ON 31 OCTOBER 1994, a turboprop airliner heading for Chicago dived into a soybean field at Roselawn in Indiana. All 68 people aboard died. Although the weather was cold that day, no one could believe it when investigators revealed that the crash was caused by a build-up of ice on the wings.
Not only did this modern plane have a fully functional de-icing system, but according to US Federal Aviation Administration (FAA) standards, the French-built ATR-72 should have had no problems flying in the cold, damp conditions. The pilots even knew their craft was icing up and attempted to clear it, following de-icing procedures to the letter.
In fact there have been a succession of crashes and close calls in which ATR-72s and similar turboprop aircraft have snapped into unexpected rolls due to heavy icing – in conditions which the rule books say the planes should have been able to handle. In 1987, for example, an iced-up Italian ATR-72 crashed in a snowstorm. A decade later another turboprop airliner – an Embraer 120 operated by US airline Comair – crashed near Monroe, Michigan after its wings began to ice up, killing all 29 people on board.
Advertisement
Last March another Embraer 120 returning from the Bahamas encountered icing and without warning went into a steep dive before the pilots regained control. And in December 2002 an ATR-72 fell into the Taiwan Strait, killing its crew. Ice is the prime suspect.
So is the design of these turboprop airliners defective in some way? The National Transportation Safety Board (NTSB), the body that investigates plane crashes in the US, has little doubt that there is a problem. Last September it updated the original findings from its Roselawn investigation, swinging some of the blame away from the aircraft’s manufacturer and emphasising its original call to the FAA for an overhaul of the design standards aircraft must meet in order to be licensed to fly in icy conditions.
Today’s standards are based on data gathered 50 years ago. And though there have been some changes to design and operational limits, turboprop planes continue to fall out of the sky. Ice and freezing rain seem to be more complex beasts than anyone ever suspected.
All aircraft designers who have to face the problem are familiar with the challenge of icing: many have been making airliners for at least 40 years. And manufacturers must demonstrate that their aircraft are capable of flying safely in cold, wet conditions where they might ice up. In the US these conditions are specified by the FAA’s “Appendix C”, which lays out the kinds of cold weather which an aircraft must be able to deal with. At the heart of Appendix C is the fact that the water in clouds is often in the form of thermodynamically unstable supercooled droplets.
Supercooled water drops remain liquid even though they may be well below 0 °C. When a cold aircraft flies into them it acts like a seed crystal: the moment the droplets touch its wings or tailplane they freeze solid. A thick layer of ice can build up quickly, usually along the leading edge of the wing, tailplane and rudder. The ice creates drag, reduces lift and can make the plane drop like a stone.
The favoured solution is to incorporate de-icing equipment directly into the leading edges of the wings and tail. This includes heaters or, in the case of turboprop airliners, rubber bladders or “boots” that can be inflated to dislodge any ice that accumulates.
However, Appendix C was drawn up way back in the early 1950s, and there is growing evidence that certain conditions can quickly overwhelm an aircraft, even if it has de-icing apparatus. Appendix C requires aircraft to be able to cope with supercooled cloud droplets with a mean diameter of 30 micrometres. But in nature you never see a uniform droplet size, says Daniel Bower of the NTSB, who has investigated icing accidents. Instead you get a wide distribution of droplet sizes that can behave in unexpected ways.
One of the conditions not fully covered by Appendix C is freezing rain, which occurs when small water droplets coalesce into large raindrops at least 1 millimetre across, and these start falling and pass through a layer of sub-zero air. Such drops will freeze the instant they strike a cold aircraft surface and will do so in such large amounts that few de-icing systems can cope. Pilots who run into freezing rain are advised to turn tail and flee.
And that’s not all. Since the 1980s the NTSB has been urging the FAA to revise Appendix C after a series of accidents suggested there were conditions other than just freezing rain which could rapidly ice up an aircraft. In the mid-1980s, as a graduate student at the University of Wyoming, Marcia Politovich helped review eight years’ worth of atmospheric testing conducted aboard a test aircraft. “We had some measurements of large drops that sort of scared us,” Politovich recalls.
