Skip to content

Research projects

For Safety

Data driven Safety improvements

In a more and more operational environment, it becomes very difficult to understand safety issues and identify safety improvements. Safety Line conducted research projects aimed at highlighting week signals from existing or new type of data.

SafetIcare / RAPID

Analysis of serious events (accidents or near accidents) is generally considered the “reactive” part of risk management, because it is triggered in response to events that we want to avoid. It is often considered to be insufficiently effective, the reasoning being that it leads especially to avoiding events that have already occurred, and therefore lags behind. We tend therefore to prefer approaches generating a lead, dealing with events deemed incidental but as far removed as possible from the accident. But these approaches, generally termed “proactive”, use the same logic as accident analyses: understanding “why it happened” by analysing the tree structure of direct and/or indirect (organisational) causes of the incident, and then ensuring that “this does not happen again” by eliminating these causes with appropriate measures. 


FlightScanner has been developed in collaboration with LSTA. Its purpose was to use Flight data as a major source of information on risks. However, with classic analysis tools, no one can browse the millions of pieces of information available. By comparing all flights, machine learning algorithm will provide continuous and accurate measurements of risks  The risks are actually well known, but not systematically monitored. A metric is used for each risk, runway excursion, hard landing, take-off performance, etc. For example, the landing distance for each flight is compared to the actual runway length. Doing so, it is possible to separate flights with long and short landing distances. The innovation is that Machine Learning algorithms are able to determine what the main parameters are between two information sets. As the learning process is updated on a regular basis, it allows to discover trends and can really perform safety assurance based on data.