12 - Artificial neural network analysis of acoustic emission data during longtime corrosion monitoring of post-tensioned concrete structures

Abstract Damage evaluation by acoustic emission (AE) long-term monitoring of reinforced concrete structures requires particular attention and the use of proper procedure and analytical tools. \AE\ source localization, source identification and noise removal from a large size database could be a very arduous task. A multistep analysis procedure can be used with the aim of identifying homogeneous clusters of \AE\ signals to be related to specific damage conditions (e.g. tensile or shear cracks, micro-cracking or macro-cracking). The procedure proposed in this chapter includes the following steps: denoising clustering, univariate analysis, multivariate analysis and damage analysis. From the multivariate analysis, three distinct stages of damage mechanisms were identified: activation, propagation and rupture. However, the damage analysis performed by means of an artificial neural network analysis allowed the correlation of each acoustic stage with unambiguous significant variables to a specific degradation mechanism.