DigiCell Explores Battery Modelling at its 4th Public Workshop
On 1 July 2026, DigiCell engaged with its stakeholders by hosting its fourth public workshop, entitled "From Data-Driven to Physics-Informed Modelling for Batteries – A View from the EU Project DigiCell".
The workshop focused on battery modelling across different scales, ranging from data-driven machine learning and physics-based finite element modelling to industrial applications at battery pack level and battery manufacturing. We were delighted to welcome 28 participants, whose active engagement and the workshop's open and collaborative atmosphere led to lively discussions and insightful questions throughout the event.
The programme featured the following presentations:
- Introduction to Statistical and Machine Learning Modelling: Selected Concepts and Pitfalls, presented by Fabian Guignard (METAS, Swiss Federal Institute of Metrology)
- Battery State-of-Health Estimation from a Minimal Set of Impedance Frequencies Using Machine Learning, also presented by Fabian Guignard (METAS)
The discussions following Fabian's presentations centred on identifying the most informative electrochemical impedance spectroscopy (EIS) frequencies for predicting battery state of health. Participants explored the reasoning behind reducing the number of frequencies, the possibility of selecting an optimal subset, and the complementary roles of statistical analysis and electrochemical interpretation.
- Multiscale Computational Framework for Degradation Mechanism Analysis in Lithium-Ion Batteries, presented by Ahmad Azizpour (Johannes Kepler University Linz, JKU)
During the discussion, Ahmad highlighted two key conclusions. First, a multiscale modelling approach—from cathode particle fracture to the charge-discharge behaviour of a complete battery cell—can successfully capture interactions across multiple length and time scales, enabling accurate prediction of capacity degradation. Second, chemo-mechanical phase-field fracture models provide valuable insights into the mechanisms of cathode particle cracking and can support the development of more durable battery electrodes.
- Reduced-Order Electrochemical Aging Modelling of Industrial Battery Packs, presented by Alberto Romero (KREISEL Electric GmbH).
- Modelling Battery Manufacturing via Physics-Informed Machine Learning, presented by Francisco Fernandez (Université de Picardie Jules Verne, UPJV).
The discussion after Francisco's presentation focused on the assumptions underlying the modelling approach and the validation of simulation results against experimental data. Participants also discussed the scope of the process models, clarifying which manufacturing steps are included and which are not. The conversation further explored how UPJV's work compares with other research activities in the field. Francisco concluded by emphasizing that well-calibrated and experimentally validated models developed at prototype and pilot-line scale have strong potential to support process optimisation in future battery gigafactories.
The final Q&A session reinforced the importance of accounting for cell-to-cell variability and applying rigorous validation methodologies to ensure reliable and robust machine learning models for battery development.