Polymers, Journal Year: 2025, Volume and Issue: 17(9), P. 1245 - 1245
Published: May 2, 2025
This study investigates the use of advanced convolutional neural networks (CNNs) to analyze and classify fracture behavior U-shaped concrete modified with polyurethane (PU) under repeated drop-weight impact loads. A total 17 specimens were tested multiple loads for each PU binder content (0%, 10%, 20%, 30%) by weight cement. By integrating digital image correlation (DIC) dynamic static mechanical testing, this research evaluates concrete’s resistance flexural varying content. Three CNN architectures, InceptionV3, MobileNet, DenseNet121, trained on a dataset comprising 1655 high-resolution crack images failure stages into no crack, initial failure. Experimental results revealed that 20% optimally enhances strength, while properties declined significantly 30% The strain localization in DIC analysis indicated reduced matrix cohesion, which was measured extent concentration material, highlighting importance maintaining below avoid compromising structural integrity. Among models, InceptionV3 demonstrated superior accuracy (96.67%), precision, recall, outperforming MobileNet (94.56%) DenseNet121 (90.03%). combination deep learning offers robust, automated framework assessment, improving efficiency over traditional methods such as visual inspections, are time-consuming reliant expert judgment.
Language: Английский