Magnetic flux leakage (MFL) detection technology provides an effective method to conduct high-speed detection of the damage suffered by rail surface. With regard to high-speed detection, there is frequently a complex noise contained in the magnetic signal of railway leakage, which is similar to the amplitude of defect and the overlaps of frequency spectrum. In this paper, an improved adaptive filtering method is proposed to solve the problem caused by filtering the MFL signal on the rail top surface. Through the characteristics of distribution shown by defects on the top surface of the railway and those of the data collected by the rail top array sensor, this method is applied to construct a virtual channel containing almost only interference signals but no defects. Then, in combination with the adaptive filtering algorithm, the virtual channel signal is taken as the reference input of the adaptive canceller, each single channel MFL signal is taken as the original input of the adaptive canceller, and the filtered MFL signal is taken as the output. Then, the MFL signal of rail top is collected by the train at the speed of 30km / h on the manual calibration line. According to the experimental results, the noise intensity of MFL signal is reduced by up to 81.44%. In addition, the filtering method is adopted to process MFL signals with different directions and varying detection speed. As indicated by the results, the noise intensity of MFL signal is reduced by more than 74%.
This study addresses the critical challenge of insufficient classification accuracy for different defect signals in rail magnetic flux leakage (MFL) detection by proposing an enhanced intelligent classification framework based on particle swarm optimized radial basis function neural network (PSO-RBF). Three key innovations drive this research: (1) A dynamic PSO algorithm incorporating adaptive learning factors and nonlinear inertia weight for precise RBF parameter optimization; (2) A hierarchical feature processing strategy combining mutual information selection with correlation-based dimensionality reduction; (3) Adaptive model architecture adjustment for small-sample scenarios. Experimental validation shows breakthrough performance: 87.5% accuracy on artificial defects (17.5% absolute improvement over conventional RBF), with macro-F1 = 0.817 and MCC = 0.733. For real-world limited samples (100 sets), adaptive optimization achieved 80% accuracy while boosting minority class (“spalling”) F1-score by 0.25 with 50% false alarm reduction. The optimized PSO-RBF demonstrates superior capability in extracting MFL signal patterns, particularly for discriminating abrasions, spalling, indentations, and shelling defects, setting a new benchmark for industrial rail inspection.
The rapid development of the railway industry has brought convenience to people’s lives. However, with the high speed, high frequency and heavy load characteristics of rail use, the safety of rail is seriously threatened. In this paper, a magnetic flux leakage testing (MFL) detection technology of rail based on a double-track flaw detection vehicle is introduced in detail, which can effectively detect the damage of rail top surface, which is the blind area of ultrasonic detection. The magnetic dipole model is used to analyze that the leakage magnetic field in the direction of Bx and Bz above the damage is related to the depth and width of the damage. The relationship between the depth of the damage and the leakage magnetic field is quantitatively studied for the damage with fixed width but varying depth. The finite element simulation tool is used to model and simulate the damage at different depths. After analyzing the different characteristic values, it is found that the peak value of magnetic leakage signal has a certain correlation with the depth of damage, and the natural logarithm function is fitted out—VBx = 0.1451ln(b) + 0.2705, VBz = 2.7787ln(b) + 0.0087. In order to verify the prediction function of the injury depth fitted by the simulation data, the human injury with different depths was processed and the dual-track flaw detector was used to carry out the experiment of high-speed detection environment. The peak-to-peak fitting of the magnetic leakage signals in the direction of Bx and Bz of the experimental results shows that the peak-to-peak variation rule is roughly in line with the natural logarithm function in the simulation. The correlation between the fitting results of the experimental data and the simulation fitting function is analyzed using the Pearson coefficient. The Pearson coefficient in the direction of Bx is ρx = 0.91386. The Pearson coefficient of the Bz direction is ρz = 0.98597, the peak-to-peak value of Bx and Bz direction is positively correlated with the depth of damage and the fitting effect of the Bz direction is better than that of the Bx direction.
The magnetic flux leakage (MFL) detection technique is applied to the detection of weld defects, such as cracks and pores. As the weld has a distinct structure, there are differences in the magnetization path and leakage field intensity under different magnetization directions. According to surveys, a suitable magnetization direction can significantly enhance detection rates of small-sized defects by stimulating a stronger field signal of the defect leakage. In this study, ANSYS finite element simulation software is used to calculate the weld defect leakage field based on the quantitative analysis. Specifically, the leakage field component strengths of circular hole defects and longitudinal rectangular groove defects are compared when the magnetization direction is perpendicular or parallel to the weld. Furthermore, the characteristic rules of the defect leakage field and its components under any magnetization direction are discussed, and a weld MFL detection platform is set up for validation. According to the experimental results, the amplitude of the magnetic leakage signal during vertical magnetization of circular hole defects is only 18.6% of that during parallel magnetization. Similarly, the amplitude of the magnetic leakage signal during parallel magnetization of longitudinal crack-type defects is only 9.2%~29.3% of that during vertical magnetization.
This study addresses the critical challenge of insufficient classification accuracy for different defect signals in rail magnetic flux leakage (MFL) detection by proposing an intelligent classification method based on particle swarm optimized radial basis function neural network (PSO-RBF). The research systematically develops a comprehensive solution through three key aspects: experimental data acquisition, model architecture optimization, and performance validation. Initially, 400 sets of MFL signal samples were collected through artificial defect experiments, with six crucial MFL signal features selected as inputs for the 6-15-4 structured RBF network. To ad-dress the nonlinear mapping relationship between MFL signal characteristics and defect categories, an enhanced PSO algorithm incorporating dynamic learning factors and nonlinear inertia weight was implemented for RBF parameter optimization. Experimental results demonstrate significant performance improvements, with classification accuracy reaching 87.5% on artificial defect datasets (17.5 percentage points higher than conventional RBF), macro-F1 score increasing to 0.817, and MCC coefficient improving to 0.733. For actual defect detection with limited samples (100 sets), adaptive model adjustments including hidden layer reduction (10 neurons) and key feature weighting achieved 80% accuracy, while particularly improving minority class performance ("spalling" F1-score increased by 0.25 with 50% false alarm reduction). The proposed PSO-RBF model effectively enhances MFL signal feature extraction capability through precise optimization of center positions and kernel width parameters, demonstrating superior discriminative ability for various defects (abrasions, spalling, indentations, shelling) and providing an effective intelligent solution for rail defect classification in MFL detection.