Learning Quality Rating of As-Cut mc-Si Wafers via Convolutional Regression Networks
- Resource Type
- Authors
- Aditya Kovvali; Stella X. Yu; Patrick Virtue; Stefan Rein; Matthias Demant
- Source
- IEEE Journal of Photovoltaics. 9:1064-1072
- Subject
- Messtechnik und Produktionskontrolle
Feature extraction
detection
02 engineering and technology
01 natural sciences
Convolutional neural network
law.invention
law
0103 physical sciences
Solar cell
Point (geometry)
Electrical and Electronic Engineering
Mathematics
010302 applied physics
Artificial neural network
Image (category theory)
DenseNet
imaging
021001 nanoscience & nanotechnology
Condensed Matter Physics
Inline-Wafer-/Prozessanalytik und Produktionskontrolle
Electronic, Optical and Magnetic Materials
Silicium-Photovoltaik
Feature (computer vision)
Photovoltaik
regression
Production (computer science)
0210 nano-technology
Charakterisierung von Prozess- und Silicium-Materialien
Algorithm
CNN
- Language
- ISSN
- 2156-3403
2156-3381
This paper investigates deep convolutional neural networks (CNNs) for the assessment of defects in multicrystalline silicon (mc-Si) and high-performance mc-Si wafers for solar cell production based on photoluminescence (PL) images. We identify and train a CNN regression model to forecast the $I\text{--}V$ parameters of passivated emitter and rear cells from given PL images of the as-cut wafers. The presented end-to-end model directly processes the PL image and does not rely on the human-designed image feature. Domain knowledge is replaced by a model based on a huge variety of empirical data. The comprehensive dataset allows for the evaluation of the generalizability of the model with test wafers from bricks and manufacturers not presented in the training set. We achieve mean absolute prediction errors as low as $\text{0.11}\%_{\text{abs}}$ in efficiency for test wafers from “unknown” bricks, which improves handcrafted feature-based methods by $\text{35}\%_{\text{rel}}$ at simultaneously lower computational costs for prediction. Samples with high prediction errors are investigated in detail showing an increased iron point defect concentration.