Acoustic emission source localisation for structural health monitoring of rail sections based on a deep learning approach

Abstract

An acoustic emission (AE) approach for non-destructive evaluation of structures has been developed over the last two decades. In complex structures, one of the limitations of AE testing is to find the location of the AE source. Time of flight and wave velocity are typically employed to localise AE sources. However, complex rail structures generate multiple wave modes travelling at varying speeds, making localisation difficult. In this paper, the challenge of localisation has been split into two parts: (a) identification of the AE source zone, i.e. head, web or foot, and (b) identification of location along the length of the rail. AE events are simulated using a pencil lead break (PLB) as the source. Three models including an artificial neural network and 1D and 2D convolutional neural networks (CNNs) are trained and tested using AE signals generated by PLB sources. The accuracy of zone identification is reported as 94.79% when using the 2DCNN algorithm. For location classification it is also found that 2DCNN performed best with 73.12%, 79.37% and 67.50% accuracy of localising the AE source along the length in the head, web and foot, respectively. For AE signal generation from actual damage in a rail, a bending test on an inverted damaged rail section was then performed with loads of 100 kN, 150 kN and 200 kN. For all loads, the 2DCNN model resulted in accurate prediction of the zone of the AE source, and it accurately predicted the AE source location along the length for the loads of higher intensity (150 kN, 200 kN). It is envisaged that the deep learning approach presented in this research work will be helpful in developing a real-time monitoring system for rail inspection based on AE.