Akademik Veri Yönetim Sistemi
International Journal of Heat and Fluid Flow, cilt.119, 2026 (SCI-Expanded, Scopus)
Accurate prediction of electrohydrodynamic (EHD) airflow is essential for advancing flow-control strategies, thermal management concepts, and low-power propulsion systems. In hybrid electric–magnetic EHD configurations, the induced flow field arises from a complex interaction among electric field gradients, charge transport, space-charge–driven momentum transfer, and Lorentz-force modulation. These nonlinear couplings are difficult for conventional analytical or numerical approaches to capture with sufficient fidelity. This study presents an experimentally validated and data-driven surrogate modeling framework for predicting airflow velocity in a multi-needle EHD thruster. A structured dataset was generated by systematically varying emitter voltage, emitter–collector spacing, and solenoid excitation. Four supervised regression models—Random Forest, Gradient Boosting, K-Nearest Neighbors, and ensemble-based techniques—were trained using standardized preprocessing and k-fold cross-validation. Among them, Gradient Boosting achieved the highest accuracy with an R2 of 0.8961, MAE of 0.0859, and MSE of 0.0115. To enhance physical interpretability, SHapley Additive exPlanations (SHAP) based analysis was performed. The results showed that emitter voltage dominates EHD-induced momentum transfer, while geometric spacing and magnetic forcing produce secondary modulation of the ion-driven flow. SHAP interactions further indicated that magnetic-field effects intensify at higher electric field strengths, aligning with expected multiphysics coupling mechanisms. Independent testing at previously unseen operating points confirmed strong agreement between predicted and measured velocities. The proposed framework provides a fast, transparent, and computationally efficient alternative to conventional simulations, offering new insight into coupled electro-fluidic behavior. Overall, the study demonstrates the potential of interpretable machine-learning-assisted modeling to accelerate the design and optimization of advanced thermo-fluidic and flow-control systems.