automated in-line defect classification and localization in solar cells for laser-based repair
TRANSCRIPT
Automated in-Line Defect Classification and Localization in Solar Cells for Laser-Based Repair
Jorge Rodríguez –Araújo, Antón García-Díaz
AIMEN Technology Center, Porriño, Spain
ISIE 2014, Istambul, 2-6-2014
www.aimen.es | [email protected] 2
Index
1. Motivation and Innovative Character.
2. Proposed Solution.
3. Diagnostic and Defect Segmentation.
4. Repair Process Decision.
5. Experimental Results.
6. Conclusions and future work.
Index
www.aimen.es | [email protected] 3
Motivation and Innovative Character
4www.aimen.es | [email protected]
Motivation and Innovative Character
• Solar cells are made of silicon (156x156mm).• Monocrystalline and Polycrystalline silicon (more common materials).
• The manufacturing process produces defects.
Photovoltaic solar cells
busbars
front back
5www.aimen.es | [email protected]
Motivation and Innovative Character
• Electroluminescence imaging.
• Shunts and Cracks (most important).(> 50%) • Lacks of metallization (less frequent) (< 20%).• Finger interruptions (reduced effects). (> 58%)
• Shunts and Cracks may be repaired.• Cutting or isolating using laser technology.
• New pieces of cells are obtained.
Defects
Detection
www.aimen.es | [email protected] 6
Proposed Solution
7www.aimen.es | [email protected]
Proposed Solution
• Laser Repair System (laser process).• Defects Inspector (defects identification and laser control).• Data Input Block (electroluminescence provider).
Solar cell repair system
8www.aimen.es | [email protected]
Proposed Solution
• Cell Diagnostic (defects segmentation from EL).• Repairing Decision (laser process decision).• Cell Alignment (laser process correction).
Defects inspector
www.aimen.es | [email protected] 9
Diagnostic and Defect Segmentation
10www.aimen.es | [email protected]
Diagnostic and Defect Segmentation
• Human experts examine EL images.– Changes on texture.– Changes on the shape of texture boundaries.
• Defect Diagnostic (texture based)– Decomposition: for automatic features generation.– Adaptation: for enhancement of features.– Pixel level classification: for multiclass identification.
Bio-inspired texture approach
Texture approach
Defects
Log Gabor filters
11www.aimen.es | [email protected]
Diagnostic and Defect Segmentation
• Trained with only 4 images.– 4 cracks presents.– 4 shunts presents.
• Type of defect and boundary contour are identified.
Some examples of identified defects.
Pixel level identification
www.aimen.es | [email protected] 12
Repair Process Decision
13www.aimen.es | [email protected]
Repair Process Decision
• Decision rules:.– Isolate shunts not on a busbar.– Cut shunts on a busbar or close to one.– Cut pieces to remove cracks.
Solar cell repair decision
Examples of repair decision.
14www.aimen.es | [email protected]
Repair Process Decision
• Closing morphological operation.• Inversion and thresholding.• Edge detection and contour.• Bounding box localization.
Cell location and alignment
Steps on cell localization.
www.aimen.es | [email protected] 15
Experimental results
Experimental Results
www.aimen.es | [email protected] 16
Laser-based repair unit
Camera perspective.
Scanner working area.
Laser repair system.
Experimental results
www.aimen.es | [email protected] 17
Laser scanner calibration
Galvo-scanner working area.
21www.aimen.es | [email protected]
Experimental Results
• 47 cells considered.• 45 cells repaired.• 69% waste reduction.
Processing results
Defect type
# of Defects
TP FP FN Repaired
Shunts 43 43 4 0 42
Cracks 38 36 6 2 36
All 81 79 10 2 45*
* This value refers to the number of cells repaired.
www.aimen.es | [email protected] 22
Conclusions and Future Work
23www.aimen.es | [email protected]
Conclusions and Future work
• Defect segmentation and classification.• In-line solar cells repair system.• Able to isolate and cut defects.• 69% of rejected cells reutilization.
• Discriminate more defects, like metallization.• More complex repair strategies.• Isolation based on efficiency estimation.
Conclusions
Future work
www.aimen.es | [email protected] 24
AIMEN – Central y Laboratoriosc/ Relva 27 A
36410 – O PORRIÑO (Pontevedra)Telf.+34 986 344 000 – Fax. +34 986 337 302
Delegación Tecnológica Madrid
Avda. del General Perón, 32, 8 A28020 – MADRID (Madrid)
Telf.+34 687 448 915
Delegación Tecnológica A Coruña
Fundación Mans – Paideia Pol. Pocomaco - Parcela D-22 - Oficina 20A
15190 – A CORUÑA (A Coruña)Telf. +34 617 395 153
Thank you for your attentionJorge Rodríguez Araujo | Research Engineer
Ph +34 986 344 000 | [email protected]