Full waveform inversion of ultrasonic phased array data for the tomographic reconstruction of heterogeneous media for NDT

Within the ultrasonic non-destructive community the benchmark imaging algorithm assumes that the host material is homogeneous. This can lead to poor flaw detection, location, and characterisation. However, when the heterogeneous nature is accounted for within an imaging algorithm there is a significant improvement in the flaw images.;This work builds upon that work by creating a full waveform inversion framework to build a tomography method that reconstructs spatially heterogeneous wave speeds maps from simulated ultrasonic phased array data. This framework consists of: a Voronoi tessellation to spatially parametrise the wave speed map; a bespoke semi-analytical model that encompasses the data within an ultrasonic A-scan while being computationally efficient; a bespoke objective function toquantify how well the semi-analytical model compares to the observed data; and a Bayesian framework, namely the reversible jump Marko chain Monte Carlo method (rj-MCMC), to perform the tomographic reconstruction in the form of a posterior distribution. The reconstructed wave speed maps are then used in conjunction with an imaging algorithm (TFM), to provide enhanced flaw imaging. The quality of the imaged flaws are then quantified by calculating the signal-to-noise ratio, the flaw location error, and probability of detection via ROC curves.;This framework is first applied to a synthetic randomly heterogeneous material with a side drilled hole present. It is then applied to a layered medium that contains three inclusions contained in different layers. The latter case being studied extensively in different scenarios and phased array set ups. A new objective function is then proposed (a time windowed Hilbert transform) which is applied to the layered medium allowing for the comparison of different objective functions.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral-pg

Grantor dc:publisher.institution

University of Strathclyde

Year dc:date.issued

2020

Author dc:creator

• Niven, Russell

Advisors dc:contributor.advisor

• Mulholland, Anthony
• Tant, Katherine

DOI

10.48730/v9hj-y070

Identifier

T15874

Author Identifier

201664364

OAI identifier oai:identifier

oai:strathclyde:vd66vz91p