Showing papers by "Bart Merci published in 2019"

Towards predictive simulations of gaseous pool fires

Abstract: Focusing on advancing predictive fire modeling, large eddy simulations using improved approaches related to thermophysical, turbulence, combustion and radiation modeling are presented. More specifically, the consideration of a non-unity Lewis number, the use of the Hirschfelder-Curtiss diffusion model, the inclusion of differential diffusion and Soret effects, the application of the dynamic Smagorinsky model with a variable turbulent Prandtl number, an eddy dissipation concept model for combustion and the weighted-sum-of-gray-gases model for radiation have been included in a modified version of fireFoam 2.2.x. A comprehensive comparison between the predictions of the modified code against the standard version of fireFoam and experimental data of a medium-scale 22.6 kW methanol pool fire is presented. Evaporation modeling is not yet included.

Development of an Integrated Risk Assessment Method to Quantify the Life Safety Risk in Buildings in Case of Fire

Abstract: An integrated probabilistic risk assessment methodology is developed for the purpose of quantifying the life safety level of people present in buildings in the context of fire safety design. Multiple risk based concepts and tools have been developed in previous research to objectify performance based design methods for simple building types and layouts. However, these available models lack an integrated approach for challenging building designs and moreover they are not adequately coupled, most often resulting in a significant computational effort. Hence, there is a need for a pratical and efficient framework for dealing with complicated building layouts and different occupancy types. Therefore, a computationally efficient quantitative risk assessment method is developed that provides a framework by combining deterministic sub-models and probabilistic techniques to quantify the fire safety level by means of failure probabilities, individual and societal risk. The deterministic framework is supported by analytical and numerical models. The probabilistic framework is supported by response surface modelling, sampling techniques and limit state design. Following the theoretical description of the model, a case study of a five storey commercial shopping mall of 25,000 m² is elaborated and discussed as proof of concept. Multiple fire, building and occupant variables are implemented in the model. Three different fire safety designs are compared, resulting in quantified risks between 10-6 and 10-8. The case study proves the validity of the newly developed integrated methodology for this type of buildings and its benefits in fire safety engineering.

LES Study of a Turbulent Spray Jet: Mesh Sensitivity, Mesh-Parcels Interaction and Injection Methodology

Abstract: This study focuses on Large Eddy Simulations (LES) in the Eulerian-Lagrangian framework of turbulent spray jets. The choice of the numerical grid relates both to turbulence level description and parcels-grid cells interaction. The objective of the present work is to determine a case set-up capable of predicting flow fields of a turbulent n-heptane spray jet for which a set of experimental data in non-reactive conditions (isothermal) is available (Shum-Kivan Proc. Combust. Inst. 36(2), 2567–2575 (2017)). The study first focuses on the LES grid, based on simulation with only gas phase, and subsequently on the choice of the suitable number of parcels to describe the spray. The droplets behavior is analyzed using the particle Stokes number at different axial locations. Furthermore, a variation in the existing injection methodology, as available in OpenFOAMⓇ, reveals a beneficial impact on the prediction of the experimental data.

Numerical simulations of a full-scale cable tray fire using small-scale test data

Abstract: This paper presents a computational fluid dynamics (CFD)-based modeling strategy for the prediction of cable tray fire development. The methodology is applied to a set of five horizontal trays (each 2.4-m long and 0.45-m wide) that are positioned with a 0.3-m vertical spacing and set up against an insulated wall. Each tray contains 49 power PVC cables. Ignition is performed with an 80-kW propane burner centrally positioned at 0.2 m below the lowest tray. A collection of four groups of cables per tray (made of one homogeneous material) is considered. These groups are separated by longitudinal slots of air to "mimic" their relatively "loose arrangement." The thermal properties and surface ignition temperature are estimated from cone calorimetry (CC). When the ignition temperature is reached, the cables burn according to a prescribed heat release rate per unit area (HRRPUA) profile obtained from CC, as is or in a modified shape. A realistic flame pattern is predicted. Furthermore, using only data from CC, the peak HRR is underpredicted, and the time to reach the peak is overpredicted. The proposed "design" for the modified HRRPUA CC-profile significantly improves the results.

