An automated long range ultrasonic rail flaw detection system based on the support vector machine algorithm

This paper discusses a flaw detection system that uses automated long range ultrasonic detectors modeled on a successful pattern recognition method. The procedure has three stages: 1. a variety of feature extraction methods to better represent the signals; 2. feature ranking based on feature selection; and 3. classification using a kernel-based support vector machine (SVM). The results of analysis indicate that the proposed algorithm is successful at detecting flaws.