PRR · Capitalisation and Business Innovation
AVIATOR
Multi-domain Artificial Intelligence Algorithm for Operational Safety in Civil and Military Aviation
PRR · Capitalisation and Business Innovation
Multi-domain Artificial Intelligence Algorithm for Operational Safety in Civil and Military Aviation
AVIATOR
Artificial Intelligence for operational safety in aviation
____
02/C05-i14-01/2025
28178
Synopsis Planet – Advance Engineering, Lda.
AEROG – University of Beira Interior
IDMEC – Instituto Superior Técnico
1 June 2026 – 31 May 2028 (24 months)
Fatigue, cognitive overload, operational stress and loss of situational awareness remain among the main causes of incidents and accidents in aviation, both civil and military. Today, detecting these factors relies mostly on manual analysis of reports, carried out after the events have happened.
AVIATOR is developing an Artificial Intelligence platform for operational safety, built on the HFACS model (Human Factors Analysis and Classification System). The platform classifies and anticipates the contributing factors associated with human error, using the operators' occurrence history and safety reports.
The aim is to move from a reactive stance to proactive prevention, integrated into the Safety Management Systems (SMS) that operators already use.
THE CHALLENGE
At most operators, human factors analysis is still done by hand. Analysts read occurrence reports, classify them and look for patterns, usually after the event has happened.
Current approach
Manual reading of safety reports and narratives
Patterns identified after events have happened
Assessment depends on each analyst's perception
Significant time spent on each occurrence
With AVIATOR
Automatic extraction of the relevant elements of each report
Early risk signals drawn from the operator's history
Consistent HFACS classification, with the evidence behind it
Target: halve the analysis time per occurrence
A platform that anticipates risk associated with human factors, instead of analysing it only after the event. The project's success is measured against three targets set out in the funding application
≥ 75%
F1-score in predicting human-factors risk
+20%
in early detection, classification and flagging of HFACS scenarios, compared with manual analysis
Targets to be demonstrated at the end of the project (milestone M.5.1).
−50%
in average analysis time per occurrence
AVIATOR turns the operator's safety history into early warning signals, through a five-step process.
Technology readiness: from TRL 5 to TRL 7, validated in a real operational environment
ACTIVITIES AND MILESTONES
A platform that anticipates risk associated with human factors, instead of analysing it only after the event. The project's success is measured against three targets set out in the funding application
Two conference papers and two articles in scientific journals are planned.
Synopsis Planet is supporting, in collaboration with universities and R&D centres, two PhD theses on the early detection of operational risks using AI, with a focus on human factors: fatigue, stress and cognitive load.