Podcast: How Drones Are Shaping Aircraft Inspection Techniques

Aviation Week editors James Pozzi and Lindsay Bjerregaard are joined by Dejan Borota, CEO and co-founder of drone-based inspection specialist Mainblades, to discuss technology's development in MRO, the role of artificial intelligence in inspections and the direction  future adoption is heading.

James Pozzi (00:16): Welcome to the MRO podcast. I’m James Pozzi, MRO editor for the EMEA Region. And today we are talking about drone technologies and its use in the aftermarkets. So joining me today, two people very well-versed on this topic. First, we have Dejan Borota, who is CEO and co-founder of Mainblades, a specialist in drone-based automated aircraft inspections, and Aviation Week's own Lindsay Bjerregaard, managing editor for MRO. Lindsay has written extensively about this topic for several years, and of course wrote an excellent feature on the drones in MRO landscape earlier this year, which you likely read in Inside MRO. So Dejan, Lindsay, thanks for joining us.

Lindsay Bjerregaard (01:03): Thanks James.

Dejan Borota (01:04): Thank you James.

James Pozzi: Yeah, Dejan, we'll start with you, of course. Maybe for some of our listeners who maybe aren't so well-versed in drones and how they're used for MRO and maintenance inspection tasks, would you also just provide a basic overview of drone inspection technology in this space?

Dejan Borota (01:25): Sure thing. Drone inspection technology has been actually around in aviation for quite a while. I think the first notions of this technology, drones used for inspections, have been introduced way back in 2014, in a time when drones were taking off in all sorts of fields and areas, right? Everybody was buying a drone, trying to fly it and trying to do something with it. Technology was way less reliable back then, right? You had a lot of do-it-yourself, build your own drone stuff. But very early on, the industry caught up to the sort of idea or the vision that you could actually use this small aircraft, small unmanned aircraft, to inspect this large aircraft. And with that, do that even with great speed, so reduce the time that's needed, but get also to a height that is sometimes difficult to reach for engineers and mechanics to do it.

(02:13): So very early on, the industry latched to it, but it took its time to sort of get it integrated and onboarded into the industry. That has a lot to do with the conservative way that the industry approaches new technology, but it also has to do with the maturity of the technology. So drones needed to grow, automation software needed to grow, robotics software needed to grow, and what you have now is a far more mature product. And then an industry segment is developing around it where you essentially get drones, and maybe it's not even about the drones. You get actually cameras or different kinds of sensors, which you mount to a drone. And with the drone and the automation software behind it, you can actually do automated inspections of an aircraft much, much faster than you would do a physical inspection. The drones fly around the aircraft, usually in an automated way, capture images.

(02:59): These images are then evaluated with the help of software tools, but mostly by mechanics who can then actually compile an overview of the asset being the aircraft, finding damages, working through task cards, inspection task cards of different kinds, paint checks, lightning strikes, general visuals. And the product has … These kind of products have a lot of benefits, right? They're a lot quicker, which helps saving costs. But also the biggest promise is actually being able to increase fleet availability by returning the aircraft to service, but also making sure that people don't necessarily have to go on heights when they don't need to. You can get a quick view of your aircraft, of different parts of your aircraft, and make assessments on how to maintain that aircraft in a proper way. So starting 10 years ago, it was a nice dream, a nice vision, an idea of how to do this. It has developed over time, and now we're really seeing a product available in industry and being adopted and the traction increasing with people looking at all sorts of implementation of drones, but also automated inspection technologies.

Lindsay Bjerregaard (04:00): So you'd mentioned how long this has kind of taken, right? I started covering the industry about eight years ago, and I've sort of seen the drone inspections go through a couple different hype cycles. At one point it seemed like there were a lot of different drone startups. It looked like some of the OEMs were really trying to pilot the technology and all of that. But now it really seems like there's been market consolidation. There kind of seem to be two main players, Mainblades, and then Donecle as well. So I'm curious what your thoughts are on how that market and how the technology has changed since you guys first launched.

Dejan Borota (04:35): Yeah, I think that's a very good observation, and to an extent it makes sense even nowadays when we walk into a hangar or maintenance environment in a new space, the guys here are maintaining aircraft. They're enthusiasts, right? They love everything that flies. Naturally, they love drones as well, most of them. So when this opportunity, when the technology of drones became available 10 years ago, a lot of people jumped on it like, oh, we can do this. It's a great idea. It's very simple, right? Inspect an aircraft with the drone faster. However, what I just mentioned, technology maturity was really not there. Not only in terms of the drones itself, the reliability of the hardware components, the cost of the hardware components. I remember the first drones were a couple of hundreds of thousands of euros, and they would crash after a couple of minutes flying.

