This paper investigates the use of neural networks as surrogate models for fast and accurate performance prediction of hybrid rocket engines (HREs). To this aim, a deep feedforward neural network i...

A computational tool able to perform a fast analysis of hybrid rocket engines is presented, describing briefly the mathematical and physical models used. Validation of the code is also shown: 16 different static firing tests available in the open literature are used to compare measured operational parameters such as chamber pressure, thrust, and specific impulse with the code’s output. The purpose of the program is to perform rapid evaluation and assessment on a possible first design of hybrid rockets, without relying on computationally expensive simulations or onerous experimental tests. The validated program considers as benchmark and study case the design of a liquid-oxygen/paraffin hybrid rocket engine to be used as the upper stage of a small launcher derived from VEGA building blocks. A full-factorial parametric analysis is performed for both pressure-fed and pump-fed systems to find a configuration that delivers the equivalent total impulse of a VEGA-like launcher third and fourth stage as a first evaluation. This parametric analysis is also useful to highlight how the oxidizer injection system, the fuel grain design, and the nozzle features affect the performance of the rocket.

This paper addresses the prediction of spontaneous self-sustained transverse combustion instabilities in multi-injector engines using a multi-dimensional non-linear low-order Eulerian model. The approach utilizes a physically aware response function that directly links pressure oscillations to fuel mass flow rate, independent of external data. This enables accurate capture of the behavior of shear coaxial injectors that have been proven to show a peculiar dynamics of cyclic fuel accumulation and release in response to acoustic waves. A NASA test case is employed as a reference, and a comprehensive analysis is conducted to investigate the behavior of the model. Firstly, a preliminary analysis is performed within a quasi-one-dimensional framework for a reduced single-injector configuration. This analysis provides insights into the influence of key parameters on the instability onset. Subsequently, the full-scale geometry is considered in a multi-dimensional framework, and the low-order solver is employed to analyze the thermo-acoustic behavior of the engine. The model successfully captures the instability dynamics, including temporal evolution, dominant frequency, and mode shapes. Sensitivity analyses are conducted to examine the impact of the model free parameters on the unstable modes. Additionally, the effect of introducing baffles as damping devices within the combustion chamber is investigated. The results highlight the model predictive capabilities and its potential for guiding the design and optimization of liquid rocket engines.

This paper investigates longitudinal high-frequency combustion instabilities using a low-order numerical solver to provide fast, reliable predictions intended to support the preliminary design of new liquid rocket engines at low computational cost. A physics-based response function is employed to model thermoacoustic interactions, linking acoustic waves in the injector recess to unsteady fuel mass flow rate. This approach mimics the cyclic fuel accumulation and release characteristic of shear coaxial injectors, as documented in experimental and numerical studies. The objective of this work is to present a significant update to a low-order model previously developed by the present research group. This update is made by introducing a new response function capable of detecting the direction of acoustic waves, and by introducing an additional transport equation for the fuel released at the injector. This new approach eliminates spurious numerical oscillations observed in the previous version of the model and leads to improved predictive capabilities. A detailed comparative analysis and validation of the updated solver are performed using the continuously variable resonance combustor test case as the benchmark. The study assesses the predictive capabilities of the solver by evaluating computed limit-cycle features against experimental data. A phenomenological investigation of fuel release events and a sensitivity analysis of model parameters shows the physical consistency and robustness of the solution, which are critical aspects for predictive modeling applications.