Problem

Unscheduled downtime is frustrating for airlines, and means delays for passengers and extra costs. An aircraft struck by lightning must remain grounded until it has been inspected and found safe. An additional problem is that finding technically skilled people for aircraft maintenance, among other things, is becoming increasingly difficult. Therefore, the question is: can a drone do these lightning strike inspections faster and not at the expense of our safety?

Solution

Customdrone has proven in a case at Regional Jet center that a drone in combination with AI can not only perform such an inspection faster, but also better than a human inspector. Using this technology an inspector can therefore inspect more aircraft faster.

Approach

Customdrone has worked in a short period of time to deliver a fully operational solution for inspecting aircraft for lightning strikes. We do this in a number of phases.  

PHASE 1 Proof of concept

This phase is aimed at putting together the right hardware and software for the specific client and situation. By equipping different drones with different cameras and sensors, it becomes clear what the best combination is for each case and client. This depends on wishes and requirements, budget, system requirements and the environment in which to fly.  

PHASE 2 Data collection

The next step is data collection. That means camera-equipped drone flying and drone image collection. By working together with the inspectors and data scientists to determine what the drone should see, search and capture, an image is created of what the drone should do. In this phase, the drone flies, but the inspector does the inspection. In this way we learn, just as with a new employee, what the drone must do and what it can do.

PHASE 3 Inspection and recognition

In this phase, the drone independently selects the drone images on which a deviation has been detected. The inspector no longer has to look at all the drone images, but only at these selected images.  

PHASE 4 Autonomous flying

The final step in fully carrying out inspections. In this phase, the drone also learns to fly itself in order to inspect the entire aircraft.

After each phase there is a go/no go moment, at which point it is determined whether and how the next phase will start.

Before we proceed to roll out, we work out the business case together with the client. In doing so, we also highlight, for example, operational and cultural consequences. How do we ensure that an inspector learns to rely on a drone? What does this mean for the organisation? This is very important for the success of this transition: a transition that does not end with drone inspections, but begins there.

A good aircraft inspection drone is, among other things, equipped with artificial intelligence. This makes these drones self-learning. This allows the inspector to teach the drone to recognise and report certain damage. After every flight, the drone becomes smarter and smarter, which means it is able to recognise the following points more and more quickly:

• Lightning damage
• Marking damages
• Paint damage
• Crushed stone
• Dents and dents

In practice, drones found damage that the naked eye had not observed so quickly.

Once the drone has learned this, further steps can be taken. For example, the drone can make predictions after recognizing a number of damages. The drone can also connect to various software/ SCADA systems, including maintenance systems in the aircraft industry, such as AMOS, so that the registration of damage is processed immediately.

Due to the combination of drones with AIkan, an aircraft can be inspected up to 4 hours faster than with the human eye. That's no future, but proven in practice at Regional Jet center.

This ensures that ground time per aircraft is reduced and thus more hours of operational availability. This means more flights per year, because the aircraft are on the ground for less time.