Multiphysics-informed machine learning for assessing battery safety risk evolution with degradation

The growing use of electric vehicles (EVs) and shift to higher-energy-density battery chemistry have made battery safety a top research priority.

The statistics show that >80% of EV fire accidents occurred without explicit abuse, where battery degradation can play a role.

Nevertheless, most existing battery safety models focus on thermal runaway triggered by various forms of abuse, largely overlooking the evolution of battery safety with degradation during normal cycling conditions.

MIRACLE therefore aims to address this gap by developing an efficient and precise approach for simulating and predicting the progression of lithium-ion batteries into unsafe states under non-abuse conditions with a focus on degradation effects.

Specifically, MIRACLE will (1) develop a physics-based model that fundamentally bridges the gap between degradation and failure mechanisms, capturing when and how batteries will transition into unsafe states as a consequence of multiple coupled degradation mechanisms over significant periods of time, (2) embed the multiphysics into machine learning to enable fast, precise and physically interpretable predictions of battery cycling to failure, and (3) conduct model-based analyses to formulate the first state variable for estimating battery safety status evolution during cycling, addressing a key missing piece in current battery management systems and thereby enhancing the reliability and safety of battery operation.

The multidisciplinary research leverages the collective expertise of the host group and the fellow, promoting effective two-way knowledge transfer.

The fellowship programme sets the stage for significant personal and professional growth, paving the way for the fellow to become an independent academic in Europe.