Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields
Inverse problems involving transport phenomena are ubiquitous in engineering practice, but their solution is often challenging. In this work, we build a data-driven deep learning model to predict the heterogeneous distribution of circle-shaped fillers in two-dimensional thermal composites using the...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
AIP Publishing LLC
2020-04-01
|
Series: | AIP Advances |
Online Access: | http://dx.doi.org/10.1063/5.0004631 |
id |
doaj-f0dc99ca2fa54f5cb57b7864950cee9b |
---|---|
record_format |
Article |
spelling |
doaj-f0dc99ca2fa54f5cb57b7864950cee9b2020-11-25T02:37:49ZengAIP Publishing LLCAIP Advances2158-32262020-04-01104045037045037-1210.1063/5.0004631Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fieldsHaiyi Wu0Hongwei Zhang1Guoqing Hu2Rui Qiao3Department of Mechanical Engineering, Virginia Tech, Blacksburg, Virginia 24061, USADepartment of Mechanical Engineering, Virginia Tech, Blacksburg, Virginia 24061, USADepartment of Engineering Mechanics, Zhejiang University, Hangzhou 310027, ChinaDepartment of Mechanical Engineering, Virginia Tech, Blacksburg, Virginia 24061, USAInverse problems involving transport phenomena are ubiquitous in engineering practice, but their solution is often challenging. In this work, we build a data-driven deep learning model to predict the heterogeneous distribution of circle-shaped fillers in two-dimensional thermal composites using the temperature field in the composite as an input. The deep learning model is based on convolutional neural networks with a U-shape architecture and encoding–decoding processes. The temperature field is cast into images of 128 × 128 pixels. When the true temperature at each pixel is given, the trained model can predict the distribution of fillers with an average accuracy of over 0.979. When the true temperature is only available at 0.88% of the pixels inside the composite, the model can predict the distribution of fillers with an average accuracy of 0.94, if the temperature at the unknown pixels is obtained through the Laplace interpolation. Even if the true temperature is only available at pixels on the boundary of the composite, the average prediction accuracy of the deep learning model can still reach 0.80; the prediction accuracy of the model can be improved by incorporating true temperature in regions where the model has low prediction confidence.http://dx.doi.org/10.1063/5.0004631 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Haiyi Wu Hongwei Zhang Guoqing Hu Rui Qiao |
spellingShingle |
Haiyi Wu Hongwei Zhang Guoqing Hu Rui Qiao Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields AIP Advances |
author_facet |
Haiyi Wu Hongwei Zhang Guoqing Hu Rui Qiao |
author_sort |
Haiyi Wu |
title |
Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields |
title_short |
Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields |
title_full |
Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields |
title_fullStr |
Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields |
title_full_unstemmed |
Deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields |
title_sort |
deep learning-based reconstruction of the structure of heterogeneous composites from their temperature fields |
publisher |
AIP Publishing LLC |
series |
AIP Advances |
issn |
2158-3226 |
publishDate |
2020-04-01 |
description |
Inverse problems involving transport phenomena are ubiquitous in engineering practice, but their solution is often challenging. In this work, we build a data-driven deep learning model to predict the heterogeneous distribution of circle-shaped fillers in two-dimensional thermal composites using the temperature field in the composite as an input. The deep learning model is based on convolutional neural networks with a U-shape architecture and encoding–decoding processes. The temperature field is cast into images of 128 × 128 pixels. When the true temperature at each pixel is given, the trained model can predict the distribution of fillers with an average accuracy of over 0.979. When the true temperature is only available at 0.88% of the pixels inside the composite, the model can predict the distribution of fillers with an average accuracy of 0.94, if the temperature at the unknown pixels is obtained through the Laplace interpolation. Even if the true temperature is only available at pixels on the boundary of the composite, the average prediction accuracy of the deep learning model can still reach 0.80; the prediction accuracy of the model can be improved by incorporating true temperature in regions where the model has low prediction confidence. |
url |
http://dx.doi.org/10.1063/5.0004631 |
work_keys_str_mv |
AT haiyiwu deeplearningbasedreconstructionofthestructureofheterogeneouscompositesfromtheirtemperaturefields AT hongweizhang deeplearningbasedreconstructionofthestructureofheterogeneouscompositesfromtheirtemperaturefields AT guoqinghu deeplearningbasedreconstructionofthestructureofheterogeneouscompositesfromtheirtemperaturefields AT ruiqiao deeplearningbasedreconstructionofthestructureofheterogeneouscompositesfromtheirtemperaturefields |
_version_ |
1724793197343801344 |