Deep learning models for analysis of non‐destructive evaluation data to evaluate reinforced concrete bridge decks: A survey

Abstract

Application of deep learning (DL) for automatic condition assessment of bridge decks has been on the raise in the last few years. From the published literature, it is evident that lot of research efforts has been done in identifying the surface defects such as cracks, potholes, spalling and so forth using supervised learning methods such as deep learning. However, the health of a reinforced concrete bridge deck is jeopardized substantially due to presence of subsurface defects. Subsurface defects in bridge decks are genearlly detected using non‐destructive evaluation (NDE) methods. Interpertation of NDE data for autonomous deck evaluation requires development of DL models; however, The task of defect detection DL has not received the proper attention for subsurface defect detection in the past. The goal of this paper is to provide a review of existing DL models for analysis of NDE data of bridge decks. The authors reviewed prominent NDE techniques for subsurface defect detection of bridge decks and explored the various DL models proposed to identify these defects. First a brief overview of the working principle of NDE techniques and DL architectures is provided, and then the information about proposed DL models and their efficacy is highlighted. Based on the existing knowledge gaps, various challenges and future prospects associated with application of DL in bridge subsurface inspection are discussed.

Study on the Sustainable Detection of Machining Surface Defects under the Influence of Environmental Interference

“…Firstly, it has been successfully applied to the classical image classification task. In the existing defect detection literature, there are three types of detection methods: classification, target detection, and segmentation [8,9].…”

“…At the same time, the Feature Pyramid Network (FPN) [27] is introduced to optimize the jump connection part of the model to improve its segmentation accuracy. To avoid the problem of gradient disappearance or bursting with the model depth, the residual structure [9] is introduced.…”