Artificial neural network analyses of AE data during lung-term corrosion monitoring of a post-tensioned concrete beam

Acoustic emission (AE) technique is well suited for real-time control and detection of active defects within a structure. We used AE to monitor a post-tensioned concrete beam for about seven months, undergoing stress corrosion cracking (SCC) assisted by hydrogen embrittlement, activated by ammonium thiocyanate solution during a laboratory corrosion test. By multivariate analysis, three stages of damage mechanisms were identified: activation, propagation andrupture. An artificial neural self-organizing map (SOM) analysis was used to identify the relationship between the AE variables and to classify AE events. This methodology has in fact proved particularly powerful in identifying the evolution and extent of damage of the monitored post-tensioned concrete beam with the use of exemplified topological maps. The SOM analysis made it possible to correlate each AE stage, with unambiguous significant variables, to a specific degradation phase.