Multi-physics Preconditioning for Thermally Activated Batteries

TL;DR

The paper tackles the computational bottleneck of solving tightly coupled electrochemical systems in thermal battery simulations by introducing a hierarchical block Gauss-Seidel preconditioner implemented with Trilinos Teko, combining physics-aware subblock solvers such as SA-AMG and domain-decomposition methods. It demonstrates strong and weak scalability up to 51.3 million degrees of freedom on 2048 processors, achieving sub-second setup and near-sub-second solves for the end-to-end electrochemical problem. The study shows that domain-decomposition alone struggles for monolithic solves, while the block-based approach delivers robust convergence and parallel scalability across subblocks including Stefan-Maxwell diffusion, liquid and solid voltages, and their coupling. The results indicate the approach enables higher-resolution, 3D thermal battery models and highlights future extensions to further parallelize setup and explore alternative Jacobi-like strategies to push scalability even further.

Abstract

Thermal batteries, also known as molten-salt batteries, are single-use reserve power systems activated by pyrotechnic heat generation, which transitions the solid electrolyte into a molten state. The simulation of these batteries relies on multiphysics modeling to evaluate performance and behavior under various conditions. This paper presents advancements in scalable preconditioning strategies for the Thermally Activated Battery Simulator (TABS) tool, enabling efficient solutions to the coupled electrochemical systems that dominate computational costs in thermal battery simulations. We propose a hierarchical block Gauss-Seidel preconditioner implemented through the Teko package in Trilinos, which effectively addresses the challenges posed by tightly coupled physics, including charge transport, porous flow, and species diffusion. The preconditioner leverages scalable subblock solvers, including smoothed aggregation algebraic multigrid (SA-AMG) methods and domain-decomposition techniques, to achieve robust convergence and parallel scalability. Strong and weak scaling studies demonstrate the solver's ability to handle problem sizes up to 51.3 million degrees of freedom on 2048 processors, achieving near sub-second setup and solve times for the end-to-end electrochemical solve. These advancements significantly improve the computational efficiency and turnaround time of thermal battery simulations, paving the way for higher-resolution models and enabling the transition from 2D axisymmetric to full 3D simulations.

Figures (7)

• Figure 1: 2D axisymmetric simulation domain for multi-physics simulations (not to scale). Note that the collector, anode, separator, and cathode layers are repeated $N=20$ times.
• Figure 1: Scalability of GMRES preconditioned via DD(0)-ILU(0) for the liquid-phase species fractions. Strong scalability is maintained up to $n/P\sim6000.0$, with sub-second setup and solve times achieved across all problem scales. The weak scaling efficiency reaches 95% for the combined setup and solve phase, highlighting the efficiency of the preconditioning strategy for this subblock.
• Figure 2: Scalability of GMRES(30) preconditioned via SA-AMG for the liquid-phase pressure. Setup and solve times remain below 0.3 seconds at larger processor counts, with strong scalability observed up to $n/P\sim11500.0$. The weak scaling efficiency of 93% demonstrates the robustness of the approach for handling the liquid-phase pressure equations.
• Figure 3: Scalability of GMRES(30) preconditioned via SA-AMG for the liquid-phase voltage. End-to-end setup and solve times remain below 0.4 seconds, with strong scalability achieved up to $n/P\sim{11500}$. The weak scaling efficiency of 95% confirms the effectiveness of the approach for solving diffusion-dominated equations in this subblock.
• Figure 4: Scalability of GMRES preconditioned via SA-AMG for the solid-phase voltage. Strong scalability is observed up to $n/P\sim16000.0$, with a combined setup plus solve time below 0.48 seconds. The weak scaling efficiency of 91% highlights the ability of the approach to handle the challenges posed by highly variable electrical conductivity in the solid-phase voltage equation.