Data Science applications are at the heart of the digital transformation happening in the aviation industry. Examples of these applications can be found in manufacturing processes, air transport operations, crew training, and safety analysis, just to name a few. It seems everyone is convinced that the benefits of extracting knowledge from data is higher than the potential risk of opening their data to others.
The present perception of open data, or at least, shared data, has significantly changed from our 1st Data Science in Aviation Workshop (DSIAW) in 2013. Just 6 years ago we noted a higher reluctance of aviation stakeholders in sharing their data than the expectations of potential value data science could bring. This drastic (and rapid) change of perception is partly due to the rise in sophisticated data protection techniques based on cryptography that have demonstrated to be an efficient solution to ensure data confidentiality; along with with the growth of successful data science initiatives in complex socio-technical systems (including, but not limited to, aviation). In the 2018 edition of the DSIAW, we will continue to share the latest initiatives with presentations on predictive maintenance, fuel efficiency, delay propagation and safety intelligence, among others.
The launch of Aviation Data Exchange programmes, such as those developed by the FAA, EASA or IATA, have also opened the door to data sharing with a trusted third party in order to obtain aggregated statistical analysis of the sector performance and blind benchmarking applications among the airspace users with a focus on safety analysis:
- ASIAS is the FAA Aviation Safety Information Analysis and Sharing System. This programme was launched in 2007 with the objective of further improving the aviation safety through the open exchange of safety data and the ASIAS is the underlying system. Presently, ASIAS collects and combines close to 200 data sources across government and industry, some of them correspond to voluntarily provided safety data from their partners, including: Commercial air carriers, general aviation operators, manufacturers, maintenance & overhaul service companies, flight training entities and governmental bodies. ASIAS has established safety metrics to monitor the risks in the operations, identify trends, propose mitigation actions and validate the impact of its implementation. To ensure the protection of proprietary data, operations data are de-identified in an irreversible manner.
- Data4Safety is the European Safety Data Exchange programme launched by EASA in 2017. The programme includes an IT infrastructure that would enable the data sharing process, including data collection, storage, security and processing. The programme started with a proof of concept, and is developed by a group of stakeholders including airlines, aircraft manufacturers, national aviation authorities and pilot unions. In the implementation phase, the programme aims to improve safety through the application of big data and data mining techniques to large amounts of data being shared by the aviation community stakeholders.
- GADM is the IATA Global Aviation Data Management programme and platform. Over 90% of IATA member carriers have agreed to participate in this programme which now receives data from more than 470 organizations. The participation in the programme allows data providers to access aggregated and de-identified reports on safety metrics and trends, including analyses on accidents and incidents, operational reports and ground damage reports.
These programmes have fostered a paradigm shift in safety analysis, moving from a reactive approach based on the analysis of accidents and incidents to a proactive approach based on data-driven safety performance analysis.
The SafeClouds.eu project is fully aligned with this proactive approach, as it is based on the analysis of securely merged aviation data sources through a long-term lens, according to its research nature. The project is focusing its research efforts in going one step beyond aggregated statistical analysis, and into the application of machine learning techniques. Applying these techniques to aviation data will help understand the precursors that lead to a safety event, predicting hazards and ultimately propose mitigation or corrective actions to avoid a safety situation. Machine Learning techniques require large datasets to train and test the algorithms which, in turn, bring additional challenges in terms of data management and data protection. Simple de-identification techniques consisting of deleting confidential fields or replacing it with a placeholder are not sufficient for the use cases as detailed in “The need for smart(er) data protection mechanisms“. To overcome these challenges, it is imperative that a holistic approach is taken. SafeClouds.edu brings together data scientists and data engineers with airlines, ANSPs and safety agencies in order to overcome these challenges and contribute to the proactive analysis of safety risks.
