Abstract
Support vector machine (SVM) is used as the classifier as well as a regressor in the machine learning techniques. However, the application of SVM for acoustic emission (AE) source localization is limited. For the composite materials, the use of guided ultrasonic waves (GUV) helps in identifying the location of defects as they can be transmitted and simultaneously received by the sparse array of piezoelectric sensors. Various methods based on minimizing the error function and Bayesian filters have already been implemented to identify the location of the acoustic emission (AE) source. The current study focuses on identifying the location of impact in the plate using guided waves through a layered support vector machine framework. Under the framework, the carbon-fiber-reinforced polymer (CFRP) plate is divided into several small regions. The SVM classifier is trained for each of these regions, and thus a set of layers containing different SVM classifiers is formed. The training dataset is generated through the centroidal Voronoi tessellations. The test point generated through Markov Chain is passed through the different layers of SVM to identify the region of AE source location. Moreover, the proposed approach is validated using an experimental program that makes use of a CFRP composite panel instrumented with a sparse array of piezoelectric transducers.
