Automated Defect Detection in Non-Destructive Testing Systems

Automated defect detection in non-destructive testing (NDT) systems has emerged as a transformative approach to inspect critical components without impairing their serviceability. By combining advanced imaging techniques—such as magnetic particle inspection (MPI), ultrasonic and acoustic microscopy—with machine learning algorithms, these systems identify cracks, inclusions and other anomalies with high speed and repeatability. Key innovations include the use of convolutional neural networks to discern subtle flaw patterns in noisy backgrounds, attention mechanisms to enhance feature extraction at multiple scales and data-driven augmentation to overcome limited sample sizes. Integration of robotics and bespoke magnetisation devices permits in situ scanning of complex geometries, while post-processing methods refine defect localisation and measurement. Taken together, these developments promise greater reliability, reduced operator fatigue and streamlined workflows across industries ranging from aerospace and rail to oil and gas.

Research from Nature Portfolio

Recent studies have advanced automated crack detection for railway components by applying instance segmentation to fluorescent magnetic particle images of rivets. A decentralised labelling approach and a Gaussian-weighted correction step were introduced to improve recall in densely cracked areas. An efficient channel–spatial attention module further enhanced multi-scale feature learning, while a multi-task feature-learning framework addressed uninstantiated crack regions by leveraging both semantic and spatial cues. Complementing the software, a universal non-contact magnetisation device was developed to accommodate diverse rivet shapes. The resulting system achieved a recall rate of 86.4% and a mean average precision of 90.3%, demonstrating robust performance in real-world rolling-stock inspection conditions.

Automated Defect Detection in Non-Destructive Testing Systems publication trend

The graph below shows the total number of articles in automated defect detection in non-destructive testing systems across all publications each year (not limited to Nature Index journals).

Technical terms

Non-Destructive Testing (NDT): Inspection of materials or components without altering their future usefulness.

Magnetic Particle Inspection (MPI): A method that uses magnetic fields and ferrous particles to reveal surface and near-surface defects.

Convolutional Neural Network (CNN): A class of deep learning models designed to automatically extract hierarchical features from images.

Instance Segmentation: A computer vision task that detects and delineates each object of interest within an image at the pixel level.

Attention Mechanism: A neural network component that adaptively weights feature maps to focus on the most informative regions.

References

1. Automated crack detection of train rivets using fluorescent magnetic particle inspection and instance segmentation. Scientific Reports (2024).
2. Intelligent Detection Method of Forgings Defects Detection Based on Improved EfficientNet and Memetic Algorithm. IEEE Access (2022).
3. Automatic Defect Identification Method for Magnetic Particle Inspection of Bearing Rings Based on Visual Characteristics and High-Level Features. Applied Sciences (2022).
4. Industrial Application of AI-Based Assistive Magnetic Particle Inspection. Applied Sciences (2024).