Drone Mid-Air Collision Study
On Saturday 22nd July 2017 the UK’s Department for Transport (DfT) and Military Aviation Authority (MAA) published a report into the outcomes of testing mid-air collision scenarios between aircraft and drones. Entitled, “Small remotely piloted aircraft systems (drones): mid-air collision study”, it was paid for by the British Air Line Pilots Association (BALPA), MAA and DfT.
The publication of the study coincided with a press release issued by BALPA entitled, “Act now on proven drone collision threat say pilots” and an announcement by Lord Callanan, the Aviation Minister, that the UK Government was set to bring in tighter drone regulations.
The following reason was given for tightening the UK’s drone regulations: “The move follows safety research that concluded drones could damage the windscreens of helicopters.”
BALPA went much further stating, “BALPA believes the results of the tests are a robust verification of the Association’s warnings over several years that drone impacts on aircraft windscreens and helicopter rotors can be catastrophic, even at relatively modest speeds with small drones, and that the industry and regulator cannot rely on birdstrike data and certification for drones.”
The most obvious surprise after years of media hysteria around drone-plane collision scenarios was the absence of any reference to engine impacts. There’s a very good reason for this and one that BALPA don’t publicise. All tests so far have shown that a drone flying into a jet engine will not cause the engine to explode or the wing of the aircraft to disintegrate. The worst case scenario is that it may cause the engine to fail, in which case the plane can still continue its journey or re-route to an alternative airport for landing.
This is in stark contrast to comments made by Steve Landells of BALPA who told the Guardian newspaper in 2016, “if one [a drone] hits a jet engine, it will not only stop it but potentially cause an uncontained engine failure, with bits of metal flying off penetrating the cabin and fuel tanks.”
This hysterical claim by BALPA’s Flight Safety Specialist is in direct contradiction to the regulations governing aircraft design. The Federal Aviation Administration (FAA) stipulate that, “turbine rotor cases must provide for the containment of damage from rotor blade failure”. This condition was introduced following incidents such as the uncontained turbine failure of Delta flight 1288 in 1996 and came into force on 1st January 2002.
The regulation applies to all jet engines in service irrespective of whether they were manufactured before or after 2002. As a result uncontained engine failures are now extremely rare and often the result of poor regulatory oversight and inspection such as those that occurred in 2016.
For BALPA’s Flight Safety Specialist to suggest that a drone could cause an uncontained engine failure is both reckless and purposefully inflammatory. It’s because of comments such as these that BALPA statements are looked upon with deep skepticism by members of the drone community.
As for what the likely outcome of a drone flying into a jet engine would be it’s more reliable to look to Virginia Tech than BALPA.
Virginia Tech’s CRASH (Crashworthiness for Aerospace Structures and Hybrids), part of its mechanical engineering department, has developed a computer simulation showing what would happen should a 4kg drone fly into a jet engine at take-off.
The results show that the engine very quickly fails – a similar outcome to a large bird flying into the engine.
If a twin engine jet, such as a Boeing 787 Dreamliner lost an engine it can still fly, land and even take-off. This is tested fully before the airliner is made available to buy as Boeing explain themselves in their 2014 article, “No Engine, No Problem.”
It’s actually far more dangerous to try and land an airliner when one of the sets of landing wheels fails rather than one of its engines.
For this reason BALPA have now focussed on the issue of drones flying into aircraft windscreens or helicopter rotors because they are aware that the risk of drones flying into airline engines isn’t a catastrophic scenario. Granted it’s not desirable, but neither is a large goose flying into the engine, both of which have similar outcomes.
Going back to the BALPA, DfT, MAA study it’s important to clarify exactly what types of drone impacts are likely to be catastrophic and how probable they are to occur. From the title of BALPA’s press release you would be forgiven for concluding the, “proven drone collision threat”, would be highly significant and the evidence beyond any doubt – not so if you read the full report.
Three aircraft structures were chosen for the test:
(1) Helicopter windscreens (one birdstrike certified and one non-certified)
(2) Helicopter tail rotors
(3) Large airliner windscreens
It seems unusual that the test would choose to focus on helicopter tail rotors yet ignore testing helicopter main rotors. It’s especially odd when you consider the nature of helicopter aviation. The tail rotor is much smaller than the main rotor and principally counters the torque effect the main rotor produces. As such it’s much smaller than the main rotor and presents a far less likely strike area.
Air flow around helicopter rotors is shown in the diagram opposite and as you can see these are overwhelmingly created by the main rotor. During hover air is sucked from the space above the helicopter and pushed downwards with forces causing it to travel between 60-100 knots, or 70-115 mph.
Below the main rotor any drone is likely to be knocked out of the sky by the force of this rushing air and above it the drone is likely to be sucked into the blades. Clearly then there is a far higher likelihood of a drone striking the main rotor than the tail rotor, so why wasn’t this included in the test?
