Bayesian Model Calibration of a Mechanical Finite Element Model of an LR61 Alkaline Battery

Abstract

The growing use of batteries in vehicular applications has raised the salience of how mechanical loads, such as those from crashes, can lead to battery failure and subsequently battery pack fires. Calibrating numerical models is an involved process because batteries are complicated devices containing many materials with complex behaviors. Many techniques in the inverse problem field can improve existing model tuning procedures by extracting more information out of experimental tests and identifying which future observations can most constrain the remaining uncertainty. This thesis applies a Bayesian Model Calibration (BMC) approach to the calibration of the plasticity and damage models for an LR61 alkaline battery. BMC produces both a tuned parameter set and an uncertainty quantification (UQ) via a probability distribution over the parameter space. The resulting calibrated LR61 simulation produced accurately reflects experimentally observed deformation behavior and illustrates the usefulness of the BMC process in mechanical battery simulation development.