Multi-criteria and real-time control of continuous battery cell production steps using deep learning

Electric vehicles driven by batteries are a key part of a sustainable mobility sector. Unfortunately, battery cell production is still associated with various negative environmental impacts, the use of critical raw materials and high manufacturing costs. The rising battery demand forces automotive original equipment manufacturers to drastically increase their capabilities over the next decades while fulfilling economical and ecological requirements. Continuous production technologies bear the potential to meet future battery cell demands by enabling higher throughputs compared to established batch processes. The control and optimization of continuous battery cell production steps with respect to product quality, manufacturing costs and environmental impacts is challenging due to high parameter spaces as well as temporal dependencies of production processes. Therefore, this study develops a controller that performs real-time optimization by proposing set parameters leading to desired quality, minimal costs and impacts of manufacturing activity. The controller is implemented using a deep learning model incorporating sequential information of the production process. A continuous mixing process with data acquired from a battery cell pilot line is used to validate the outlined controller. As result, the implementation for this use case achieves a relative error of 7.63% across all controllable parameters.