Outside the envelope
Politovich found that a short flight would sometimes collect much more ice than longer flights under other weather conditions, and it seemed that large drops of water were turning to ice on the wing just behind the de-icing equipment. “They made a rough icing texture, which created a lot of drag,” she says. “We wrote some papers on it and said hey, this stuff exists, be careful, it’s outside the envelope.”
These droplets, which Politovich and her colleagues estimated to be between 50 and 500 micrometres across, were smaller than those encountered in freezing rain and could remain suspended in clouds for long periods. The team dubbed them “large supercooled drizzle drops”. However, their discovery failed to elicit any real concern until the 1994 crash at Roselawn. The day after the crash, an NTSB meteorologist telephoned Politovich, now at the National Center for Atmospheric Research at Boulder, Colorado, and asked for help in understanding the weather conditions likely to have existed the day before. Everyone knew that freezing rain could be bad, but what the researchers learned was frightening.
They worked out that over Roselawn the ATR-72 had met mid-cloud water drops larger than provided for in Appendix C, but still smaller than freezing raindrops. These created an ice ridge beyond the reach of the de-icing system. Although just a centimetre or two high, it was enough to break up the smooth airflow over the wing (see Diagram) and reduce its lift. The ridge also created a vacuum that sucked the ailerons upwards. Unable to cope, the autopilot tripped out, the plane rolled and the pilots lost control.
“We found that larger droplets are going to impinge further back on the airfoil,” says Politovich. “Some run back and freeze further aft than the de-icing boot. That’s how we got the ice ridge on the ATR at Roselawn.”
Why should turboprop aircraft be more vulnerable than jet airliners? There seem to be a variety of reasons. First, regional turboprops tend to fly shorter routes at lower altitude, and so spend more time within clouds where ice can form. Longer-range jets can often fly above the weather. Small turboprop aircraft also have less spare energy in the form of hot air or electrical power with which to warm wing surfaces and remove ice, or prevent it forming in the first place. Finally, the overhead wing design on planes such as the ATR-72 could add to the problem, since the pilots cannot easily see ice developing on its upper surface.
In the wake of the accidents at Roselawn and Monroe, the de-icing boots on some turboprop aircraft were enlarged, and the FAA also limited the flight conditions in which they could operate. This move was mirrored by other regulators, including Britain’s Civil Aviation Authority. At the same time, forecasting was improved and flight manuals were rewritten to give pilots a better chance of recognising the danger signs.
The FAA backed new research into the problem and even organised a test in which a US Air Force mid-air refuelling tanker was converted to spray larger droplets at an ATR-72 flying behind it. This confirmed that ice could indeed form aft of the de-icing boots, which was news to many pilots.
“Airplanes were certified according to Appendix C, but it was never conveyed to pilots that there were conditions out there which they were not certified to fly in,” says Mike Bragg of the University of Illinois at Urbana-Champaign. “Everyone assumed that Appendix C meant all icing. There were people in the atmospheric-science community saying that there are worse things out there, but until Roselawn, they were ignored.”
Yet almost a decade later the problem has still not been solved. Ave Bransford from Atlanta, Georgia, whose husband died in the crash at Monroe, says she wants to see changes to regulations right now. She believes the FAA has delayed far too long in revising Appendix C, especially as it relates to turboprop airliners. Bransford is also calling for automatic ice-detection systems to be fitted to all passenger aircraft. “The technology is available,” she says. That call is echoed by the NTSB and researchers alike (see “Time to open the box”). But, the FAA insists, it cannot revise the rules until the research programme it is funding has been completed.
Even now, the full range of icing conditions is not understood. Bragg, for example, regularly flies in freezing conditions aboard a NASA flying lab to see first-hand how ice forms on a wing. But there is a limit to how much risk researchers are prepared to take, and work is increasingly being done in wind tunnels.
