Air Flow and Liquid Water Concentration Simulations of the 2012 NASA Glenn Icing Research Tunnel

Abstract

Three-dimensional Reynolds-Averaged Navier-Stokes (RANS) simulations using theMenter-SST turbulence model were performed on the new 2012 configuration of the NASA Glenn Icing Research Tunnel (IRT). The IRT was simulated from the exit of the heat exchanger to the test section. A two-dimensional simulation was first performed on a crosssection of the heat exchanger to provide initial conditions for three-dimensional flow predictions through the turning vanes, spray bars, tunnel contraction and test section. The simulations showed a general increase in turbulence intensity within the test section as compared to previous IRT configurations (in 2000 and 2009) which can be attributed to wake effects from both the heat exchanger and spray bars. In addition, the heat exchanger produced variations in the yaw flow angles after the turning vanes which are consistent with experiments while the corner geometry resulted in higher flow turbulence gradients for the inner wall regions. Using these simulated time-averaged flowfields, droplet trajectories were predicted using Lagrangian calculations with an unsteady Discrete Random Walk (DRW) model to mimic turbulent fluctuations. A transfer map was developed by mapping the water droplet locations at the test section by tracking peak concentration contours associated with both nozzle rows and nozzle columns. In addition, the liquid water concentration (LWC) distribution at the test section was predicted using the 2012 calibrated nozzle locations. The simulated transfer map and LWC distribution were both qualitatively similar to experiments but demonstrate that the RANS model may not be capturing unsteadiness associated with the spray bar wakes and the jets. An appendix was provided with data on a hybrid RANS/LES simulation of a section of the IRT which captured the unsteady behavior of the airflow. The hybrid model predicts significantly higher turbulence in the spray bar wake than the RANS model which would explain the disparities between with the droplet trajectory calculations and experiments. Nomenclature α = liquid water concentration C(d) = cumulative distribution function d = droplet diameter d rr = Rosin-Rammler reference diameter h rr = Rosin-Rammler spread parameter k = turbulent kinetic energy k av = average turbulent kinetic energy l = turbulent length scale u = air flow velocity v = water droplet velocity ω = specific dissipation rate x = streamwise distance y = vertical distance y+ = dimensionless wall distance z = spanwise distance

Flow Quality Surveys in the Settling Chamber of the NASA Glenn Icing Research Tunnel (2011 Tests)

“…This was investigated further by presenting the problem Clark, et.al. at the University of Virginia, who are computing some of the flow physics within NASA Glenn's IRT using their Virtual Icing Research Tunnel (VIRT) [4]. Clark et.al.…”

“…Horizontal surveys at the "ceiling" and "centerline" locations. 4. Horizontal surveys at the "centerline" and "floor" locations.…”

“…Specifically, these include the results of the icing cloud calibration [2] and of the aero-thermal characteristics of the test section [3]. The University of Virginia also contributed to the calibration of the tunnel using three-dimensional Reynold-Average NavierStokes (RANS) simulations on the 2012 configuration of the IRT [4]. Their results were compared to the settling-chamber survey results that are presented in this paper.…”

NASA Glenn Icing Research Tunnel: 2012 Cloud Calibration Procedure and Results

“…From the computational results of Clark 4 , the Cal Team expected to see much greater turbulence mixing with the struts than without, and to expect better mixing at the inner wall (North side) than outer wall. We in fact saw this with the grid.…”

Superhydrophobic Surfaces to Resist Droplet Sprays, Simulated Rain, and Water Jets

“…The discretization scheme, governing equations, and turbulence model for free jet calculations (i.e. without a downstream wall) were the same as used by Clark, et al 31 The temporal discretization of the RANS equations The droplet trajectories were computed with a Lagrangian discrete-phase model with a particle time step of 100 µs. For improvement in the predictive ability of turbulent droplet diffusion, a Discrete Random Walk (DRW) model was used.…”