Multi-Objective Design Optimization of Shock-Induced Mixing Enhancement via Evolutionary Algorithms Assisted by Data-Driven Approaches

Supersonic combustion ramjet (scramjet) engine stands out as a prime candidate for propulsion technology to meet the increasing demand for space transportation due to its superior efficiency. Sophisticated design methodologies are essential, particularly for fuel mixing, to achieve the necessary performance under the challenging hypersonic operating conditions which are accompanied by high thermal and structural load, short residence time in the combustor, and complicated aerothermodynamics. The present study has conducted a multi-objective design optimization of shock-induced mixing enhancement in a 2-dimensional scramjet combustor using evolutionary algorithms that incorporate deep-learning-based flowfield prediction. Flowfield prediction enables accurate evaluations of objective functions and provides an informative dataset containing predicted flowfield data at a minimal computational cost during the optimization process. This optimization approach has significantly reduced the number of computational fluid dynamics (CFD) simulations needed both in the optimization process and post-analysis. However, it has been found that further efforts are required to enhance the reliability of results obtained through flowfield-prediction-based design exploration and knowledge discovery.