A deep learning augmented vision-based method for measuring dynamic displacements of structures in harsh environments

Abstract

In this paper, a deep learning augmented vision-based method (DAVIM) is proposed for measuring structural displacement in harsh environments. DAVIM incorporates deep learning-based algorithms that include a Convolutional Neural Network (CNN) and a Generative Adversarial Network (GAN) for more accurate and robust sensing of the dynamics of structural models in wind tunnel tests or full-scale field measurements. The proposed method is first validated through numerical test including a discussion on the efficiency of CNN and GAN in really harsh environments. This is followed by three different kinds of experiments involving periodic vibration tests, motion of an aeroelastic model in a wind tunnel test and field measurements to effectively validate the proposed method from practical applications perspective. The proposed DAVIM exhibits a robust and superior performance compared to the traditional sensors (e.g., a laser displacement sensor and an accelerometer) and the Vision based vibration measurement (VVM) method. This is particularly the case when measuring large displacements with rigid-body rotational components and in long-term field measurements.