Multiphysics nature of Lithium-ion Battery safety issues
Abstract
Safety issues of lithium-ion batteries (LIBs) are usually initiated from an internal short circuit (ISC) that can be triggered by external accidental abusive loadings. The generated heat and the increased temperature would lead to several complicated physio-chemical changes of the batteries, e.g., thermal runaway (TR). Thus, investigation of the multiphysics behaviors of lithium-ion batteries becomes a vital task to understand the battery safety issues. Experimental characterization and numerical simulation are essential ways to understand the underlying nature of the multiphysics behavior of batteries. However, experimental observation may only provide insufficient data due to the limitation of experimental technology, particularly for in-situ or operando experiment methodologies. Herein, a multiphysics modeling framework is developed. The framework provides a fundamental description and understanding of the safety behaviors of the battery cells. Next, the multiphysics framework is used to evaluate the safety risk of LIB cells. Two kinds of key safety risks evaluation problems are defined: ISC triggering risk and safety risk level. Machine learning (ML) models with data-driven algorithms are used to fast and accurately solve the related regression and classification problems. In terms of ISC triggering risk, an ISC risk evaluation model is first developed based on the training dataset generated by the combination of experimental data and simulation data. A Representative Volume Element (RVE) based mechanical model, which can predict accurate mechanical behaviors at a much lower calculation time cost, is established to assist the data generation. The safety risk prediction high-level performance of the Support Vector Regression (SVR) predictors are indicated by various testing cases and scenarios. In terms of safety risk level evaluation, a safety risk level classification model is developed to classify the cell’s safety levels under various work conditions. Random Forest (RF) classifiers are used to construct the model and realize the cell state classification based on only a short period of voltage and current signals. The multiphysics model is used as a surrogate model to generate as much as training samples, that cover various State of Charges (SOCs), short circuit resistances, and Charging/discharing-rates (C-rates), are generated. The prediction results show that the classifiers have a good performance and robustness. Finally, two typical safety issues: cell defect and TR propagation are systematically studied. The defective cells are characterized both electrochemically and mechanically to discuss the consequent ISC and thermal runaway triggering behaviors and modes. The multiphysics model will be used to provide necessary auxiliary instructions of the related mechanisms. Possible defect detection and identification indicators are also summarized. TR propagation behaviors of battery packs are also experimentally and numerically investigated. The pack level TR model is developed based on the multiphysics model via doing some simplifications and adding the coupling relationships among cells. Two major thermal spread modes are discovered, and their governing factors are discussed. Thus, the TR propagation mechanism is revealed to some extent. This study comprehensively investigates the multiphysics behavior of LIB cells under mechanical abusive loadings, highlights the promise of the LIB cell multiphysics modeling framework. The multiphysics models provide fundamental understanding and insights of battery safety issues, as well as an innovative solution for risk evaluation and safety risk recognition.