Four years ago, we embarked on a journey to develop an Evidence-Based Training (hashtagEBT) solution for flight crew training. With our foundation in safety management and analytics, we believe that training is the most effective and efficient way to mitigate safety risks in flight operations. Evidence-Based Training, in particular, provides a structured and flexible approach to integrating data-driven insights into the training process.
Thanks to the support of Etihad, collaboration with industry experts NLR – Netherlands Aerospace Centre, and our partner and launch customer AeroLogic, avialytics has developed an integrated training management and grading application. This system leverages Observable Behavior (OB) enhanced Threat and Error Management (TEM) markers to not only assess crew performance across various competencies but also to refine individual competencies within specific training program elements.
In essence, our goal is to utilize objective crew performance data from Flight Data Monitoring (hashtagFDM) and operator-specific characteristics to tailor individual simulator sessions. We then measure the effectiveness of the training by comparing performance before and after completing various modules.
Regarding decision support integration for trainers and flight safety departments, we distinguish three levels:
• Deterministic: These elements are essential for meeting regulatory requirements, defining safety triggers, and maintaining transparency and ease of management.
• Probabilistic: We introduce an element of surprise to enhance resilience and shift focus toward competencies rather than a fixed syllabus. Both deterministic and probabilistic elements (Markov chains) are well understood and auditable.
• AI-Enhanced: At this level, we introduce a degree of uncertainty by utilizing Azure AI services, which combine various models with domain-specific knowledge. This AI suggests training element combinations tailored to individual crew members, influencing their training flow. With generative AI, we aim to enhance concordance by offering trainers and trainees a user-friendly interface for feedback based on statistical data and procedural knowledge.
While we are not fully at the AI-enhanced level yet, evaluating this approach requires objective performance criteria. This closes the loop by comparing a crew member’s performance before and after specific training interventions. Grading and observable behaviors play a role here, but the most objective source will be FDM data—considering operational contexts like weather, ATC, and operational conditions—while ensuring the protection of personal rights through anonymization and training clustering.
There’s much more to this than what I’ve touched on in this brief post. I know many of you in my network are experts in these areas, and I would greatly appreciate engaging in a discussion with anyone interested.
hashtagiata hashtagsafety Heidelberg Innovation Park
