Abstract
Corrosion in reinforced concrete (RC) structures is a major durability issue that requires attention in terms of monitoring, in order to assess the degraded condition and reduce financial costs for maintenance and repair. The acoustic emission (AE) technique has been found to be useful to monitor damage due to steel corrosion in RC. However, further development of monitoring protocols is still necessary towards on-site application. In this paper, a hierarchical clustering algorithm based on cross-correlation is developed and applied to automatically distinguish damage sources during the corrosion process. The algorithm is verified on dummy samples and corroding RC prisms. It is able to distinguish two clusters of which the first one containing AE signals due to corrosion, absorption, hydration, and micro-cracking, and the second one AE signals due to macro-cracking. Electromagnetic interference can be distinguished as a third cluster and filtered subsequently. Due to overlapping characteristics, further differentiation of the first cluster is not possible. Afterwards, the algorithm is scaled up to two sets of RC beams, one set with a uniform corrosion zone, and the other set with a local corrosion zone. In addition, on this sample scale, the algorithm is able to successfully differentiate macro-cracking from corrosion and micro-cracking. It can therefore serve as an additional tool to assess the extent of corrosion-induced damage.
Highlights
Steel corrosion in reinforced concrete (RC) structures causes large economic losses and jeopardizes the structural safety
This paper aims to scale up the clustering algorithm to automatically differentiate acoustic emission (AE) signals recorded during the accelerated corrosion process of RC prisms
The main objective of the AE clustering is to distinguish between the corrosion process at the anode, and the concrete cracking
Summary
Steel corrosion in reinforced concrete (RC) structures causes large economic losses and jeopardizes the structural safety. The process starts internally which means that damage can be observed by visual inspection after a certain time period when the damage has progressed to the surface Electrochemical techniques such as half-cell potential, galvanostatic pulse method, and linear polarisation resistance, allow focusing on the probability, the rate, and location of ongoing corrosion processes from an early stage [1]. Clustering or unsupervised learning tries to group a data set into a number of clusters, usually based on a distance measure. It can find an internal structure in the data that was not known beforehand. Both clustering and classification can be performed based on AE parameters or based on the entire waveform
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
