Optimizing Machine Learning with SSA and PSO for Anchor Bolt-Grout Bond Strength Prediction

Area of Science:

• Civil Engineering
• Materials Science
• Computational Mechanics

Background:

• Anchor bolt bond strength (τ) is vital for anchorage engineering capacity and durability, especially with rebar corrosion.
• Traditional experimental measurement of τ is complex, time-consuming, and labor-intensive.
• Developing efficient and accurate prediction methods for τ is essential for reliable structural assessment.

Purpose of the Study:

• To develop a generalized machine learning model for predicting anchor bolt bond strength (τ).
• To optimize machine learning models using metaheuristic algorithms for improved accuracy and reduced overfitting.
• To compare the performance of optimized models against empirical formulas and analyze feature importance.

Main Methods:

• Utilized pullout test data from 429 rebar-concrete specimens.
• Developed and optimized Random Forest (RF), Least Squares Boosting (LSBoost), and Generalized Additive Model (GAM) using Sparrow Search Algorithm (SSA) and Particle Swarm Optimization (PSO).
• Performed comparative error analysis and SHAP analysis to evaluate model performance and feature contributions.

Main Results:

• Unoptimized models showed low accuracy and overfitting.
• Optimized models, particularly PSO-LSBoost, demonstrated significantly improved prediction accuracy (R² = 0.93) and reduced overfitting.
• The PSO-LSBoost model's predictions for τ substantially surpassed those of three empirical formulas.
• SHAP analysis identified corrosion rate (Cw) as the most influential factor and rebar type (ST) as the least.

Conclusions:

• Machine learning, optimized with SSA and PSO, provides a novel and efficient approach for predicting anchorage bond strength.
• The PSO-LSBoost model offers superior accuracy and generalization ability compared to traditional methods and empirical formulas.
• This approach enhances the reliability of anchorage structures by improving the assessment of bearing capacity and bolt durability.