Shockwave Interaction Enhanced by Computational Modeling

Supersonic shockwave interaction is an underexplored topic of aerodynamic research. Few studies have been done that address a computational model of these interactions in a way that balances physical accuracy with computational efficiency. This study asks the question: Can computational modeling of oblique shockwave interactions enhance the aerodynamic design of supersonic aircraft? Four objectives were established in order to answer this research question. The first objective was determining the air flow properties before, during, and after two shockwaves interact by including calculations for pressure, temperature, and Mach number. The second objective was developing a visual computational model to simulate and animate the formation and interaction of oblique shockwaves. The third objective was analyzing the weak and strong shockwave interactions by comparing their effects on air properties and aircraft aerodynamics through entropy-based methods of calculation. Finally, the fourth objective was to incorporate machine learning (ML) techniques to predict oblique shockwave behavior and to improve the performance of the modeled interaction. The classical compressible flow equations formed the theoretical basis of this study. The Prandtl-Meyer function and θ-β-M relations were the foundation for the computation of the shockwave angle, after shockwave temperature, and pressure conditions. Oblique shockwaves can be either weak or strong, depending on the shock angles and flow conditions. A Python model was created to evaluate the air flow properties before, during, and after the oblique shockwave interaction. The model produced a visualization of two interacting supersonic objects in a 3D space, along with their generated oblique shockwaves. Additionally, a ML code was developed to enhance the shockwave modeling process, reducing computational cost as compared to physics-based models. Recurrent Neural Networks (RNN) and Physics-Informed Neural Networks (PINN) were used to account for spatial laws. The final system produced rapid predictions of shockwave interaction behavior for use in supersonic design applications by iteratively training the ML model on simulated shockwave interactions.