Back to Search
Start Over
Data mining in lithium-ion battery cell production.
- Source :
-
Journal of Power Sources . Feb2019, Vol. 413, p360-366. 7p. - Publication Year :
- 2019
-
Abstract
- Abstract Data mining methods are used to analyze and improve production processes in a lithium-ion cell manufacturing line. The CRISP-DM methodology is applied to the data captured during the manufacturing processes. Key goals include the identification of process dependencies and key quality drivers as well as the prediction of the product quality before the cumbersome and costly formation and aging procedure. Several Data mining methods, such as Generalized Linear Model (GLM), Artificial Neural Networks (ANN), Support Vector Regression (SVR), Decision Trees (DT), Random Forest (RF), and Gradient Boosted Trees (GBT) are compared and evaluated. Best results are yielded by an application of GLM, RF, and GBT for prediction of battery cell capacity before the expensive formation process. Key quality drivers identified are the electrode fabrication processes, as well as the electrolyte filling process during cell assembly. This is, to our knowledge, the first time data from a real battery production line has been systematically processed and analyzed along the whole process chain. The results of this paper can assist to manufacture better batteries and to reduce costs of lithium-ion cells by providing a systematic procedure for data acquisition and by lowering scrap rates during production. Graphical abstract Image 1 Highlights • Data mining approaches were applied to a real battery production line. • A systematic procedure for data acquisition, processing, and analysis is given. • Electrode fabrication and electrolyte filling are identified as key quality drivers. • The results can help to decrease battery production cost by reducing scrap rates. [ABSTRACT FROM AUTHOR]
Details
- Language :
- English
- ISSN :
- 03787753
- Volume :
- 413
- Database :
- Academic Search Index
- Journal :
- Journal of Power Sources
- Publication Type :
- Academic Journal
- Accession number :
- 134184738
- Full Text :
- https://doi.org/10.1016/j.jpowsour.2018.12.062