However, the classification-based CNN and the edge-detection algorithm only used part of the crack structure information, which reduced the reliability of the method. Previously, the authors investigated a crack propagation detection method using classification-based CNN and an improved canny edge-detection algorithm, and verified the accuracy and efficiency of the method via central hole specimens. Although these approaches can detect the cracks with high reliability even surrounded by crack-like noises, it would be difficult to accomplish quantitative detection of crack length. Other researchers used trained CNN to divide every crack pixel from the background in the image. Cha used trained CNN and sliding window techniques to detect cracks from images. Many researchers have carried out studies for crack detection based on CNN. Yu put forward a DCNN-based method to localize damages to smart building structures with high accuracy on raw noisy signals. For example, Dong established a CNN model to identify microseismic events and blasts accurately. ![]() In recent years, machine learning and convolutional neural networks have shown a strong capability for feature extraction and target detection, and they have been used in structural health monitoring (SHM). At the same time, accurate detection of crack tips, which is key for crack length measurement, remains challenging. As a result, there are difficulties in distinguishing true cracks from crack-like noises such as scratches and structure boundaries. A general limitation of these approaches, however, is that most of these vision-based methods detect cracks by searching over the entire area of an image. Many methods have been established based on image processing techniques including edge detection, Hough transform, image segmentation, identification and detection of feature points, the digital image correlation (DIC) method, and photogrammetry. Vision-based crack detection methods have been widely studied over recent decades for their advantages of non-contact, high precision, and good real-time performance. Therefore, the application of acoustic emission technology for monitoring the length of cracks is limited. However, this technology relies on a good correlation between acoustic emission data and the damage mechanism, while the experimental results will be affected by environmental noise. It records and analyzes the signals released by materials or structures when deformation or damage occurs to detect the damage location and predict the time to failure. Acoustic emission technology is a new nondestructive testing method. Complicated calibration is needed when applying the compliance method. The accuracy of the electrical method is easily affected by the experiment conditions and environment. The human inspection method requires large testing times since it needs to interrupt the fatigue test to manually measure crack lengths. Existing approaches including the human inspection method, the electrical method, the compliance method, and acoustic emission technology all have limitations for application during a fatigue test. In recent decades, extensive research on monitoring crack length propagation has been carried out. ![]() Crack length is one of the most relevant parameters that needs to be recorded during laboratorial tests. Fatigue crack propagation testing is an essential method of studying metallic or structural fatigue life prediction in fracture mechanics. ![]() Furthermore, crack length could be measured with submillimeter accuracy.įatigue cracks caused by repetitive loads, which are of great concern for structural safety, always exist in old structures including airplane and highway bridges. The results demonstrated that the proposed approach could robustly identify a fatigue crack surrounded by crack-like noises and locate the crack tip accurately. The effectiveness and precision of the proposed approach were validated through conducting fatigue experiments. Then, a crack tip-detection algorithm was established to accurately locate the crack tip and was used to calculate the length of the crack. Convolutional neural networks were first applied to robustly detect the location of cracks with the interference of scratch and edges. In this paper, a new framework based on convolutional neural networks (CNN) and digital image processing is proposed to monitor crack propagation length. Traditional vision-based methods are insufficient in distinguishing cracks from noises and detecting crack tips. Human inspection is the most widely used approach for fatigue failure detection, which is time consuming and subjective. Fatigue failure is a significant problem in the structural safety of engineering structures.
0 Comments
Leave a Reply. |