Interpretable & Explainable Machine Learning for Ultrasonic Defect Sizing - accepted author manuscript (AAM)

Aug 5, 2026

Upload by Logic

Interpretable and explainable machine learning is applied to ultrasonic nondestructive evaluation to overcome the “black-box” limitation common in many industrial ML models. The work introduces Gaussian feature approximation (GFA), which fits a 2D elliptical Gaussian to each ultrasonic image and encodes seven Gaussian parameters as compact, analysis-ready inputs for a defect sizing neural network. For inline pipe inspection, GFA features are compared with neural-network sizing from raw images, principal component analysis, and 6 dB drop-box parameters, and are evaluated using SHAP to identify how each feature affects predicted defect length.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Abstract

Introduction

Method overview (GFA and neural network)

Comparison of dimensionality reduction methods

Explainability using SHAP

Pyle, R. , Hughes, R. R. , & Wilcox, P. D. (2023) . Interpretable and Explainable Machine Learning for Ultrasonic Defect Sizing. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 70(4), 277-290 . [https://doi.org/10.1109/TUFFC.2023.3248968](https://doi.org/10.1109/TUFFC.2023.3248968) Peer reviewed version Link to published version (if available): 10.1109/TUFFC.2023.3248968 Link to publication record on the Bristol Research Portal PDF-document This is the accepted author manuscript (AAM) . The final published version (version of record) is available online via IEEE at [https://ieeexplore.ieee.org/document/10052715.Please](https://ieeexplore.ieee.org/document/10052715.Please) refer to any applicable terms of use of the publisher. University of Bristol – Bristol Research Portal General rights This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: [http://www.bristol.ac.uk/red/research-policy/pure/user-guides/brp-terms/](http://www.bristol.ac.uk/red/research-policy/pure/user-guides/brp-terms/) This article has been accepted for publication in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control. This is the author's version which has not been fully edited and content may change prior to final publication. Citation information: DOI 10. 1109/TUFFC.2023.3248968 1 IEEE TRANSACTIONS ON ULTRASONICS, FERROELECTRICS, AND FREQUENCY CONTROL Interpretable & Explainable Machine Learning for Ultrasonic Defect Sizing Richard J. Pyle, Robert R. Hughes, Paul D. Wilcox, Member, IEEE Abstract—Despite its popularity in literature, there are few examples of machine learning (ML) being used for industrial nondestructive evaluation (NDE) applications. A significant barrier is the ‘black box’ nature of most ML algorithms. This paper aims to improve the interpretability and explainability of ML for ultrasonic NDE by presenting a novel dimensionality reduction method: Gaussian feature approximation (GFA) . GFA involves fitting a 2D elliptical Gaussian function an ultrasonic image and storing the seven parameters that describe each Gaussian. These seven parameters can then be used as inputs to data analysis methods such as the defect sizing neural network presented in this paper. GFA is applied to ultrasonic defect sizing for inline pipe inspection as an example application. This approach is compared to sizing with the same neural network, and two other dimensionality reduction methods (the parameters of 6 dB drop boxes and principal component analysis), as well as a convolutional neural network applied to raw ultrasonic images. Of the dimensionality reduction methods tested, GFA features produce the closest sizing accuracy to sizing from the raw images, with only a 23% increase in RMSE, despite a 96.5% reduction in the dimensionality of the input data. Implementing ML with GFA is implicitly more interpretable than doing so with principal component analysis or raw images as inputs, and gives significantly more sizing accuracy than 6 dB drop boxes. Shapley additive explanations (SHAP) are used to calculate how each feature contributes to the prediction of an individual defect’s length. Analysis of SHAP values demonstrates that the GFA-based neural network proposed displays many of the same relationships between defect indications and their predicted size as occur in traditional NDE sizing methods. Index Terms—Interpretability, machine learning ultrasound, defect characterization, neural network, plane wave imaging, simulation I. INTRODUCTION Inferring the structural integrity of components without damaging them is an essential task for many industries, especially those with high-value or safety critical components. Non-destructive evaluation (NDE) techniques solve this challenge through analysis of a component’s response to stimuli such as X-ray or ultrasound. Most NDE

FAQs

What problem does the paper address in ultrasonic defect sizing?

It addresses the limited interpretability and explainability of most ML models used in industrial ultrasonic nondestructive evaluation, where trust and certification are difficult due to “black-box” behavior.

How does Gaussian feature approximation (GFA) represent ultrasonic data?

GFA fits a 2D elliptical Gaussian to an ultrasonic image and stores seven parameters describing each Gaussian, which then serve as inputs to downstream data analysis such as a defect sizing neural network.

Which dimensionality reduction approach performed best in defect sizing accuracy?

Among the tested methods, GFA features achieved the closest sizing accuracy to using raw images, with only a 23% increase in RMSE despite a 96.5% reduction in input dimensionality.

Aug 5, 2026

Upload by Logic