by   |    |  Estimated reading time: 8 minutes  |  in Aerospace & Defense, AI, Aviation   |  tagged , , , ,

Rob Mather, Vice President, Aerospace and Defense Industries at IFS

Since the release of ChatGPT, it seems like all anyone can talk about is Artificial Intelligence (AI). However, beyond the public domain, AI has the potential to offer vast opportunities for increased efficiency and productivity in various industries.

How, though, can we apply AI in an industry like aviation which is, for good reason, extremely cautious and safety-conscious, and also requires complete visibility and traceability, when these requirements are not typically aligned with our understanding of AI? Well, AI constitutes way more than just ChatGPT-style generative AI and Large Language Models (LLM). Incorporating AI into aviation presents a unique opportunity to enhance safety, efficiency, and accuracy by employing a wide range of tools and technologies.

Two possible paths for the future use of AI in aviation are only using AI to surface options and enable human decisions, or through explainable AI.


Information surfacing for decision support:

A foundational principle in aviation is the idea of a trained and certified individual who takes responsibility for decisions, under appropriate authority and licensing. AI can be used to improve and assist rapid decision-making by providing relevant information and options quickly while leaving the ultimate decision with the authorized person. By using AI as a supporting tool, people can make decisions faster, and more accurately without compromising control.

Explainable AI must be responsible AI:

Another path to using AI in aviation lies in the concept of explainable AI. Explainable AI comprises tools and frameworks that help humans to understand an AI model’s predictions. When deploying a model, these tools help improve model accuracy. When looking at the outcome of a model, they help to understand how the model arrived at a particular conclusion or outcome.

Many machine learning-derived algorithms are regarded as “black boxes” that magically produce the correct outcomes.  This can be perceived as a problem in an industry like aviation where actions must comply with regulatory requirements, and the reasons those actions were taken needs to be understood by the regulator. Claiming “the AI decided because of magic” won’t be an acceptable explanation. Ensuring AI is explainable is critical to its widespread adoption within the aviation industry.

Making the use of AI acceptable in the industry is one thing, but actually achieving real-world benefits is a whole other issue.  Four practical ways AI can offer real value in aviation maintenance come immediately to mind:

Maintenance Scheduling & Supply Chain Optimization:

There are many different types of AI models that can perform optimizations but some of the most common involve iterative approaches, running hundreds of thousands of scenarios in an instant and regressing the results to an optimal outcome. AI-powered optimization engines aren’t as prominent in the zeitgeist as Generative AI, but there are a multitude of applications for optimization in aviation. Since we’re talking aviation maintenance, maintenance scheduling optimization is the obvious standout.

Giving up maintenance yield means performing more maintenance over time – which means additional costs – but also, having the aircraft out of service for longer means lost revenue. An optimization engine that can schedule maintenance at the best possible time at the best possible location has the potential to greatly reduce maintenance costs and improve maintenance yield fleet wide.

At the same time, to ensure cost-effective and timely completion of a given maintenance visit, optimizing the order in which tasks are performed and how they are assigned numerous benefits. Doing so can significantly enhance the efficiency of the process, reduce costs, and improve turnaround time (TAT). In turn, this will help get the aircraft back in the air faster, leading to increased revenue generation.

Furthermore, by applying optimization to the maintenance supply chain, you can minimize material delays by ensuring that the right parts are always where you need them when you need them to get your aircraft back in the air as quickly as possible.

Error Detection & Reclassification:

AI can also assist in identifying data entry errors and reclassifying data to improve accuracy and the overall quality of datasets.

The misclassification of a fault’s failed ATA system is a common issue in the airline industry. Sometimes this is the result of human error, but often the fault first gets classified at the time it is raised, based purely on what amounts to symptoms. Then, when the fault is eventually resolved, it is found that the originally identified system was not the actual culprit – that another was the actual root cause. These misclassifications can have a significant impact on data quality. Airlines may assign their technical records team or reliability engineers to review records, identify errors, and perform reclassifications to address the issue. However, this process can be time-consuming, costly, and requires painstaking attention to detail.

IFS customer Southwest Airlines, has recently launched an AI-based solution that can detect misclassified faults by leveraging an aviation-specific language model to identify patterns in text. This approach allows for improved data quality by detecting and presenting potential errors to a reliability engineer, who still has the final decision-making authority, resulting in a more efficient process that maintains human oversight.

Automated Failure, Troubleshooting, and Repair Identification:

When working a fault, a technician often needs to spend significant time examining the fault-isolation manual, researching the correct origin of the fault, and determining the appropriate troubleshooting steps and repairs.

On top of fault classification, an aviation-specific language model could be used, in real-time, when a fault is being raised, to identify potential sources of the failure, suggest troubleshooting activities, and propose repairs. By presenting these options to the technician, the model would pare down the noise and allowing them to use their time more efficiently. By including the past success rate for each option, the model would enable the technician to select options that save time and resolve the issue more quickly. Resolving the fault on the first try can help the aircraft to get back in the air sooner and even prevent recurrences in the future. Avoiding or reducing delays has real value to airlines. According to Airlines for America, in 2023, delays have a direct cost of $101.18 for every minute a flight is delayed.

Predictive Maintenance & Anomaly Detection

The concept of predictive maintenance is nothing new. However, what is new is the application of newer types of AI, namely Anomaly Detection and Pattern Recognition, which is making predictive maintenance much more accessible.

Early predictive maintenance models could only look backwards at historical data. The introduction of IoT and live data feeds from sensors made it possible to include current, up-to-the-minute data. However, the sheer amount of data being generated required highly trained data scientists. With the advent of Machine Learning, data scientists could focus on creating the learning model for the AI, rather than processing the data themselves.

Unsupervised learning models are lowering the barrier to entry even further for the use of AI in predictive maintenance applications. Although to a lay person, the concept of “unsupervised” learning models for AI might seem scary, it actually means is that you can plug the AI into a set of data, and it can essentially figure out its own algorithm. Not only does this reduce the time and cost of implementing a solution, but it also has the power to strip out bias from the process, particularly when dealing with large amounts of unlabeled data like the flood of sensor data generated from a modern aircraft or jet engine—management consulting firm Oliver Wyman expects the newest generation aircraft will be generating between five and eight terabytes of data per flight by 2026.

Anomaly detection enables the integration of AI with your sensor data. By identifying what is considered “normal” it can alert you when deviations from that normal state occur. This serves as an early warning system. When combined with pattern recognition, AI can learn to detect patterns in the sensor data that indicate certain events are likely to occur. This results in an accurate and reliable early warning system that can predict upcoming events. and Falkonry AI’s anomaly detection capabilities have the ability to democratize AI, meaning that vast amounts of data generated over long periods of time can be made easy to access and interpret, finally making truly predictive maintenance available to all.

The future is AI and the future is now

AI undoubtedly needs to be deployed with great care – nowhere more so than in the realm of aviation. The aviation industry has the ability to make significant strides in terms of efficiency and accuracy by leveraging AI to streamline human jobs and provide decision support. This not only helps reduce the noise of information, but also ensures that the human is still in the loop and remains the ultimate decision-maker. By combining the best of both worlds, AI and human expertise, organizations can achieve optimal outcomes and gain a competitive advantage in their industry. These improvements can represent real value to airlines & air operators and, ultimately, their customers. AI is here and it’s here to stay, and early adopters have the chance to get an edge unlike any other.

Leave a Reply

Your email address will not be published. Required fields are marked *