Fracture characterisation and performance evaluation of corroded RC members by AE-based data analysis
Steel reinforcement corrosion has been regarded as one of the major causes of reinforced concrete (RC) structures failing prematurely, posing a serious structural durability problem worldwide. Detailed assessment of corrosion-induced damage and its effects on RC structures is critical for sustaining structural reliability and safety. This study develops and examines the feasibility of acoustic emission (AE) monitoring and data analysis methodologies to characterise corrosion-induced damage in RC members, followed by an evaluation of the effect of corrosion on load behaviour. Experimental investigations were conducted on a series of specimens of different configurations, namely concrete cubes with steel bars for pull-out tests and RC beams of different dimensions to be subjected to static and cyclic loading regimes. Focusing on developing evaluation methods based on AE monitoring and data analysis, a summary of work completed, and the associated findings are given as follows. Characterisation of the concrete cracking using parametric and waveform analysis was conducted to investigate the effect of corrosion on steel-concrete bond behaviour in the pull-out tests of concrete cubes. It was found that a small amount of corrosion (approximately 6%) could slightly increase the bond strength as a result of the rust expansion and reactionary confinement of concrete. Corrosion was also found to be able to mitigate the damage caused by cyclic loading. AE signal analysis indicates that the concrete cracking mode during the steel-concrete de-bonding process has changed as a result of steel corrosion. Characterisation of load behaviour and failure mode of corroded RC beams was conducted by flexural load tests aided by AE monitoring and digital image correlation (DIC). The DIC strain mapping results and AE signal features revealed that corrosion has an influence on the concrete cracking mechanism of the beam specimens. Corrosion has also altered the failure mode of a shear-critical beam specimen series to flexure owing to the change of steel-concrete bond behaviour. Numerical simulation of AE wave front propagation in RC media and tomographic evaluation of internal damage was implemented on one group of RC beam specimens tested in this study. The numerical model of the specimens was discretised using finite-difference grid meshing, and the different acoustic properties of steel and concrete were defined. On this basis, simulation of AE wave front propagation considering concrete cover cracking and steel rust layer formation was carried out using the fast-marching method. The effect of corrosion-induced damage on the AE rays was studied by examining non-linear ray tracing in the simulation. A tomographic reconstruction approach that solved by the quasi-Newton method provided a potential way to quantitatively evaluate the internal damage of RC beams using AE monitoring data. A novel method was developed for assessing the corrosion level in RC beams using a data-driven approach. Normalization of AE data was applied using principal component analysis to minimise variations in AE signal features caused by differences in the geometrical and material properties of RC beams as well as in the AE monitoring instrumentation setup. The machine learning models, including k-nearest neighbours (KNN) and support vector machines (SVM), were trained using the normalised AE features. The trained KNN models were found effective at predicting the corrosion level in RC beams using the secondary AE signals as input, which could be acquired from the cyclic loading of beams. Key words: Steel Corrosion, Concrete cracking, Steel-Concrete Bond, Reinforced Concrete Beam, Load Behaviour, Acoustic Emission, Digital Image Correlation, Tomographic Reconstruction, Data-driven.