Transfer Learning for Battery Cell Manufacturing: Review On Applications, Challenges, and Benefits

The increasing demand for battery cells in the automotive sector forces battery cell manufacturers to accelerate product development and scale‐up of their production processes. To achieve this, digitalization expands data acquisition and enables the use of machine learning methods. These methods can be utilized for quality assurance, process optimization, and more. However, the application of machine learning requires a large amount of data, which is especially difficult to acquire in the pre‐series production due to the vast number of parameter variations, the complex process chain and the small production quantities. Additionally, these obstacles lead to high costs in pre‐series production of battery cells. Therefore, there is a need for methods, which are able to train machine learning models on small datasets and increase their generalization abilities. A possible method for this is transfer learning which can use parts of previous models on a new, but similar problem. This method has scarcely been applied in the context of battery cell manufacturing and motivates a structured literature review about its applicability, challenges, and benefits. This study compares transfer learning methods applied to other industries with machine learning approaches of battery cell manufacturing to identify and evaluate potential use cases.