Research on semantic understanding and representativeness of contemporary ceramic materials based on deep learning.
Saved in:
| Title: | Research on semantic understanding and representativeness of contemporary ceramic materials based on deep learning. |
|---|---|
| Authors: | Liu, Yingrui1 (AUTHOR) liuyingrui-tafa@outlook.com |
| Source: | Journal of Computational Methods in Sciences & Engineering (Sage Publications Inc.). Nov2025, Vol. 25 Issue 6, p5574-5587. 14p. |
| Subjects: | Ceramic materials, Deep learning, Three-dimensional imaging, Convolutional neural networks, Microstructure, Recurrent neural networks |
| Abstract: | The semantic understanding and representativeness of contemporary ceramic materials are critical to advancing their artistic and functional applications in modern design. However, analyzing subtle characteristics of ceramic materials—such as texture, composition, and form—has traditionally relied on subjective techniques. This study introduces an advanced deep learning (DL) framework that combines Convolutional Neural Networks (CNNs) and Elman Recurrent Neural Networks (ERNNs) to systematically assess and quantify the semantic features of modern ceramic materials. Data collected from 3D surface texture maps and high-resolution Scanning Electron Microscopy (SEM) images are used to train models capable of recognizing and segmenting complex microstructural patterns, including surface imperfections, crystallization stages, and fault types. The CNN model extracts hierarchical features from SEM images, while the ERNN generates synthetic high-resolution images for data augmentation, thereby improving segmentation accuracy. The results demonstrate that the proposed models outperform traditional methods, achieving an Intersection over Union (IoU) score of 97.8%, an accuracy of 97.69%, and a precision of 95.32%. Additionally, the trained models facilitate precise reconstruction of 3D microstructures, revealing spatial distributions of phases that are challenging to capture with conventional imaging techniques. When integrated with simulation tools, this approach enhances semantic insights into ceramic materials, enabling real-time applications in ceramic design and manufacturing—thus promoting higher quality control, material innovation, and process efficiency. [ABSTRACT FROM AUTHOR] |
| Copyright of Journal of Computational Methods in Sciences & Engineering (Sage Publications Inc.) is the property of Sage Publications Inc. and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | The semantic understanding and representativeness of contemporary ceramic materials are critical to advancing their artistic and functional applications in modern design. However, analyzing subtle characteristics of ceramic materials—such as texture, composition, and form—has traditionally relied on subjective techniques. This study introduces an advanced deep learning (DL) framework that combines Convolutional Neural Networks (CNNs) and Elman Recurrent Neural Networks (ERNNs) to systematically assess and quantify the semantic features of modern ceramic materials. Data collected from 3D surface texture maps and high-resolution Scanning Electron Microscopy (SEM) images are used to train models capable of recognizing and segmenting complex microstructural patterns, including surface imperfections, crystallization stages, and fault types. The CNN model extracts hierarchical features from SEM images, while the ERNN generates synthetic high-resolution images for data augmentation, thereby improving segmentation accuracy. The results demonstrate that the proposed models outperform traditional methods, achieving an Intersection over Union (IoU) score of 97.8%, an accuracy of 97.69%, and a precision of 95.32%. Additionally, the trained models facilitate precise reconstruction of 3D microstructures, revealing spatial distributions of phases that are challenging to capture with conventional imaging techniques. When integrated with simulation tools, this approach enhances semantic insights into ceramic materials, enabling real-time applications in ceramic design and manufacturing—thus promoting higher quality control, material innovation, and process efficiency. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 14727978 |
| DOI: | 10.1177/14727978251346050 |
