Automatic determination of mechanical properties from Molecular Dynamic Stress-Strain curves using Regression Fringe Response.

Molecular dynamics (MD) simulations have become a powerful tool for studying polymers, designing new materials, and composites. Mechanical behavior in MD simulations are typically evaluated through stress-strain curves. However, MD-generated stress-strain curves are often obscured by thermal noise from atomic vibrations. This noise complicates the reliable derivation of mechanical properties and hinders objective and reproducible analysis. The noise in the stress-strain curves has been removed by using a Butterworth low-pass filter, enabling clearer interpretation of stress-strain data. Building on this, a novel analysis framework the Regression Fringe Response (RFR) is introduced to systematically determine the Young's modulus, yield strength, and Poisson's ratio from filtered stress-strain curves. The RFR-method is explained, considerations for phenomena such as toe regions and plasticity are incorporated, and validation is performed across a range of polymer systems. The results demonstrate that RFR-method provides an accurate, robust, and reproducible approach for quantifying mechanical properties from MD simulations.