Artificial Intelligence for Lithium-Ion Battery Failure Analysis in Electric Vehicles: A Critical Review

Lithium-ion batteries (LIBs) in electric vehicles can fail in two ways: slowly losing capacity over time, or suddenly overheating (thermal runaway), which threatens safety and reliability. Older physics-based models struggle to stay accurate once manufacturing differences and real-world driving conditions come into play, so researchers are turning to artificial intelligence (AI) to detect failures using data like voltage, current, and temperature. This review organizes recent AI research on diagnosing LIB failures in EVs. It first explains the main causes of battery degradation and failure, then reviews AI techniques used to detect them—including traditional machine learning, deep learning models (like LSTM and CNN), and newer approaches that combine physics knowledge with multiple data sources. While these AI methods work well in lab settings, several problems prevent real-world use: not enough failure data, differences between lab and real conditions, poor explainability, and no standard way to test and compare models. The review concludes that combining physics with AI, using multiple sensor types, digital twins, and continuous testing/improvement cycles are the most promising paths to make AI-based battery diagnostics practical for real EVs, not just research labs.