Printed headline: Inspecting by Drone
EasyJet and unmanned-aircraft developer Blue Bear Systems Research made headlines in 2015 when they demonstrated inspecting an airliner by using a drone. Fast forward to 2018, and the partners are close to fielding a system to assist with visual inspections for hail, lightning strike and other damage.
“There are now other people out there, but we still think we are fairly far ahead with a solution that is integrated with an existing MRO database,” says Gavin Goudie, operations director at UK company Blue Bear. “Access to EasyJet engineering has been a big boost.”
Toulouse-based Donecle delivered its first aircraft-inspection drone systems to operators at the end of 2017, says co-founder Josselin Bequet. The company began flight testing in 2015 and since 2016 has been developing the system with Air France Industries–KLM Engineering & Maintenance, one of the early customers now using the system for general visual inspections, checking markings and paint quality.
Back in 2015, EasyJet demonstrated inspecting an Airbus A320 inside a hangar at Luton, England, using Blue Bear’s Riser drone. This is a specialized air vehicle with four enclosed rotors designed to operate in confined spaces. “Riser is designed for nuclear inspection. We needed a dedicated system for aircraft inspection so we created Rapid,” Goudie says. “Now we are close to commercial release.”
Rapid, which stands for Remote Automated Plane Inspection and Dissemination, has been developed with UK company Output42, which provides software for 3D modeling and damage reporting. They have formed a joint venture, MRO Drone, to offer a drone-based inspection system. “A number of fairly big UK, European and U.S. clients want to see demonstrations in 2018,” Goudie says.
Rapid uses an available commercial drone, rather than the purpose-designed Riser. “We are keen not to have just one platform, but a plug-in system that works with any off-the-shelf platform,” says Goudie. “It’s about the ability to integrate our algorithms and payloads.” The initial payload is high-definition imaging. “We are working to integrate 3D scanning to automate damage detection,” he says.
The drone flies a preplanned path around the aircraft to image the surface and help engineers locate damage. “This is not about circumventing inspection requirements and manufacturer procedures. It’s about making their job easier by pointing out areas where there may be an issue,” Goudie says.
Aircraft have to be checked for lightning-strike damage twice a year and a drone could reduce inspection time to 2 hr. from 6 hr. “That means you return an aircraft to service in 2 hr. or begin a repair more quickly,” he says. For a low-cost carrier like EasyJet with a big fleet, turnaround time savings can make drone inspection worth the outlay, and it can provide more consistent inspections with fewer errors.
Donecle has developed a specialized drone, with four protected rotors. “Available drones were not reliable enough to use around aircraft,” says Bequet. A laser-positioning system and multiple onboard sensors allow the drone to calculate its relative position and navigate automatically around the aircraft being inspected, without having to be recalibrated for every new aircraft. “The operator puts in the registration of the aircraft, which enables the drone to know the type of aircraft and engines, then the system generates a navigation trajectory around the aircraft,” he says.
Initial use is for internal quality processes, including checking regulatory markings, because this is more easy to deploy within existing regulations. “There are a couple of hundred of these around the aircraft that need to be checked to ensure they are in good condition,” Bequet says. The drone images the markings, checks against their original condition via an internal database and generates an automated report. The system can also detect and report rivet rash and other paint damage.
For Blue Bear, most of the work in developing Rapid has focused on the front-end operating system and payload. “We have developed intuitive software to allow someone with little training to take the system, plug it together and inspect parts of the aircraft without very much knowledge of how a drone works,” Goudie says. Using the inspection planning station, the operator can map out the flight, dispatch the drone, view real-time imagery and flag visible defects for closer inspection.
Output42’s Mark English, who is leading development of the system’s front end, took Aviation Week through an online demonstration. An inspection begins by identifying the aircraft to be checked. Entering its registration queries, a database calls up previous inspection records.
The operator then identifies the drone to be flown and battery to be used, the inspection type and reason for flying. The aircraft’s position and orientation inside the hangar are determined by tagging three known locations—the nose and left and right wingtips—using a pole held next to the aircraft. Four landing zones the drone can use are then identified. “We like to keep them clear, but it’s not mandatory,” English says. The drone will only try to land when a zone is clear of obstacles.
Next comes the planning phase. The default is to inspect the entire aircraft, but the operator can select a part of the airframe to be checked—for example, the wing leading edge for a birdstrike or fuselage crown for a lightning strike.
Hitting the “plan” button creates a flightpath for the drone. Displayed as a yellow line, this is always a safe distance from the aircraft. Red dots on the flightpath show where the drone will pause to take pictures; dotted lines show the direction of the camera and blue areas show the parts of the aircraft covered by the camera. “Based on past records, the operator can add inspection points,” says English.
