Machine Learning-Based Prediction Models for Punching Shear Strength of Fiber-Reinforced Polymer Reinforced Concrete...

Area of Science:

• Civil Engineering
• Materials Science
• Machine Learning

Background:

• Fiber-reinforced polymers (FRPs) offer advantages in concrete slabs but are prone to punching shear failure.
• Existing empirical models for predicting punching shear strength in FRP-reinforced concrete slabs exhibit significant inaccuracies.
• There is a critical need for more reliable predictive models for structural integrity.

Purpose of the Study:

• To develop and validate gradient-boosted regression tree (GBRT) models for accurate punching shear strength prediction in FRP-reinforced concrete slabs.
• To address the limitations and improve upon existing empirical predictive models.
• To utilize machine learning for enhanced structural analysis in civil engineering.

Main Methods:

• Compiled a comprehensive database of 238 experimental results for FRP-reinforced concrete slabs.
• Developed and evaluated Gradient-Boosted Regression Tree (GBRT) models against other machine learning algorithms.
• Employed the SHapley Additive exPlanation (SHAP) method for model interpretability.

Main Results:

• The GBRT model demonstrated high prediction accuracy with R²=0.955, RMSE=64.85, and MAE=42.89.
• GBRT models significantly outperformed traditional empirical models in predicting punching shear strength.
• SHAP analysis identified slab thickness, FRP reinforcement ratio, and critical section perimeter as key influencing variables.

Conclusions:

• The GBRT model provides a highly accurate and reliable method for predicting the punching shear strength of FRP-reinforced concrete slabs.
• Machine learning offers a powerful approach to overcome limitations of traditional empirical models in structural engineering.
• Understanding key influencing factors through SHAP analysis can guide future research and design optimization for FRP-concrete structures.