Title: HPCI Seminar Series: FEM with GPUs, with application to NDE: “Implementing the Finite Element Method on Graphics Cards for Ultrasound Simulations”
Speaker: Peter Huthwaite, Senior Lecturer, Mechanical Engineering, Imperial College London
Date: Friday, July 13, 2018
Time: 9:00am
Location: NASA/LaRC, Bldg. 2102, Room 114/115
Host: Cara Leckey, NASA/LaRC
Abstract: The finite element method is a very powerful, general tool for simulating ultrasound. The main advantage is that it is straightforward to deal with complex geometry; the underlying approximation of the finite element method does not rely on uniform grids, so the mesh can conform to arbitrary geometry. It has therefore found a range of uses across both bulk wave and guided wave simulation. Graphics cards are highly parallel architectures, designed to process images, represented by pixels, with many lightweight threads. This architecture is also very well suited to many simulation techniques, since these commonly rely extensively on parallel processing for speed. The finite element method is one technique which would benefit from the speed improvement, but its very advantage of not relying on structured grids makes implementation on graphics cards less straightforward. This talk will cover three aspects: the basics of the finite element method and how it can be used for ultrasound simulation, the graphics card architecture and how to exploit it, and finally how the finite element method can actually be implemented successfully on graphics cards.
Bio: Peter Huthwaite is a senior lecturer in Mechanical Engineering at Imperial College London. He leads imaging research within the Non-Destructive Testing group. His main research area is limited data imaging for ultrasonics and radiography, to reduce issues associated with acquisition time, system complexity and physical restrictions which occur in practice. His interests include the development of quantitative image reconstruction algorithms, with applications in medicine for breast cancer diagnosis and NDT for guided wave tomography. He also has interests in the numerical modelling of waves and has developed the GPU-based finite element software package Pogo.