Inspector Gadget

The hangar is busy but among the buzz of tools and engineers there’s a different sound.  A drone appears, steadily working around a fuselage.  Nobody pays any notice, their industry continuing unabated.  There is no pilot visible but, as the drone lands, survey data is already being scrutinised.  It could be a vision from an airport of the future but it’s December 2017 at Stansted.  The aircraft is A319 G-EZFH, recently returned from Reykjavik and scheduled for major winter servicing.  This is a perfect opportunity for the latest trial of the drone, a heavily modified DJI Matrice.

 

Pressurised environment

At summer peak, easyJet will operate nearly 2,000 flights daily, 284 aircraft flying from 0600-0200.  All routine maintenance is performed in a brief, overnight window with bigger tasks scheduled over winter.  Maintenance is also unscheduled.  Aircraft accumulate occasional dents and scrapes around cabin doors and cargo hatches, rubbed daily by airbridges, steps and loading equipment.  Any birdstrikes, hailstone damage or lightning strikes (in the lightning strike case alone, an aircraft is typically struck twice yearly) can cause visible dents and burns.  All need assessment and logging for treatment, according to severity and location, before the aircraft can be returned to service.  Much is cosmetic but nothing is left to chance.  The fallout can be extensive – and expensive – for all parties, particularly if occurring at remote destinations without engineering cover.  This is the job of the Maintenance Operations Centre (MOC) working around both clock and continent to co-ordinate staffing, tooling and minimise aircraft downtime.  They work to a key performance indicator (KPI) of downtime hours prior to ‘wheels up’.

 

Before every flight or service, pilots and engineers will walk a circuit of the aircraft looking for external damage: what the latter call a General Visual Inspection (GVI).  A GVI of the lower fuselage can be completed quickly on foot, but the upper is very different.  This requires a crew roped into a cherry picker, inching along the fuselage.  Could the aircraft be surveyed more quicker?  How to quantify and prioritise unexpected damage over such a large fleet? How to ensure the safety of your staff in wintry conditions atop a 12m tail?  What if they cause additional, accidental damage? It was these questions that prompted easyJet and Blue Bear Systems Research, a world leader in Unmanned Air Systems research and development, to collaborate.  They believe that instead of eight hours currently, a GVI by drone could be achieved inside two.  ‘The aim is quick, reliable and consistent damage mapping to accelerate AOG recovery’ explains easyJet Avionics Fleet Technical Management Engineer and drone enthusiast Rupert Harvey-Lee.  Gavin Goudie, Blue Bear Operations Director adds: “Our intent has always been threefold – automation, simplifi cation and no additional training burden to the client.”

 

Genesis
News first broke of easyJet’s intention to use drones for automated inspection in 2014, followed by a demonstration of proof of concept in 2015.  RAPID (Remote Automated Plane Inspection and Dissemination) originated from a previous Blue Bear collaboration, the Remote Intelligent Survey Equipment for Radiation (RISER) deployed inside contaminated sites, such as Fukushima.  Meanwhile, a joint venture between Blue Bear and Output42, an Irish specialist in software solutions for aircraft engineering, led to the foundation of MRO Drone Ltd.  Goudie explains: ‘The vision of MRO Drone is to produce innovative, integrated hardware and software tools to allow airlines and MROs to conduct safer engineering inspections faster and more efficiently.’  Progress has accelerated over the past two winters.  “The pressure is high over summer to even find an aircraft; the best possibilities always occur over winter,” explains Harvey-Lee.

 

Navigational priority

RAPID needs to orient itself in relation to the aircraft and autonomously fly circuits above fuselage, wings and tail at a standoff distance of between 1-3m (3m from wingtips due to regulations on the fuel tank vents).  This will compliment an equivalent by eye from below: the threshold between the two is the cabin floor level.  “We call this the waterline,” says Harvey-Lee “…anything above this is what the drone is interested in.”

 

Real Time Kinematic (RTK) GPS would be the ideal candidate, but can’t be guaranteed inside a hangar.  Initial work found that LIDAR was susceptible to the busy hangar environment.  Ultrasonic positioning also fell short of the precision required, crucial given aircraft proximity and value.  These initial trials were carried out without an aircraft, a virtual model in the software only.  Ultra-wideband (UWB), as used on automated, rolling production lines, was eventually chosen.  Transmitting above 5Ghz it offers a high data transfer rate over short distances.  In reality, that’s a precision of 10-15cm; ideal at a distance of 1-3m from the aircraft when combined with precise aircraft reference models.  A network of UWB masts surrounds the aircraft, currently a setup that is completed in just over an hour.  A fixed installation in the hangar roof would be ideal and would be implemented where required but the final system also has to be deployable anywhere in Europe.

 

RAPID must be immune to the potential interference of personnel, cables, vehicles and tooling.  Says Harvey-Lee: “This is what we’d call a challenging environment but it’s reality.”  Goudie reinforces: “It’s vital for us to test in this environment, you couldn’t buy this sort of access and exposure to the real engineering operation.  A perfect example: in order to test landing gear, G-EZFH rests upon jacks, sitting 50cm higher than otherwise and at a nose-low angle – not as the model would normally expect.  This won’t hold up the trial but is a real-world factor that has to be considered during initial setup.  A digital model is now aligned to the aircraft by marking wingtips and nose using a positioning sensor: the system knows aircraft dimensions and position.  This flexibility is critical to accommodate A319/A320, with the A321 arriving this year.  RAPID transmits position to a nearby control station, allowing it to autonomously fly preset paths. The team likens this to Pacman eating a string of waypoints.

