Three-dimensional acoustic emission source localisation in concrete based on sparse least-squares support vector regression

Abstract

In order to improve the accuracy of concrete damage localisation based on acoustic emission (AE) monitoring, a multi-output sparse least-squares support vector regression (S-LS-SVR) method is attempted for AE source localisation in concrete. The AE events are produced by pencil lead breaks and the response wave is received by piezoelectric sensors. A Newton iterative method, an improved exhaustive method and two S-LS-SVR approaches (S-LS-SVR(A) and S-LS-SVR(B)) are used to locate the AE source, then the positioning accuracies of the methods in the three coordinate directions are compared and analysed. The results show that the accuracy of AE source localisation using the S-LS-SVR(B) model is higher than that of the other methods. The accuracy of the S-LS-SVR model using the time difference of arrival (TDOA) and the sequential number of sensors that arrive successively as input parameters is higher than that of the other AE signal combination trialled as the input. This shows that the S-LS-SVR(B) model is better than the S-LS-SVR(A) model. The intelligent S-LS-SVR(B)-based localisation method provides a basis for application in actual damage detection.