Assessment of the Numerical Performance of the Single-step Simplified Lattice Boltzmann Method

Abstract

The advances of computing methods and computing hardware progress hand in hand. Prior to the invention of the general purpose graphic processing unit (GPGPU), compu- tations were primarily relied on central process unit (CPU). The prevailing numerical methods were conventional techniques that addressed nonlinear equations. After the in- troduction of GPGPU, massive parallelisation becomes common; shifting the focus of numerical method investigation towards leveraging GPUs for computational resources. The lattice Boltzmann method (LBM) is one of the ”advanced” methods, which is nat- urally suitable for parallel computing and has become a widely applied robust method. One of the drawbacks of LBM is the high consumption of running memory. To overcome this drawback, a single-step simplified LBM (SSLBM) was recently proposed which theoretically reduces the memory consumption of LBM while keeping the efficiency and parallelism of LBM. Due to the lack of comprehensive analysis, in this thesis, through the implementation of SSLBM on GPU, a systematic numerical analysis of SSLBM perfor- mance in accuracy, stability, memory reduction, and computing efficiency is carried out to fill the gap. A comprehensive explanation and implementation of boundary conditions in SSLBM is stated and validated with different benchmark cases. A new outlet bound- ary condition is derived and implemented in SSLBM. At the end, based on the detailed numerical analysis, a GPU accelerated numerical solver is developed.

Date of Award	6 Jan 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Alistair Revell (Supervisor) & M Nabawy BSc, MSc, PhD, MRAeS, SMAIAA, FHEA (Supervisor)