moving flame front, comprised of two or three sections.  At ignition, there is a portion which is burning and a portion which has not yet ignited. Then, as the flame front moves into an area of fresh fuel, it becomes a three-phase event. Behind the flame front is an area where the fuel has been consumed and the fire is out, then an area of flame, and ahead, an area of unignited fuel into which the flame is moving. The cube array sensors were able to confirm this observation. 

A moving flame front is, by definition, directional. That is, if it is moving toward your face rapidly, and self-extinguishing as it moves by consuming all the fuel, then you would not predict burns to your back.  If you are sprayed with a fire hose for a second or two while standing still, one side of you will be wet and one side of you will be dry.  Similarly, if a flash fire is a moving flame front, it is directional and would be predicted to show a high heat flux on the sensor surface facing the flash, and a low heat flux on the surface on the back of the cube, in the “shadow” of the unit. This is exactly what the data showed. In each case, the side of the cube facing the oncoming flash fire recorded elevated heat flux consistent with the single sensor units, but the side of the cube facing away from the flash (a mere six inches from the high heat sensor) recorded very little or no elevated heat flux. 

NFPA and CGSB each created standards to address the flash fire hazard in the mid and late 1990s. These standards committees were staffed by subject matter experts and highly experienced industry personnel, resulting in excellent non-commercial guidance and test protocol. The parameters of that testing were based on the best available science and accident investigations. They intended to require passing performance against a worst-case flash fire, which they defined as a rapidly moving flame front lasting typically three seconds or less.  They set the heat flux at 2 cal/cm2*sec(84kW/m2),  because that is the average for hydrocarbon flash fire in air.
 
The research discussed above was driven by marketing, which has created a climate of confusion against these standards. The experiments were designed and executed by personnel from University of Alberta’s PCERF, one of the top flash fire labs in the world, at the Texas A and M Brayton Fire Field and the conclusions are clear. Hydrocarbon flash fires are moving flame fronts, with average heat flux of about 2 cal/cm2*sec(84kW/m2), and durations below three seconds.

The standards are correct.  

-----------
Scott Margolin is the international technical director at Westex, a global developer of flame resistance fabrics based in Chicago.