Simulation of Air Flow and Drop Trajectories in the NASA Glenn Icing Research Tunnel

The NASA Glenn Icing Research Tunnel (IRT) was constructed in 1944 to replicate aircraft icing conditions seen during flight in a controlled environment for analysis and certifications. The tunnel is periodically updated and modified to improve air flow quality and droplet distribution uniformity. In the first portion of this study (Part I), Computational Fluid Dynamics (CFD) was applied to three IRT geometries corresponding to configurations in 2000, 2009 and 2012. The simulations employed three-dimensional Reynolds-Averaged Navier-Stokes (RANS) modeling for the turbulent air flow combined with Lagrangian trajectories for the water droplets. These trajectories diffuse stochastically based on the turbulent kinetic energy of the air flow. Their distribution is important for predicting and understanding Liquid Water Content (LWC) uniformity.

The Year 2000 tunnel configuration was simulated from the spray bars to the test section and the results indicated that the RANS model predicted reasonable test section conditions. The 2009 tunnel configuration also was initiated just upstream of the spray bars but included vertical struts that were installed to increase LWC uniformity as well as multiple Mod-1 air jets implemented using embedded velocity profiles. These changes were found to increase the turbulent kinetic energy throughout the IRT. The 2012 tunnel configuration simulation included a new heat exchanger installed in 2011 as well as the ensuing Corner D which is upstream of the spray bars. The heat exchanger was simulated using a two-dimensional RANS model to provide an inlet boundary condition for the 3-D tunnel flow simulation in to Corner D. The results indicated an increase in turbulent kinetic energy from the 2009 tunnel configuration, especially near the inner wall. A transfer map was developed to show the droplet locations in the test section for specific water nozzles and was compared to maps developed experimentally. The result indicated good qualitative agreement, but under-predicted the droplet diffusion. Future studies should investigate the 2012 calibrated nozzle positions and the unsteadiness emanating from the spray bars and vertical struts (as discussed in Part II) to improve modeling fidelity and investigate opportunities to improve the LWC uniformity at the test section. Similarly, modifications to the spray bar and tunnel walls should also be considered to improve uniformity.

For the second portion of this study (Part II), a hybrid RANS and Large Eddy Simulation (LES) model was utilized to capture the unsteady phenomenon, e.g. wake shredding and flow recirculation, observed in jets released from certain water nozzles in the IRT. Since the RANS/LES model is computationally expensive, the domain was restricted to the region near a spray bar and its spray nozzles. A new boundary condition method was developed to translate the unsteady near-field air jet flow generated by a nozzle into a larger domain (far-field). The technique was denoted the Recorded Interface Boundary Condition and was found to reasonably reproduce the mean and turbulent velocities in the far-field without requiring the small time-steps and high resolution domain associated with the near-field. A single spray bar without any vertical struts or active nozzles was simulated with the RANS/LES model to study the unsteady flow in its wake. Results indicated high vortex shedding and flow separation which create a large unsteady wake downstream with greater turbulence intensity and wake spread than that predicted with the RANS description. The RANS/LES single spray bar simulation and new methodology for modeling jet flows form a foundation which spray bars with vertical struts and active water nozzles should be integrated. The RANS/LES model is recommended to be incorporated into the full IRT model and is expected to improve the fidelity of the droplet trajectories and LWC predictions.