Wind Tunnel Modeling of Dense Gas Dispersion

This paper will summarize several years of research and applications of the research regarding heavy gas dispersion funded by The American Petroleum Institute and various petroleum and chemical companies. This research was conducted utilizing wind tunnel modeling with an overall purpose to better understand heavy gas dispersion in complex environments, test then current numerical modeling techniques and to assess various mitigation measures (vapor barriers and water spays) to minimize off-site consequence. The similarity requirements for wind tunnel modeling are derived from the basic equations governing atmospheric and plume motion (conservation of mass, momentum and energy). For an exact solution to the basic equations, the dimensionless quantities and boundary conditions must be the same in the wind tunnel and in the plume as they are in the corresponding full-scale case. The complete set of requirements for similarity is: undistorted geometry; equal Rossby number; equal gross Richardson number; equal Reynolds number; equal Prandtl number; equal Eckert number; similar surface-boundary conditions; and similar approach-flow characteristics. Not all of the dimensionless quantities can be matched in model and full scale. Research has been conducted, however, to determine which parameters are important for certain flow situations. EPA and the Gas Research Institute have prepared guidelines that aid in the selection of the appropriate scaling parameters. In general, the procedure allows for establishing accurate simulations of heavy gas releases in complex settings. The EPA guideline is primarily used for fluid modeling of passive pollutants and establishing appropriate atmospheric boundary layers. The primary purpose of the Gas Research Institute guideline was to discuss the capabilities and limitations of fluid modeling for dense gas cloud behavior and suggest standards to be followed when conducting such studies. The guideline compares data from 26 dense gas field experiments with physical model simulations. In general, the physical model clouds were found to be very similar in appearance, they spread and traveled at the correct rates, measured concentrations compared very well and peak concentrations were often predicted to within a factor of 2 or better. Model simulations where specific gravity, volume flux ratio and Froude number equality were maintained produced the most successful predictions of field concentrations. When only volume flux and Froude number equality were maintained, peak concentrations compared well but cloud arrival and departure times were distorted. The report also discusses ranges over which various simulation parameters are appropriate. The Gas Research Institute's guideline for use of wind tunnel modeling concluded that there are three reasons why fluid modeling has value in engineering analysis and health and safety evaluations. First, fluid modeling does some things better than analytic and numerical modeling alternatives. Second, wind tunnels are, in effect, analog computers which have the advantage of “near-infinitesimal” resolution and “near-infinite” memory. A fluid model employs real fluids; hence, the fluid model is implicitly non-hydrostatic, non-Boussinesqu, compressible, includes variable fluid properties, non-slip boundary conditions, and dissipation, permits flow separations and recirculation. Third, the fluid model bridges the gap between analytic or numeric models of turbulence and dispersion, and their applications in the field. Fluid modeling may be used to plan field experiments, provide conservative estimates of plume transport, and validate modules of numerical code. It is most useful for near source dispersion estimates where mechanically induced turbulence is present from structures such as trucks, complex structures, tanks, vapor detention systems, water sprays, etc., and where uncertainties in mathematical modeling of the complex dispersion process is greatest. The wind tunnel modeling methodology consists of first establishing model operating conditions for the full scale heavy gas release scenario using the similarity requirements discussed previously. Next a scale model of the site structures is constructed (see the Figure below for an example). The scale model is then installed in an atmospheric boundary layer wind tunnel with the appropriate roughness upwind and downwind. An appropriate boundary layer is then established (neutral, unstable or stable). A dense gas mixture is then released from the location of concern and the resulting concentration levels are measured at the locations of interest. This process is then repeated to assess the effect of wind speed, wind direction and/or mitigation. Average or time varying concentrations are measured depending upon the release type (continuous, instantaneous, time varying) and goals of the evaluations. The resulting concentration data that is collected can be utilized to test numerical models or used directly for consequence analysis purposes. In summary, this presentation will discuss the similarity requirements for wind tunnel modeling of heavy gas dispersion and will provide an overview of the following research regarding heavy gas dispersion: 1) effect of surface roughness; 2) effect of structures; 3) effect of vapor barriers on decreasing concentrations; 4) effect of water sprays on decreasing concentrations. A summary of past wind tunnel/field comparisons will also be provided.