Call for participation in the second workshop organized by the IAFSS Working Group on measurement and computation of fire phenomena

Abstract: Early 2015, a new initiative called ‘‘the IAFSS Working Group on Measurement and Computation of Fire Phenomena’’ (aka the MaCFP Working Group) was launched (http://www.iafss.org/macfp/). This initiative is endorsed and supported by the International Association for Fire Safety Science (IAFSS, http://www.iafss. org). The first workshop organized by the MaCFP Working Group was held in June 2017 as a pre-event to the 12th IAFSS Symposium in Lund, Sweden. Details are found on https://iafss.org/3770-2/ and in Ref. [1]. The primary objective of this letter is to engage the members of the fire research community to participate in the second MaCFP workshop, scheduled on April 25–26 2020 as a pre-event to the 13th IAFSS Symposium in Waterloo, Canada (http://iafss2020.ca). Continued updated information on the MaCFP Working Group effort is found at http:// www.iafss.org/macfp/.

Large Eddy Simulations of a Set of Experiments with Water Spray-Hot Air Jet Plume Interactions

Abstract: Large eddy simulations of water spray-hot air jet plume interactions, as obtained with FireFOAM 2.2.x, are presented. Three hot air jet plumes, with thermal powers of 1.6, 2.1 and 2.6 kW, are examined, interacting with a water spray with discharge rate of 0.084 lpm. A systematic comparison between simulations and experiments involving only the hot air jet plumes, the water spray alone and the combination of the two has been performed in order to evaluate the predictive capabilities of FireFOAM. Overall, the code is capable of predicting well the mean values of the hot air jet plumes but deviations are evident for the rms values. Discrepancies in the predictions of the volume fluxes in the near-field for the water spray alone case are observed if the experimentally reported injection angle is used. Improvements are observed if the injection angle is modified based on the experimentally reported data in the near-field. The interactions between the hot air jet plumes and water sprays, are characterized by the location of the interaction region. The interaction boundary moves up from the base of the plume by increasing the convective heat release rates. The simulation results follow the experimental trend but deviate up to 26% due to the differences in the predicted hot air jet plumes and spray characteristics.

Multimodale datafusie voor tijdsruimtelijke analyse van het brandgedrag

Abstract: This appendix provides an overview of basic fire behavior concepts and physics, to give some background and to guide the reader through this dissertation. Subsequently an introduction is given into enclosure fires and finally a preface is given into CFD and zone modeling for fire forecasting concept. Major parts of this chapter are inspired by the work of Merci et Beji. [1].

Analytical Modelling of the Effect of In-Depth Radiation within a Liquid Layer in the Case of a Pool Fire

Abstract: In this paper, we present a ‘simplified’ approach for the numerical modelling of the convective currents that occur within a liquid fuel in the case of a pool fire and which are induced by in-depth thermal radiation. This approach is based on the concept of ‘effective’ thermal conductivity, which is calculated herein based on the analytical solution of a steady-state one-dimensional heat conduction equation including a source term for in-depth radiation. This solution leads to a temperature profile which displays a horizontal liquid layer (of a given depth) that is bounded by a temperature that is higher at its bottom than its top. This thermal structure generates Rayleigh-Benard instabilities which enhance heat transfer within the liquid. This effect is modeled via an increase of the ‘actual’ thermal conductivity of the liquid by a dimensionless heat transfer number, namely the Nusselt number. The Nusselt number is calculated based on the ‘classical expression’ of the Rayleigh number for the case of a ‘horizontal cavity heated from below’. The paper provides the details of the derived solution for the ‘effective’ thermal conductivity along with examples of application to several fuels.

Synergy between mesh refinement and parcels number in LES-CMC of turbulent reactive sprays

