University of North Dakota
Uncertainty And Sensitivity Analysis Of Lithium-Ion Battery Performance With Recycled Materials: A Physics-Based Computational Framework Approach
Abstract
dc:description.abstractThe global transition to a net-zero energy economy is increasing demand for lithium-ion batteries (LIBs) and positioning battery recycling as an essential component of future supply chains. However, recycled battery materials exhibit variability due to mixed feedstocks, heterogeneous black-mass compositions, degradation histories, and recycling processes. The resulting uncertainty in material properties can affect battery performance, yet physics-based methods for quantifying these effects remain limited.This dissertation presents a novel computational framework for uncertainty quantification (UQ) and global sensitivity analysis (GSA) of lithium-ion battery performance using recycled cathode materials. The framework combines the Doyle-Fuller-Newman (DFN) electrochemical model, implemented in PyBaMM, with a Random Forest surrogate trained on 32 high-fidelity DFN simulations spanning nine uncertain recycled-material parameters. The surrogate achieved a root-mean-squared error (RMSE) of 0.0025 Ah (approximately 0.05% of nominal cell capacity), enabling efficient evaluation of 5,000 Monte Carlo realizations. A two-stage sensitivity-analysis workflow consisting of Morris screening and Sobol variance decomposition was then applied to identify the parameters most responsible for performance variability. The results show that uncertainty in recycled-material properties propagates to measurable differences in battery performance over a 100-cycle horizon. The median capacity retention declined to approximately 85% after 100 cycles, while the 5th–95th percentile interval ranged from 84.29% to 87.16%, corresponding to a total spread of approximately 2.9 percentage points. The resulting end-of-life capacity distribution exhibited a lower-performance tail that is not captured by deterministic mean-parameter simulations. Morris and Sobol analyses consistently identified solid-phase diffusivity (Ds), particle radius (Rp), and electrode porosity (ε) as the dominant contributors to performance variability, providing a quantitative basis for prioritizing quality-control efforts within recycled-material supply chains. This work advances the field from qualitative recognition of recycled-material variability toward quantitative, physics-based prediction. The framework establishes a direct link between uncertainty in recycled materials and battery performance outcomes, providing a foundation for performance-oriented material qualification, improved quality-control strategies, and broader adoption of recycled materials within a circular battery economy.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy (PhD)
- Level thesis:degree_level
- Dissertation
- Discipline thesis:degree_discipline
- Energy Engineering
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ajumobi, Abdurahman
- Contributors dc:contributor
-
- Olusegun S. Tomomewo
Subjects
dc:subject × 1Identifiers
dc:identifier.*- Repository record dc:identifier
- https://commons.und.edu/theses/13860