Ejectors are passive devices for mixing, compression, or vacuum generation, widely used in aeronautics, refrigeration, and process industry. They operate by expanding a high-pressure primary flow to entrain a lower-pressure secondary flow, reaching an intermediate pressure after mixing. Their performance is governed by the maximal secondary mass flow under choked conditions, yet the choking mechanism remains only partly understood. Two theories exist: compound choking, based on both streams and the assumption of uniform static pressure, and Fabri choking, based on each stream reaching its own sonic point.
Conventional 0D models are too coarse to resolve these mechanisms, while RANS simulations reproduce choking without revealing the underlying physics. This thesis therefore develops one-dimensional models that retain local axial information at similar cost to 0D models. Single- and double-stream formulations are built, calibrated, and analysed, with explicit comparison of compound and Fabri choking in the double-stream case. Early quasi-2D developments address the limitations of purely 1D approaches.
A general calibration framework is proposed, combining adjoint-based sensitivities with machine-learning regression to identify closure relations. Finally, an extensive numerical and experimental database validates the models. Together, these tools provide a consistent low-dimensional representation of ejector flows and clarify the mechanisms governing choking and overall performance.
Membres du jury :
- Prof. Yann Bartosiewicz (UCLouvain) (Promoteur)
- Prof. Miguel A. Mendez (VKI) (Promoteur)
- Prof. Nicolas Moës (UCLouvain) (Président)
- Prof. Philippe Chatelain (UCLouvain) (Secrétaire)
- Prof. Adriano Milazzo (università di Firenze)
- Prof. Paola Cinnella (Sorbonne)
- Dr. Krzysztof Banasiak (SINTEF)
Soutenance publique également accessible par visio-conférence via le lien (TEAMS) :
https://teams.microsoft.com/l/meetup-join/19%3ameeting_YWQwY2U0ZDItMGM5Yy00MzNjLThkN2EtNzVhMGMyZTU2Y2Rm%40thread.v2/0?context=%7b%22Tid%22%3a%227ab090d4-fa2e-4ecf-bc7c-4127b4d582ec%22%2c%22Oid%22%3a%223be4c7c3-56d9-4142-ba00-8f9b561a6d55%22%7d
