Predicting the Lateral Load Carrying Capacity of Reinforced Concrete Rectangular Columns: Gene Expression Programming

Area of Science:

• Structural Engineering
• Artificial Intelligence
• Seismic Analysis

Background:

• Reinforced concrete (RC) columns are critical structural components, especially under seismic loading.
• Accurate prediction of lateral load capacity is essential for seismic design and performance evaluation.
• Existing design codes may have limitations in predicting the complex behavior of RC columns under earthquake conditions.

Purpose of the Study:

• To develop novel artificial intelligence (AI) based gene expression programming (GEP) models for predicting the lateral load carrying capacity of RC rectangular columns under seismic loading.
• To compare the performance of the developed AI models with established design codes, specifically ACI 318-19.
• To provide more accurate analytical models for flexural and shear capacity estimation.

Main Methods:

• Utilized an experimental database of 250 cyclic tested RC rectangular columns from the PEER center.
• Selected seven key input variables based on linear regression and cosine amplitude methods.
• Developed and validated AI-based GEP models for predicting flexural and shear capacities.

Main Results:

• The GEP models achieved high accuracy, with R² values of 0.96 for flexural capacity and 0.95 for shear capacity.
• The proposed models significantly outperformed the ACI 318-19 code, showing lower RMSE, MAE, and RRSE values.
• Flexural capacity model fitness indicators: R²=0.96, RMSE=53.41, MAE=38.12, RRSE=0.20.
• Shear capacity model fitness indicators: R²=0.95, RMSE=39.47, MAE=28.77, RRSE=0.22.

Conclusions:

• AI-based GEP offers a powerful and accurate approach for predicting the seismic lateral load capacity of RC rectangular columns.
• The developed GEP models provide a more reliable alternative to traditional code-based methods like ACI 318-19.
• This research contributes to enhanced seismic resilience in structural engineering through advanced predictive modeling.