Structural displacement measurement is crucial for the safety and integrity of civil structures. Traditional methods often face challenges such as limited measurement points and vulnerability to environmental conditions. Vision-based methods, while promising, have struggled with accuracy and robustness. Due to these issues, a novel approach that provides more reliable and comprehensive displacement data with high spatial resolution is necessary to advance structural health monitoring.
Researchers from the Harbin Institute of Technology have developed a novel deep-learning model, Nodes2STRNet, for structural dense displacement recognition. The study (DOI: 10.1002/msd2.12083), published in the International Journal of Mechanical System Dynamics in 2023, demonstrates how this model outperforms traditional methods. Nodes2STRNet utilizes a combination of a control node estimation subnetwork (NodesEstimate) and a pose parameter recognition subnetwork (Nodes2PoseNet) to deliver accurate and robust dense structural displacement measurements.
The core innovation of Nodes2STRNet lies in its ability to accurately recognize structural displacement using a deformable 3D mesh model and dense optical flow. The NodesEstimate subnetwork calculates the 2D position of control nodes from video frames, while Nodes2PoseNet maps these coordinates to structural pose parameters. This method overcomes the limitations of sparse point measurement by providing a dense displacement field. The self-supervised learning strategy further enhances its efficiency, eliminating the need for extensive manual annotations. Experimental validation through seismic shaking table tests on a four-story building model confirmed the model's high accuracy and robustness to light condition variations. The results showed that Nodes2STRNet consistently outperformed existing methods, in recognizing displacement under different peak ground accelerations and light conditions. This improvement is particularly significant for applications requiring high precision and reliability in dynamic environments.
Dr. Yang Xu, from the Harbin Institute of Technology, stated, "Nodes2STRNet represents a significant advancement for structural displacement recognition in the field of structural health monitoring. Its ability to provide accurate and dense displacement data, even under challenging conditions, will greatly enhance our ability to monitor and maintain the condition and integrity of critical structures."
The development of Nodes2STRNet has significant implications for structural health monitoring and disaster prevention. Its robust and accurate displacement measurements can be applied to various structures, from buildings to bridges, ensuring their safety and reliability. Additionally, this technology can aid in the early detection of structural health issues, potentially preventing catastrophic failures and improving the overall resilience of civil infrastructure.