Classification of Cast Iron Alloys through Convolutional Neural Networks Applied on Optical Microscopy Images DOI Creative Commons

Marta Bárcena,

L. Lloret Iglesias, Diego Ferreño

et al.

steel research international, Journal Year: 2024, Volume and Issue: 95(12)

Published: Aug. 28, 2024

Classification of cast iron alloys based on graphite morphology plays a crucial role in materials science and engineering. Traditionally, this classification has relied visual analysis, method that is not only time‐consuming but also suffers from subjectivity, leading to inconsistencies. This study introduces novel approach utilizing convolutional neural networks—MobileNet for image U‐Net semantic segmentation—to automate the process alloys. A significant challenge domain limited availability diverse comprehensive datasets necessary training effective machine learning models. addressed by generating synthetic dataset, creating rich collection 2400 pure 1500 mixed images ISO 945‐1:2019 standard. ensures robust process, enhancing model's ability generalize across various morphologies particles. The findings showcase remarkable accuracy classifying (achieving an overall 98.9 ± 0.4%—and exceeding 97% all six classes—for ranging between 84% 93% segmentation images) demonstrate consistently identify with level precision speed unattainable through manual methods.

Language: Английский

Process defect analysis and visual detection of aluminum/copper cable joints with magnetic pulse crimping DOI
Hao Jiang, Weixingyu Zhou,

Ming Yong Lai

et al.

Thin-Walled Structures, Journal Year: 2024, Volume and Issue: 202, P. 112110 - 112110

Published: June 12, 2024

Language: Английский

Citations

11
Deep alloys: Metal materials empowered by deep learning DOI
Kaiyuan Zheng, Zhongping He, Lun Che

et al.

Materials Science in Semiconductor Processing, Journal Year: 2024, Volume and Issue: 179, P. 108514 - 108514

Published: May 18, 2024

Language: Английский

Citations

6
Real-time detection of steel corrosion defects using semantic and instance segmentation models based on deep learning DOI
Yılmaz Yılmaz, Safa Nayır, Şakir Erdoğdu

et al.

Materials Today Communications, Journal Year: 2025, Volume and Issue: unknown, P. 112050 - 112050

Published: Feb. 1, 2025

Language: Английский

Citations

0
Machine Learning Assisted Design of High Thermal Conductivity and High Strength Mg Alloys DOI

Huafeng Liu,

T. Nakata, Chao Xu

et al.

Metallurgical and Materials Transactions A, Journal Year: 2025, Volume and Issue: unknown

Published: March 3, 2025

Language: Английский

Citations

0
A lightweight algorithm for steel surface defect detection using improved YOLOv8 DOI Creative Commons
Shuangbao Ma, Xin Zhao, Li Wan

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 15, 2025

In response to the issues of low precision, a large number parameters and high model complexity in steel surface defect detection, lightweight algorithm using improved YOLOv8 is proposed. Firstly, GhostNet utilized as backbone network order reduce computational complexity. Secondly, MPCA (MultiPath Coordinate Attention) attention mechanism integrated enhance feature extraction capabilities. Finally, SIoU (Simplified IoU ) used replace traditional CIoU loss function, which can make anchor frame more fast accurate regression process, improve stability robustness detection. The experimental results indicate that these enhancements have led reduction 37% calculation amount for YOLOv8n algorithm, decrease 32% parameter count, an increase average detection accuracy ( mAP by 1.2%. This achieves balance between lightweighting while providing viable solution deployment computationally resource-constrained edge computing environments such embedded systems mobile devices.

Language: Английский

Citations

0
Metallographic Spheroidization Rate Classification by Using Deep Learning DOI Creative Commons

Lin Chiu‐Chin,

Pei-Ying Chiang,

K Chen

et al.

Engineering Reports, Journal Year: 2025, Volume and Issue: 7(4)

Published: March 27, 2025

ABSTRACT In the steel manufacturing process, spheroidizing annealing is a crucial heat treatment step primarily aimed at improving ductility and machinability of material. Currently, determination spheroidization rate in metals mainly relies on manual inspection through microscope. These methods are time‐consuming subject to inconsistent subjective judgments. To overcome these challenges, this paper proposes deep learning method for classifying metallographic rates using an improved YOLOv8 model, referred as YOLOv8‐DFFN. This model integrates channel attention (CA) vital feature fusion (VFF) techniques, effectively increasing classification accuracy different levels. Experimental results show that YOLOv8‐DFFN achieves mean average precision (mAP) 98.17% across datasets various alloy compositions. represents improvement 1.42% over baseline model. Additionally, surpasses performance original algorithm. innovative technology expected not only enhance production efficiency material quality but also significantly reduce costs human resource investment. It will contribute continuous innovation advancement metal processing industry.

Language: Английский

Citations

0
Toward enhancing concrete crack segmentation accuracy under complex scenarios: a novel modified U-Net network DOI
Feng Qu, Bokun Wang, Qing Zhu

et al.

Multimedia Tools and Applications, Journal Year: 2024, Volume and Issue: 83(31), P. 76935 - 76952

Published: Feb. 19, 2024

Language: Английский

Citations

1
Lightweight-Detection: The strip steel surface defect identification based on improved YOLOv5 DOI

Yan Lu,

Zhichao Huang, Yu‐Qiang Jiang

et al.

Materials Today Communications, Journal Year: 2024, Volume and Issue: 40, P. 109814 - 109814

Published: July 10, 2024

Language: Английский

Citations

1
Development of an Open-Source Software Tool for Microstructure Analysis of Materials Using Artificial Intelligence DOI
Gia Khanh Pham,

Kerim Yalcin,

Alvin Wan

et al.

Key engineering materials, Journal Year: 2024, Volume and Issue: 1004, P. 103 - 110

Published: Dec. 23, 2024

Investigating the microstructures of materials with microscopy is a key task in quality assurance, development new materials, and optimization manufacturing processes. However, conventional image analysis often demands significant time for large volume images, predictions produced are commonly constrained. Applying deep learning, models can be trained to analyze material quickly greater accuracy. The objective this study provide method automatic segmentation microstructural images obtained from microscopes or scanning electron using Convolutional Neural Networks. For purpose, two software scripts were developed Python employing OpenCV fastai library. first script designed generate reference while second utilized training model predicting microstructure an image. test tools demonstrates that robust prediction results attainable by high-quality images. This tool has been made available as open-source on GitHub public use enhanced further if required.

Language: Английский

Citations

1
Classification of Cast Iron Alloys through Convolutional Neural Networks Applied on Optical Microscopy Images DOI Creative Commons

Marta Bárcena,

L. Lloret Iglesias, Diego Ferreño

et al.

steel research international, Journal Year: 2024, Volume and Issue: 95(12)

Published: Aug. 28, 2024

Classification of cast iron alloys based on graphite morphology plays a crucial role in materials science and engineering. Traditionally, this classification has relied visual analysis, method that is not only time‐consuming but also suffers from subjectivity, leading to inconsistencies. This study introduces novel approach utilizing convolutional neural networks—MobileNet for image U‐Net semantic segmentation—to automate the process alloys. A significant challenge domain limited availability diverse comprehensive datasets necessary training effective machine learning models. addressed by generating synthetic dataset, creating rich collection 2400 pure 1500 mixed images ISO 945‐1:2019 standard. ensures robust process, enhancing model's ability generalize across various morphologies particles. The findings showcase remarkable accuracy classifying (achieving an overall 98.9 ± 0.4%—and exceeding 97% all six classes—for ranging between 84% 93% segmentation images) demonstrate consistently identify with level precision speed unattainable through manual methods.

Language: Английский

Citations

0