Enhancing Seismic Damage Detection and Assessment in Highway Bridge Systems: A Pattern Recognition Approach with...

Area of Science:

• Civil Engineering
• Structural Engineering
• Earthquake Engineering

Background:

• Highway bridges are critical infrastructure requiring rapid recovery after extreme events like earthquakes.
• Near real-time structural damage diagnosis is essential for efficient post-event recovery and community resilience.
• Bridge columns are vital components, and their damage assessment is key to overall structural integrity.

Purpose of the Study:

• To develop and validate a data-driven approach for seismic damage detection and assessment in bridge columns.
• To introduce novel cumulative intensity-based damage features for improved analysis.
• To leverage machine learning for accurate classification of damage states.

Main Methods:

• Nonlinear time history analysis simulations were used to generate structural response data.
• Cumulative intensity-based damage features were developed and evaluated.
• Unsupervised and supervised learning techniques, including Support Vector Machines and Bayesian optimization with Gaussian processes, were applied.

Main Results:

• The proposed methodology demonstrated high accuracy in detecting seismic damage.
• The approach successfully classified both the presence/absence of damage (binary) and the severity of damage (multi-class).
• The cumulative intensity-based features proved effective in damage assessment.

Conclusions:

• The data-driven approach offers a reliable method for near real-time seismic damage detection and assessment in bridge columns.
• This research supports the practical implementation of on-board sensor computing for structural health monitoring.
• The findings contribute to enhancing the resilience of transportation infrastructure against seismic hazards.