Evaluation of residual flexural strength of corroded reinforced concrete beams using convolutional long short-term memory neural networks

The artificial corrosion process is distinct from the natural environment; thus, posing a challenge in the evaluation of the structural performance of reinforced concrete (RC) structures in reality. This study aims at bridging this gap by addressing the effect of the natural corrosion process on the residual flexural strength of beam structures in many aging RC buildings. In particular, the field inspection data on RC beams are collected from various aging RC structures in Ho Chi Minh city. From this surveyed field data, the corrosion rate at the beginning of corrosion propagation is interpreted from the amount of diameter reduction of longitudinal rebar; the corrosion activity index is determined using a selected empirical model; the remaining moment-carrying capacity is calculated by integrating the surveyed data into empirical models. The reliability of the remaining flexural capacity proposed by this study is assessed by comparison with published experimental results. To facilitate automation in the assessment of the service life of RC beams, a robust soft computing model to predict the remaining moment-carrying capacity of corroded reinforced concrete (CRC) beams will be developed. To cope with the spatial and temporal variations of the natural corrosion process, a convolutional long short-term memory neural network (CLNN) is adopted thereof. In essence, the research gap between research outputs from the laboratory accelerated corrosion process to the actual project under the effect of the natural corrosion process is bridged; hence, a reliable and effective framework for evaluating the effect of the natural corrosion process on the moment-carrying capacity of RC beams is provided for engineering practice.