13 December 2023
Artificial Intelligence and Aviation
IBM defines Artificial Intelligence (AI) it its simplest form as is a field, which combines computer science and robust datasets, to enable problem-solving. McKinsey (2023) state that artificial intelligence is a machine’s ability to perform the cognitive functions we usually associate with human minds. AI technology with complex algorithms can process large amounts of data and recognise patterns, make decisions, and make judgements from the analysed datasets. In essence humans are building machines to judge and make decisions like humans.
AI is evolving quickly and is in the process of being adopted in the aviation domain. It is already in use with some large carriers in various aspects of their operation to manage and analyse their operational data supporting management risk-based decision making. While the concept of AI has been around for decades, its accelerated development in recent years has been due principally to advances in technologies around data storage and management, computing power and the design, development and learning capability of AI algorithms. AI is in its 4th generation of evolution where the capability extends to understanding context; abstracting knowledge and learning without the need for large amounts of training data. The 5th generation capability extends to parallel processor hardware and AI applications being used in the development of computers.
AI algorithms are at the heart of emerging autonomy technologies and application of this technology that can enhance existing airline data management processes and also open new possibilities across the aviation value chain.
The problem is that AI can be perceived as a ‘black box’ where operational data is analysed as a result of complex algorithms and system learning that is not clearly visible or understandable to the end-users and is less likely to be accepted without challenge.
Large carriers amass volumes of data every day both digital data and hardcopy (Market data, Revenue Management, Flight Data Monitoring, Aviation Safety Reports, Audit reports, Customer Surveys etc..) and the effective analysis of this data offers new business opportunities and can represent a competitive edge. Also, the continual improvement and evolution of algorithms and AI to collate, manage and exploit operational data increases in capability as analysis tools and technologies advance.
The Fedral Aviation Administration (FAA) is utilising data-driven safety approaches as part of the U.S. State Safety Program and FAA Safety Management Systems (SMS). With data analysis, the FAA seeks to identify causes of accidents and fatalities, mitigate risks, and encourage positive and progressive infrastructure and behavioural changes.
As aviation applications evolve using digital data improvements in cybersecurity will be paramount to secure methods of data transmission, collection and analysis.
The following illustration from IATA demonstrates the mapping of different AI capabilities against the business capabilities and needs of airlines in 3 clusters (Operational, Support & Management and Customer touch-points).
Artificial Intelligence and Aviation Safety
AI has the potential to streamline processes, operate 24/7 without human limitations, and significantly reduce the human error component in aviation safety. AI can continually monitor aircraft systems and crew performance for anomalies and provide immediate communication to key stakeholders to ensure safe operations. , AI, working alongside humans can augment and increase safe performance. IATA (2018) outlines some of the potential future characteristics, capabilities and use cases of AI systems in the aviation context include:
- 24/7 AI Services: AI never gets tired and has a multi-layer parallel redundant architecture, i.e. unlikely to fail and if failure does occur there are many backups.
- 100% up-to-date: AI systems can be loaded with complete competence packages for all the systems, tools and equipment used in an airline; with updates as and when there are changes.
- More comfortable flight experience: While IATA and airlines are already working on a real-time turbulence database, AI could enhance this, for example with a real-time multidimensional pressure data feed of the airspace within a radius of certain value, which can be used for optimisation purposes as well as avoidance of turbulence.
- Safety & Real-time Monitoring: AI could make it possible to have an Aircraft Real-time Health Monitoring System (ARHMS), driven by sensors at the atomic level on every part of the aircraft, its content, and surroundings. Any anomalies (e.g. stress, pressure, magnetic, temperature, humidity fluctuations) can be measured and acted upon. Upon landing, AI could conduct a physical safety check, inspecting the entire aircraft through the ARHMS and using an autonomous multi-drone system.
AI has the potential to streamline processes, operate 24/7 without human limitations, and significantly reduce the human error component in aviation safety.
An aviation SMS, by design, is required to be systematic, explicit and predictive. Due to the significant volume of disparate data, the challenge lies in detecting and understanding unknown vulnerabilities and anomalies, as compared to current application of safety performance metrics (leading and lagging). Predictive SMS capability will require the development of predictive safety models based on the analysis of system performance data, including data from different aviation stakeholders. Data science and machine learning approaches can generate new knowledge on how to help aviation stakeholders further improve safety. AI and machine learning techniques need to be developed into complex, data-driven models to predict actions, recognise patterns, and uncover hidden insights. AI techniques and algorithms can be designed, adapted and validated against safety issues derived from operator Safety & Compliance database event categories, Video feed data (Ramp and apron) and Flight Data Monitoring data for example.
Airline safety databases are structured around a number of isolated data silos with limited interaction due to their organisation, legal and/or technical issues. During the last 30 years however information sharing and collaboration between airlines, regulators, manufacturers and aviation operators (International Air Transport Association, Flight Safety Foundation, International Civil Aviation Organisation, Helicopter Association International, HeliOffshore, Manufacturers) has led to significant improvements in aviation safety.
McKinsey, in their article ‘What is AI’ (2023), state that AI is a machine’s ability to perform the cognitive functions we associate with human minds, such as perceiving, reasoning, learning, interacting with an environment, problem solving, and even exercising creativity. We can expect going forward that AI augmentation in aviation is here to stay and a future AI capability being able to deal with real-time data management that will enable real time risk management.
AI can be used directly to support airline safety Management through:
- Support structuring (data fusion) and classification (text analysis and auto classification) of safety events and data,
- Detection of emerging atypical risk patterns in the data through identifying correlations and anomalies;
- Supporting the development of Risk Mitigation plans through data driven decision making by management teams.
McKinsey, in their article ‘What is AI’ (2023), state that AI is a machine’s ability to perform the cognitive functions we associate with human minds, such as perceiving, reasoning, learning, interacting with an environment, problem solving, and even exercising creativity. We can expect going forward that AI augmentation in aviation is here to stay and a future AI capability being able to deal with real-time data management that will enable real time risk management.
Let's talk
Eduardo Dueri
Senior Partner, Safety & Operations Aviation Risk Solutions
Gallagher Specialty
Arthur J. Gallagher (UK) Limited is authorised and regulated by the Financial Conduct Authority. Registered Office: The Walbrook Building, 25 Walbrook, London EC4N 8AW. Registered in England and Wales. Company Number: 119013.