A Pioneering Approach to Predicting the Shear Strength of RC Beams by Employing Artificial Intelligence Techniques

• Main Authors: Hakim, S. J. S., Mhaya, A. M., Ibrahim, M. H. W., Mohammadhasani, M., Mokhatar, S. N., Paknahad, M., Kamarudin, A. F.
• Format: Article
• Language: English
• Published: 2024
• Subjects: TA401-492 Materials of engineering and construction. Mechanics of materials
• Online Access: http://eprints.uthm.edu.my/11726/1/J17072_f9fef400877e90b6f0272ebcb894f021.pdf; http://eprints.uthm.edu.my/11726/; https://doi.org/10.13189/cea.2024.120118

Shear failures exhibit a brittle nature, often resulting in catastrophic collapse without sufficient advance warning or the capacity to redistribute internal stresses. Consequently, shear failures pose a greater risk and require more attention from structural engineers. It is crucial to incorporate preventive measures in structural design to avoid abrupt shear failures. The work presented in this article attempts to predict the shear strength of reinforced concrete beams as a complex structural engineering problem without the need for extra computational resources by employing the capabilities of Artificial Intelligence (AI) techniques. In recent decades, significant amounts of research have been done on the AI methods such as artificial neural networks (ANNs), fuzzy logic and genetic algorithms to predict the shear strength of RC beams. In this research, adaptive neuro-fuzzy inference system (ANFIS) and ANNs are developed to predict the shear capacity of RC beams. The required data in the form of major factors affecting the shear capacity of RC beams lacking stirrups are compressive strength of concrete, beam depth, effective width, shear span-to-depth ratio, proportion of longitudinal steel and the yield strength of the reinforced longitudinal steel have been considered in this study. Also, in the context of this investigation, a comparison was conducted between the techniques of ANNs and ANFIS. The outcomes demonstrated that both methods exhibited favourable predictive capabilities. Nevertheless, the ANFIS architecture proposed, which incorporates a hybrid learning algorithm, outperformed the multilayer feedforward ANN that utilizes the backpropagation algorithm. The findings indicated that ANFIS is a suitable technique for predicting intricate relationships between input and output parameters, thus making it a valuable tool in predicting the shear strength of RC beams.