The answer must be that BALPA, the DfT and MAA all knew that a drone striking a helicopter main rotor would cause no significant damage in much the same way that birds flying into the main rotor don’t. This indicates that the researchers intentionally selected aircraft structures that were most likely to be damaged in a collision, not those that were most likely to be involved in a collision, which raises some questions about the intention behind the study.
By far the most likely parts of aircraft to be involved in collisions with drones are jet engines, wings, fuselage, and helicopter main rotors yet all of these were excluded from this research.
The European Aviation Safety Agency (EASA) commissioned a report into birdstrike damage in 2009. In the final report they found that only 1.8% of all reported bird strikes involved helicopters. Over 98% of bird strikes involved fixed wing aircraft, the vast majority of which were airliners.
Was there such a large focus of this study on helicopters with the intention to try and prove a “drone collision threat” at all costs? Is that also why they excluded engines and main rotors from the study?
On further examination of the report there are a lot of factors that appear to indicate this study was biased from the outset, which is very troubling, especially when it’s used as justification for increased legislation.
One of the proven threats involved the testing of non-birdstrike certified helicopter windscreens. These were shown to fail when hit by drones, which is no great surprise since they also fail when struck by birds, as their name would suggest.
In general aviation (GA), windscreens don’t need to be certified to withstand birdstrikes, so the findings of this study are applicable to a wide variety of aircraft within the GA field. In short they found that if your windscreen won’t withstand a birdstrike it won’t withstand a drone strike either.
To put this within context there were 1835 confirmed birdstrikes reported to the Civil Aviation Authority (CAA) in 2016. To date there has never been a single confirmed drone strike reported to the CAA.
When testing helicopter windscreens that were certified to withstand birdstrikes it was found that these were only penetrated by quadcopter drones when the helicopter was simulated at travelling close to its cruising speed.
The study didn’t specify what speed this was, but the average helicopter cruising speed is around 130 knots/150 mph. They worked on the assumption that the drones would be flying at their top speed directly into the aircraft, so this would give a combined collision speed of around 160 knots/185mph.
The study also failed to disclose which quadcopter drones caused this critical damage to the birdstrike certified windscreens. As such it would be the safest option to conclude that the critical damage occurred with both the 1.2kg and 4kg quadcopters travelling at 160 knots/185mph towards the windscreen.
They also tested fixed wing drones and concluded that these penetrated the certified windscreens when the helicopter was travelling at much lower speeds. However, this is based on the fixed wing drone flying at a faster top speed than a quadcopter, so the combined collision speed could have been in the same region of 160 knots/185 mph.
The inclusion of fixed wing drones in the study again raises the question of the intention behind the research.
Fixed wing drones are incredibly rare in the hobbyist market and weigh much less than 1kg. In the professional market they weigh around 2kg and cost in excess of £20,000. As such they are only ever flown by very competent pilots who would never operate them in the vicinity of any aircraft.
Despite this the study chose a 3.5kg class fixed wing drone with a front mounted propellor.
The most popular hobbyist fixed wing drone is the Parrot Disco which has a rear mounted propellor and weighs just 750g. Even a professional drone such as the Datahawk by Quest UAV only weighs 2.2kg and also features a rear mounted propellor.
Despite the fact that almost all fixed wing drones have rear mounted propellors the researchers chose to use one with a front mounted propellor. They did this specifically because it would cause more damage and would be more likely to penetrate windscreens.
“The fixed wing type with the nose-mounted propeller was used as this was considered to represent a greater impact risk than tail-mounted propeller.” [page 10]
Why would independent scientific research choose the most damaging and rare type of drone to test rather than the most likely type of fixed wing drone for aircraft to collide with? It’s the equivalent of choosing a golden eagle to test birdstrike resilience rather than a seagull.
The authors of the study could not be clearer that they primarily wanted to show damaging results and this was far more of a motivating factor than realism.
Why is this desire for a pre-determined outcome acceptable for research that will be used to shape UK legislation?
The study also fails to make clear the certification standards that apply to helicopter windscreens for them to pass birdstrike tests.
The European Aviation Safety Agency (EASA) state in their Certification Specifications for Large Rotorcraft CS-29, “The rotorcraft must be designed to assure capability of continued safe flight and landing (for Category A) or safe landing (for Category B) after impact with a 1 kg bird” [CS 29.631]
Sikorsky developed a birdstrike resistant windshield to meet British Civil Airworthiness Requirements Section G (BCAR Section G) standards when introducing the S-76A to the UK market. These were tested in 1978, 1982 and 1985 and in each case the testing consisted of firing a 1kg object towards the windshield at 160 knots/185mph.
