Deep learning-based planar crack damage evaluation using convolutional neural networks. (1st April 2021)

• Record Type: Journal Article
• Authors: Long, X.Y.; Zhao, S.K.; Jiang, C.; Li, W.P.; Liu, C.H.
• Is Part Of: Engineering fracture mechanics. Volume 246(2021)
• Publication Date: 2021-04-01
• Subjects: Crack damage evaluation -- Deep learning -- Computational vision -- Deep convolutional neural network -- Stress intensity factor; Fracture mechanics -- Periodicals; Rupture, Mécanique de la -- Périodiques; Fracture mechanics; Periodicals; 620.112605
• DOI: 10.1016/j.engfracmech.2021.107604
• Languages: English
• ISSNs: 0013-7944

Abstract

Highlights: Deep learning-based damage evaluation method is proposed by computer vision. A deep convolutional neural network for predicting stress intensity factors is designed. The data consisting of reference and deformed speckled images is prepared. Remarkable results are obtained with prediction accuracy larger than 96%. Abstract: This article presents a novel deep learning-based damage evaluation approach by using speckled images. A deep convolutional neural network (DCNN) for predicting the stress intensity factor (SIF) at the crack tip is designed. Based on the proposed DCNN, the SIF can be automatically predicted through computational vision. The data bank consisting of a reference speckled image and lots of deformed speckled images is prepared by a camera and an MTS testing machine. Experiments were performed to verify the method, and the achieved results are quite remarkable with larger than 96% of predicted SIF values falling within 5% of true SIF values when sufficient training images are available. The results also confirm that the appropriate subset size of images within the field of view is 400 × 400 pixel resolutions.