 

On-the-fly

RAPID is run via a ground control system (GCS – the work of MRO Drone) on a single laptop.  Development flights are recorded - invaluable when paired with onboard logged data to compare demanded inputs with control outputs.  The navigational finesse is most evident when RAPID passes metal stanchions in the hangar wall half a metre away.  Viewed from a nearby, elevated gallery, itself a mass of signal-disrupting metal, it’s rock solid.  A circuit of either wing, tail or crown skin (the top of the fuselage) takes less than ten minutes, high-resolution imagery assessed on the fly via the GCS.  Black and white gives best contrast for fi ne fuselage details; stacked and sharpened by the GCS to throw features into sharper relief.  Goudie states: “The drone is interesting, the tech is interesting but the usage of the data is what really adds value.  Anyone could fly a drone around an aircraft and take pictures; it’s what happens next that really matters.”

 

RAPID

The DJI Matrice is a powerful, flexible platform for enterprise-level development, offering engineers a chance to get under the bonnet.  A second battery can be fitted, doubling endurance or powering ‘smarts’: avionics, control algorithms and payloads.  Hardware can be slid around on the chassis to optimise balance.  A software development kit (SDK) permits quick integration with the programming of an application such as RAPID.  Users can program commands to modify flight, raw operating data fed back for assessment.  The RAPID architecture has been designed to be plug-and-play, not platform specific; a key requirement is easy migration onto other platforms.  This enables the developer to focus on the ultimate aim: high accuracy, position-referenced inspection images that can then be shared within MRO or airline engineering systems.

 

Crucially, the entire survey flight from take-off to landing is made automatically. Engineer Luke Cowan explains how he acts as safety pilot for testing and has been involved from the start to develop RAPID to become pilotless.  The engineers using it will not have to become drone pilots, it purely gathers the evidence they need.  “RAPID doesn’t need a pilot and will ultimately just have a single operator controlling it from the GCS,” explains Goudie.  Circuit complete, RAPID climbs above the height of the tail to return overhead the launch point.  An optical/ultrasound collision avoidance functions on four sides and downwards.  “Only minor interventions from the pilot are required, typically for issues with the collision avoidance system, often down to ambient hangar lighting,” says Harvey-Lee.  ‘There’s no safety risk, the drone will pause when unhappy and wait for a nudge, which can then be resolved by analysis and re-coding of the control algorithms.’  Several times flights are paused as the software is interrogated and recoded in real time.  As with any flight there’s much pre- and post-flight activity.  Goudie comments: “During our development process it takes a few days to effectively step back through the dataset and establish what we’ve learnt so that the user can get results in a matter of hours.”  In this way the team are close to releasing an initial product, a huge ratio of development hours per flight hour.

 

Future systems and airports

The ultimate system would be completely autonomous and portable, run at the touch of a button and use 3D mapping for navigation and damage measurement – not just observation.  Data can already be instantaneously sent to various stakeholders (engineering headquarters, manufacturer and outstations worldwide).  When all parties have the same evidence, repair decisions can be made quickly and consistently.  The final system would ideally integrate images onto an electronic technical log, perhaps even using the aircraft WiFi, part of the steady drive towards paperless operation.  Alongside us are aircraft being prepared for end-of-lease and onward sale to other airlines; an integrated system would facilitate transferring a detailed service history onto subsequent owners.  The released product would be tested at the main UK maintenance centres, then rolled out to other key stations across Europe.  “Ultimately this may well become a service that is scalable worldwide,” says easyJet Head of Engineering Gary Smith: “We are pushing it forward.”

 

A handheld version might also be considered for the lower fuselage where flight of the drone is extremely challenging.  Goudie notes: “There would be potential in a deployable system for less known and uncontrolled environments.”  However, he’s realistic: “Portability is a key design requirement but we have to match our intent to the capability we’re adding.  We’re not going to comply with our weight limitations (the drone is bound by a 3.5kg mass) if we want to start 3D scanning with the currently available technology.”  Considering two recent industry developments, drones might become fundamental to future airport operations.  Gatwick Airport has expressed interest in implementing drone runway inspections.  In 2019, London City will have the UK’s first digital air traffic control tower.  Harvey-Lee is optimistic about the potential here: “Airports might be the ideal location for legitimate drones as everyone is aware of the degree of control and regulation required.”  Goudie adds: “The regulator and airport operators like the work contained in a closed hangar but there is no reason with appropriate permissions, why drone surveys cannot be carried out on the ramp or at the gate.”  RAPID could also have use on any large, mobile asset, such as inspecting the sides of a cargo ship.

 

Getting airborne again

With navigation nearly perfected, progress now moves towards two final phases: damage detection and payload integration.  This means perfecting the visual acuity RAPID can resolve in a passing sweep and combine as a usable output; and how to miniaturise the payload and streamline the necessary algorithms.  Goudie summarises: “We detect industry opinion has gone from ‘cool idea, how will you do it?’ to ‘cool idea, when will you do it?’ and drones are at the forefront of this.”