I-Corps: Machine learning-driven design of lithium-ion batteries

Indiana University, Bloomington IN

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is the development of high-performance, cost-effective, and safe lithium-ion batteries (LIBs) based on components optimized with a novel computational platform that can minimize the discovery and design-cycle time scales. The new simulation and design technologies have a direct impact on segments including developers in LIB companies and electric vehicle companies, and university/laboratory researchers. This technology can accelerate adoption of electric vehicles, which can reduce emissions and achieve better fuel economy. This I-Corps project is based on the development of geometrically accurate numerical models of particles, binders, and conductivity enhancers in LIB electrodes. Finding optimal LIB electrode architectures and corresponding LIB configurations that maximize energy and power densities, rate capability, and safety during operation presents significant challenges. This project proposes use of Bayesian Machine Learning (BML) algorithms. The proposed BML-driven design algorithms incorporate statistical microstructure characterization and reconstruction methods to support the LIB electrode design, multitask Gaussian Processes (GPs) to integrate experimental data with multi-fidelity and multiscale simulations, and a multi-objective acquisition function to achieve the LIB multi-objective design. The use of a Bayesian optimization approach allows the systematic exploration of the design space. The result is an LIB electrode architecture optimizing the trade space of performance, cost, and safety. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.