In the 1990s, researchers at the University of Illinois replicated the Roselawn conditions in a wind tunnel and showed how even a low ridge of ice forming beyond the de-icing boots could dramatically reduce lift and increase the wing’s stalling speed. Bragg also showed that such ice can create a dangerous vacuum near the wing’s trailing edge, unexpectedly hoisting the aileron into the airflow.
When the airflow is disrupted, the other controls become much less effective, making the aircraft extremely difficult to manage. Bragg’s team found that even a thick coating of ice protruding forwards on the wing’s leading edge has much less of an impact than a relatively thin layer further aft.
Wind tunnels are useful but they don’t always reflect exactly what happens in the real world, warns Myron Oleskiw, an aerodynamics researcher at Canada’s National Research Council in Ottawa. While white rime ice created by small drops accreting on the leading edge is relatively easy to scale up to the real world, other forms are more difficult.
Mark Potapczuk of NASA’s Glenn Research Center in Cleveland, Ohio, is equally cautious. The physics of icing sounds simple, he says – supercooled droplets hitting the surface and freezing – but in many cases they don’t freeze entirely on impact and there are complex interactions at the aerofoil surface. “It’s like taking all the difficult issues in fluid mechanics and putting them together in one problem,” he says. “Computing aerodynamic flow is difficult enough. It gets even more complicated when you put ice on it.”
Engineers at NASA reckon they will need several more years of flights and wind-tunnel work before they come close to creating a simulation of real icing conditions in the lab wind tunnel. One problem they face is the fact that NASA’s wind tunnel is designed to study small droplets striking a test wing. When larger droplets are used, they tend to fall out of the airflow before they reach the wing.
One possible solution comes from Britain, where David Hammond of Cranfield College of Aeronautics at Cranfield University in Bedfordshire is designing a vertical wind tunnel in collaboration with NASA and Wichita State University in Kansas. It is designed to measure the mass of ice that collects on a wing, as well as how much water splashes off.
Hammond plans to examine the fate of the droplets using a high-speed camera like those used to view explosions in slow motion. He needs to get images of the droplets at the moment they approach the wing, since they may be distorted by the aerodynamic forces near the surface, and this could affect how efficiently they form ice. “We want to see the impact, if there is a splash, or a bounce or a range of things. We want to see exactly what happens and measure the amount of mass going in different directions,” he says.
The Cranfield vertical icing tunnel is due for trials in the next few months, and Oleskiw says it could solve many problems for researchers. “It has been done before in hailstone research but never for aircraft icing.”
Roselawn triggered a lot of research, Bragg says. Since then, icing research has changed from an academic exercise into an applied science providing life and death solutions for aviators. “There was also a change in attitude, on things that would directly impact flight operations,” he says. “Beforehand, research was fundamental, without trying to get it out into the field. Now, even when I’m doing fundamental work, I’m wondering how it can apply directly to the cockpit.”
Time to open the box
Michael Bragg of the University of Illinois at Urbana-Champaign is always struck by the fact that the crash investigators who examined the Roselawn aircraft’s black boxes had far more data than was available to the doomed pilots. It is information that might have alerted them to the danger and perhaps saved many lives.
The first the pilots knew was when the autopilot, unable to keep them level, tripped out and the aircraft banked sharply. “From that moment, the Roselawn crew became test pilots because they were flying an aeroplane with different characteristics than they had ever flown before,” says Bragg.
Now Bragg’s department is collaborating with NASA on a project called the Smart Icing Systems Program, which will take the information going into the black box and also route it to the aircraft’s flight-management computer. In addition, ice detectors in the wings will alert the pilots and the computer to what is happening outside the cockpit and automatically switch on the de-icing systems. These are manually activated at the moment.
Bragg also wants to adapt the controls so that even a heavily iced aircraft can be kept in the air; an extra minute of flight time could make the difference between life and death. “Obviously, we can’t defy physics. The stall speed would still be dangerously high, but we want to be able to automatically compensate [for the ice] with other controls like engine power and angle of attack.”