An intelligent crack damage assessment method by integrating information and physics

Abstract

This paper proposes a novel intelligent crack damage assessment method by integrating information and physics, in which automatic crack detection, real-time crack damage assessment, and external load optimization are integrated into a system. The pixel-level information of cracks is automatically identified and segmented by the mask region convolutional neural network (Mask R-CNN). A crack digital model is reconstructed by fusing perceived information and geometric models. An offline physical model and an online deep learning model are integrated for assessing crack damage status. Experiments show that the method can effectively detect and localize fatigue surface cracks together with a load optimization to ensure safe conditions when visible fatigue cracks appear.