The use of deep learning and machine learning for the optimisation of space propulsion technologies

Deep learning has established itself as a useful tool for the analysis and processing of big data, used extensively in computer science and the gaming industry, it has only had relatively limited (yet important) advances in the aerospace engineering sphere. This PhD will develop the neural network architectures being developed at Kingston University for the space propulsion industry and try to expand them. The current architectures are:
• PPTNET – This neural network is based on old experimental data and has been developed to predict the performance of a Pulsed Plasma Thruster [1]. The network requires significantly more data to generalise trends.
• PlasmaNET – This neural network is an attempt to generalise particle interactions as simulated by a PIC code simulation, with the intention of being able to simulate large plasma formations found in Electric Propulsion devices, but able to investigate at the particle scale in a reasonable timescale.
• RocketNET – This network is based on the firing data of rockets developed at Kingston University and is aimed at creating a tool that can optimise the design of a rocket and predict its firing data.

The successful candidate will look at developing one of these neural networks and apply it to a practical scenario. The ideal candidate will be competent in Python, have experience with machine learning or deep learning and have working knowledge of spacecraft propulsion systems or have an Aerospace background.