This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

The design and certification process of aerospace and automotive composite structures is notably time-consuming and costly due to the necessity of thousands of physical tests to obtain design allowables. These tests significantly limit specimen configurations to a few stacking sequences and geometric parameters. In this presentation, we will introduce an efficient finite element (FE) - based framework designed to perform high-fidelity progressive failure analyses of fiber-reinforced polymers. This framework extends beyond the micro-scale fiber-matrix unit cell to encompass composite laminates at the macro-scale. Initially, cure-induced residual stresses are calculated using a coupled chemo-thermo-mechanical analysis. Following the determination of residual stresses and deformed geometry, a progressive failure analysis (PFA) step is conducted. The FE framework for PFA is based on a semi-discrete modeling approach, which aims to achieve a compromise between continuum and discrete methods. The laminate is modeled in a layer-wise manner (meso-scale), where each lamina is connected via cohesive elements or contact. The enhanced semi-discrete damage model (ESD2M) toolset includes an automated smart meshing strategy with failure mode separation, the enhanced Schapery theory with a novel generalized mixed-mode law, and various probabilistic modeling strategies. These combined modules make the computational tool highly effective in accurately capturing various failure modes, such as matrix cracks, fiber tensile failure, and delamination. The underlying principle of the semi-discrete mesh is to partition each lamina layer into strips and bulk regions. The finite width of the strips is chosen to be much smaller than the bulk distance for higher accuracy. While the bulk elements only capture fiber failure modes, the strip elements can capture both fiber and matrix failure. Each strip is assigned with random values of failure strengths associated with matrix failure, while the axial strength and delamination (cohesive) properties are randomized using a global field function. Since the model can be run in a probabilistic manner, Monte Carlo simulations can be performed to predict the structural response and its scatter. The material damage and failure is modeled using a non-linear constitutive model, combining Schapery theory and the crack band method, which is implemented in a VUMAT user subroutine. The model is validated using various test cases, including unnotched and open-hole laminates subjected to quasi-static tensile and compressive loading.