Mark Potapczuk, Ph.D.
Aerospace Engineer
NASA Glenn Research Center
Cleveland, Ohio

NASA uses aircraft such as this Twin Otter for icing research, including for the verification and validation of software codes that simulate icing.

NASA uses aircraft such as this Twin Otter for icing research, including for the verification and validation of software codes that simulate icing.


NASA is currently upgrading and modifying a Navy S-3 Viking into an icing research platform or IRA (icing-research aircraft).

NASA is currently upgrading and modifying a Navy S-3 Viking into an icing research platform or IRA (icing-research aircraft).


Ice (in white) formed on the leading edge of an airfoil changes the Machnumber contours. The ice shape was obtained by tracing a real ice shape generated in the Icing Research Tunnel at the NASA Glenn Research Center. The computational simulation shows large regions of flow separation aft of the two horns and a smaller separated flow region between the two horns.

Ice (in white) formed on the leading edge of an airfoil changes the Machnumber contours. The ice shape was obtained by tracing a real ice shape generated in the Icing Research Tunnel at the NASA Glenn Research Center. The computational simulation shows large regions of flow separation aft of the two horns and a smaller separated flow region between the two horns.


Example of an ice shape generated on a wing section in the NASA Icing Research Tunnel.

Example of an ice shape generated on a wing section in the NASA Icing Research Tunnel.


Streamline colors correspond to temperature and show a hot-air iceprotection system in operation at the leading edge of a wing. Only a short segment of the wing is displayed.

Streamline colors correspond to temperature and show a hot-air iceprotection system in operation at the leading edge of a wing. Only a short segment of the wing is displayed.


The grid around a 2D model of a clean airfoil changes as ice forms. The grid changes to accommodate ice growth generated during a simulated 14-min exposure to icing conditions. The simulation was run using 1-min time steps.

The grid around a 2D model of a clean airfoil changes as ice forms. The grid changes to accommodate ice growth generated during a simulated 14-min exposure to icing conditions. The simulation was run using 1-min time steps.


Every winter, passengers, airlines, and the FAA are forced to deal with disruptions, delays, and danger due to icing. In fact, since 1993 icing has brought down 135 aircraft and killed 171 people, according to the National Transportation Safety Board. Icing affects aircraft by adding weight and, more importantly, by altering the shape, and hence the aerodynamics, of the wings.

Long before pilots face the hazards of flying through icing conditions, aircraft manufacturers evaluate how icing affects the performance of the planes they design and build. They have several ways to do this. The most direct and costly approach is to fly the aircraft through actual icing conditions. This lets a manufacturer assess how ice accretion or buildup affects the vehicle's aerodynamics and how well its iceprotection system works. Alternatives to flight testing include wind-tunnel testing and computational simulation. Increasingly, manufacturers use a combination of all three to evaluate how aircraft behave in icing conditions.

The method that has seen the most advances in recent years is computer simulation. Faster computers, better algorithms, and advanced software let engineers and scientists more accurately model and understand ice buildup, evaluate changes in aircraft performance, and analyze ice-protection systems.

SIMULATING ICE GROWTH
Engineers simulate ice buildup to predict how much ice will grow on an aircraft surface due to exposure to supercooled water droplets and the shape that ice will take. There are four basic steps to the simulation. Engineers first calculate the flow field surrounding the aircraft using standard computational fluid-dynamics (CFD) methods. They use this data to then determine the trajectories of water droplets in that field.

Next, they figure out how much of that water will freeze. And finally, they forecast the shape of the ice. Ice growth is inherently unsteady, so computer simulation normally consists of iteratively going through these four steps. How the iteration is done varies with the particular computer program used, but the following description covers the basics of icegrowth simulation.

Flow-field calculations that predict ice buildup are similar to those used to determine airflow in standard aircraft simulations, but they can be complicated by the presence of the ice itself. So engineers must reevaluate the flow field at the start of each iteration. Another complication is that the simulation should take into account ice roughness and regions of local flow separation. Additionally, as ice forms a shape, which can be complex, researchers must adjust the computational grid or mesh required to account for the new shape.

To sidestep the issue of changing the grid with each iteration, aerospace engineers use potential flow codes to evaluate airflow. Potential flow simulation assumes the flow field conforms to a particular set of conditions. Although these conditions can be violated during an icing encounter, the resulting discrepancies do not always preclude using these codes.

In particular, these codes do not need a grid for the field surrounding the body, so it's relatively easy updating geometry during ice-growth simulation. Although there are still issues about applying such methods to flows over ice-shape geometries, results so far indicate potential flow codes often can accurately simulate ice growth.

Determining the trajectories of water droplets in the previously calculated airflow around the body of interest tells researchers where those drops hit the body. This can be done from one of two perspectives, either from a Lagrangian or Eulerian reference frame. In the Lagrangian approach, the frame of reference is attached to the droplet and representative droplets are tracked from some upstream location to the point where they hit the body. This calls for routines that determine whether or not a droplet hits the body or passes it by. Additional logic optimizes the number of droplet trajectories needed to properly evaluate water loading on the surface of the body.

In the Eulerian approach, the frame of reference is a fixed point in space and water is tracked passing through particular locations. This is the same approach that calculates airflow in gridbased methods and thus requires solving only a few additional equations along with the airflow. The Eulerian approach also solves for water flow throughout the computational domain and thus avoids the need for specialized logic to choose which trajectories need to be calculated.

