Detection and avoidance of low probability phenomena using probabilistic graphical models in electromechanical actuators.

Current condition monitoring methods for electromechanical actuators are lacking robustness. This is a consequence of the large number of external (e.g. load) and internal (e.g. manufacturing tolerances) factors that influence the measurements on which the condition monitoring is based. This research project aims to improve the robustness by augmenting the models used for condition monitoring with a probabilistic model which avoids false conclusions. Additionally, the capabilities of the use of probabilistic models for root-cause analyses that can be used to avoid harmful condition degradations will be explored.