Comparative Analysis of Reinforced Concrete Beam Behaviour: Conventional Model vs. Artificial Neural Network...

Area of Science:

• Structural Engineering
• Computational Intelligence
• Materials Science

Background:

• Reinforced concrete beams are critical structural elements.
• Current design codes (ACI) provide conventional methods for analysis.
• Predicting the behavior of reinforced concrete beams, especially without stirrups, presents challenges.

Purpose of the Study:

• To compare the predictive accuracy of Artificial Neural Networks (ANNs) against conventional ACI code models for reinforced concrete beams without stirrups.
• To develop a knowledge-based structural analysis model using ANNs.
• To evaluate the performance of ANNs in predicting first crack load, flexural strength, and shear strength.

Main Methods:

• A dataset of 110 reinforced concrete beams without stirrups was compiled.
• A Multilayer Backpropagation Neural Network was trained using MATLAB.
• ANN predictions were compared with results from the conventional ACI building code model.

Main Results:

• The ANN model achieved high coefficients of determination: 0.9404 for first cracking load, 0.9756 for flexural strength, and 0.9787 for shear strength.
• ANN model results closely aligned with the conventional model.
• Linear regression analysis confirmed the strong correlation between ANN predictions and experimental/conventional results.

Conclusions:

• Artificial Neural Networks demonstrate significant potential for accurately predicting the structural behavior of reinforced concrete beams.
• The ANN model provides a reliable and precise alternative for structural analysis, complementing conventional design codes.
• This research establishes a robust knowledge-based model for enhanced structural analysis and design.