FPEeXTRA Issue 79

Considerations for the Use of Heat Release Rate Data in Engineering Analysis

By: David Morrisset, Ian Ojwang, Jonny Reep, Rory Hadden, PhD and Angus Law, PhD, CEng

View full PDF here

This article is the first in a series of publications involving work sponsored by the SFPE Foundation investigating the uncertainty and repeatability of HRR data. Further details of this experimental campaign involving upwards of 25 repeat trials of upholstered furniture HRR will be presented in subsequent publications; this preliminary article serves to outline particular considerations when using experimental HRR data in fire engineering analysis.

HRR as an input

Heat release rate is an essential input for many engineering analyses, with design fire scenarios often characterized using the HRR of a fuel package as a function of time [1,4]. Typical methods which rely on heat release data include ceiling jet correlations commonly used in determining sprinkler activation time [5], plume correlations used in applications such as zone model calculations [6], and correlations used to approximate the flashover potential of a compartment [7,8].  HRR curves are also integral input to other computational models such as the Fire Dynamics Simulator (FDS) [9].

When using HRR in engineering analysis, the user may choose whether to use HRR data from an actual experiment or a generic HRR curve such as an idealized  fire [1,10]. These two options require the user to make a decision regarding the precision with which a given fire scenario should be modelled. This article is focused primarily on the use of experimentally determined HRR data.

It is widely acknowledged that the HRR for complex objects will be strongly dependent on the mode of ignition and the environment in which the object is burning. Thus, a specific HRR curve is inherently tied to the particular experiment and the events that occurred in that instance. Understanding the extent to which different factors will affect the measured HRR can help fire engineers gain confidence in their design and establish bounds for a sensitivity analysis when using HRR data. The following experimental results will identify the influence of factors such as ignition location and the placement of the item relative to obstructions such as walls.

Considerations for Experimentally Determined Data

The resulting HRR curve for a fuel package such as an upholstered chair is a product of various stochastic events that influence the burning rate. These observable events manifest themselves in the measured HRR as seen in the data presented in Figure 1. The HRR data seen in Figure 1 was measured for an upholstered chair that was ignited beneath the seat cushion using a small (<1 kW) Bunsen burner flame. Several key events linked to changes in the heat release rate are marked, showing that changes in the HRR can be aligned to observed events. In other words, determining key physical occurrences that manifest in the HRR data provides essential context to the observed HRR curve.

Figure 1. The transient HRR response of an upholstered chair ignited with a Bunsen burner located beneath the seat cushion. Individual characteristics of the HRR curve can be linked to specific physical occurrences (e.g., the ignition of the backrest at approximately 275 seconds).

The key events in the burning history are as follows:

• Backrest ignition:

The position of the burner first ignites the seat cushion. Once the fabric of the backrest ignites, flames spread up the backrest leading to a rapid increase in HRR.

• Burn-through of the backrest:

Between the backrest igniting and the peak HRR, flames are observed to have burned through the backrest. At this point all of the foam in the backrest becomes involved in the fire and there is another rapid growth to the peak HRR.

• Peak HRR:

The peak HRR is achieved once all of the foam in the backrest becomes involved.

• Burnout of foam:

Shortly after the peak, the HRR rapidly decreases as the foam is consumed leaving only the oriented strand board (OSB) frame of the chair remaining.

• Backrest collapse:

The OSB frame supporting the chair ultimately chars and weakens to a point at which the backrest frame collapses, exposing fresh material in the frame to burn.

• Localized flame and smoldering:

After the frame collapses, the chair displays the characteristics of a bonfire showing localized area of flaming and smoldering – no longer displaying any resemblance to the original chair.

The time at which each of these key events occur can be found in Table 1. The significance of each event is not simply the time at which it occurred, but the subsequent effect each event had on the HRR curve. The observation of the backrest igniting, for example, showed a transition to a rapid growth in HRR from the flames spreading up the surface of the backrest.

