Numerical analysis and reduced-order modeling of a clustered aerospike engine

The last decade has seen a growing interest in reusable launchers. The clustered aerospike engine consists of multiple thrusters, referred to as cells in the following, arranged in an annular fashion around a central aerospike nozzle. This technology appears particularly promising for reusable launch vehicles due to its continuous altitude adaptation and differential throttling capabilities. These features can be in principle exploited twice, during both the ascent and landing phases of flight. In this context, accurately predicting engine performance while maintaining low computational effort is challenging, especially considering the complex and non-linear plume interactions that arise during differential throttling operation. In this paper, a procedure to generate a non-linear reduced-order model of a clustered aerospike engine has been implemented. This model is capable of predicting, with an error below 10 % for most test cases analyzed, axial and lateral components of thrust in different configurations. A CFD analysis has been carried out, considering a 28-cells aerospike scale model in a cold-flow configuration. Different RANS simulations were performed, aiming for a good compromise between accuracy and computational cost, to construct two different databases. The training database consists of 14 simulations with two active cells at different angular distances, plus one simulation with a single active cell. From this dataset, plume interactions were analyzed and characterized through the definition of appropriate interaction coefficients. Using these coefficients, a reduced-order model of the aerospike has been proposed and subsequently validated against a testing database, that instead comprises various symmetrical and asymmetrical configurations with a larger number of active cells. The proposed reduced-order modeling procedure is formulated in such a way that, in principle, it could be applied to any arbitrary aerospike engine and, at least for axial thrust predictions, even in a scenario where only experimental data would be available. Furthermore, this model could be applied as a first-order approximation to predict thrust components and their variations in the event of random failure modes, where one or more cells may become inactive during flight. The proposed model demonstrates good accuracy in predicting axial thrust, while further refinement together with a larger testing database is needed to improve its predictions of lateral loads.