Cracks, Currents, and Code: Unifying Mechanistic Insight With Machine Learning for Battery Analytics

Abstract Battery analytics is emerging as a critical enabler for understanding, predicting, and optimizing the performance and safety of lithium-ion batteries. However, the development of robust analytics is complicated by the inherently complex, coupled nature of battery systems, where electrochemical, thermal, and mechanical processes interact across multiple scales. This perspective explores the origins and propagation of defects, the influence of physical and manufacturing parameters, and the need for cross-disciplinary modeling approaches. There is a need for scalable, physics-informed frameworks that can bridge the gap between mechanistic understanding and data-driven inference. By leveraging synthetic data generation and modular modeling, such approaches can support diagnostics, accelerate design, and enhance reliability across diverse chemistries and formats. In the absence of a universal model, anchoring analytics in first-principles physics enables model reduction without compromising mechanistic fidelity, thereby facilitating the derivation of generalizable insights across the battery value chain.