Reduced order design space analysis of for ramjet engines with data mining techniques

Ramjet propulsion is commonly preferred for powering supersonic vehicles for cruising faster than Mach 3. Although the geometrical layout of these engines is simple compared to turbo-based counterparts, the complexity of flow physics involved with their operation drives design and optimization procedures to be challenging. Accordingly, a design and analysis methodology based on zero-and-one dimensional approaches utilizing Busemann intakes and detail chemistry solutions is employed to conduct design space exploration of ramjet engines. The investigation of design parameters in a wide range of operational conditions results in the generation of a multi-dimensional data matrix. In this paper, characterization of the propulsive performance of various operational and design conditions were discussed with the aid of sensitivity analysis such as Shapley Additive Explanations (SHAP) framework. Moreover, artificial neural network is applied to each operational and design variable in order to make a comprehensive exploration of the design space and build machine learning models to represent the propulsive performance of the ramjet.