Pattern recognition enabled acoustic emission signatures for crack characterization during damage progression in large concrete structures

Abstract

The present study focuses on the investigations on technique for assessing damage progression and localization in concrete structure using acoustic emission (AE) technique. Damage is introduced in a girder-deck system of reinforced concrete (RC) bridge by monotonically applied load in terms of strain in reinforcement, at defined intervals. AE signals emitted at different damage stages are recorded to detect crack initiation and progression. Acoustic parameters such as energy, signal strength are considered to examine their efficacy in identifying the initiation and propagation of crack in concrete structures. Few of the frequency parameters of AE signal are identified to be very effective and able to clearly differentiate between the initiation of new crack and progression of existing crack(s) in concrete. AE waveform characteristics, as identified in the present study, can be used to classify the damage progression of in-service concrete structures. Further, unsupervised- and supervised- pattern recognition algorithms are used to classify the AE signal dataset recorded at different damage stages. To validate the effectiveness of feature selection and support vector machine (SVM) classifier, SVM classified locations of AE events are compared with the experimentally observed damage pattern at different damage stages. It is found that, SVM can effectively be able to classify two types of AE sources appropriately, enabling the potential application of AE technique for initiation and its progression, and localization of damage in critical in-service structures such as bridges.