In isolation, a user has very little basis to attribute transient increases or decreases in HRR to any physical phenomenon. The significance of these key events on the HRR becomes further evident when comparing two repeat trials of the same upholstered chair. Figure 2a illustrates two trials of identical upholstered chairs ignited using the same testing procedure (<1 kW burner impinging beneath the seat cushion). Table 1 outlines the occurrence of these key events for both trials seen in Figure 2a. The first trial (the same data presented previously in Figure 1) can be seen in black and the repeat trial under notionally identical conditions can be seen in red (Trial 2).  

Comparing the results of the two trials shows:

• The initial growth rate for Trial 2 is much greater, displaying an initial peak in HRR followed by a short decay period.
• The peak HRR of the Trial 2 is nearly 30% lower than in Trial 1.
• Both trials show a local peak in HRR when the backrest collapses.
• Both trials show very similar HRR decay during the localized flaming and smoldering phase.

Figure 2. a) A comparison of two repeat trials of the same upholstered chair ignited beneath the seat cushion using the same ignition source. b) Two identical upholstered chairs ignited with the same ignition source; one trial was ignited from beneath the seat cushion (black) and another trial was ignited from the front of the seat cushion (red).

Even using identical chairs under notionally identical experimental conditions, these two “repeats” exhibited stochastic variations in aspects such as growth rates and peak HRR. Physical meaning can be attributed to the variations between the trials once the HRR data is compared with the key events described in Figure 1. For example, the variation in the growth rate between the two trials can be explained by observing the time at which the chair backrest ignited. Both trials therefore show notable similarities in HRR behavior when considering these key events.

A third trial was conducted using the same ignition source. For this case the burner impinged on the front corner of the seat cushion for 20 minutes, rather than being located beneath the seat. Images comparing these two ignition sources can be seen in Figure 4. This configuration was compared against Trial 1 as shown in Figure 2b (black line is Trial 1, red line is Trial 3). The timing for the key events which occur across all three trials is shown in Table 1.

Table 1. A comparison of the key events across the trials discussed in Figure 3.

Comparing the Trial 3 against Trials 1 and 2 it was found that:

• Unlike the trials ignited from beneath, Trial 3 displayed an immediate jump to the peak HRR after the burner was removed.
• All three trials showed a localized HRR peak when the backrest collapsed.
• The HRR behaviors of all three trials converge in the localized flaming and smoldering phase.

The data presented in Figure 1 and Figure 2 were determined in the open (i.e., no obstructions to influence entrainment to the burning item). However, discrete fuel packages in real fires are often obstructed by nearby items or are often situated alongside external walls. Figure 4 compares repeat trials conducted using identical upholstered chairs (same chairs used in Figure 2 and Figure 3) conducted both in the open under the calorimetry hood and placed in a wall corner constructed from gypsum plasterboard.

Figure 3. A comparison of the experimental configurations discussed in this work. From left to right: ignited from beneath, ignited from the front of the cushion, and ignited beneath while placed in a corner configuration.

Figure 4. A comparison of results from an upholstered chair trial both in the open and a trial ignited in an identical fashion but placed in the corner of two gypsum lined walls: a) showing the HRR for both the open and corner configurations. b) showing total heat flux measurements taken 0.5 m from the chair at two different heights (0.4 m and 0.9 m) for both the open and corner configurations.

Comparing the open and corner configurations found that:

• The observed trends in HRR time histories for both the open and corner configurations follow similar behaviors to one another. There are slight variation in peak HRR, but these occurred within the scatter seen between repeats for identical conditions in Figure 1
• The corner configuration resulted in reduced entrainment to the burning item, which led to larger flame heights as seen in Figure 3.
• The reduction in entrainment is not immediately evident in the HRR data, but is clearly seen in the total heat flux from the fire in Figure 4

The HRR curves seen between the open and corner configurations appear to behave similarly in this case. The increase in heat flux, however, can be directly linked to the presence of the wall behind the chair. The corner configuration led to increased flame heights from reduced entrainment and additional radiation from the walls themselves being heated by the fire. These results therefore warrant careful consideration of environmental factors when applying analysis such as heat flux from a target fuel package.

