Abstract:This study introduces a novel self-supervised learning approach for volumetric segmentation of defect indications captured by phased array ultrasonic testing data from Carbon Fiber Reinforced Polymers (CFRPs). By employing this self-supervised method, defect segmentation is achieved automatically without the need for labelled training data or examples of defects. The approach has been tested using artificially induced defects, including back-drilled holes and Polytetrafluoroethylene (PTFE) inserts, to mimic different defect responses. Additionally, it has been evaluated on stepped geometries with varying thickness, demonstrating impressive generalization across various test scenarios. Minimal preprocessing requirements are needed, with no removal of geometric features or Time-Compensated Gain (TCG) necessary for applying the methodology. The model's performance was evaluated for defect detection, in-plane and through-thickness localisation, as well as defect sizing. All defects were consistently detected with thresholding and different processing steps able to remove false positive indications for a 100% detection accuracy. Defect sizing aligns with the industrial standard 6 dB amplitude drop method, with a Mean Absolute Error (MAE) of 1.41 mm. In-plane and through-thickness localisation yielded comparable results, with MAEs of 0.37 and 0.26 mm, respectively. Visualisations are provided to illustrate how this approach can be utilised to generate digital twins of components.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in the ultrasonic testing of composite materials (especially carbon - fiber - reinforced polymers, CFRP), how to achieve automatic defect segmentation and localization without relying on labeled training data. Specifically, the authors propose a new self - supervised learning method for volume segmentation of three - dimensional ultrasonic testing data. ### Problem Background 1. **Applications and Challenges of Composite Materials** - CFRP is widely used in aerospace, automotive, shipbuilding, and civil engineering fields, and is highly favored for its high strength and lightweight characteristics. - However, the complex manufacturing process may lead to various defects (such as delamination, cracks, foreign matter inclusions, etc.), which will affect the structural integrity and performance. 2. **Importance of Non - Destructive Evaluation (NDE)** - NDE techniques (such as radiography, thermography, electromagnetic methods, and ultrasonic testing) are used to inspect components without compromising their integrity. - Ultrasonic testing (UT) has become the standard method for aerospace composite material inspection because it can detect multiple volume defects and is relatively safe and easy to implement. 3. **Limitations of Traditional Defect Detection** - Traditional defect detection methods rely on a large amount of manually labeled data, which is not only time - consuming and labor - intensive, but also difficult to obtain sufficient labeled data. - Manual interpretation of results has the risks of low efficiency and being error - prone. ### The Method Proposed in the Paper To overcome the above problems, this paper proposes a new method based on self - supervised learning (SSL), combined with one - dimensional convolutional neural network (1D CNN), to achieve three - dimensional defect segmentation without labeled data. The main features of this method include: - **No Need for Labeled Data**: Through self - supervised learning, the model can learn feature representations from unlabeled clean samples, avoiding the dependence on a large amount of labeled data. - **Minimal Pre - processing Requirements**: There is no need to remove geometric features or apply time - compensation gain (TCG), simplifying the data pre - processing steps. - **Geometric Feature Preservation**: The geometric features in the original data are preserved, enhancing the generalization ability of the model. - **Anomaly Detection Instead of Classification**: The problem is redefined as anomaly detection instead of classification of specific defect categories, thereby improving the adaptability to new types of defects. ### Specific Application of the Method This method is implemented through the following steps: 1. **Data Acquisition**: Use an automated robotic phased - array ultrasonic system to obtain ultrasonic data of CFRP samples. 2. **Pre - training Task**: Use clean samples for pre - training to predict the distribution of the next value. 3. **Inference Stage**: Apply the pre - trained 1D CNN to each 64 - long input sequence, predict the distribution in the subsequent B - scans, and compare it with the actual measurement results to identify abnormal areas. ### Experimental Results The experimental results show that this method can accurately detect all defects, and the in - plane and thickness - direction positioning errors are 0.37 mm and 0.26 mm respectively, and the average absolute error of defect size measurement is 1.41 mm. In addition, the segmentation masks generated by this method can be used to create digital twins of components, further supporting subsequent analysis tasks. In conclusion, this paper proposes an innovative self - supervised learning method, which solves the problem of relying on a large amount of labeled data in traditional defect detection and improves the detection efficiency and accuracy.

3-DUSSS: 3-Dimensional Ultrasonic Self Supervised Segmentation

Advancing Carbon Fiber Composite Inspection: Deep Learning-Enabled Defect Localization and Sizing via 3-D U-Net Segmentation of Ultrasonic Data

A Novel 3D Evaluation Method for Surface Defects Using Broadband Laser-Generated Rayleigh Waves with Wavenumber Analysis.

Automatic three-dimensional reconstruction of subsurface defects by segmenting ultrasonic point cloud

Unsupervised machine learning for flaw detection in automated ultrasonic testing of carbon fibre reinforced plastic composites

Three-Dimensional Residual Neural Architecture Search for Ultrasonic Defect Detection

3-Dimensional residual neural architecture search for ultrasonic defect detection

A study of machine learning object detection performance for phased array ultrasonic testing of carbon fibre reinforced plastics

Automated Defect Recognition for Additive Manufactured Parts Using Machine Perception and Visual Saliency

Fully Noncontact Hybrid NDT for 3D Defect Reconstruction Using SAFT Algorithm and 2D Apodization Window

Stable 3D Deep Convolutional Autoencoder Method for Ultrasonic Testing of Defects in Polymer Composites

Identification of Location and Geometry of Invisible Internal Defects in Structures using Deep Learning and Surface Deformation Field

Visual Detection of Surface Defects Based on Self-Feature Comparison in Robot 3-D Printing

Machine learning based segmentation of delamination patterns from sparse ultrasound data of barely visible impact damage in composites

Manifold learning and segmentation for ultrasonic inspection of defects in polymer composites

Spatial and Temporal Deep Learning in Air-Coupled Ultrasonic Testing for Enabling NDE 4.0

Automatic flaw detection in sectoral scans using machine learning

Automated ultrasonic testing for near-surface flaws in CFRP

Head-to-tail and heteroleptic pentanuclear circular helicates.

Depth extraction of 3D defects on curved surfaces with multi-line lasers

Weld defect segmentation and feature extraction from the acquired phased array scan images