Deep Turnaround: Artificial Intelligence is revolutionising turnaround management

posted on 20th August 2019
Deep Turnaround: Artificial Intelligence is revolutionising turnaround management

Artificial Intelligence and other new technology are changing turnaround management, writes Michael Muzik, Senior Product Manager and Consultant at Lufthansa Systems

Air traffic is growing at an average annual rate of more than three per cent. A total of 46,460 aircraft are forecast to be delivered between 2018-2037 – almost twice as many as today.

Passenger numbers will also double. With more congested air and ground spaces, increased delay rates and flight cancellations will result, leading to negative passenger sentiments and higher costs for airlines.

Punctuality and profitability of airlines, airports and ground handlers depend largely on turnaround management efficiency. They all need to find new strategies to cope with these challenges. Digitalization, artificial intelligence (AI) and other new technology are the answer for it. They have the potential to improve ground operations. The use of AI and other new technologies within aircraft turnarounds are discussed therefore here.

Turnaround Management as core business process

A turnaround is the time the aircraft spends between parking on the stand and starting pushback for the next departure. During this period, a large number of service providers (cleaning, catering, etc.) execute multiple cross-functional processes.

Punctual turnarounds have become therefore a critical factor for success for the airline’s on-time performance, they remain a relatively uncharted territory for many airlines – even there is great potential for improvement: The airport may use the gates more efficiently. The service provider benefits from better utilization of personnel and incentives. Finally, airlines can increase their on-time performance by 3-9% and reduce costs due to 35% less ground handling related delays.

Turnaround Management today

An optimized turnaround is teamwork at its best, comparable to a pit stop: a seamless interplay between airline, airport, and ground handlers. Turnaround Management supports this complex interplay, as the time window for a turnaround is short – especially for short-haul flights sometimes down to 25 minutes.

Most software solutions permit users to real-time monitor turnarounds. Alerting mechanisms will inform in case of process delays allowing the Turnaround Manager to manage these delays actively. To set-up up such real-time alerts, two steps are necessary:

  1. Set up of reference models with pre-defined target times
  2. Real-time monitoring of the turnaround and steering in case of irregularities

The prerequisite for a real-time turnaround monitoring are actual time stamps for each turnaround process to match them against the pre-defined target times. These actual time stamps can originate from various sources such as airline, aircraft, or airport systems – and / or service providers. Some of them are fed automatically, for instance, if the cargo door opens, (= indicating de-loading starts).

However, in most of the cases a person must insert the actual time stamps for servicing manually, indicating for example: catering is (1) at position, (2) has started, and (3) has finished.

Precisely this manual gathering of the time stamps is the biggest obstacle for implementing Turnaround Management most of the time. AI is the perfect solution for this problem.

Deep Learning in Turnaround Management: Deep Turnaround

With AI technology, the manual data collection can be fully automated. The term AI describes machines that mimic “cognitive” functions that humans associate with the human mind, such as “learning” and “problem solving”. AI is therefore the basis of the innovative Turnaround Management capability from Lufthansa Systems and its subsidiary zeroG called “Deep Turnaround”. It helps identify unusual circumstances both outside and inside the aircraft during turnaround without manual help.

In practice, the “Deep Turnaround” approach is quite easy: this AI solution is based on the data that is very often already available at many airlines, but unfortunately still not used. It means the use of the cameras installed at the airport (gates). These cameras are equipped with deep-learning video classification algorithms that bring the necessary automation into the manual collection of turnaround data leveraging airlines, airports and ground operations providers from this sort of work. Concretely they transform unstructured data of the apron video stream into structured actionable data and insights.

The term “Deep Turnaround” specifically refers to “deep learning”, as subset of AI. It means, that the systems’ algorithm recognizes objects (aircraft, fuel or catering truck “at position”) and actions (passengers de-boarding, start and stop of fueling or (de-)loading) received from the video stream and automatically sets the necessary timestamps, which formerly had to be inserted manually. Additionally, “Deep Turnaround” has the advantage that all the above can be done by leveraging airports´ and airlines´ already existing CCTV infrastructure.

Machine Learning: Predictive Turnaround Management

Using AI technology in turnaround management will bring airlines even one step beyond: the switch from descriptive into predictive real-time monitoring by “machine learning”. Besides deep learning, machine learning is the second subset of AI. It is a term that describes algorithms and statistical models – based on a vast of (historical data) – used in order to perform specific tasks effectively without using explicit instructions, relying on patterns and inference instead[1].

The benefits are clear: first, machine learning can predict expected delays according to the actual operational situation, because the algorithms improve as they are exposed to more data over time. Just imagine how beneficial information, such as “remaining fueling time: 11 min” or “remaining loading time: 5 min for 85 bags” will be for the turnaround coordinator.

Second, such predictions can also be used not only for the day of operations, but also for mid- to long-term adjustments of the reference models and their target times. By using such a vast of data existent, the AI algorithm can analyse this (historical) data of the turnaround and use the findings to adjust turnaround times that are planned too tight or with too much buffers in a systematical way. With machine learning, the turnaround quasi optimizes itself: it might for instance be that over a significant period of time the cleaning for an A330 in reality takes only 10 minutes to clean, as opposed to the originally planned 12 minutes

With this finding, the turnaround time can be minimized to the optimum. Instead of programming fixed or parameterized business processes, the algorithms of the future will be derived from data. AI-based technology is paving the way for this digital future.

Other new technology for ground operations

When speaking about the digitalization of airline ground services to improve efficiency, it is important to accomplish the picture with other new technologies already available that support this business goal as well. Here is a short outline:

GPS location: For the special load completion check the GPS location data from special load is the key. If the respective container is not at the aircraft, but the turnaround coordinator can retrieve the exact location on the mobile device, time-consuming calls with the cargo department will be history. The same principle applies for all ground handling equipment, for instance staircases, loading belts, etc. The automated feed of GPS data from ground support equipment on mobile airport maps will prevent delays, because nowadays the ground staff is not always aware of the exact location for the equipment needed. Such GPS location technology will also help to better monitor in real-time the actual location of moving crew and passenger buses. Predictions about estimated arrival times of passenger or crew buses based on AI will complement this information to perfection.

Autonomous vehicles (AV): AV will change the ramp environment — in particular, the adoption of fully automated vehicles. In the future vehicles can manage most aspects of driving (under human surveillance) or even operate without human input with high automation. This will have an impact on promptly stairways or jet-ways for instance and prevent delays because of missing or incorrect planned human resources on the ground.

Virtual & Augmented Reality (VR/AR): The use of VR/AR with special glasses will reduce turnaround delays, for instance due to packaging errors. With AR glasses for instance the turnaround coordinator can instantly check, whether a pallet or cargo load item is tied up correctly or not. Moreover, the pallet or ULD might also indicate automatically to the loading personnel, on which position it should be loaded and even confirm by itself, it was loaded correctly.

The “future” is already available – invest now!

To meet the challenges of the fast-growing air traffic, the airline industry must start to switch to solutions based on digitalization, AI and other new technologies – especially in ground operations. AI solutions are based on the data that very often is already available at many airlines and airports. The strategic aim should therefore be to make effective use of this golden treasure.

The same applies for GPS, autonomous vehicles and AR/VR reality: these technologies are already available on the market. For some reason – maybe because the ramp is still perceived very much as a “hands-on world” – it still hasn´t found its way there.

Therefore, airlines should start to consider their ground operations as being a substantial part of their digitalization strategies today. It must become a top business objective in this hyper-competitive market as it provides a high degree of automation, innovation and cross-departmental integration. All this adds up to an increased on-time performance, less delay costs and – most important – an improved passenger travel experience.