Ensemble machine learning-based sensitivity and parametric assessment of headed stud shear connectors behavior in...

Area of Science:

• Civil Engineering
• Structural Engineering
• Materials Science

Background:

• Headed stud shear connectors are critical components in composite steel and concrete structures, influencing overall structural integrity.
• Understanding their behavior under various conditions is essential for safe and efficient design.
• Existing analytical methods may not fully capture the complex interactions affecting connector performance.

Purpose of the Study:

• To perform a sensitivity and parametric assessment of headed stud shear connector behavior in composite construction.
• To uncover hidden correlations and patterns influencing connector performance using machine learning.
• To provide insights for optimizing the design and evaluation of shear connectors.

Main Methods:

• Utilized a comprehensive database of 464 push tests on shear connectors.
• Applied eight types of ensemble machine learning models.
• Conducted feature importance and partial dependence analyses to understand variable influence.

Main Results:

• Identified key parameters affecting the shear capacity of headed stud connectors.
• Quantified the sensitivity of connector behavior to different variables, including welding methods.
• Revealed complex relationships between design parameters and connector performance.

Conclusions:

• Ensemble machine learning provides a powerful approach to understanding shear connector behavior in composite construction.
• The findings offer valuable, data-driven insights for structural engineers and designers.
• This research bridges the gap between theoretical knowledge and practical application in composite construction.