At the international conference 20th CIRP International Conference on Intelligent Computation in Manufacturing Engineering (CIRP ICME '26), held in Ischia, Italy, our researcher and PhD student Tesa Herceg participated in and presented research conducted as part of the project INSIGHT (Intelligent data models for the production and recycling of battery cells) between the University of Zagreb, Faculty of Mechanical Engineering and Naval Architecture and the BMW Group Battery Cell Competence Centre (BCCC). <\/p>\n\n\n\n
CIRP ICME '26 brings together researchers and industry experts from around the world to present and discuss advances in computational methods, machine learning, simulation, and data-driven approaches applied to modern manufacturing engineering. The technical programme of the conference covered a broad range of topics, from intelligent process control and quality prediction to digital twins and sustainable manufacturing <\/p>\n\n\n\n
Tesa presented her paper in the session dedicated to data-driven approaches in manufacturing, specifically in the Battery Production session. <\/p>\n\n\n\n
Her work, Machine Learning-Based Prediction of Electrolyte Wetting Rate in the Lithium-ion Battery Cell Filling Process, addresses the electrolyte filling and wetting stage which is a critical and time-intensive step in the production of lithium-ion battery cells. Uniform and complete wetting of the electrode layers by the electrolyte is essential for achieving high cell quality and electrochemical performance, yet the process is difficult to monitor and control directly. <\/p>\n\n\n\n
The paper proposes a machine learning approach to predict the electrolyte wetting rate from real production data. The wetting rate is derived from process measurements using a Washburn-based equation, and the input features include electrode and jelly roll characteristics from preceding production steps, together with filling process parameters identified as most influential through Design of Experiments (DoE). Several machine learning models were trained and compared, while Random Forest and Gradient Boosting achieved the best predictive performance, with R\u00b2 \u2248 0.88. Feature importance analysis further identified electrode length, jelly roll diameter, and filling pressure as the most influential factors governing the wetting rate. These findings support a deeper understanding of the filling process and can reduce the need for repeated experimental investigations. <\/p>\n\n\n\n
These results represent a further contribution to the INSIGHT project's overarching goal of developing intelligent data models that support and optimise every stage of battery cell production and recycling. In addition to the scientific programme, the conference offered numerous opportunities for networking and knowledge exchange among researchers, professors, and industry professionals, providing valuable insight into the latest trends and open challenges in intelligent manufacturing. <\/p>","protected":false},"excerpt":{"rendered":"