Planning complete, the operator is given instructions to load the battery, power up the drone and check communications with the ground control station. “Hit launch, it takes off, reaches altitude and goes to mission status,” he says. The operator’s display now shows a video feed from the drone’s camera on the left and on the right a 3D representation of the flight plan and the drone’s position.
The mission cannot be replanned in flight, but the operator can flag any point on the aircraft and go back to take a look either with another drone mission or in post-flight analysis.
A full inspection cannot be accomplished on one battery with the current drone, a DJI Matrice 100. Instead, according to English, the vehicle flies a portion of the flight, lands, its battery is changed, and it takes off and carries on as if it were one flight. “At any time the operator can stop, land the drone, then resume,” he says. “We will increase the endurance of the vehicle and, in time, streamline inspections.”
On completing the inspection, the drone requests a safe zone in which to land, even if that means traversing the aircraft. On landing, the images are compressed and uploaded to the cloud for use in post-flight analysis. “We then have a complete record in the database, and all images have metadata, so we can go back and perform the same inspection again,” says English, adding “We are thinking about automated analysis tools to compare images.”
The post-flight analysis tools were demonstrated by Paul Crabb, director at Dent & Buckle, a provider of software for reporting of aircraft damage. From a strip of thumbnail images, the operator can go to an area of interest and click on an image. In addition to the high-definition picture, the display shows the drone and image-capture area. “That way they know what part of the aircraft they are looking at, rather than just large white expanses,” he says.
A sequence of status buttons allows the operator to mark the image as pass, investigate, fail or unusable. Zooming in on a possible paint chip above the cockpit window, the operator can annotate the area as potential paint damage, create a report and send it to the MRO engineering system so that engineers can make a repair decision. This can be achieved using Dent & Buckle or some other software.
In Dent & Buckle, the engineer sees a 3D view of the aircraft showing the damage reports. “It is effectively an end-to-end system, going from the drone flight to the MRO system—all within minutes,” says Goudie. “This is not a replacement for standard processes. There is no certification basis, although we would like to get there. This is an aid to reduce the time required to do damage inspections.”
Most of Donecle’s experience so far has been in inspecting widebody aircraft. “We divide the aircraft into four sections and conduct four flights for each quadrant,” says Bequet. This allows maintenance technicians to work on other areas of the aircraft while drone inspections are underway. The drone is mostly used in C checks, “but it is starting to be used in light A checks,” he says.
Most inspections will be performed in the hangar, avoiding regulatory restrictions on flying drones in airport airspace. But the goal is to conduct outside inspections. “We have done a lot of flight tests outdoors, but regulation is harder,” says Bequet. Rain or condensation on the aircraft can make it hard to see defects, and wind is a concern. The drone tries to maintain a safe 1-m distance from the aircraft. If winds are too strong, there is a risk it will hit the aircraft or have to fly farther away, reducing image resolution.
“Once we prove we will not put a drone through an engine, we will come out of the hangar to the engine running bay, which is a bit isolated, then move closer to where aircraft are operating,” Goudie says. “Ultimately, we want a system we can use at the gate to constantly survey aircraft as they come in.”
For Blue Bear, collision avoidance to prevent the drone hitting the aircraft has two parts. First, both the aircraft and the drone’s flightpath are modeled accurately, and the imagery collected is referenced to datums of the aircraft model. The path planner keeps the drone at a distance from the aircraft that is safe and provides the best data. “It flies at a known safe distance,” says Goudie.
Second, optical and ultrasonic sensors onboard the drone can detect people or equipment not in the model. “People can wander into the safe zone, so we have both an ideal model and dynamic collision avoidance,” he says. On detecting an obstruction, the drone will stop. “It will then attempt to carry on, and it if can’t, it will recover to a landing,” he says.
High-definition imaging can detect paint chips and other visible damage. Analyzing dents is harder and requires 3D scanning, either optical or lidar. “The technology is slightly on the large side at the moment, and there is a weight limit Airbus and Boeing are comfortable with dropping on their aircraft,” says Goudie. “For now there is no technology we can plug in and do dent detection, but by next year it will be onboard.”
The MRO Drone team is looking at different business models for aircraft inspections. “Some operators want to purchase systems outright. Some will tell their MRO they want the system, and the MROs are very interested. Others would like it as a managed service,” he says. “It depends on operator requirements, and the system is designed to be as flexible as possible.”
Donecle’s model is to lease the system and not to operate the drones itself. Initial operators are airlines with internal MRO divisions, Bequet says, mostly European, with the first U.S. customer expected to be announced in a month or two.