Ultrasonic guided wave-based switch rail monitoring using independent component analysis

Switch rails are indispensable components of high speed railway systems, which have stringent nondestructive testing requirements owing to the severe operating conditions. In this article, an ultrasonic guided wave method is proposed for defect detection and localization using independent component analysis (ICA). The temperature effect is included in the data matrix by a random selection of the signals measured at different temperatures. A damage index named the average standard Euclidian distance (ASED) is used to evaluate the deviations of the test signals from the baseline signals in the feature space consisting of the independent components for the defect detection. Once the defect existence is found, defect localization is conducted by another ICA-based decomposition of a new data matrix with additional test signals for the same defect. Independent components whose coefficient vectors show a high correlation with the standard step change vector are chosen to construct the ICA-based residual signal. Then the time instance and location of the defect is determined by observing the first very high peak occurring in the residual signals. A detectability index for defect location (DIDL) is proposed. Experimental validations are performed for the defects on the foot and web of a switch rail. The results of the ASED curves clearly indicate the existence of artificial defects, and the ICA-based residual signals show the location of the defects. The proposed method is found to be superior to conventional methods such as simple baseline subtraction and optimal baseline subtraction regarding the DIDL.