Abstract:Defect depth is an essential indicator in magnetic flux leakage (MFL) detection and estimation. The quantification errors for defect depth are closely related to length and width errors, and this feature has always been used to support the operator's judgment in defect identification. However, the existing defect quantification algorithms based on shallow and deep neural networks only employed simple general network structures inspired by the field of artificial intelligence; consequently, these network structures lack the support of physical concepts and result in large quantification errors regarding defect size, especially depth. In this article, to describe and integrate the above theory into a deep neural network, we propose a physics-informed doubly fed cross-residual network (DfedResNet) suitable for MFL defect detection based on deep learning. Physics-based MFL defect quantification theory is studied and integrated into loss functions during the neural network training. DfedResNet quantifies defects in MFL data and automatically extracts deep features of defects. The experimental results show that it effectively achieves high-precision quantification of defect length, width, and depth simultaneously, especially defect depth. Moreover, it considers data from all three dimensions during network training, and use the originally measured magnetic signal data in place of recognized images to avoid defect information loss and further improve the quantification accuracy. The deep DfedResNet model proposed in this article reduces defect length and width quantification errors to within 0.3 mm and defect depth quantification errors to within 0.4% t. In addition, compared with other network structures and traditional algorithms, DfedResNet improves defect quantification accuracy by 1–2 orders of magnitude and thus achieves a high quantification performance.

Three-Dimensional Defect Inversion from Magnetic Flux Leakage Signals Using Iterative Neural Network

Reconstruction Of 3-D Defect Profiles From Mfl Signals Using Radial Wavelet Basis Function Neural Network

Quantitative Interpretation For The Magnetic Flux Leakage Testing Data Based On Neural Network

Three-dimensional Defect Reconstruction from Magnetic Flux Leakage Signals in Pipeline Inspection Based on a Dynamic Taboo Search Procedure

Equivalent MFL Model of Pipelines for 3-D Defect Reconstruction Using Simulated Annealing Inversion Procedure

A Pipeline Defect Inversion Method With Erratic MFL Signals Based on Cascading Abstract Features

An Iterative Stacking Method for Pipeline Defect Inversion With Complex MFL Signals

Reconstruction of Arbitrary Defect Profiles from Three-Axial Mfl Signals Based on Metaheuristic Optimization Method

A Novel Defect Quantification Method Utilizing Multi-Sensor Magnetic Flux Leakage Signal Fusion

Magnetic Anomaly Inversion Based on Convolutional Neural Network

Fast Reconstruction of 3-D Defect Profile from MFL Signals Using Key Physics-Based Parameters and SVM

Injurious or Noninjurious Defect Identification From MFL Images in Pipeline Inspection Using Convolutional Neural Network

Deep Learning for 3-D Magnetic Inversion

Magnetic flux leakage defect size estimation method based on physics-informed neural network

Experiment and simulation study of 3D magnetic field sensing for defect characterisation

A genetic algorithm with tabu search method for 3 dimensional detect profile reconstruction using MFL measurement

Experiment and Simulation Study of 3D Magnetic Field Sensing for Magnetic Flux Leakage Defect Characterisation

Development of a Physics-Informed Doubly Fed Cross-Residual Deep Neural Network for High-Precision Magnetic Flux Leakage Defect Size Estimation

A Streamlined Neural Network Architecture for Magnetic Data Inversion

Deep Defects Evaluation of Conductive Materials Based on Eddy Current Nondestructive Testing with GMR sensor

An Estimation Method of Defect Size From MFL Image Using Visual Transformation Convolutional Neural Network