Quantitative Investigation of Acoustic Emission Waveform Parameters from Crack Opening in a Rail Section Using Clustering Algorithms and Advanced Signal Processing.

Abstract

Acoustic emission (AE) is an emerging technology for real-time non-destructive testing of structures. While research on a simulated AE source in rail and testing on rail material using small beam samples have been conducted, a study is required in lab environment to investigate AE waveform characteristics generated by crack in rail. In this paper, a three-point bending test is conducted on an actual rail section of 1500 mm with transverse damage of 38% head area to simulate AE source due to crack opening in the rail. AE signals are recorded for three different loads. For data analysis, unsupervised machine learning algorithms such as k-means, fuzzy-C mean and gaussian mixture model are used to cluster and filter out usable signals from the whole dataset corrupted by noisy signals from various sources. k-mean with principal component was observed to be best technique based on silhouette score. The frequency and amplitude of waveform have been discussed in relation to load and crack opening displacement. This study establishes a baseline for linking load, crack opening, and AE wave characteristics. This work can ultimately aid in the development of robust denoising, and damage analysis algorithms based on the frequency content and dispersion of the AE waveform.