Optimization and Predictive Modeling of Reinforced Concrete Circular Columns

Area of Science:

• Structural Engineering
• Optimization Algorithms
• Machine Learning

Background:

• Metaheuristic optimization is crucial for structural design.
• Reinforced concrete columns require efficient dimensioning for cost-effectiveness.

Purpose of the Study:

• To apply the harmony search algorithm for optimal dimensioning of reinforced concrete circular columns.
• To minimize the total cost of steel and concrete in column construction.

Main Methods:

• Harmony search algorithm for optimization.
• Generation of 3125 optimal design samples.
• SHapley Additive exPlanations (SHAP) and ensemble learning for variable impact analysis.
• Design of experiments for visualization.

Main Results:

• Highly accurate predictive models demonstrated using statistical measures (e.g., coefficient of determination).
• Identification of key design variables influencing optimal column dimensions and cost.
• Successful visualization of variable-objective function relationships.

Conclusions:

• A novel technique for generating large datasets for data-driven structural design models was developed.
• The methodology facilitates the application of high-performance machine learning in optimal structural design.
• The harmony search algorithm effectively optimizes reinforced concrete column dimensions.