Thermal runaway (TR) of lithium-ion batteries (LIBs) represents a critical safety challenge in EV applications. This study explores the potential of data-driven safety management strategies for mitigating TR risks in EVs. To minimize the impact of external environmental factors on the degradation of LIBs, experiments were conducted using an accelerating rate calorimeter (ARC). The intrinsic thermal behavior of six nickel–cobalt–manganese (NCM) cells at different states of health (SOH) and operating temperatures has been captured in created adiabatic conditions. Multiple sensors were deployed to monitor the temperature and electrochemical and environmental parameters throughout the degradation process until TR occurred. The results show that both the thermal and electrochemical stability of LIBs have been affected, exhibiting consistent thermal patterns and early electrochemical instability. Furthermore, even under adiabatic conditions, the degradation of LIBs show synergistic effects with environmental parameters such as chamber temperature and pressure. Correlation analysis further revealed the coupling relationships between the monitored parameters. Through calculating their correlation coefficients, the results indicate advantages of combining thermal, electrochemical, and environmental parameters as being to characterize the degradation of LIBs and enhance the identification of TR precursors. These findings stress the importance of considering the battery-environment system as a whole in safety management of EVs. They also provide insights into the development of data-driven safety management strategies, highlighting the potential for achievement and integration of anomaly detection, diagnosis, and prognostics functions in current EV management frameworks.
The rapid progress of large language models (LLMs) has fostered the development of domain-specific variants in law, medicine, and finance. However, existing legal LLMs still struggle to generate contextually grounded and regulation-compliant responses in complex scenarios of fair competition review. To address this, we present FairAgent, a collaborative multi-agent framework that unifies data refinement and reinforcement learning for legal reasoning. FairAgent integrates two core modules: (1) EchoCourt, a closed-loop data generation and refinement pipeline that constructs high-quality question–answer pairs through generation, critique, and optimization guided by a hierarchical Fairness Knowledge Forest; and (2) a two-stage outcome-based reinforcement learning mechanism that progressively teaches the model to invoke and integrate external retrieval in reasoning. We further enhance learning stability through a RAG-based rollout and retrieval-mask loss. Extensive evaluations demonstrate that FairAgent significantly improves reasoning accuracy, interpretability, and compliance in fair competition review compared with state-of-the-art baselines, establishing a scalable framework for retrieval-augmented legal intelligence.