Blog

Read our articles to learn more about data visualisation and its applications

Digital Twin for Aircraft Maintenance

2021-10-04

Predictive maintenance is a hot topic. The overarching goal is to predict when an asset will fail by collecting thousands of measurements from sensors. Interpreting this sensor data is not always easy. Together with EPCOR, we created a digital twin that visualises relevant sensor data on top of a virtual 3D model. This made it possible to explain the reasoning behind predictive maintenance decisions to non-technical stakeholders such as management, clients and domain experts.

The world of the APU

This project started when we were approached by Methylium, who were making a predictive maintenance system for a company called EPCOR. Being part of KLM Engineering and Maintenance, EPCOR repairs so-called APUs. Since we had limited understanding of what an APU is and how it works, EPCOR engineers gave us a tour of their workshop.

An APU is a complex machine that provides an airplane with power and compressed air for the airconditioning while the aircraft is at the gate. As with any modern piece of equipment, it has countless of sensors that read internal temperatures, oil pressure, rotations per minute, etc.

Motivation

After each flight, an APU sends its sensor data to EPCOR's predictive maintenance system. EPCOR engineers are automatically alerted if there are any anomalies. The predictive maintenance system shows the data in around 20-30 graphs so engineers can find the root cause of a potential issue. Based on their years-long experience, engineers can also use these graphs to predict roughly when an APU will fail.

The ideal predictive maintenance scenario is to service an APU just before it breaks down. When an APU breaks down completely, internal components are often damaged and need to be replaced. These new components can cost hundreds of thousands of dollars per APU. If the APU hadn't broken down completely, the existing components could have been deep-cleaned and reused. Predicting when an APU will fail is therefore crucial in reducing maintenance costs. EPCOR's motto: repair before failure.

The Challenge

If the predictive maintenance systems indicate that an APU is likely to fail in the next months, this need to be communicated to the aircraft planners at the corresponding airline so that the APU can be taken out and serviced. The planners at airlines, however, are often not convinced by an engineer showing them various graphs of oil pressures and temperatures. Their main objective is to keep the aircraft flying for as long as possible. If an APU still works, then why take the aircraft out of service?

The challenge here is that engineers have years of hands-on experience with APUs and know them inside out. They can mentally visualise what happens when the oil pressure rises beyond a certain level, which parts are affected and how this influences the working of the entire APU. Aircraft planners don't have this experience and the dozens of graphs that an engineer will show them are meaningless to them without context.

The goal of this project is to bring back some of that context and make planners better understand how we arrive at certain conclusions. When everyone has the same reference point, it is much easier to make joint decisions.

Concepts and References

EPCOR owns a high-resolution 3D scanner that they use to scan each part that they service to make sure it has exactly the right dimensions. We have used these 3D scans to create a virtual version of an APU. Since the position of each sensor is know, we can accurately position the data readouts on the virtual version. This makes it easier to see relationships between sensor values.

The first step in such a project is to draw some sketches of what the final product might become. Some initial sketches can be seen below.

Based on these sketches, we've searched for existing visuals that roughly match the look and feel that we are going for. These visuals will serve as a rough guide when we're developing our own application and give the client an idea about what to expect.

Scope and Story

Every visualisation needs a good story. Instead of directly trying to make a generic application that could handle all APUs, we focused on a specific use case that had to do with a metric that EPCOR had devised as a predictive maintenance KPI. By following this story, we were able to keep the scope small while creating a product that could prove the principle of visualising data in its natural context. We created a large 3D version of an APU with buttons along the bottom to navigate the story. We brought the engine to live with animated particle effects to represent combustion and airflow as well as animating the moving parts of the engine.

Dubai Air Show Presentation

The resulting application has been used internally and externally at the Dubai Air Show to convince clients of the use of predictive maintenance. The visualisation gave people a better understanding of how an APU worked and made it easier for them to visualise the real-life impact of an abnormal sensor reading. The application has also been shown to Pieter Elbers, CEO of KLM.

categories

client

EPCOR, a wholly-owned subsidiary of KLM Engineering & Maintenance.

Deliverable

A 3D Digital Twin that visualises predictive maintenance data in an engaging and accessible way.