Towards the Integration of Surrogate Models in the Robust Optimization of Hybrid Rocket Engines

Hybrid rocket engines combine the appealing features of solid and liquid propulsion systems, offering advantages such as cost-effectiveness, safety, and eco-friendliness. Their potential role in sustainable space transportation aligns with the goals of the NewSpace Economy, particularly in light of environmental concerns surrounding traditional propulsion systems. While hydrazine-based liquid rocket engines face potential bans in Europe, and solid rocket motors contribute significantly to pollution, hybrids using liquid oxygen and paraffin-based wax are regarded as sufficiently green. Additionally, advances in liquefying fuels have led to increased global interest, as demonstrated by numerous recent development programs. Design optimization for hybrid rocket engines is inherently multidisciplinary, with propulsion system and trajectory optimization closely coupled. While deterministic methods have been widely used, real-world uncertainties, especially in fuel regression rates, can significantly jeopardize system performance and reliability. Robust optimization approaches mitigate these effects, but they are computationally expensive. This paper presents a preliminary investigation into the capabilities of surrogate models in these scenarios. A binary classifier is coupled with a neural network to predict system performance in the presence of uncertainties. The design optimization of a hybrid-powered upper stage is selected as a case study to test the proposed method’s effectiveness.