University of Michigan × Toyota Research
Solid-State Battery Degradation Research
X. Zhang, T. Gupta, Z. Wang, A. Trewartha, A. Anapolsky & K. Garikipati
Computational research on battery aging using multiphysics simulation, finite elements, image-based datasets, and HPC workflows.
- research collaboration
- Toyota
- peer-reviewed publication
- JMPS
Overview
Before product management, I worked in computational research at the University of Michigan on a Toyota Research collaboration focused on solid-state battery degradation.
The work sat at the intersection of mechanics, numerical methods, image processing, automated meshing, finite elements, and high-performance computing. I co-developed particle-scale simulation workflows and data pipelines to reason about how battery materials degrade over repeated operating cycles.
The project taught me to respect the full stack of simulation work: the physics, the numerical method, the data pipeline, and the judgment needed to decide whether a result is meaningful.
Research visuals
From cell structure and mesh preparation to lithium transport and interface fracture.
Physical system
Multi-particle cell configuration
The model represents active particles, solid electrolyte, and additive interfaces where degradation can affect transport.
Data preparation
Image-to-simulation workflow
Particle boundaries and material regions were converted into solver-ready inputs instead of hand-built toy geometry.
Numerical method
Cartesian mesh abstraction
The method avoids rebuilding a body-fitted mesh for every particle shape by representing interfaces inside a regular mesh.
Simulation sequence
Lithium during charging and discharging
A simulation of how lithium concentration changes inside battery particles over repeated charge and discharge cycles.
Simulation result
Interface fracture and degradation
Comparison of no-fracture and fractured interfaces under stress-dependent diffusion and reaction behavior.
The challenge
- Solid-state battery degradation involves coupled physical behavior across materials, interfaces, geometry, and operating conditions.
- Experimental microstructure imagery had to be converted into simulation-ready datasets before the physics could be studied computationally.
- The computational workflow needed to scale on HPC infrastructure using parallel solvers while preserving enough physical detail to be meaningful.
Research approach
Build simulation-ready data
Converted experimental microstructure imagery into datasets that could support automated meshing and particle-scale simulation workflows.
Model coupled degradation behavior
Built multiphysics simulation workflows for battery aging, including interface failure behavior and repeated operating-cycle analysis.
Scale computation
Used HPC infrastructure and parallel solvers so the simulations could move beyond small examples into larger computational studies.
Connect research to engineering judgment
Connected the numerical results to the physical mechanisms behind battery degradation.
Outcomes
Research contribution
Contributed to a Toyota-backed battery aging simulation project and a peer-reviewed JMPS publication on interface fracture in solid-state batteries.
Computational framework
Built particle-scale simulation workflows, image-processing pipelines, automated meshing inputs, and HPC-enabled analysis loops.
Product foundation
Developed the computational engineering intuition that now helps me work with simulation, robotics, and SDK customers.
What I took away
Turning experimental images into useful simulations required judgment at every step: which material boundaries to preserve, how finely to resolve them, and whether the numerical results explained the failure mechanisms.