Publications

A subgrid model is presented for use in CFD fire simulations to account for thermal suppressants and strain. The extinguishment criteria is based on the ratio of a local fluid-mechanics time-scale to a local chemical time-scale compared to an empirically-determined critical Damkohler number. Local extinction occurs if this time scale is exceeded, global fire extinguishment occurs when local extinction has occurred for all combusting cells. The fluid mechanics time scale is based on the Kolmogorov time scale and the chemical time scale is based on blowout of a perfectly stirred reactor. The input to the reactor is based on cell averaged temperatures, assumed stoichiometric fuel/air composition, and cell averaged suppressant concentrations including combustion products. A detailed chemical mechanism is employed. The chemical time-scale is precalculated and mixing rules are used to reduce the composition space that must be parameterized. Comparisons with experimental data for fire extinguishment in a flame-stabilizing, backward-facing step geometry indicates that the model is conservative for this condition.

A partial non-dimensionalization of the Navier-Stokes equations is used to obtain order of magnitude estimates of the rate-controlling transport processes in the reacting portion of a fire plume as a function of length scale. Over continuum length scales, buoyant times scales vary as the square root of the length scale; advection time scales vary as the length scale, and diffusion time scales vary as the square of the length scale. Due to the variation with length scale, each process is dominant over a given range. The relationship of buoyancy and baroclinc vorticity generation is highlighted. For numerical simulation, first principles solution for fire problems is not possible with foreseeable computational hardware in the near future. Filtered transport equations with subgrid modeling will be required as two to three decades of length scale are captured by solution of discretized conservation equations. By whatever filtering process one employs, one must have humble expectations for the accuracy obtainable by numerical simulation for practical fire problems that contain important multi-physics/multi-length-scale coupling with up to 10 orders of magnitude in length scale.

This paper outlines general physical and computational issues associated with performing numerical simulation of fire suppression. Fire suppression encompasses a broad range of chemistry and physics over a large range of time and length scales. The authors discuss the dominant physical/chemical processes important to fire suppression that must be captured by a fire suppression model to be of engineering usefulness. First-principles solutions are not possible due to computational limitations, even with the new generation of tera-flop computers. A basic strategy combining computational fluid dynamics (CFD) simulation techniques with sub-grid model approximations for processes that have length scales unresolvable by gridding is presented.

This report describes a Laboratory Directed Research and Development (LDRD) activity to develop a diagnostic technique for simultaneous temporal and spatial resolution of fluid flows. The goal is to obtain two orders of magnitude resolution in two spatial dimensions and time simultaneously. The approach used in this study is to scale up Particle Image Velocimetry (PIV) and Planar Laser Induced Fluorescence (PLIF) to acquire meter-size images at up to 200 frames/sec. Experiments were conducted in buoyant, fully turbulent, non-reacting and reacting plumes with a base diameter of one meter. The PIV results were successful in the ambient gas for all flows, and in the plume for non-reacting helium and reacting methane, but not reacting hydrogen. No PIV was obtained in the hot combustion product region as the seed particles chosen vaporized. Weak signals prevented PLIF in the helium. However, in reacting methane flows, PLIF images speculated to be from Poly-Aromatic-Hydrocarbons were obtained which mark the flame sheets. The results were unexpected and very insightful. A natural fluorescence from the seed particle vapor was also noted in the hydrogen tests.

This paper is a brief overview of work over the last several decades in understanding what occurs to jet fuel stored in aircraft fuel tanks on impact with the ground. Fuel dispersal is discussed in terms of the overall crash dynamics process and impact regimes are identified. In a generic sense, the types of flow regimes which can occur are identified and general descriptions of the processes are given. Examples of engineering level tools, both computational and experimental, which have applicability to analyzing the complex environments are presented. Finally, risk based decision is discussed as a quick means of identifying requirements for development of preventative or mitigation strategies, such as further work on the development of an anti-misting agent.

While all fires are complex and involve many phenomena, this report is limited to large, turbulent liquid-hydrocarbon pool fires. Large, liquid-hydrocarbon pool fires present a risk in petrochemical storage and processing facilities and transportation systems that contain large amounts of liquid hydrocarbons. This report describes observations, speculations, and numerical simulations of vortical structures in pool fires. Vortical structures are observed in fires with length scales ranging from those that bend millimeter-thick flame zones to those that entrain air many meters from the edge of the fire to its centerline. The authors propose that baroclinic vorticity generation is primarily responsible for production of rotational motion at small scale and that amalgamation is responsible for the production of large-scale rotational structures from the myriad of small-scale structures. Numerical simulations show that vortical structures having time-mean definitions can be resolved with a Reynolds-Average Navier-Stokes (RANS) approach. However, for vortical structures without time-mean definition, RANS is inappropriate, and another technique, such as Large Eddy Simulation (LES), should be employed. 39 refs., 52 figs., 3 tabs.

A fuel dispersal model for C-141 transport accidents was developed for the Defense Nuclear Agency`s Fuel Fire Technology Base Program to support Weapon System Safety Assessments. The spectrum of accidents resulting from aircraft impact on a runway was divided into three fuel dispersal regimes: low, intermediate, and high-velocity impact. Sufficient data existed in the accident, crash test, and fuel-filled bomb literature to support development of a qualitative framework for dispersal models, but not quantitative models for all regimes. Therefore, a test series at intermediate scale was conducted to generate data on which to base the model for the high-velocity regime. Tests were conducted over an impact velocity range from 12 m/s to 91 m/s and angles of impact from 22.5{degrees} to 67.5{degrees}. Dependent variables were area covered by dispersed fuel, amount of mass in that area, and location of the area relative to the impact line. Test results showed that no liquid pooling occurred for impact velocities greater than 61 m/s, independent of the angle of impact. Some pooling did occur at lower velocities, but in no test was the liquid-layer thickness greater than 5.25 mm.

Flame acceleration and deflagration-to-detonation transition (DDT) studies have been conducted in a 19.4-cm high, 14.5-cm wide, and 2. 242-m long channel (MINIFLAME) that is a 1:12.6 scale model of the 136-m{sup 3} FLAME facility. Tests were conducted with two levels of hydrogen concentration -- 20% and 30%, with and without obstacles in the channel, and with three levels of transverse top venting -- 0%, 13%, and 50%. The flame acceleration results in MINIFLAME are qualitatively similar to those in FLAME; however, the small-scale results are more benign quantitatively. The results show that insufficient venting, 13% venting in this case, can promote flame acceleration due to turbulence produced by the flow through the vents in smooth channels. However, with obstacle-generated turbulence in the channel, 13% top venting was found to be beneficial. Flame acceleration resulting in DDT was shown to occur in as little as 35 liters of mixture. Comparison of the DDT data with obstacles in MINIFLAME and FLAME supports d/{lambda} scaling of DDT, where {lambda} is the detonation cell width of the mixture and d is the characteristic open diameter of the channel. In the MINIFLAME and FLAME tests, DDT occurred for d/{lambda} greater than approximately three. Comparison with other experiments shows that the value of d/{lambda} for DDT is not constant but depends on the obstacle type, spacing, and channel geometry. The comparison of MINIFLAME and FLAME experiments extends the use of d/{lambda} scaling to different geometries and larger scales than previous studies. Small-scale-model testing of flame acceleration and DDT with the same combustible mixture as the full-scale prototype underpredicts flame speeds, overpressures, and the possibility of DDT. 18 refs., 16 figs.