Multi-scale Damage Analysis of Composites

Abstract

A multi-scale approach for progressive damage analysis is developed for various types of composites including braided textile composites, short fiber reinforced composites, and woven fabric composites. Finite element (FE) models of the aforementioned composites are developed using Python scripts which can automatically generate FE models and run the analysis together with the multi-scale progressive damage model. Biaxial and triaxial braided composites are modeled using finite element methods (FEM) to predict the effective material properties: for biaxial braids, diamond pat-tern (1/1), regular pattern (2/2), and Hercules pattern (3/3) are modeled; for triaxial braids, regular pattern (2/2) is modeled. A micromechanical approach is adopted to calculate material properties of the tows composing the braided composites. By applying the periodical boundary conditions (PBC), four representative unit cells (RUC) in correspondence to the four patterns of braids, are analyzed and compared. Subsequently, effective material properties are obtained with the braiding angle varying from 15° to 75° in an increment of 5°, from which the variation of the engineering constants with the braiding angle is studied. The prediction results are compared with the experimental values for two material systems, and good agreement is achieved. The strength of braided textile composites is predicted using a multi-scale approach bridging the mesoscale and microscale regimes. Mesoscale finite element models of representative unit cells of biaxial and triaxial braided composites are developed for predicting strength. The constituent stresses of tows inside the braided unit cell are calculated using micromechanics. Correlations between mesoscale stresses and microscale constituent stresses are established by using stress amplifica-tion factors (SAF). After calculating microscale stresses, a micromechanics-based progressive dam-age model is employed to determine the damage statues of braided composites. A volume-averaging homogenization method is utilized to eliminate damage localization in the matrix of tows, and a parametric study is performed to evaluate the effects of dam-age homogenization. Subsequently, the ultimate strength is predicted for braided composites in which the braiding angle ranges from 15° to 75°. The prediction results are compared with the experimental stress-strain curves, and good agreement is observed. The dynamic material response of braided fabrics with thermosetting and thermoplastic resin under impact loading is predicted using a multi-scale approach bridging the mesoscale and microscale regimes. Mesoscale finite element models of braided composites are developed by repeating the corresponding representative unit cells for impact analysis. Microscale constituent stresses are calculated from mesoscale stresses using stress amplification factors. After calculating microscale stresses, a micromechanics-based progressive damage model is employed to detect the damages inside braided composites. Different progressive damage models are established for both thermosetting and thermoplastic resin systems. The impact analysis is performed for both biaxial braided composites with various braiding angles and triaxial braided composites at various impact velocities. Element deletion method is adopted to remove elements meeting the element deletion criteria during the impact analysis. The effects of different resin systems on the impact response of braided composites are observed from the predictions. A hybrid multi-scale approach combining a virtual representative volume element and a finite element unit cell is developed for progressive failure prediction of short fiber rein-forced composites. The virtual representative element volume represents the fiber orientation and distribution of the whole composites, from which the global mechanical behavior can be obtained. The finite element unit cell captures the local mechanical responses of each short fiber by transforming global strains to local strains. The constituent strains of the fiber, matrix and interface are calculated from the local strain using micromechanics. Correlations between mesoscale local strains and microscale constituent strains are established by using strain amplification factors. After computing microscale stresses, a progressive damage model is employed to determine the damage statues of all constituents. A homogenization method is then employed to eliminate damage localization in the matrix and interface of the unit cell. The stress field given by the hybrid approach is compared with that of the conventional finite element model, and good agreement is observed. The stress-strain prediction with progressive failure using the hybrid approach is com-pared with the experimental values, and good agreement is also achieved. Mesoscale finite element representative unit cell models are established for various types of woven fabrics: plain-weave, twill weave, and satin weave. The mechanical behavior of woven fabrics is predicted using a multi-scale approach bridging the mesoscale regime and microscale regime which provides the failure observation inside the constituents. The constituent stresses of the fiber and matrix in the warp and fill tows of the woven fabric unit cell are calculated using micromechanics. Correlations between mesoscale tow stresses and microscale constituent stresses are established by using stress amplification factors. After calculating microscale stress-es, a micromechanics-based progressive damage model is employed to determine the damage statuses in each constituent of woven fabric composites. For the matrix of tows, a volume-averaging homogenization method is utilized to eliminate damage localization by smearing local damages over the whole matrix region of the unit cell. Subsequently, the ultimate strength is predicted for woven composites with different tow architectures. The prediction results are compared with the experimental data, and good agreement is observed, which verified the validity and flexibility of the multi-scale approach.