Dresden 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 5: Poster

AKPIK 5.7: Poster

Thursday, March 12, 2026, 15:00–16:30, P5

Hybrid Machine Learning Framework for Predicting Cycling-Induced Ageing in Lithium-Ion Batteries — •Sandhra Ganesh — University of Münster Institute of Physical Chemistry AK Heuer 48149 Münster, Germany

Predicting cyclical capacity fade is critical for assessing the long-term reliability and second-life potential of lithium-ion batteries. Traditional physics-based ageing models provide valuable interpretability but are often computationally expensive and depend on parameters that are difficult to obtain experimentally. Conversely, purely data-driven methods offer efficiency but typically struggle to generalise across operating conditions and lack physical grounding.This work proposes a hybrid modelling framework that integrates physics domain knowledge with deep learning to more accurately capture cycling-induced degradation. The framework incorporates physically meaningful feature extraction from voltage, capacity, and operational profiles, together with physics-guided constraints that ensure realistic degradation behaviour without requiring detailed mechanistic models. The approach aims to improve predictive accuracy, interpretability, and transferability across varying conditions and datasets. Its effectiveness will be evaluated through cross-condition generalisation studies and assessments of practical cycle-life prediction accuracy.

Keywords: Hybrid; Machine learning; Physics; Battery; Aging