Abstract: The interaction of spray droplets with the surrounding turbulent flow is crucial in spray combustion, because it affects the evaporation rate and, consequently, the process of vapor mixing between air and fuel. Depending on the accuracy of the adopted computational model, some intrinsic dynamics of the spray flame, which are experimentally observed, might not be adequately captured. In the present work this issue is investigated for an n-heptane hollow cone spray (dp ∈ [0.5 − 80 μm]), issuing from a central nozzle surrounded by an annular turbulent non-swirling jet. The set-up belongs to the Coria Rouen Spray Burner (CRSB) database [1], for which a set of experimental data is available both in reactive and non-reactive (isothermal) conditions. Recently, a set of Large Eddy Simulations (LES) with the Eulerian-Lagrangian (EL) point particle approach of the CRSB was performed for isothermal conditions [2]. In [2] four different meshes (M1-M4), in order of growing refinement were evaluated and a threshold for the cell size at the injection plane above which the LES is reliable (∆ ≤ 0.2 mm) was found. Furthermore, once the LES cell size was set, another threshold for the minimum number of computational parcels (N?̇? ≥ 2.2 ∙ 10 6 parcels/s) was identified. The present work aims to assess whether or not the sensitivity analysis performed in isothermal conditions can be extended to the reactive case. LES, coupled with Conditional Moment Closure (CMC) method as combustion model, are performed. The CMC modelling follows the approach adopted in [3]: a three-dimensional unstructured code with a first order closure for chemical source term is used. A modified one-step chemistry model for n-heptane air combustion, developed by [4], is used to reduce the computational cost associated with the CMC model. The LES equations are solved using the reactingParcelFoam solver available in OpenFOAM2.4.0, whereas the CMC equations are solved using an in-house unstructured finite volume solver, built as an extension of the LES solver [3]. The CMC structured grid consists of 1.4∙ 10 cells and is held constant for the four LES grids under investigation. The mixture fraction space is discretized in 50 bins clustered around the stoichiometric mixture fraction (ξst=0.062). For the grid M3, three simulations, with different number of computational parcels per second are performed: Np=0.5∙10 parcels/s (M3-05), Np=2.2∙10 parcels/s (M3-2.2) and Np=4∙10 parcels/s (M3-4). For all the other cases the number of computational parcels is equal to Np=2.2 ∙ 10 parcels/s. The simulations start from the same burning solution achieved using M3. The fields with M3 grid are mapped onto the other grids using the mapFieldsDict OF utility. This is considered as time t=0 s. The parameter chosen for this sensitivity study is the total amount of liquid (md) available in the system during each time step. This quantity is directly linked to the evaporation rate, implicitly connected to the flame leading edge position and to flame instabilities (i.e., global extinction). Figure 1 provides the total liquid mass in the system as function of the physical time for the four grids. For each of the cases, with an exception for M1, the simulations cover a time-span of 0.8 s. Figure 1. Total liquid mass [kg] in the system as function of time [s] for grid M1, M2, M3. The corresponding centered moving average with period N=2500 is indicated for case M1 (red line), M2 (blue line), M3 (green line) and M4 (cyan line). Np=2.2∙10 6 parcels/s. From 0 s up to 0.12 s the liquid mass increases for all the cases; the slopes of the centered moving averages emphasize the slower evaporation rate from M1 to M4. From t=0.12 s onward, the behaviors of M1 and M2 are different from those of M3 and M4. M1 suddenly decreases, the flame leading edge lowers until it reaches the bottom of the domain and the flame blows off. The mass in M2 increases until t=0.32 s after which remains steady around m̅d=5.5∙10 -7 kg. It has been observed that the higher liquid mass compared to those obtained with M3 and M4 corresponds to a higher lift-off height. The further refinement from M3 to M4 does not have a significant impact on the liquid mass. Combura Symposium 2019 October 9-10, 2019, Soesterberg, The Netherlands Due to the thermal expansion of hot flue gases, in the reactive scenario the jet undergoes a broadening and higher axial velocities than the isothermal case are measured. This is reported in Figure 2. Results with both grid M3 and M4 were in excellent agreement with results in non-reactive conditions [2], whereas with both grids the velocity is slightly over-predicted at the outer radius in reactive conditions. The numerical results seem to be comparable to the experimental ones at 0.01 m above the measured location. This requires further investigation. Figure 2. Mean gas axial velocity plotted versus the radius r [m] at x=0.04 m. Grid M3 and M4 with 2.2∙10 6 parcels/s. Time averaging period is 0.8 s. Left: Isothermal conditions. Right: Reactive conditions. If the grid size is important, the number of computational parcels sensitivity is no less. The total liquid mass in the system as function of the physical time for M3-05, M3-2.2, M3-4 is reported in Figure 3. Figure 3. Total liquid mass [kg] in the system as function of time [s] for grid M3 and Np=0.5∙ 10 parcels/s (M3-05), Np=2.2∙ 10 parcels/s (M3-2.2) and Np=04∙ 10 parcels/s (M3-4). The corresponding centered moving average with period N=2500 is indicated for case M3-0.5 (green line), M3-2.2 (blue line) and, M3-4 (red line). Cases M3-2.2 and M3-4 show almost identical trends which is in agreement with that of the LES sensitivity study in non-reactive conditions [2]. The most surprising aspect is related to M3-05, where the mass is always 20% lower than M3-2.2 and M3-4. It has been observed that this difference in liquid mass drastically changes the flow-field, the mixing and the flame shape. Funding This research has been funded by Ghent University (Belgium) through GOA project; [BOF16/GOA/004] (PREdiction of Turbulent REactive Flows http://www.pretref.ugent.be/)

Numerical study on the importance of the turbulent inlet boundary condition and differential diffusion in a turbulent H2/N2/air jet diffusion flame

Abstract: This work concerns numerical simulations of the turbulent H2/N2/air jet diffusion flame (Meier et al., 1996a), using large eddy simulations with conditional moment closure as the combustion model. ...