Estimation and Uncertainty Quantification of Yield Via Strain Recovery Simulations

Abstract

In computational materials science, predicting the yield strain of crosslinked polymers remains a challenging task. A common approach is to identify yield via the first critical point of stress-strain curves produced by molecular dynamics simulations. However, the simulated data can be excessively noisy, making it difficult to extract meaningful results. In this work, we discuss an alternate method for identifying yield on the basis of residual strain computations. Notably, the associated raw data produce a sharper signal for yield through a transition in their global behavior. As we show, this transition can be analyzed in terms of simple functions (e.g. hyperbolas) that admit straightforward uncertainty quantification techniques.