Experimental and numerical study of fire dynamics in air-tight buildings

Air-tight compartments are encountered in both residential and industrial buildings. The demand for energy-saving promotes the use of air-tight residential buildings to limit the energy consumption. Besides, a high confinement level is also applied in industrial constructions such as nuclear power plant buildings to prevent the radioactive substance from leaving the building. However, from the standpoint of fire safety, the airtightness will introduce new fire risks. Unlike in compartments with sufficiently large openings, the fire-induced pressure can rise substantially in air-tight compartments due to thermal gas expansion and the difficulty in releasing hot combustion products through small openings. The overall objective of the study is to investigate fire dynamics in air-tight compartments, with a specific focus on fire-induced pressure variations, which is generally less considered a concern in fire research. The thesis contains experimental research, theoretical analysis and numerical studies. In terms of the reduced-scale NYX experiments: an extensive experimental campaign was designed and conducted at IRSN (French Institut de radioprotection et de Sûreté Nucléaire) to systematically study the influence of heat release rate (HRR) evolutions, mechanical ventilation configurations and boundary conditions on fire dynamics, focusing on the pressure variation. In general, the overall fire-induced pressure evolution trends have been shown to be similar in the tests. Qualitatively, the pressure follows the evolution trends of the fire HRR. The pressure evolution is characterized by an over-pressure peak followed by an evolution to a quasi-steady state, then an under-pressure peak and a relaxation to the initial pressure level. Four fire HRR evolution parameters are considered in the experiments, namely the fire growth/decay rate coefficient, the maximum steady state heat release rate, the fire growth/decay exponent and the steady state fire duration. In general, the higher the fire growth/decay rate coefficient, the higher the obtained over-pressure peak and under-pressure peak. For a given maximum fire HRR, the magnitudes of the over-pressure peak and under-pressure peak increase with increasing fire growth/decay rate coefficient, until a plateau value is reached. The magnitudes of the over-pressure peak and under-pressure peak increase linearly with increasing maximum steady state HRR. For a set fire growth/decay rate coefficient, the higher the maximum steady state HRR, the longer the time for the pressure to reach the peak value. For cases with fire growth/decay exponent below 1, the pressure reaches peak values before the HRR reaches a steady state (maximum value or zero) because the net heat gained in the gas phase already reaches its peak values then. In cases with fire growth/decay exponent above 1, the pressure reaches peak values at the moment where the HRR reaches the steady state. The fire growth/decay exponent has an impact on the under-pressure peak, while it hardly affects the over-pressure peak. The over-pressure is not affected by the steady state fire duration because the temperature, and hence heat loss, is small at the fire growth phase and the HRR evolution is the dominant influencing factor. The under-pressure does depend on the fire duration, through the heat losses, but becomes almost independent of the steady state fire duration when the fire lasts long enough. The influence of mechanical ventilation configurations is studied with different ventilation flow resistances and initial ventilation flow rates. The magnitudes of pressure peaks are strongly affected by the mechanical ventilation settings. Correspondingly, the fire-induced pressure has an influence on the ventilation flow rates. Higher pressure peaks are obtained with increased ventilation flow resistance. It has been illustrated that not only the total, or ‘equivalent’, flow resistance is important, but also the individual flow resistances in the admission and extraction ducts separately. A higher imposed initial flow rate leads to a higher equivalent flow resistance and hence higher over-pressure and under-pressure peak values. The enhancement and reduction of ventilation flow rates depend on both the fire-induced pressure and ventilation flow resistances. Experimental results show that, with the same HRR evolution, the ventilation configurations do not strongly affect gas temperatures. The change of boundary conditions (i.e., wall properties) in the current experiments affects the temperature evolution but hardly influences the pressure variation and ventilation flow rate. Full-scale experiments were conducted at RPA (Régie Provinciale Autonome) in Belgium in cooperation with the University of Mons. The leakage area of the compartment is measured under different pressure conditions and is proved to increase with pressure rise. In the methanol pool fire tests with different square fuel pan sizes (i.e., 0.3 m × 0.3 m and 0.5 m × 0.5 m), the pressure inside the fullscale compartment reaches peak values of 133 Pa and 697 Pa for the small and the large pool fire cases, respectively. The pressure is high enough to prevent evacuation. In order to better interpret the fire-induced pressure variation in air-tight compartment fires, the energy balance equation is discussed starting from a commonly used form, taking into account the mutual influence between compartment pressure and ventilation behavior. The obtained pressure formulation, under certain assumptions, can explain the observed linear relationship between pressure variation and fire evolution. Based on the mathematical development, the difference between fire-induced pressure and initial compartment pressure is expected to depend on the net heat gained in the gas phase. It is also found from the theoretical reasoning that the mass flow rate difference between admission and extraction ducts depends on the net heat gained in the gas phase, and is not strongly related to the ventilation resistance. These are subsequently validated by experimental results. Numerical studies were conducted using FDS (Fire Dynamics Simulator). Different versions were employed, i.e., 6.5.2, 6.5.3 and 6.7.5. The quantitative assessments of the simulation results (compared with reduced-scale experiments) indicate that the fire-induced pressure and ventilation flow rates are less sensitive to mesh cell size and can well be reproduced, while the temperature is strongly dependent on mesh cell size. The predictions of CO2 and O2 are better than those of CO. In the full-scale methanol pool fire simulations, the discrepancies between experimental and numerical over-pressure (under-pressure) peaks are 6.5 % (40 %) and 1.6 % (21.7 %) for the small pool fire case and large pool fire case, respectively. The magnitudes of the simulated leakage flow rate reach a maximum of 65 m3/h for the small pool fire case and a maximum of 200 m3/h for the large pool fire case, which are comparable to mechanical ventilation flow rates. The parametric numerical study based on full-scale experiments (which have been conducted by UMons) with mechanical ventilation indicates that the combination of the real fan curve and dampers, as in the set-up of the full-scale tests, provides better results than a surrogate. The settings of reference leakage area and leakage pressure exponent have an impact on the pressure peaks. The adjacent room pressure also reaches a high level although the door between rooms is closed. The scaling analysis indicates that the scaled-up fires corresponding to the t2 fires in reduced-scale NYX experiments are classified as medium fires to fires that grow faster than the ultra-fast classification. A preliminary scaling analysis of pressure based on energy balance illustrates that the fire-induced pressure should be scaled as the geometry scale, which is confirmed by simulations of cases in different scales. The present study is expected to: 1) give more insight into the fire dynamics, especially fire-induced pressure variations, in air-tight compartments; 2) provide an assessment of numerical modeling of the fire-induced pressure with FDS and to provide extended analyses using FDS; 3) raise attention in considering the potential risk associated with fire-induced pressure during fire safety design. Besides, the experimental results presented in this thesis serve to provide data for further model validations