Accurate Measurement of Bridge Vibration Displacement via Deep Convolutional Neural Network

Abstract

Displacement measurement is an essential method for structural safety assessment and health monitoring, and the static and dynamic characteristics of the structure can be obtained through displacement. In order to overcome the limitations of sensors in vibration measurement of large structures, as well as the poor adaptability of visual measurement algorithms, such as machine learning and digital image processing, this article takes the bridge structure as the research object and introduces deep learning into the field of visual vibration measurement. Moreover, based on the deep convolutional neural network (CNN), a new high-precision displacement measurement algorithm of multiscale feature extraction and fusion is proposed to solve the inaccurate measurement of existing neural networks. Experiments are carried out on bridge models in the laboratory environment and bridges in the real world to verify the adaptability and reliability of the proposed method. At the same time, the time–frequency characteristic curves of different deep learning models, template matching algorithms, and acceleration sensors are compared. The result analysis shows that the vibration displacement trajectory of the algorithm in this article has the best coincidence with the standard displacement signal. Three experiments have fully verified that the algorithm in this article has good application potential and implementation space in the field of structural state monitoring.