Computing Profession

Hitting the Heights

A drone impacting an aircraft windshield.
Assessing the affect of drone collisions on aircraft windshields.

In scrubland to the west of Shanghai, in China, a bizarre vehicle has been plying the landscape with a curious regularity: the severed nose section of an airliner has been repeatedly zooming across the countryside at many hundreds of miles per hour, on the back of a rocket-powered sled. It only ends its journey when it smashes into a drone strategically suspended above the rails on which it travels.

It might sound like a form of rocket-assisted vandalism, but the aim of engineers at the Shanghai Aircraft Airworthiness Certification Center is a deadly serious one: to boost air safety by filling a gap in the aviation industry's collective knowledge of precisely what happens when five different types of consumer drone collide with the windshield of an airliner.

Once the physical damage to the windshield has been fully assessed, a dataset describing it, for each of five popular types of drone design, is being used to boost the accuracy of a software-based finite-element collision simulator that models drone-on-airliner impacts.

Such simulations are needed because drone impacts on  airliner cockpits could be lethal. Unlike soft, lightweight, hollow-boned birds, against which aircraft components are reinforced by airworthiness standards in manufacture, drones comprise a rigid chassis carrying heavy batteries, motors, and cameras. All of these might crack or even penetrate an airliner's reinforced triple-layer glass windshield. If that happened, depressurizing the cockpit too quickly for the crew to don oxygen masks, control of the aircraft could be lost.

Although no such critical event has yet occurred on a commercial jet, post-flight forensics show that one airliner landed safely after a suspected collision with a drone in late 2018. The Drone Database of the Alexandria, Virginia-based Flight Safety Foundation Aviation Safety Network, which collates accident and incident data from the world's air accident authorities, lists 20 suspected drone collisions with powered airplanes, helicopters, gliders and balloons.

Near-misses between passenger airplanes and drones have become rife, with the U.S. Federal Aviation Administration (FAA) warning that drone sightings by civilian pilots have increased dramatically in the last couple of years. In the U.K., pilots are reporting two near-misses every week. Last September, for instance, a jetliner with 50 passengers on board preparing to land at Manchester Airport in the U.K. came within 16 feet of hitting a stationary, hovering drone, with the U.K. Airprox Board (which defines "airprox" as "a situation in which, in the opinion of a pilot or air traffic services personnel, the distance between aircraft as well as their relative positions and speed have been such that the safety of the aircraft involved may have been compromised") saying only providence prevented a collision.

Such risks cannot be ignored; national aviation authorities like the FAA, the Civil Aviation Administration of China (CAAC), and the European Aviation Safety Agency (EASA) need to know the true effects drone impacts could have on airliners; that is, on key structure like the wing and tail assembly, the windshield, and the engines. Of these, wing structure has been tested and found to be alarmingly penetratable by drones at a high closing speed of 238 mph, in experiments conducted by a team led by Kevin Poormon of the Impact Physics Group at the University of Dayton's Research Center in 2018.

Jet engines, however, costing circa $12 million each, have not been subjected to drone ingestion tests. Some think a drone sucked into a jet engine could cause the engine to suffer an "uncontained failure," (also known as a disk burst), in which superhot, fast-rotating engine components explode outward, piercing wings, fuel tanks, and even the aircraft cabin. However, until somebody is willing to sacrifice an engine for such testing, we won't know its results for certain. "I am not aware of any drone ingestion tests that have been performed, but there are a few agencies that are trying to procure funding to do this. It is only a matter of time until this takes place," says Poormon.

Some testing has been carried out on airplane windshields, although not with intact drones flown at them. As the largest windward surface on an aircraft, the windshield  is also the one that probabilistic analysis says is most likely to be impacted by airborne flying objects. So the British Airline Pilots Association (BALPA), the U.K. Department of Transport, and the U.K. Military Aviation Authority jointly ran tests in 2017 in which drone components (motors, batteries, cameras, chassis arms, etc.) were fired out of a compressed gas cannon at a static airliner windshield. Components were fired because the team did not have a gas gun big enough to fire an intact drone.

Across three drone types weighing 0.4kg, 1.2 kg and 4kg, BALPA et al found that at slow take-off and landing speeds (around 130 m.p.h.), the sloped, reinforced airliner windshields held up and retained their integrity during impacts. However, at greater speeds at higher altitudes, the 4kg quadcopter components caused severe damage to the windshield, with cracks risking structural failure of the aircrew's last line of defense.

