Return to search

FE-PML Modeling of Guided Elastic Waves and its Applications to Ultrasonic NDE

This thesis investigates the use of a combined finite element and perfectly matched layer approach in modeling guided elastic wave motion in infinite plates and cylinders and its potential applications to non-destructive evaluation. Underlying principles of the per-fectly-matched, absorbing layer are demonstrated on one-dimensional wave propagation in a semi-infinite elastic rod.
Feasibility of using the perfectly matched layer as absorbing boundary condition in the finite-element modeling of guided elastic wave propagation and scattering is studied for the canonical problem of shear horizontal wave motion in isotropic plates. Numerical re-sults in this study are validated against exact analytical solutions. Excellent agreement has motivated the endeavour to take the technique to the next level of pressure, shear-vertical wave motion in isotropic and transversely isotropic plates.
Time-domain, finite-element formulation of the perfectly matched layer for pressure, shear-vertical wave motion was validated through comparisons with semi-analytical lit-erature data and reciprocity checks. Numerical implementation of the model was em-ployed in studying the effect of crack presence on the time of arrival in a pitch-catch, non-destructive inspection arrangement. Predictions made confirmed previously-reported experimental findings.
Extensions into three-dimensional, Cartesian and cylindrical spaces were validated against reported data. Practical examples of wave scattering in damaged concrete beams, oil and gas pipelines, and composite shells demonstrated the potential use of the proposed model in simulating elastic-wave based non-destructive inspection. Up to 80 % of the computational time needed to run an extended-mesh, finite-element model can be saved by introducing the perfectly-matched, absorbing layer to the finite-element model as the current thesis proposes. This significant saving in computational time by the proposed FE-PML model can accelerate the production of artificial neural network training data or help tackle complicated non-destructive testing applications.

Identiferoai:union.ndltd.org:MANITOBA/oai:mspace.lib.umanitoba.ca:1993/4152
Date10 September 2010
CreatorsMahmoud, Abdel-Rahman
ContributorsWang, Quan Abraham (Mechanical and Manufacturing Engineering), Luo, Yunhua (Mechanical and Manufacturing Engineering) Rattanawangchareon, Nipon (Civil Engineering) Pan, Ernian (University of Akron)
Source SetsUniversity of Manitoba Canada
Languageen_US
Detected LanguageEnglish

Page generated in 0.0018 seconds