Modeling for detecting weld defects based on magneto-optical imaging.

Area of Science:

• Materials Science
• Non-Destructive Testing
• Electromagnetism

Background:

• Ferromagnetic weldments require reliable nondestructive testing (NDT) methods.
• Magneto-optical (MO) imaging principles, including the Faraday effect and magnetic domain theory, are crucial for understanding material properties.
• Existing NDT methods may have limitations in detecting subtle weld defects.

Purpose of the Study:

• To propose and validate a magneto-optical (MO) imaging technique for nondestructive testing of ferromagnetic weldments.
• To investigate the relationship between MO imaging parameters and weld defect detection.
• To develop a defect classification model for weldments using MO imaging data.

Main Methods:

• Analysis of the Faraday magneto-optical effect, magnetic domain theory, and magnetic hysteresis loops.
• Development of magnetic domain distribution models for various welding states under external magnetic fields and remanence fields.
• Experimental application of MO imaging to high-strength steel (HSS) weldments to detect cracks, sag, and incomplete penetration.
• Implementation of a defect classification model using principal component analysis and an error backpropagation neural network.

Main Results:

• The MO imaging system's performance was correlated with excitation voltages and spacer plate thickness.
• The proposed NDT method demonstrated effectiveness in detecting incomplete penetration defects.
• Experimental results confirmed the capability of MO imaging to identify cracks, sag, and incomplete penetration in HSS weldments.
• The developed classification model achieved a high accuracy of 92.8% for weld defect identification.

Conclusions:

• Magneto-optical (MO) imaging is a viable and effective nondestructive testing (NDT) method for ferromagnetic weldments.
• The technique can accurately detect critical weld defects such as cracks, sag, and incomplete penetration.
• A machine learning model based on MO imaging data provides a highly accurate approach for weld defect classification.