• A novel damage detection method integrates model reduction and response reconstruction to address challenges in large-scale civil structures.
• The two-step model updating framework (substructure-level localization and element-level detection) enhances damage localization and quantification accuracy.
• The improved sensitivity algorithm incorporates reconstructed responses, reducing sensor requirements while maintaining detection reliability.
• Numerical validation on a truss bridge demonstrates improved efficiency and accuracy over traditional methods, offering a practical solution for structural health monitoring.