Camera-Based Real-Time Damage Identification of Building Structures through Deep Learning

Abstract

Real-time damage identification (DI) augments smart structures with instant damage information. Capturing the severity and location of the damage via real-time DI will allow for effective scheduling of preventive measures and action plans to isolate the damage and replace affected elements. It also improves structural safety, especially against extreme events unknown at the design stage. There is a need to overcome the difficulties and limitations of model-based approaches and train supervised machine-learning classifiers in the absence of measured damaged data. This paper proposes an image-based DI methodology using deep neural networks to provide real-time data-driven damage information for structural systems. The proposed methodology is evaluated experimentally using a three-dimensional (3D) moment-resisting frame structure subjected to dynamic loading. Two data acquisition configurations are studied simultaneously to measure the dynamic response and compare the accuracy between sensors and video recording. Video processing techniques track the floor levels to capture structural response. The deep learner outputs provide real-time DI describing the damage’s severity and location. Results show the effectiveness of the proposed nondestructive and model-free methodology for real-time DI.