Full-field displacement reconstruction in structural health monitoring using machine learning approach: case study with experimental validation

The following paper focuses on structural displacement tracking that is a significant process, inter alia, for evaluating safety of structures, load classification, or structural control applications. The authors presented a method for the full-field displacement identification based on strain sensor measurements and machine learning. Using this method, it is possible to recreate full-field displacement maps of the entire structure or its parts, even for different load cases. An example is given in which a typical aerostructure (composite hat-stiffened panel) is subjected to displacement monitoring. Two neural networks were trained to identify full-field displacement maps of the panel, based on strain gauges measurements. The accuracy of the predictions was experimentally tested using digital image correlation (DIC). The predicted displacement maps were qualitatively and quantitively compared with the results of finite element simulation and experimental DIC measurements.