In inverse design, the goal is to determine the geometry corresponding to a specific pressure or velocity distribution on the boundary. In the aerodynamic design of airfoils via inverse design, this distribution is modified to increase lift, decrease drag, or reduce flow losses. One inverse design method for airfoils is the elastic surface algorithm, which models the airfoil wall as an elastic bent beam. The shape of this beam changes according to the difference between the target and existing pressure distributions until this difference becomes zero, yielding the final shape corresponding to the target pressure distribution. In this research, the capability of changing the angle of attack while modifying the geometry was added to the elastic surface algorithm. This was done by changing the boundary condition of the elastic beam at its starting point from a hinge to a vertical groove. This adjustment addressed issues of non-convergence and unrealistic geometries where the suction and pressure surfaces intersect. Several validation and design examples demonstrated the robustness and flexibility of the enhanced method. The second part of the research discussed the limitations of pressure-based inverse design in the ultra-low Reynolds number regime. It was shown that the thick boundary layer and laminar bubble separation in this regime decrease the pressure parameter's sensitivity to changes in airfoil geometry. This results in instability, unrealistic geometries, and divergence in pressure-based inverse design. To address this, a shear-stress-based elastic surface algorithm was developed and validated for inverse design in ultra-low Reynolds number flow regimes. The shear stress distribution of the NACA 4404 airfoil was then modified using several approaches, and the corresponding geometry was obtained using the improved inverse design method. These approaches included reducing drag, increasing lift, controlling the laminar bubble separation at the optimal angle of attack, and controlling the fully separated flow area during the aerodynamic stall conditions. The results indicated the accuracy, fast convergence, and robustness of the shear-stress-based inverse design method for various design objectives in this flow regime. The performance of the designed geometries was eva‎luated and compared using the Fluent solver, showing a 13.72% and 16.09% increase in lift-to-drag ratio at the optimal and stall angles of attack, respectively. The designed airfoil in the stall conditions also delayed the stall angle by 4 degrees. In the final part of the research, two deep-learning models were trained to predict aerodynamic coefficients using shear stress distribution data from different inverse designs at the optimal angle of attack. By coupling genetic algorithms, the trained deep-learning models, and inverse designs across several optimization cycles, an optimal shear stress distribution was obtained to maximize the lift-to-drag ratio, with the corresponding geometries calculated using inverse design. The performance of different optimization cycles was eva‎luated and compared using the Fluent solver, resulting in a 22.42% increase in the lift-to-drag ratio at the optimal angle of attack. Keywords: Airfoil, Angle Of Attack, Elastic Surface Algorithm, Laminar Sepration Bubble, Pressure distribution, Shear stress distribution, ultra-low Reynolds number regime

مهدي نيلي احمدآبادي

احمدرضا پيشه وراصفهاني , ايمان چيت ساز