(05:22): But also the automation, the biggest thing is that even if you have a drone flying around an aircraft to be able to conduct a reliable, repeatable, good inspection, this has to be fully automated. So all the robotics algorithms, all the automated flight algorithms, even the evaluation algorithms, machine learning and artificial intelligence, they all needed to develop. And quite frankly, from my point of view, this technology has only been mature maybe for three or four years to be reliable and well enough to actually be able to be used by airlines, MROs and other organizations in a production environment. From the get-go, there were a lot of parties, contractors, employees, organizations. We've heard everything. We've heard MROs, airlines, OEMs built something, but you do need some kind of quite specialized technological knowledge for it, which is each quite specific. I always try to say, Hey, we're doing similar work in terms of automation like a Tesla does, but we're doing it at 3D, which makes it quite a bit more challenging.

(06:20): That's a specific skill you need. And over time, people saw these projects rising up, but not latching, how would I say, not catching traction in the industry. And they winded down. The second barrier here, of course, was not only the technology maturity, but the regulatory hurdles, which were anywhere from, can I fly a drone indoor, outdoor, at an aerodrome? Can I use images to inspect? Can I use artificial intelligence tools to help me with that? And all of that has also took its sweet time to be reviewed, evaluated, reviewed, adopted by industry players. I have to say that the people that are still involved with us now, ourselves, Donecle, but also a couple of other organizations that are leading mostly airlines and OEMs, we're supporting efforts. They've had to sit through or stick through the hurdles of proving that this is safe, this is reliable, this is something that can be used in a production environment that is the most important thing, that eventually all this technology can be used in a classical MRO environment where it's busy, where it's messy, where it's windy, where it's a lot of stuff is happening and albeit remaining reliable.

(07:30): So I think while that has progressed, the industry has become more and more open to it, and regulatory hurdles have, I wouldn't say disappeared, but they have allowed more steady adoption of the technology.

James Pozzi (07:43): So it's gathered a bit of pace, and it's developed and come on and progressed in say, the five- to 10-year span.

Dejan Borota (07:50): Yeah, and I think I would argue that the biggest progression has been in the last four or five years when ourselves and others have had equipment that is able to do the work and we engaged regulators with the right partners, be it airlines or OEMs, to actually find a way to operationalize this technology, which is still a giant hurdle. I think there are still a lot of airlines and MROs who are trialing this technology or doing something with it, but they're still finding difficulty to operationalize it. So to actually use it in that scenario and actually generate the benefits that it promises.

James Pozzi (08:25): Yeah. Artificial intelligence. Now it's becoming more omnipresent in our daily lives and naturally in the aftermarket. And certainly AI is going to have a huge impact as it is already, but even more so in the coming years, other existing technologies being used, no doubt. But what role do you see this potentially playing in drone inspections going forward and AI?

Dejan Borota (08:46): I get the question a lot, and I think it's a very valid question. I love talking about the promise of AI. I consider the third hurdle for adoption of technology. So the first one being, can I fly a drone around an aircraft on an aerodrome? That's the first hurdle which has overall been mitigated. We can fly in a lot of different environments, even outdoor and active airports. It's proven with every civil aviation authority, but it's there. The second question then is can we use this kind of inspection technology to actually do task cards, do inspection task cards? That has also been proven with several civil aviation authority with several of our airline partners and with OEMs as well. And then the third hurdle is can we use AI? And it's a funny thing. So the second hurdle is very much similar, like the third hurdle, because in both scenarios, the drone is used in the second scenario, the drone is used as a tool.

(09:39): You're gathering data images to help the mechanics make an assessment. Now, AI is being sometimes reviewed as something entirely different, but it's not. It's equally a tool. It's equally a system that is deployed to help the mechanics use the data to actually come to a quicker evaluation. I think there's a huge role for it. I think there's still a little bit of, people are a little bit sort of frightened or what's going to mean often. Most often I get the question, how well does it perform? Is it better than humans? Which is quite frankly the wrong question. So for example, if you look at the medical sector as a comparison, AI is being used or utilized quite a while earlier there. And are doctors afraid to be replaced by that? Hell no. Are they afraid that they're going to be outsmarted? No. It helps them make an assessment faster, better. It helps 'em do their job better. And that is exactly the same way aviation should be looking at this as well. Right?

James Pozzi (10:33): I saw a list the other day. I think Microsoft did a hundred jobs most likely to be replaced by AI and a hundred jobs least likely to be replaced by AI. It's quite interesting this last week. So worth looking at. Doctors weren't on that list.

