Objective and automated calibration of progressive damage models for finite element simulation of fiber reinforced composites

The determination of suitable input parameters for the computational analysis of progressive damage in composites is mostly based on trial-and-error attempts leading to subjective models with limited general use. This study explores the application of genetic algorithms for an objective and automated calibration of continuum damage models in finite element simulations. The general applicability and robustness are demonstrated in three case studies containing carbon and glass fiber-reinforced laminates subjected to progressive tensile and compressive fracture tests. The load–displacement curves of these fracture tests build the basis for optimizing the input parameters of the damage models. The validation in independent load cases and a good correlation of damage patterns between experimental observations and simulations show that the optimized parameters can produce accurate and physically meaningful results. The optimization process requires up to 250 finite element simulations which is significantly less than comparable data-driven approaches incorporating machine learning.