Surface Crack Size Estimation Based on Quantification and Decoupling of Magnetic Flux Leakage (MFL) Signals of Circular Array Sensors

• Published in: IEEE sensors journal Vol. 24; no. 10; pp. 16752 - 16762
• Main Authors: Gong, Wendong, Akbar, Muhammad Firdaus, Jawad, Ghassan Nihad, Zhang, Fengqin
• Format: Journal Article
• Language: English
• Published: New York IEEE 15-05-2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Circular sensor array; Decoupling; Fatigue cracks; Fatigue failure; Finite element analysis; Flaw detection; Flux density; Magnetic flux; Magnetic flux leakage; magnetic flux leakage (MFL) detection; Magnetic sensors; rail cracks; Rails; Sensor arrays; Sensor phenomena and characterization; Sensors; signal features; Surface cracks
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2024.3379401

It is of great importance to improve the detection accuracy of initial fatigue cracks on rail top surfaces to provide an early warning of rail damage. However, it is a challenge to accurately quantify such cracks due to their usual narrow width and various angular variations. This article proposes a quantification method for crack surface dimensions (i.e., angle, width, and length) by decoupling the magnetic flux leakage (MFL) resulting from circular sensor arrays. First, the correlation between the distribution characteristics of the MFL signals and the crack surface size is analyzed by finite element simulation to determine the quantification sequence of the surface crack dimensions. Next, the crack angle is quantified based on the temporal relationship of the MFL circular sensor array signals. Then, the gradient edge of the magnetic leakage flux density modulus is used to quantify the width of the surface crack. Finally, the crack length is quantified by utilizing the number of sensors that detect a valid MFL signal in the circular array. Experimental results show that the maximum quantification accuracy for the crack's angle, width, and length are about 95.0%, 91.2%, and 88.4%, respectively. This indicates that the proposed method can highly improve the quantification accuracy of crack surface dimensions.