Dejan Borota (10:45): It's very funny. If you look at AI tools, if you are in aviation when it's making a decision, people get very anxious because of course you have licensed aircraft engineers and they have certificates. So how can a software always make a decision for them? It won't. It will help them make a decision. It'll make the better decision, a faster decision. And if we have that conversation, then we can actually do something about how to bring this into the industry. I think that's very important to sort of shift that conversation scope of whether it's better or not than the current, because we already know it can be better than the current, but it's the way you utilize the tool to help the current process, the current mechanic do his or her job better.

Lindsay Bjerregaard (11:26): Yeah and when I spoke with Donecle CEO about the same thing, he had said that AI is good at complementing technicians. It helps them find very small defects. It's good at that. Whereas the technicians are good at finding unexpected or out of the ordinary damages and defects. So they see it as a combination of technician and machine. And he had mentioned as well that AI could basically automatically assess “the easy images” and remove those from the scope. So it would reduce the amount of images pretty significantly that need to be reviewed. So instead of having to look at hundreds or thousands of images, you can kind of cut that in half. So speaking of our recent conversations about this Dejan, when we spoke at MRO Europe last year, we talked a little bit about how some companies are kind of looking at leveling up drone inspections with additional technologies. So there's companies like Korean Air that are using multiple drones at once to do a swarm inspection, or there's some companies that are looking at integrating things like nondestructive testing equipment onto drones. And at the time you were pretty skeptical about the benefits of doing that. So could you explain why or if your opinion has changed about that?

Dejan Borota (12:41): I'm open to actually looking at all kinds of technologies, but it has to be very, you have to weigh off to what the customer can actually use. So a couple of your examples, I think swarm technologies are interesting, but I have a very different view on that. If you look at the basic operations, you're narrowing down an inspection time of about 10 hours for a narrowbody or 20 hours for widebody into one or two hours. So you're cutting away 70, 80, 90% of the time that was needed previously. Now you can do two drones and then you can cut it a little bit more. But is that really going to be your bottleneck? Is that sort of, because the drone inspection is the data gathering, if you cut the data gathering process by 80, 90%, would it help them to have that from one hour to half an hour?

(13:25): Not really, because you still get on the other side, the images to review and the report to evaluate. So actually I think you should weigh in on what is needed. So my own view on, for example, swarms is to use swarms or use smaller drones in difficult to reach areas if you would need that, for example, in some edges or corners to make very high details photos of specific areas. You could also look at different kind of tools, robotics technologies, ground vehicles, to actually get them to gather data from different areas. So I would like to look at it more as a sort of a operation where you have several machines, drones, robot vehicles, maybe smaller drones. But you have to look at, okay, what makes sense from an operational point of view, deploying 10 drones and then you do your inspection in 15 minutes, but then you have to do two hours of reviewing of images?

(14:12): That doesn't help anymore. So you have to look at what tool to build. Actually brings us back to the AI tool because that can be used to help on the second part, on the review part, and I think what you just mentioned with indeed, AI being used to narrow down the scope of inspection that is often spoken about. I would like to shift that scope or the framework. Again, you don't want to necessarily discard images. That's where CAAs get very concerned, but you're supposed to look at the whole aircraft. You are. So what would you do? You can actually use AI technology or advanced software technologies to look at orders. What parts should we look at first? Usually when there's a lightning strike, there's a path along an aircraft. AI can be utilized to actually help a mechanic find that lightning strike path, right? There's an entry point of the lightning strike.

(14:59): It has a path around the aircraft and it's an exit point. That's the path the mechanic has to find. He doesn't have to cover 95% of the aircraft to find it because the algorithm can help in that. And then he can cover the task card in a different way. So I would look at what tool helps our customer most. To the second technology. I opened actually with the sensor, the drone is a tool, right? It's supposed to bring a sensor near to an aircraft to perform an evaluation. Currently, the most common method is a general visual inspection using cameras. We are looking at all sorts of tools, but usually we want it to be a specific customer request. What do they need? And maybe a bit rude to say, but a lot of customers don't know what they need. In the very early stages of when we talk to people and do this discovery, we get a lot of questions.

(15:47): Can you put this on a drone? Can you put that on a drone? Can you put X, Y, Z on a drone? Sure everything you can put everything. Why do you want to do that? What do you want to gain from it? What do you want to measure? Most often when people are in a very early stage of assessing drone inspection technologies, they have sort of an own idea vision of how this could be. And they ask for all sorts of, can we put this on a drone and then measure this? But then you ask 'em, Hey, but that sensor is available already in another form factor. Have you used that before? The most common answer I get is no. So why not? Because if you use the other form factor, which might not be drone-based, you can already evaluate whether that helps you improve your process.

