Scalable Impact Detection and Localization Using Deep Learning and Information Fusion

Abstract

Due to their unpredictable nature, many impact events (e.g., overheight vehicles striking on low-clearance bridges) go unnoticed or get reported hours or days later. However, they can induce structural damage or even failure. Therefore, prompt impact detection and localization strategies are essential for early warning of impact events and rapid inspection of structures. Most existing strategies are developed for aircraft composites panels utilizing high rate synchronized measurement from densely deployed sensors. Limited efforts are made for other applications, such as infrastructure systems or extraterrestrial human habitats, which require large-scale measurement and scalable detection strategies. Particularly in harsh environments, structural impact localization must be robust to limited number of sensors and multi-source errors. In this study, an effective impact localization strategy is proposed to identify impact locations using limited number of vibration measurements. Convolutional neural networks are trained for each sensor node and are fused using Bayesian theory to improve the accuracy of impact localization. Special considerations are paid to address both measurement and modeling errors. The proposed strategy is illustrated using a 1D structure, and numerically validated for a 2D dome-shaped structure. The results demonstrate that the proposed method detects and localizes impact events accurately and robustly.