Precise Impulse Design for Hybrid Rockets Using Machine-Learning Informed Digital Throttle Input

Salt Lake Community College Westpointe Campus, Salt Lake City, UT

Hybrid rocket motors are known to be safer and greener than solid motor rockets. Hybrid rockets provide additional benefits to mission design with the ability to shutdown, restart, and throttle reliably. Capabilities well suited to applications such as sounding rockets, orbital insertion of small satellites, and surface launch systems. The implementation of hybrid rockets does not come without its own complications. A significant uncertainty in hybrid rocket thrust output is the Oxidizer-to-Fuel ratio (O/F) producing variable thrust, also known as the O/F shift. The effect of O/F shift on thrust has the potential to limit the performance and reliability of hybrid motors. Potentially, the O/F shift can be compensated through the oxidizer injection rate, by changing the discharge coefficient. This study proposes the modeling of one major control parameter of hybrid rockets, injector discharge coefficient (CD), using known thrust profiles and Machine Learning models. The two ML methods tested in the current study are, Deep Neural Networks (DNN) and Long Short-Term Memory (LSTM). The modeling of CD profiles for given thrust profiles, will facilitate creation of a compensated injector schedule to correct for O/F shifts and deliver a more consistent commanded thrust over the burn lifetime of a hybrid motor.