Computer Methods in Materials Science
Loading...
ISSN 2720-4081
e-ISSN: 2720-3948
Issue Date
2024
Volume
Vol. 24
Number
No. 3
Description
Journal Volume
Computer Methods in Materials Science
Vol. 24 (2024)
Projects
Pages
Articles
Shear strength estimation of a FRP-strengthened RC beam: A comparison between an artificial neural network and guideline equations
(Wydawnictwa AGH, 2024) Nezaminia, Hamid
In recent years, several experimental tests have been conducted on the shear strengthening of reinforced concrete (RC) beams strengthened by fiber-reinforced polymer (FRP) systems. In this regard, some equations have also been proposed to estimate the shear strength of beams reinforced with FRP systems. The aim of this study is to investigate the estimation of the shear strength of beams reinforced with FRP systems using an artificial neural network model. For this purpose, a comprehensive and extensive review of forty published articles has been carried out to compile data on 304 RC beams strengthened with externally bonded FRP systems to improve their shear strength. These laboratory results have been used to provide a database for the ANN model to evaluate the shear behavior. The input to the ANN model consists of the 11 variables, including the sectional geometry, reinforcement ratio, FRP ratio, and the characteristics of concrete, steel reinforcement, and composite material, while the output variable is the shear strength of the FRP-strengthened RC beam. In order to evaluate the effectiveness of the neural network model in estimating the shear capacity of RC beams, the results obtained from the neural network model are compared with the equations from the Publication No. 345 and ACI 440.2R guidelines. The comparison of the results shows that the predictive power of the proposed model is much better than the experimental guidelines. Specifically, the mean absolute relative error (MARE) criteria for the studied data is 13%, 34% and 39% for the ANN model, ACI 440.2R guideline and the Publication No. 345 guideline, respectively.
An evaluation of the capabilities of image-based metal component defect recognition with deep learning techniques
(Wydawnictwa AGH, 2024) Wójcik, Michał P.; Pawlikowski, Kacper; Madej, Łukasz
In the era of Industry 4.0, deploying highly specialised machine learning models trained on unique and often scarce datasets is an attractive solution for advancing automated quality control and minimising production costs. Therefore, the main aim of this research is to evaluate the capabilities of three deep learning models (ResNet-18, ResNet-50 and SE-ResNeXt-101 (32 × 4d)) in the identification of surface defects in forged products. Leveraging advanced photography techniques, including studio lighting and a shadowless box, high-quality images of complex product surfaces were acquired for the training data set. Given the relatively small size of the image dataset, aggressive data augmentation techniques were introduced during the training and evaluation process to ensure robust model generalisation ability. The results obtained demonstrate the significant impact of data augmentation on model performance, highlighting its importance in training and evaluating deep learning models with limited data. This research also emphasises the need for innovative data pre-processing strategies in an efficient and robust machine learning model delivery to the industrial environment.
Detecting dents in car bodies using machine learning and structured light projection
(Wydawnictwa AGH, 2024) Potasz, Izabela; Potasz, Sławomir; Laska, Michał
This article discusses feasible methods for detecting dents in car bodies caused by transportation damage, commuting collisions, and hail. The authors review existing approaches exploiting their limitations, including smartphone-based ML detection algorithms and drive-through tunnels. The paper details the setup for capturing dents using computer vision with industry-grade cameras and structured light projection, emphasizing optimized data acquisition and computer vision setup. A particular emphasis is placed on acquiring high-quality input data thanks to the proper calibration and alignment of cameras, structured light, and the synchronization between them. Challenges related to obtaining high-quality footage in real-life conditions, such as car speed, body color, and lighting conditions, are thoroughly discussed. The method covers algorithms for detecting car paint, optimizing camera parameters, and identifying dents. Data annotation methods are described in detail, ensuring robust training datasets. Validation of the method is based on comparing the results of an inspection by professional car appraisers with algorithm detection outcomes. The results demonstrate the effectiveness of the proposed methods. Additionally, the article explores future research opportunities, such as scratch detection, damage severity estimation, and integrating these systems into automated production lines. The potential for enhancing vehicle inspection processes through advanced computer vision and structured light techniques is also considered.