werkenbijnlr · Amsterdam, Noord-Holland, Nederland
Skills: Python
Background Aircraft engines and battery systems operate under changing and demanding conditions, which lead to complex degradation over time. Factors such as operating conditions, mission profiles, environmental effects, and component aging all influence the performance of these systems. Recent advances in machine learning have enabled powerful methods for health monitoring and prognostics using measurement sensor data. In particular, latent dynamical models and probabilistic deep learning methods have shown strong potential for modelling complex temporal behaviour. However, many existing approaches mainly focus on prediction accuracy, while offering limited interpretability. In safety-critical aerospace applications, this can result in overconfident predictions, unrealistic degradation behaviour, and reduced trustworthiness for maintenance decision-making. A promising direction is the integration of reliability engineering principles and degradation physics into latent dynamical learning frameworks. By constraining or guiding the evolution of latent health representations using reliability-informed stochastic processes, degradation-aware priors, or physically meaningful dynamics, it becomes possible to improve model interpretability, uncertainty quantification, robustness, and extrapolation capabilities under changing operational conditions. Despite its strong potential, this integration remains relatively underexplored in current AI-based prognostics research. This project aims to bridge reliability engineering, stochastic degradation modelling, and modern AI-based prognostics. The proposed framework will investigate learned latent health representations whose evolution follows reliability-aware stochastic dynamics. References https://neurips.cc/virtual/2022/59892 T. Christen and F. Macedo, "Theory of Fokker–Planck Equations for Reliability and Remaining Useful Lifetime Prognostics," in IEEE Transactions on Reliability , vol. 73, no. 1, pp. 344-356, March 2024, do…
Sign in to see the original posting
Browse more developer jobs on RoleBurst