| dc.description.abstract | Corrosion in reinforced concrete (RC) structures is one of the foremost and most severe causes of early degradation. This deterioration leads to structural failure, impacting human life as well as the environment. Furthermore, corrosion-related damage necessitates frequent and costly repair and maintenance work, imposing a financial burden on society. Therefore, it is vital to estimate the remaining capacity of corroded reinforced concrete (CRC) structures to perform necessary preventive maintenance work before structural collapse or required expansive rehabilitation techniques. This study employed artificial neural network (ANN), Gaussian process regressor, and linear regression-based machine learning (ML) models to develop a more dependable and precise model for estimating the residual flexural capacity (RFC) of CRC beams. Levenberg–Marquardt, scaled conjugate gradient, and Bayesian regularization training techniques were used to train the ANN models. The performance and results of the developed ANN models were evaluated and compared with a design guideline (ACI-318), analytical model (one), and empirical models (eight). The results demonstrated that the developed ANN model with four neurons in the hidden layer (DANN-4) was trained using the Bayesian regularization algorithm with 80% of the dataset for training and the remaining 20% for testing, outperforming other developed ML models, design guidelines, analytical models, and empirical models. The comparative study indicated that the developed DANN-4 model had the highest correlation between the actual and predicted values with the lowest errors, demonstrating the efficacy of the developed ANN model in predicting the RFC of CRC beams compared to existing ML models, design guidelines, analytical models, and empirical models. The developed ANN-based model can be used by structural engineers, researchers, and rehabilitation industry experts to estimate the RFC of corrosion-damaged RC beams. The research on estimating the RFC of CRC beams through ANN modeling offers a significant advancement in structural engineering practice. By employing ANN models trained with various algorithms and comparing their performance with existing design guideline, analytical models, and empirical approaches, the study demonstrates a more reliable and accurate method for assessing the structural integrity of CRC beams. This research holds practical implications for infrastructure maintenance and management, allowing engineers to better prioritize preventive maintenance efforts and allocate resources efficiently. With the ability to predict the RFC of CRC beams more precisely, maintenance teams can mitigate the risk of structural failure, reduce costly emergency repairs, and extend the service life of RC structures. | |
