journal article·2019-07Open access

Visualizing Material Quality and Similarity of mc-Si Wafers Learned by Convolutional Regression Networks

Matthias Demant, Patrick Virtue, Aditya Kovvali, Stella X. Yu, Stefan Rein

IEEE Journal of Photovoltaics· Volume 9· Issue 4

Abstract

Convolutional neural networks can be trained to assess the material quality of multicrystalline silicon wafers. A successful rating model has been presented in a related work, which directly evaluates the photoluminescence (PL) image of the wafer to predict the current-voltage parameters after solar cell production. This paper presents the results of two visualization techniques to understand what has been learned in the network. First, we reveal what has been learned in the PL image by visualizing the spatial quality distribution of the wafers based on the activation maps of the network. The method is denoted as regression activation mapping. We compare regression activation maps with J 0 images of solar cells to show the semantically meaningful representation of the trained features. Second, we show what has been learned in the data by mapping the learned network representation of all wafers into a low-dimensional subspace. Visualizations reveal the smoothness of our representation with respect to the PL input and measured quality. This technique can be used to detect material anomalies or process faults for samples with high prediction errors.
Source: openalex_inverted_index · Quality: candidate unverified · Source record ↗

Research context

OpenAlex topics: Industrial Vision Systems and Defect Detection · Silicon and Solar Cell Technologies · Thin-Film Transistor Technologies

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

Fraunhofer Institute for Solar Energy, Freiburg, Germany

International Computer Science Institute, University of California, Berkeley, CA, USA

Funding

German Federal Ministry

References (22)

Show source-provided references
1. Diabetic retinopathy detection via deep convolutional networks for discriminative localization and visual explanation
2. Visualizing high-dimensional data using t-SNE
3. DeCAF: A deep convolutional activation feature for generic visual recognition
4. 10.1109/CVPR.2017.24310.1109/CVPR.2017.243 ↗
5. 10.1016/j.jcrysgro.2012.01.00710.1016/j.jcrysgro.2012.01.007 ↗
6. 10.1016/j.egypro.2017.09.34110.1016/j.egypro.2017.09.341 ↗
7. 10.1016/j.solmat.2015.06.05510.1016/j.solmat.2015.06.055 ↗
8. 10.1002/pip.243710.1002/pip.2437 ↗
9. 10.1145/2647868.265488910.1145/2647868.2654889 ↗
10. demant · Quality Rating of Silicon Wafers—A Pattern Recognition Approach · 2016
11. 10.1002/pip.270610.1002/pip.2706 ↗
12. Analysis of multicrystalline wafers originating from corner and edge bricks and forecast of cell properties
13. About the relevance of features in as-cut multicrystalline silicon wafers on solar cell performance
14. Wafer imaging and processing method and apparatus
15. Understanding neural networks through deep visualization
16. Visualizing and understanding convolutional networks
17. 10.1109/JPHOTOV.2019.290603610.1109/JPHOTOV.2019.2906036 ↗
18. 10.1063/1.223474710.1063/1.2234747 ↗
19. 10.1109/CVPR.2016.31910.1109/CVPR.2016.319 ↗
20. 10.1063/1.344343810.1063/1.3443438 ↗
21. 10.1109/JPHOTOV.2014.235879510.1109/JPHOTOV.2014.2358795 ↗
22. 10.1109/JPHOTOV.2011.216994210.1109/JPHOTOV.2011.2169942 ↗