A hybrid methodology for structural damage detection uniting FEM and 1D-CNNs: Demonstration on typical high-pile wharf

Vibration-based structural damage detection (SDD) has been a subject of intense research in health monitoring (SHM) for large civil engineering structures over the decades. The performance conventional SDD approaches predominantly relies on rational choices feature and classifier. Hand-crafted features or fixed classifiers would not be optimal choice all damaged scenarios. This paper proposes novel, quick precise real-time framework high-pile wharf foundations using combination finite element modeling 1D convolutional neural networks (CNNs). distinct advantage this method lies extracting damage-related from raw displacement response directly automatically, computational complexity compact CNNs is significantly lower because data processing involves only simple operations. results show that presented have superior ability to accurately identify occurrence location real time. In addition, comprehensive trained by dataset component form better than based absolute value form. also demonstrated although proposed are more sensitive longitudinal lateral responses structure, vertical still positive effect improvement generalization robustness CNNs.