The Ministry of Defence (MoD) Defence Standard 00-970 Part 7 Section 2 “Design and Airworthiness Requirements for Service Aircraft Part 7: Rotorcraft”, defines the requirement for meeting birdstrike certification in helicopters in relation to a bird with a mass of 1kg:
“The threat for the purpose of these requirements is a single strike by a bird of 1.0 kg mass.”
It is clear that in order for a helicopter windscreen to be certified to withstand birdstrikes they are tested to withstand impacts from birds no heavier than 1kg at speeds of around 160 knots/185mph.
Yet in this study the researchers were firing 1.2kg – 4kg class drones at the windscreens at roughly the same speeds. It’s therefore no surprise that these windscreens failed in test conditions that exceed their certification.
The researchers should have made clear the specifications relating to helicopter birdstrike certified windscreens because if they had fired 1.2kg – 4kg birds at the same windscreens at 160 knots/185 mph they would also have failed.
What is alarming is that the authors hid this aspect from the final report.
Why didn’t they test a 1kg drone with the birdstrike certified windscreens as this would have given a side-by-side comparison with birdstrike testing?
Is it because they were only interested in finding results that proved drones could damage aircraft?
Helicopter Tail Rotors
The reason for the inclusion of this aircraft structure becomes immediately obvious with the following statement:
“The modelling of helicopter tail rotors showed that they would be vulnerable to impacts with all types of drones. Due to the very high speed of a rotating tail rotor blade, it could be critically damaged by an impact with any drone.” [page 15]
This provided the study with their, “proven drone collision threat”, and “catastrophic” impact results. However these results were obtained only through untested computer modelling.
The report explains the modelling process on page 11:
“Use of impact-modelling software can provide additional insight and can enable a wider range of impact conditions to be considered, but to do so reliably requires that the models be validated by experimental tests. This calibration and validation activity was at the centre of the study’s requirements, to ensure that an accurate model was created which could be used for this and future work.”
To use modelling reliably it requires that the models be validated by experimental tests, yet no such tests were used with the tail rotors. By the report’s own admission all of their conclusions and data in relation to helicopter tail rotor drone strikes is therefore unreliable.
The study stated that, “No live testing was performed for the tail rotor blades due to difficulty in acquiring blades and the limited resources available for the study.”
Bearing in mind this study was commissioned by the British Air Line Pilots Association, the UK government’s Department for Transport and their Military Aviation Authority it seems ludicrous that tail rotor blades couldn’t be obtained for live testing.
In their 2015 annual return BALPA declared over £12.5 million in total assets and stated that they had just under 9,000 members who provide just under £6 million a year in subscription income. Today BALPA state that they represent 10,000 members which means their annual income is in excess of £6.5 million from members alone.
According to their annual accounts the DfT had an operating income of over £329 million for 2016-17 and the MAA is part of the Ministry of Defence with a comparably significant annual budget.
We are therefore expected to believe that the combined finances of the three commissioning organisations and their industry access and connections meant they were unable to source a single helicopter tail rotor.
I don’t accept there is a significant difficulty in obtaining rotor blades. For example a quick Google search reveals one company, Ross Aviation, that carries over 20,000 helicopter parts in stock.
Additionally the study was conducted by QinetiQ and Natural Impacts who are described as follows: “Both organisations have a wealth of experience in their respective fields, and are highly regarded organisations who conduct studies for defence, international companies and regulators. They also have first-hand experience with birdstrike testing and impact modelling.”
Given that the helicopter tail rotor was shown to be, “vulnerable to impacts with all types of drones”, it is simply unacceptable that this data was only obtained through computer simulations that were unreliable and uncalibrated.
It is likewise unbelievable that they couldn’t afford to buy a single tail rotor or that they couldn’t find any that were available for purchase. Their statement is patently untrue and as such casts a long shadow of doubt over the veracity of these findings.
The absence of live testing of tail rotors is even more troubling when you consider EASA’s certification standard for large rotorcraft CS 29.631 quoted above. Both category A and category B helicopters are tested with 1kg projectiles launched at their tail rotors without causing them to fail.
Furthermore research shows that the S-76D helicopter tail rotor passed ice strike tests with a 1lb ice block causing only minor non-structural damage.
Could their reason for only using uncalibrated modelling rather than live testing be because they knew drones wouldn’t cause tail rotors to fail?
The study found that airliner windscreens could withstand impacts from all drones on approach and landing which clearly wasn’t the result that they were hoping for. They went on to state:
“At higher altitudes and speeds, modelling and testing showed that severe damage to the Airliner-A windscreen, including complete structural failure of the windscreen, did not occur with the 1.2 kilogram class quadcopter components, but could occur during impacts with the 4 kilogram class quadcopter components. Additionally, during one high speed live test with the Airliner-B windscreen, the 3.5 kilogram class fixed-wing drone components penetrated the windscreen.”
