Characteristics of Externally Venting Flames and Their Effect on the Façade: A Detailed Experimental Study

Abstract

In a compartment fire, externally venting flames (EVF) may significantly increase the risk of fire spreading to adjacent floors or buildings; EVF-induced risks are constantly growing due to the ever-increasing trend of using combustible materials in building facades. The main aim of this work is to investigate the fundamental physical phenomena associated with EVF and the factors influencing their dynamic development. In this context, a series of fire tests is conducted in a medium-scale compartment-façade configuration; an n-hexane liquid pool fire is employed, aiming to realistically simulate an “expendable” fire source. A parametric study is performed by varying the fire load density (127.75, 255.5 and 511 MJ/m²) and ventilation factor (0.071 and 0.033 m^3/2). Emphasis is given to characterization of the thermal field developing adjacent to the façade wall. Experimental results suggest that the three characteristic EVF phases, namely “internal flaming”, “intermittent flame ejection” and “consistent external flaming”, are mainly affected by the opening dimensions, whereas the fuel load has a notable impact on the fuel consumption rate and heat flux to the façade. Fuel consumption rates were found to increase with increasing fire load and opening area, whereas the global equivalence ratio increases with decreasing opening factor. The obtained extensive set of experimental data can be used to validate CFD fire models as well as to evaluate the accuracy of available fire design correlations.

Appendix: Estimation of Experimental Errors

The ASME methodology [48] has been used to estimate the uncertainty of the measurements presented in this work. The total (expanded) measurement uncertainty (U _t ) is estimated using Eq. (A.1), where B _t is the root sum square of the elemental uncertainty estimated by statistical methods (e.g., systematic or bias uncertainties) and S _t is the root sum square of the elemental uncertainty estimated using non-statistical methods (e.g., scientific judgment, manufacturer’s specifications, calibration reports, random uncertainties). The expanded uncertainty is obtained by multiplying the combined standard uncertainty by a coverage factor k; in this work, a coverage factor value of k = 2 was used, aiming to achieve a 95% confidence interval (2σ) for the total uncertainty [49]. The values of the root sum square (B _t or S _t ) of the various types of elemental uncertainties (B _i or S _i ) are calculated using Eq. (A.2).

$$ U_{t} = \, \pm k\left[ {\left( {B_{t} /2} \right)^{2} + S_{t}^{2} } \right]^{1/2} $$

(A.1)

$$ B_{t} = \, \left( {B_{1}^{2} + B_{2}^{2} + B_{3}^{2} + \ldots } \right)^{1/2} $$

(A.2)

Gas and wall surface temperatures were measured using K-type thermocouples connected to the signal acquisition system via extension wires. Bare-bead thermocouples are assumed to have negligible statistical uncertainties (B _t  = 0) [49]. In terms of the S _t uncertainties, the K-type thermocouples used in the tests exhibited a standard calibration uncertainty ±2.2°C or ± 0.75% of the measurement reading (the largest value is taken into account) for a 99% confidence interval (3σ); in this work, the respective values of ±1.5°C or ±0.5% have been used to achieve a 95% confidence interval (2σ). Another source of error in bare bead thermocouple measurements is radiative losses. Attempts of estimating the error due to radiation can be found in the literature [50]; in the current study radiative loss errors were assumed to be 0% to −6%. In addition, a ±3% random error was introduced in the analysis.

Measurements using heat flux sensors are subjected to errors due to sensor properties, calibration traceability, quality assurance and measurement-related uncertainties. The initial calibration accuracy of the heat flux sensors used in the tests was ±3%. In addition to this error, the most significant error sources are non-linearity, convection and the radiative heat transfer balance. A ±1% sensitivity error was employed, based on the assumption that the mean air speed was 0.5 m/s; also, a ±6% convective error and a ±3% random error were also taken into account.

In terms of the gas concentration measurements, the real-time gas analyser used in the tests exhibited an equipment uncertainty of ±1%. The calibration gases used for zero and span calibrations exhibited negligible uncertainties, as reported by the manufacturers. Zero and span calibrations performed prior to each experiment showed a systematic standard uncertainty of ±1%. Random and mixing/averaging errors due to the sampling lines, equal to ±3% and ±5%, respectively, were the larger sources of uncertainty. Estimated values for all components of measurement uncertainty are tabulated in Table A1. Indicative values of the estimated experimental errors for the gas temperature inside and outside the fire compartment and the heat flux to the façade surface are presented in Figures 6 and 14.

Table A1 Summary of Measurement Uncertainty Components