Vision-based model updating and evaluation of miter gates on inland waterways

Abstract

Miter gates are critical components of inland waterways infrastructure, typically used as both the door for and damming surface of locks. Therefore, knowing the condition of miter gates is especially important to ensure their continued operation. Measured displacements of the gates under load are known to be informative of their condition. Such measurements can also be used to update numerical models of the structure that can serve as part of a digital twin, allowing for more comprehensive management of the infrastructure. However, model updating of large-scale infrastructure requires sufficient displacement data that are often challenging to obtain, due in part to inaccessibility and a lack of a viable reference location for contact-type sensors. Vision-based displacement measurements have been proposed to inform a model-updating scheme of the miter gates; however, several challenges have prevented field implementation of these technologies. First, determining the appropriate location and orientation of cameras is critical to obtaining useful displacement information for model updating. In addition, environmental factors that can negatively affect vision-based structural displacement measurements such as lighting changes and camera motion need to be remediated. Moreover, interpreting displacement from photographic imagery is abstract because 2D images can only capture a projection of displacements of the 3D structure or the 3D numerical models. This research addresses each of the challenges through the use of a graphics-based digital twin (GBDT). Prior to conducting the field survey, optimal camera locations and orientations are explored. Additionally, a robust optical flow method is developed to mitigate the negative influence from environment factors on displacement measurements. Finally, displacement from the photographic survey is interpreted by projecting displacement from the numerical models onto the 2D image plane using the GBDT; the difference between these displacements is used to inform a heuristic-based model updating strategy. The proposed approach is demonstrated on the miter gates at The Dalles Lock and Dam. This initial study focuses on determining degradation parameters to which the displacement measurements of the miter gate are most sensitive, laying the groundwork for future rigorous model updating to provide the necessary information regarding repairs and retrofits that would otherwise be difficult to obtain. The main novelty and contribution of this research lie in how the practical challenges of field implementation of existing technologies regarding vision-based model updating are effectively resolved, thus bridging the gap between theory and application to large-scale civil infrastructure.