Reduced-order prediction of multi-jet flow control effects on vortex breakdown over delta wings from single-jet training

Vortex breakdown over delta wings strongly influences aerodynamic performance at moderate to high angles of attack. When active jet control is introduced, the flow response becomes highly sensitive to jet placement, making systematic exploration of multi-jet configurations computationally expensive. This study develops a reduced-order CFD-machine-learning framework to estimate multi-jet vortex behaviour using training data obtained solely from steady single-jet simulations. Analysis of the single-jet database shows that jets placed near the mid-span and close to the leading edge promote upstream migration of the breakdown, whereas jets positioned closer to the wing apex delay vortex breakdown. A reduced-dimensional representation of the flow is constructed to capture the dominant vortex structures, and the baseline CFD predictions are assessed against Particle Image Velocimetry measurements. Jet parameters are then mapped to this modal space to enable rapid reconstruction of planar velocity fields. Without retraining on combined-actuation cases, the single-jet-trained model is used to generate rapid superposition-based estimates for selected two-, three-, and four-jet configurations. The reconstructed fields preserve the main vortex topology, with error increasing gradually but remaining bounded as jet count increases. Full multi-jet CFD simulations are then used to verify the selected configurations and show that the physical response is cumulative but not strictly additive, exhibiting saturation and partial cancellation depending on the actuator arrangement. The proposed framework reduces the computational cost of steady RANS-based screening from 20.85- 83.40 core-hours, depending on mesh resolution, to approximately one minute on a single CPU core, enabling rapid screening of complex multi-jet flow-control strategies.