The next step is calculating how much ice accumulates in a given time. To do this, engineers divide the surface of the body into small surface elements and perform a mass and energy balance to calculate how much water in each element has turned to ice. This mass and energy balance calculation was first formulated by B.L. Messinger in 1953 and is shown in the following equation:

where k= the thermal conductivity of the body in a direction normal to the surface, W/m/°K; T = temperature at a given location on the body geometry, °K; f = a direction normal to body surface, m; q"nc = the net convective heat lost from body, W/m2; q"evap = evaporative heat lost from surface water, W/m2; q"ke = the kinetic heat gain to the body from impinging droplets, W/m2; q"lat = the latent heat gain to body from freezing, W/m2; and q"sens = the sensible heat loss (or gain) from surface water W/m2.

The final step calculates the shape of the ice. The mass and energy balance calculation in the previous step reveals how much ice forms in a time step. Ice density is then used to calculate how much ice must be added to each surface element in each time step. This generates the next ice surface for the next time step. This four-step process is repeated until it replicates the entire time the surface is exposed to icing conditions.

Engineers then use the shapes generated from such simulations to evaluate changes in aircraft performance using flight test or wind-tunnel testing. For flights, simulated ice shapes are turned into physical models, attached to a wing, and flown. Ice shapes can also be attached to wing models and checked out in wind tunnels. In flight tests, however, designers can evaluate complete aircraft performance, providing valuable information for aircraft certification. Engineers also simulate the effects of ice on performance using the same CFD tools that predict ice growth.

ICE AND FLIGHT
Analyzing iced airfoils or wings using CFD is complicated by the geometry of the ice shape and by the flow around these shapes. Unlike the rather loose requirements on accuracy for flow calculations used to simulate ice growth, determining the changes in aerodynamic parameters such as lift and drag requires a higher degree of fidelity. As in all CFD analyses, accuracy stems from creating a well-defined grid and modeling all the appropriate physics for the flow field of interest. For iced airfoils and wings, engineers must pay attention to details of the grid surrounding the highly irregular geometry, including flow over rough surfaces and flow separation at low angles of attack in some regions. They must also evaluate unsteady, completely separated flow at or near maximum lift conditions. These requirements push the limits of most CFD tools and are the subject of current research.

Computational methods are also used to simulate ice-protection systems. There are three types of ice-protection systems, thermal, mechanical, and chemical. Thermal systems heat aircraft surfaces to either prevent ice from forming (anti-icing) or melt ice that has already accumulated (deicing). The most common type of thermal ice protection is the hot-air bleed system. In these systems, heat from the engine is piped to the area behind the wing's leading edge or engine cowling, prime spots for ice formation. The heat prevents ice from forming and water can be heated to the point that it does not run back and freeze on aft portions of the aircraft.

Mechanical systems are strictly for deicing and use various schemes to mechanically debond ice from the wing and let aerodynamic forces sweep it away. For example, some aircraft have rubber boots on the leading edge of their wings. Inflating the boot makes it expand, cracking the ice loose, and the airflow blows the ice away.

Chemical systems apply freezingpoint depressant fluids to the wings, thus preventing ice from forming if temperatures aren't too low. Currently, computational methods only cover thermal ice-protection systems.

To model thermal systems, researchers add heat-transfer analysis to the tools used to predict ice buildup. Hot-air system models are based either on empirical relationships for the heat transfer coefficient on the inside surface of the protected region or by using a Navier-Stokes internal flow code. In addition to modeling hot airflow, heat conduction through the surface can be calculated and then connected to icegrowth simulation, thus completing an analysis of the entire ice-protection system.

MAKING SURE IT'S GOOD TO GO
All the simulation methods described here require a good deal of verification and validation to determine how well they represent the physics being modeled. (Verification is typically described as determining whether the equations are properly represented in the computer code. Validation is evaluating whether the equations properly simulate the physics.) Complete and extensive validation efforts are difficult and expensive. A complete validation effort consists of verification and validation of individual modules that make up the code as well as the complete code itself.

As researchers developed icingsimulation software, they discovered that the type of data needed for validation was not generally available in the open literature. As a result, researchers largely under the auspices of NASA and the FAA in the U.S. and similar agencies in Canada and Europe began developing such databases over the past 10 to 20 years.

At NASA, the development of validation databases has been undertaken using two specialized facilities, the Icing Research Tunnel (IRT) and an Icing Research Aircraft (IRA). Engineers have used the IRT to validate data for waterdroplet trajectories and ice shapes, and to evaluate iced aircraft performance and various ice-protection systems.

The IRA is a specially equipped DeHaviland DHC-6 Twin Otter. It has been used for many different purposes over the years including: characterizing clouds that tend to create ice; evaluating cloud instrumentation; and iceprotection systems, gathering data on ice shapes, and cataloging performance changes due to ice buildup. Cloud instrumentation carried on the IRA measures water-droplet size and quantity, ice particle size and quantity, as well as total water content of the cloud. NASA will soon retire the Twin Otter and replace it with a modified Navy S3 Viking aircraft as the next-generation IRA.

Research organizations, alone and with industry partners, continue to explore new methods for computationally simulating all aspects of aircraft encounters with icing in flight. The work involves fundamental research into icing, developing theoretical models of complex physical processes, and building and using databases for validating computer models. These efforts increase flight safety, decrease certification costs, and let engineers more fully analyze all elements of flying in icing conditions.