This study introduces a real-time probabilistic safety assessment of a 18650 cylindrical battery. The physics-based failure scenarios from battery abuse are mapped onto Bayesian Network framework, enabling a quantitative estimation of monitoring parameters, health indicators and loss of control events. The model (encompassing multiple states, weightage factors, interdependencies, and uncertainties) is validated with 70 experimental data points of the same geometry, comprising 4 categories of abuses (surface heating, fast overcharging, external short circuit, and internal deformation). Through testing, probabilistic safety thresholds are established for 3 observed events: gas formation, venting, and fire and explosion.Sensitivity analysis reveals state of charge (at SOC > 50 %) and surface temperature (at Tsurf > 90°C) as the most influential parameters while inadequate cooling system and current interrupt device (CID) as influential safeguards for thermal runaway and fire to occur. Moreover, the identified safety thresholds are subjected to validation and testing against multiple health indicators to draw important insights. Probability, P(fire)deformation>P(fire)surface heat as internal deformation lowers onset to thermal runaway. In culmination, a safer operating area (SOA) is proposed, leveraging the identified thresholds and most sensitive parameters. The proposed SOA with 4 distinct boundaries is not only consistent with the temporal variations in cell temperature during thermal abuse test, but also provides a robust approach to monitor the safety of a battery.
Read full abstract