To give some idea of the speeds and altitudes they used to get these results they state these, “impact speeds would usually be encountered when the aircraft is at higher altitudes, 10,000 feet or above”.
The most likely scenario for a 4kg quadcopter to reach 10,000ft would be for it to be of the DJI Inspire 2 type. This has a mass of 3.5kg and a maximum take off weight of 4kg. Its ascent rate is 5m/s and 10,000ft is around 3,000m. That means it would take 10 minutes for someone to fly an Inspire 2 to that altitude. Its maximum flight time is 25 minutes which will reduce as the temperature drops on ascent, but it’s just about possible for someone to fly it to that height and back down again.
The reason for the 10,000ft altitude is because airliners have a 250 knot speed limit when below 10,000ft, so the study was clearly using speeds well in excess of 250 knots to produce these results.
To put that in layman’s terms it means that when they fired 3.5kg and 4kg class drone components at the airliner windscreen they were only critically damaged at speeds well in excess of 300mph. They don’t specify what speeds they were, it could easily have been 500 knots or 575mph which is the average cruising speed of an airliner.
The fact that this study went to such extreme testing conditions in a desperate bid to show drones can damage airliner windscreens is yet further evidence that this study was not about testing a hypothesis, it was about proving a point, no matter how absurd.
Rather than being a damning study showing a clearly “proven drone collision threat”, to airliners this report does the opposite. It shows that there is no threat to airliners from drones on approach and landing and that any threat would only occur in the most extreme and, by definition, rare of circumstances, if ever.
It shows that there is a threat posed by drones to the general aviation (GA) community, but that is the same threat that birds also pose to them. So it tells the GA community what they already knew – avoid drones and birds.
The report did raise some concerns about the possible danger to helicopter windscreens that are certified to withstand birdstrikes, but it failed to ensure a comparable test within the windscreen certification limits. As such it leaves helicopter pilots unsure as to the likelihood and possible severity of drone collisions, which is a significant failing on the part of the commissioning organisations.
Likewise the data on helicopter tail rotors is obtained in such an unreliable manner that it doesn’t justify its inclusion. The authors are fully aware that the data is unreliable, uncalibrated and invalid by their own definition of the modelling process, yet they fail to admit this.
BALPA go further by using this knowingly unreliable data to claim collisions between drones and, “helicopter rotors can be catastrophic, even at relatively modest speeds with small drones”. The claim is clearly misrepresenting the data and is inflammatory. The fact BALPA are so keen to create and fan the flames of media hysteria towards drones should raise immediate concerns and exclude them from participation in future government research.
I am a huge supporter of increased drone regulations and registration for private owners, but I am appalled by the low standard of research this study presents.
I find it unacceptable to conclude that QinetiQ and Natural Impacts were incapable of more rigorous real-world testing procedures and can only surmise that this study intended to demonstrate drones could cause critical damage to aircraft at absolutely any cost.
It seems the result of this study was clearly specified prior to its commencement and as such this sets the arguments for tighter regulations back rather than moving them forwards. This is pseudoscience that reeks of confirmation bias which only provides ammunition for those opposed to drone registration.
Where this study fails the most is that it seeks to try and prove that threats are posed to aviation by hobbyist drone operators. The rationale for this seems to be so that it can be used to justify Lord Callanan’s announcement of tighter regulations.
This bias from the outset blinds the researchers and its commissioning organisations from a somewhat different but rather obvious conclusion. Namely that the study highlights a significant danger to the GA community that drone delivery services such as Amazon Prime Air pose.
Non-certified windscreens, as found on most general aviation aircraft, were susceptible to critical damage from all drones, including penetration of the windscreen when at speeds well below the cruising speed of a helicopter. It can therefore be concluded that GA aircraft operating above stall speed will suffer catastrophic windscreen failure should they collide with a drone.
At present the UK government is poised to give Amazon permission to operate delivery drones at altitudes up to 400ft throughout the UK. Amazon drones will be flying autonomously and will have sense and avoid systems. However it is impossible for them, or any drone for that matter, to detect an aircraft travelling at 60 knots or 70mph and move out of its path in time to avoid a collision.
This study therefore demonstrates there is a clear and present danger to the GA community should Amazon be allowed to implement its delivery drone concept. For some reason though this rather obvious conclusion has missed BALPA, the DfT, the MAA, the CAA, and Lord Callanan. Instead once more hobbyist drone pilots are targeted, their potential threat overstated and the real threat to aviation missed entirely.
Maybe instead of looking to the UK for the answers to the problems posed by this emerging technology we should look to the US and truly independent scientists such as those at Virginia Tech?