Prototype of a rail wear measuring system using digital image processing

Abstract

The rail is one of the most important components of railways and it serves directly as a running track of the rolling stock in this transportation system, suffering a lot of wear because of this fact. Most of the manual equipment for measuring this problem does not allow continuous data acquisition, requiring point-to-point measurements, besides it usually have a high price. As an alternative to collecting data more efficiently and detecting wear on rails, it was proposed to build a equipment prototype for continuous image collection and, through the processing of these images using computational techniques, to extract the desired measure of the wear. It was developed a mobile rail track cart for capturing images of the rail with the profile demarked by a laser bean. All the photos of the bench were processed in the laboratory using the implemented processing routine, being possible to extract the values of the vertical wear measurement and compare it with the measurements made with a template. The results showed a maximum absolute mean error of 0,80 milimeters (comparing the average of the measurements with the measurement of the manual template) and maximum standard deviation of 0.9794mm for vertical wear, in addition to the values of absolute mean error of 0.43mm and maximum standard deviation among the experiments of 0.6046mm for lateral wear. It was possible to verify that the built prototype and the processing algorithm were able to acquire the data efficiently and portray the wear.