However, some commercial drone makers and users have criticized those gas-gun tests as unrealistic: intact drones would not, they argued, have the same effect as a gun spitting a fusillade of components at a windshield.

In an experiment written up in the March 2020 edition of the journal Aerospace Science & Technology, a team led by Xiaohua Lu at Nanjing University of Aeronautics and Astronautics attempted to quash that criticism by getting rid of the gas gun and detailing the results of rocket-powered windshield impacts with intact drones.

Working with the Shanghai Aircraft Airworthiness Certification Center, Lu and his colleagues used a rocket sled to accelerate an airliner cockpit section to three different flight speeds at which airplanes might meet a drone: 259 m.p.h., 318 m.p.h., and 337 m.p.h. At each of these airspeeds, the cockpit was impacted by one of five of the most popular quadrotor unmanned aerial vehicles (UAVs), all donated for the experiments by drone maker DJ-Innovations (DJI) of Shenzhen, China; the drones were suspended on a high-speed-camera-populated gantry in front of the rocket sled at a height that would intercept the windshield. The DJI drones tested were the Spark (weight: 300 grams), Mavic (700g), Phantom 4 Pro (1.36kg), the twin-battery Phantom 4 Pro (1.82kg), and the Inspire 1 (3.33kg).

One of the most interesting findings from these tests was that drone mass alone was not in itself an indicator of risk to windshield airworthiness: the drone's impact orientation, or "attitude" in aviation speak – mattered too, as did the windshield impact position (such as the center or corner of the pane), and the way the mass of the camera, battery, and motor was distributed about the drone chassis.

The tests show that the super-lightweight Spark had no effect on the windshield that concerned the experimenters, while the Mavic broke only the outermost layer of the three-layer reinforced windshield, at top speed and in one impact orientation. Indeed, none of the drones actually penetrated the windshield at any speed.

However, the Phantom 4 Pro, at one incident angle and at top impact speed, did crack every layer of the three-layer windshield, constituting an unairworthy condition, Lu's team says. "Flying windshield fragments of the innermost glass would threaten pilot safety," they say in their paper. The twin-battery version broke both the outer and middle layer, a condition regarded as "dangerous," say the researchers. Yet the heavier Inspire 1 only broke the outermost layer of glass, a condition considered "safe."

"This independent research corroborates our own research findings, primarily being that a drone strike is likely to cause significantly more damage than a birdstrike, and therefore that aircraft certification standards required for resilience against bird strikes does not provide any guarantee of resilience against a drone collision," says Joji Waites, flight safety specialist at BALPA in London.

"While it is encouraging that the research shows similar results, and therefore might be taken onboard by some naysayers in the drone community, it is concerning to see the level of damage that could be caused, and we believe the wider aviation industry needs to take ownership of this threat."

The wider industry now has improved software to do just that: all the actual physical damage data from these tests is being used to refine the impact simulation software, so future drone designs and windshield structures can be tested better. Although the Nanjing team found that its finite-element model fell down when it came to predicting interactions between the three layers of the windshield, these physical tests have given them the data they need to improve the model so that in the future, it will be a more versatile piece of predictive code.

Drone maker DJI sees the test methodology as valid. "The scientific analysis seems sound, and it's certainly worth noting that the two smallest drones – the Spark and Mavic Pro – did not seem able to hit an airplane windshield strongly enough to make it unairworthy," says a DJI spokesperson in Shenzhen.

"We believe this is a sound foundation for further safety research and product design consideration, for scenarios that remain extremely rare."

However, the real technological prize in this arena will be the development of software and systems that avoid drone impacts. "Collision avoidance software development is important. We need to try to avoid the drones, but we also need to understand what happens if a collision does occur so we can assess flight safety risks before we populate the skies with millions of drones," says Poormon.

Waites agrees. "Avoiding a collision in the first case is preferable, so efforts need to be redoubled to establish safe separation standards and requirements for the use of geo-fencing and technology to facilitate remote drone ID."

It is fitting, since it was computer science that gave us the astonishing capabilities of drones, providing the breakneck-speed real-time computing capabilities that fuels their dynamically gyro-stabilized flight, that computer science, through simulations and collision avoidance tech, should now also help alleviate the risks drones have introduced.

Paul Marks is a technology journalist, writer, and editor based in London, U.K.

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More