After I obtained my BEng from the University of Manchester, I completed an MSc in Computational & Software Techniques in Engineering at Cranfield University. I then decided to re-join Manchester to carry out a PhD titled ‘Automated de-featuring of CAD geometries for simulation of complex systems’ under the supervision of Prof. Lee Margetts and Prof. Hujun Yin.
A fusion reactor comprises hundreds of thousands of components that each contain geometrical features such as bolt holes or fillets. During simulation, such features may not necessarily affect actual physical results or engineering decision-making, however they have an adverse impact on the run time and complexity of the simulation. Currently, it is up to the keen eye of the simulation engineer to carry out the time-consuming task of manually de-featuring the geometry.
This project seeks to investigate whether de-featuring can be carried out automatically using machine learning techniques. The project will involve the use of NVIDIA’s Omniverse platform, which not only allows users to collaborate across platforms, but also do away with the broad range of proprietary CAD formats currently in use by engineers. Furthermore, Omniverse offers connectors to most common CAD and machine learning packages. I will assess the feasibility of the proposed system by attempting to automatically de-feature the CAD geometry of the MAST-U Spherical Tokamak.