(16:23): If it helps you and you have a drone, which you have value as well, then you could look at combining technologies and seeing as does that additionally improve our processes or not? So I'm open, we're experimenting now with two or three different payloads, we call them, right? To see what makes sense. But it's very important to not develop something out of a technical capability like, oh, this works, it's cool. Let's try to push. It has to be a commercial need, right? There has to be a market need asking for specific measurement on an aircraft, which requires a specific payload. And then the drone solution should be actually an added factor of improving that payload's ability to do a quick assessment. And in a lot of cases, I see a big technology push, right? It's possible. So let's try it. But then the question is what does it help? We're always very critical about that point of view. You should make the development once it's clear that it can help a customer in their use.

James Pozzi (17:16): Sure. And finally, the last few minutes, obviously had quite a lot of use cases. We've looked at, Korean Air was mentioned, of course. I remember several years ago in EasyJet here in the UK, was doing drone inspections. Of course, AAR is another one that comes to mind in North America. But throwing this out to both of you, the regions where drone inspections are seeing the most traction growth, is there a particular region in the world where this is the case, or is it all more balanced between region to region?

Dejan Borota (17:46): I could comment on that with an anecdote. Last week, I flew a circle around the globe, literally full circle, visiting customers in each continent, customers and partners. I would argue at this point in time, it is everywhere. Everybody has heard about it, saw something about it and is now really interested in it. I think maybe a year or two or three ago, you would see a big bubble around in the North Americas, right? They were focused on it. Europe. Europe might've even led for a while. I think it started in Europe, quite coincidentally, both ourselves and our competitor Donecle are European-based. Europe has been the region where it sort of started first. Same is EasyJet. They kicked it off that 10-plus years ago. The Americas took a little bit to it. They progress it. I would think with our American partners. We progressed it towards a place where others were, Hey, this is now really interesting.

(18:36): But now I see it literally everywhere. We are flying around the globe constantly. Americas, Europe, Gulf, Asia, Southeast Asia, but also northern parts of Asia and everybody's there sniffing around, okay, what can we do? How can we integrate this? There's approvals in place. There is a precedent by both CAAs and OEMs to actually allow these inspections. And now everybody wants a bite of it. I think it's only a matter of time before we're going to start talking about, Hey, why are you actually not using this? If it's available and if it's as good as we see, why is this not being used overall?

Lindsay Bjerregaard (19:10): And just to chime into a little bit, I pretty much just agree with what you said, Dejan, and from my perspective, the countries where it's been most publicized or the regions where it's been most publicized have obviously been Europe and North America. We're starting to see more being talked about in Asia. South America actually it seems like has had quite a bit of drone inspection activity happening there. Donecle claims that they've really dominated that market. Matthieu Claybrough, the CEO, said they have drones in most of the countries down there. They're in talks with other airlines, so he's expecting they might soon have drones across the entire continent, but like Mainblades, they're also targeting the Middle East and Asia. And I just thought that this was personally interesting, Donecle's planning to open a new U.S. subsidiary in Chicago to help with their production ramp up. So I'm based in Chicago. Hopefully I'll finally get to see a drone inspection in person. But, I dunno, Dejan, maybe you can get to it first and let me see.

Dejan Borota (20:09): You're always welcome to come visit us in Amsterdam. We'll happily host you here, show you around. In the U.S. as well. We're actually also ramping up our operations and it's looking like in the coming six months, we'll open a couple of offices, not necessarily to produce drones. Production is not a bottleneck, but we do see that the proximity with operators, with customers, with users is something we would like to have at least in a somewhat similar time zone, right? You can imagine if you talk to the Australias of the world with a 10- or 11-hour time difference, it's nice if you have something on the east, in the Southeast Asia. Same with America. So I think now that it's all getting traction across the globe, we also need to start operating from everywhere in the globe. So we're opening likely couple of offices around the globe in the foreseeable time as well. But again, I want to reiterate, you're always welcome in Amsterdam. We'll happily host you here, show you around, or any of our other partners around the globe.

Lindsay Bjerregaard (21:01): Thank you.

James Pozzi (21:02): Excellent. Well, this brings this really fascinating discussion to its conclusion. But Dejan and Lindsay, thank you so much for your insights today on the drone market and giving our listeners that insight into maybe how that'll evolve in the coming years or so.

Dejan Borota (21:19): Thank you for the opportunity. It was a pleasure to be with you here today.

James Pozzi (21:23): Excellent. And just a shout out as well to Andrea Copley-Smith for producing this podcast, making us sound good. Don't miss the next episode by subscribing to the MRO podcast wherever you listen to podcasts. And one last request. If you're listening in Apple Podcasts or Spotify and want to support this podcast, please leave us a star rating or write a review. Thank you.