Quantum mechanics

Random forest

Engineering

Psychological resilience

Computer security

Reinforced concrete

Telecommunications

Computer science

Geology

Boosting (machine learning)

Mechanical engineering

Vulnerability (computing)

Physics

Seismic loading

Vulnerability assessment

Psychology

Psychotherapist

Gradient boosting

Structural engineering

Idealization

Forensic engineering

Machine learning

Frame (networking)

Journal of Building Engineering

Robust seismic vulnerability assessment for a building under expected earthquake ground motions necessitates explicit consideration of all-important sources of uncertainty in structural model idealization. This paper presents a machine learning-based methodology for reliably predicting the seismic response and structural collapse classification of ductile reinforced concrete frame buildings under future earthquake events by accounting for component- and system-level modeling uncertainties. The proposed methodology uses two different types of machine learning methods—regression-based and classification-based methods—to achieve the goal of this study. Machine learning techniques with boosting algorithms (i.e., adaptive boosting and extreme gradient boosting) are the best methods for both response prediction and collapse status classification of modern code-compliant reinforced concrete frame buildings. Finally, the effect of uncertain modeling parameters on the response and collapse identification is examined. The reinforced concrete beam modeling-related parameters (i.e., plastic deformation properties) of ductile, low-to mid-rise frame buildings are significant predictors of seismic response due to capacity design principles.

Machine learning-based approaches for seismic demand and collapse of ductile reinforced concrete building frames