Assessing the fatigue damage of concrete structures using automatically classified crack severity level information

Abstract

The aim of this paper is to assess the fatigue damage of concrete structures using automatically classified crack severity level information. To solve the problem that the label data used for the automatic classification of regional complex crack severity levels can only be labeled by intuition, this paper proposes a quantitative index of regional complex crack severity levels that can consider the regional crack width, height, and density information. The testing set F1 score of the Swin Transformer classification model trained by the label data reaches 0.9903. To investigate the feasibility of assessing the concrete structure fatigue damage based on the crack change information, the paper adopted the scanning shooting method to acquire the concrete beam panoramas at different cyclic loading times in the fatigue test, and automatically classified the crack severity levels in different regions of the panoramas using the Swin transformer. The test results show that the information on the change in the number of regions with the same crack severity level on the concrete structure surface can be used to estimate approximately the fatigue life percentage interval that the concrete structure is currently in, which provides a reference for the development of maintenance measures and safety warnings.