Author: Li, Feilong
Title: Acceleration of ultrasonic wave propagation modelling using spectral elements and multi-processing unit-based parallel computing
Advisors: Zou, Fangxin (AAE)
Su, Zhongqing (ME)
Degree: Ph.D.
Year: 2023
Subject: Ultrasonic waves
Ultrasonic testing
Hong Kong Polytechnic University -- Dissertations
Department: Department of Aeronautical and Aviation Engineering
Pages: xviii, 239 pages : color illustrations
Language: English
Abstract: Ultrasonic non-destructive testing (NDT) can detect flaws and defects, and even check the growth of known damages, without destroying the object being inspected. Thus, this method can potentially prevent the failure of parts, components, or entire assets. Ultrasonic NDT largely depends on the properties of wave propagations in inspection targets. Due to the complexity of inspection targets in which the waves propagate, extensive wave modes, and the complex interaction between them, ultrasonic wave propagation problems are complicated and deeply rely on numerical simulations.
The classical finite element method (FEM), which is widely applicable for modelling various mechanical problems, generally uses linear elements, while accurate solutions can only be obtained by dense meshes, leading to high computational expense, especially for large-scale structures. In contrast, the spectral element method (SEM), which employs high-order elements formulated by high-order shape functions, converges fast with much coarser meshes than the classical low-order FEM. SEM may be more promising in large-scale modelling of wave propagation problems.
Parallel computing uses more than one processor to handle a computation by dividing the workload between different processors, all of which run simultaneously. A GPU features rich processing units and an extensive memory hierarchy, which equip the GPU with a powerful computational ability. However, the capability of a single GPU is also very limited, especially for large-scale modelling. Multiple GPUs would help to address those problems.
In this thesis, SEM and parallel computing are used to accelerate simulations of ultrasonic wave propagations in solid media. Firstly, a 2D hybrid spectral/finite element scheme for numerically resolving crack-induced contact acoustic nonlinearity is proposed. In it, spectral elements (SEs) are used to discretize the large intact region of a structure, while finite elements (FEs) are used to precisely depict the shapes of cracks within a small region. Then, a novel multi-GPU and CUDA-aware MPI-based SE formulation for simulating linear ultrasonic wave propagations is introduced. Based on CUDA-aware MPI, two novel message exchange strategies are developed to achieve communication between different GPUs. Considering the non-ignorable ability of CPUs in GPU computing, collaborations between CPUs and GPUs would have the potential to further improve computational performance. Therefore, simulations of linear ultrasonic wave propagations were further accelerated by a hybrid multi-core CPU/multi-GPU-based SE formulation. At the end of the thesis, the multi-GPU-based SE formulation is extended to model the interaction between material nonlinearity and ultrasonic waves. To do so, material nonlinearity is formulated by third-order elastic constants which establish a nonlinear relationship between strain and stress in a solid media.
The accuracy and the efficiency of the methods proposed in this thesis for simulating ultrasonic wave propagation problems have been deeply investigated. Throughout this thesis, the practicability of the novel methods has been extensively demonstrated through modelling realistic scientific and engineering problems. At last, a summary of the research capability and the problem-solving skills that the author has gained throughout the entire research journey is reported in this thesis for further reference.
Rights: All rights reserved
Access: open access

Files in This Item:
File Description SizeFormat 
6704.pdfFor All Users8.32 MBAdobe PDFView/Open

Copyright Undertaking

As a bona fide Library user, I declare that:

  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.

By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

Show full item record

Please use this identifier to cite or link to this item: