Finite-element modeling and machine learning for efficient thermal management in battery packs

Efficient thermal management is crucial for applications ranging from electronics and automotive industries. In this talk, we present a machine learning-based platform that overcomes existing challenges in estimating the thermal conductivity of various parts of a given system (e.g., a battery pack in an electric vehicle). We present a versatile and automated method for measuring thermal conductivity and diffusivity using steady-state infrared imaging. We validate our approach through experimental and finite-element modeling, predicting the thermal conductivity of pristine and nanocomposite materials over a wide range of values (0.1 to 400 W/mK) and comparing it to conventional methods. Additionally, we investigate the role of thermal conductivity in electric vehicle battery managment, focusing on polymer-nanocomposite thermal interface materials (TIMs) and their impact on minimizing temperature variations and peak temperatures in Li-ion battery packs. Using both experimental and numerical approaches, we examine TIMs synthesized from polylactic acid (PLA), polyimide (PI), polyethylene (PE), graphene, and boron nitride nanoplatelets, integrated with liquid cooling systems. Our results demonstrate how novel TIMs and real-time thermal imaging, coupled with detailed electrochemical and fluid dynamic simulations, can enhance thermal management in battery modules, thereby preventing thermal runaways and ensuring safer, high-performance operation.