Conclusions

Data presented here illustrate the variation that can exist between repeat trials of identical fuel packages under the same experimental conditions as well as the variation observed when aspects of those experimental conditions are changed (e.g., changing the ignition location dramatically changing the HRR curve). The observed variations in this study are only for a single item of furniture – the stochastic variations in HRR behavior can be even more significant for more complex assemblies. Discussions have been limited to stochastic variation in observed results while users may also need to considered measurement uncertainty in HRR data, which has been discussed in detail elsewhere [3,10]. The results presented here illustrate the significance in using both experimental data and the observation of key events in the interpretation of HRR data and in the application of such data in further analysis. Engineers must also consider how relevant environmental conditions may influence the HRR for the specific fire scenario in question.

Experimentally determined HRR data still provides critical insight into the burning behavior of complex fuel packages. The use of experimental HRR data can be balanced between striving for a realistic scenario and the various uncertainties, including variability, involved in the data. A higher degree of insight into the HRR data can be gained through observation of key physical phenomenon that drive the burning process. These key physical drivers will dictate the resulting HRR curve and indicate which aspects of a given HRR curve are more or less sensitive to the occurrence of such events.

Therefore, the authors would suggest that the HRR curves for design fires be selected with consideration of the assumptions built into any given design fire scenario (e.g., the precision of experimentally determined data, or the lack of precision of an  fire). As a fire engineer, using experimental data may benefit from the following questions:

• How sensitive are the results of any analysis (e.g., sprinkler activation, two zone models, FDS) to the assumed HRR curve?
• What assumptions have been made using this data (e.g., would this item behave differently in a compartment fire scenario)?
• What uncertainty might be placed on the assumed HRR curve (both measurement uncertainty and stochastic variation)?
• Do those uncertainties fundamentally change the outcome of the analysis?

Acknowledgments

The authors would like to thank the SFPE Foundation for their generous support on this project. We would also like to acknowledge both the financial and technical contributions of the Edinburgh Fire Research Centre.

David Morrisset, Ian Ojwang, Jonny Reep, Rory Hadden, PhD ,and Angus Law, PhD, CEng are with the University of Edinburgh, School of Engineering

References

[1]         V. Babrauskas, R.D. Peacock, “Heat release rate: the single most important variable in fire hazard” . In: Fire Saf. J. 18 (1992) , pp.255–272.

[2]         V. Babrauskas, Heat release rates, in: SFPE Handb. Fire Prot. Eng., Springer, 2016: pp. 799–904.

[3]         “NIST Fire Calorimetry Database (FCD)” (n.d.). https://doi.org/10.18434/mds2-2314.

[4]         G. V Hadjisophocleous, J.R. Mehaffey, Fire scenarios, in: SFPE Handb. Fire Prot. Eng., Springer, 2016: pp. 1262–1288.

[5]         R.L. Alpert, Ceiling jet flows, in: SFPE Handb. Fire Prot. Eng., Springer, 2016: pp. 429–454.

[6]         G. Heskestad, Fire plumes, flame height, and air entrainment, in: SFPE Handb. Fire Prot. Eng., Springer, 2016: pp. 396–428.

[7]         R.D. Peacock, P.A. Reneke, R.W. Bukowski, V. Babrauskas, “Defining flashover for fire hazard calculations” . In: Fire Saf. J. 32 (1999) , pp.331–345.

[8]         V. Babrauskas, R.D. Peacock, P.A. Reneke, “Defining flashover for fire hazard calculations: Part II” . In: Fire Saf. J. 38 (2003) , pp.613–622.

[9]         K. McGrattan, S. Hostikka, R. McDermott, J. Floyd, C. Weinschenk, K. Overholt, “Fire dynamics simulator user’s guide” . In: NIST Spec. Publ. 1019 (2013) , pp.1–339.

[10]      R.A. Bryant, G.W. Mulholland, “A guide to characterizing heat release rate measurement uncertainty for full-scale fire tests” . In: Fire Mater. An Int. J. 32 (2008) , pp.121–139.