Rail bottom rust localisation imaging monitoring using ultrasonic-guided wave combining with variational modal decomposition and principal component analysis

This paper presents a high-precision ultrasonic guided wave imaging monitoring technique for rust monitoring at the bottom of rails, which mainly utilizes piezoelectric sensing arrays combined with an ellipsoidal probability-weighted imaging method. A combination of Variational Mode Decomposition (VMD) and Principal Component Analysis (PCA) was utilized to enhance imaging monitoring accuracy. The feasibility and effectiveness of the method were validated through the establishment of simulation models with varying rust depths and the construction of experimental systems. The region with the highest probability value in the monitoring image is extracted, and the relative distance between the predicted and actual rust locations（RD_pa） is calculated to assess the accuracy of damage localization. The results of the simulation and experimental studies show that the RD_pa error values after VMD-PCA treatment are 18mm and 20mm, respectively, and these values exhibit significantly smaller numerical errors compared to those of the untreated, VMD-only, and PCA-only treatments. The probability value extracted from the predicted position image, processed using the VMD-PCA feature enhancement algorithm, increases concurrently with the depth of the rust pits. This increase reflects the severity of rust damage to the rail bottom, facilitating a quantitative assessment of the rust extent on the rail.