- Thanks to Artificial Intelligence, OSAC is improving its civil aviation airworthiness monitoring
Designed in tandem with the SixFoisSept team, a decision support tool is now making it possible to adopt a risk-based approach for effective civil aviation inspections.
France possesses one of the busiest airspaces in the world. From single engine two-seaters to the ultra-high capacity A380, not forgetting helicopters, between 10,000 and 15,000 aircraft fly over the country every day, according to the magazine ‘Aviation Civile’. Our country boasts 172,000 pilot licence-holders and over 500 airport runways according to the French Ministry of Europe and Foreign Affairs (ministère de l’Europe et des Affaires Etrangères), which adds that France has aviation in its blood, by dint of its history, culture and manufacturing tradition.
Perhaps more than in any other field, safety is a matter of constant concern in aviation. OSAC, Organisme pour la sécurité de l’aviation civile (France’s civil aviation safety inspectorate), is authorized by Ministerial decree to inspect aircraft on French territory and award airworthiness certificates.
SixFoisSept team members have worked with OSAC to help them improve their oversight programme with a risk-based approach. ‘We had to start by deciding what we meant by risk’, recalls Sophie Guérin, Head of Data Projects at SixFoisSept.
A shared definition of risk
Two factors were seen as key: the gravity of the consequences of an accident – the damage resulting from a commercial airliner crash is naturally greater than that caused by the crash of a banner-towing plane – and the likelihood that an accident will happen in the near future especially in light of past issues. ‘All accidents are recorded, for instance, a collision with a hangar door whilst taxiing after landing or a runway excursion’, says Sophie Guérin.
Other data is taken into account, such as the performance of the body that manages the aircraft. Other information may be examined, like the finances of the body in question or operational conditions. ‘An aircraft that is more exposed to salt and the wind due to frequent flights over the sea will rust more quickly’, she adds. When in use, the platform will have access to all data collected by OSAC inspectors. It has been designed so that it can easily make use of new sources of data (including open data), these being numerous in the field of civil aviation.
No black box effect
The solution put forward is a data visualization tool. Every year, the head of the oversight programme can personalize it by configuring the business rules as they specifically relate to aviation risks. During an inspection, the inspector simply has to enter the aircraft’s registration number to obtain its risk index, how it was calculated, along with basic statistics (date the design first flew, number of passengers carried etc). To make the tool simple and user-friendly, a UX designer was brought onboard the project to design more effective dashboards.
There is no ‘black box’ effect as regards the risk index. The algorithmic model applied is transparent and the factors used to calculate risk are known and explicable. SixFoisSept has epitomized the ethical face of data science since its foundation, as CEO and co-founder Erwan Prud’homme reminds us in this opinion piece.
For Sophie Guérin, the other key factor in the success of this project was how their industry partners embraced and participated in it. ‘They were involved in the tool’s design, then again to help assess the weighting given to different risk factors. We needed their knowledge to create a tool that did the job that was required of it.’ She explains that the adoption of decision support tools is heavily dependent on education: explaining to the inspectors that the tool being rolled out did not call their expertise into question, rather that it would act as an aid. ‘Their on-the-ground experience is irreplaceable.’
She also paid tribute to the successful pairing of statisticians and industry experts. The Head of Data Projects feels that the role of consultant Data scientist is all about immersing themselves in a field of activity and quickly picking up its unwritten rules and conventions. You don’t have to become a civil aviation expert, but it is essential to be able to communicate easily with the professionals in this field.
An approach that can be extended to other fields of activity
After having demonstrated its usefulness at the OSAC, the risk-based approach to oversight can be applied to any supervisory body performing audits focused on procedures or products.
Allocating a risk index or score makes it possible to shift the emphasis on monitoring and inspections onto the areas of greatest risk as well as changing how inspections are carried out – remotely, through answering a questionnaire for example, or on-site. The frequency of inspections may also be varied.
Fields which might be suited to this approach include inspections of buildings, hospitals, boats and vehicles. The bodies that are in charge of inspecting these regulated activities either perform random inspections, as was the case with the OSAC, or more systematic inspections, as applies to the accreditation of healthcare facilities or MOT tests.
In the latter case, it would be possible – depending on the risk associated with individual vehicles – to adjust the methodology or the frequency of the MOT test. ‘The length of time between inspections could be shortened or increased and the test could be made more exhaustive or simpler’, suggests Sophie Guérin.