USE OF CLUSTER ANALYSIS OF ACOUSTIC EMISSION SIGNALS IN EVALUATING DAMAGE SEVERITY IN CONCRETE STRUCTURES

摘要

AE technique gained increasing interest in the last decade in the field of civil engineering as a monitoring methodology and as an assessment tool for safety and reliability evaluation of reinforced concrete structures, historic and masonry buildings. There are several established statistical methods (Z, RA, b and Ib value), which can be used in analyzing AE data to evaluate damage status of a structure subjected to a particular loading condition. Artificial neural networks (ANN) have recently been applied as a tool to reduce data redundancy and to optimize feature set of AE signals. Cluster analysis was generally used to separate a set of parameters into several classes reflecting dataset internal structure. In this paper such analytical procedure was applied in evaluating acoustic emission data obtained during 4-point bending tests on concrete beams under cycling and constant load condition and at increasing loads. Two kinds of unsupervised clustering methods were used: the principal component analysis (PCA) and the self-organized map (Kohonen map). Combining both methods, it has been possible to quantify the damage severity and to identify the evolution of the damage itself during the test.