Generation of periodic representative volume elements based on high-resolution computed tomography images
Topic(s) : Experimental techniques
Co-authors :
Rui GUO (BELGIUM), Mahoor MEHDIKHANI (BELGIUM), Christian BREITE (BELGIUM), Yentl SWOLFS (BELGIUM)Abstract :
Representative Volume Elements (RVEs) are a common method in microstructural analyses and multiscale simulations for their cost-effective computational requirements. Their accuracy heavily depends on realistic microstructural features, such as randomly distributed fibres [1] and resin-rich pockets [2]. Advanced imaging techniques like computed tomography (CT) and scanning electron microscopy are extensively used to capture these complex microstructural features. However, the conversion of the acquired images into periodic geometric RVEs requires addressing the following critical issues:
High-resolution CT imaging, despite its robustness, results in measured fibre diameters slightly smaller than those measured by a laser diffraction system (LDS). This difference can affect the calculated fibre volume fraction (V_f) and hence the mechanical response of image-based models [3].
To correct fibre diameters measured by CT images, a correction factor can be applied via fitting with LDS measurements [3]. Additionally, segmentation can create errors regarding the measured position of fibre centres. These two sources of error may lead to unrealistic instances of fibres touching each other, or too small inter-fibre distances, resulting in difficulties in creating image-based finite element models.
The process of creating periodic RVEs from images involves moving and shrinking the fibres close to the image border, which demands a careful balance between the two: maintaining the original fibre position as much as possible while minimizing the deviation from the original V_f. This balance becomes challenging when processing images with dense fibre packings near the image border.
Our new proposed algorithm uses an iterative process, which can automatically adjust the inter-fibre distance for those cases of touching- or too close fibres without losing representativity for the fibre diameter and V_f. Furthermore, it creates periodicity at the boundaries of the microstructures for high V_f (as shown in Fig. 1). This work is highly useful for the microstructural analysis and RVE representation of composites, helping to gain a deeper understanding of the impact of their microstructure on mechanical properties. The validity of our algorithm will be demonstrated by measuring homogeneous elastic properties under the periodic boundary condition and comparing them with analytical solutions.
Fig. 1. The overlay of fibre centre and greyscale images: (a) Original non-periodic microstructures and (b) periodic microstructures: the image size is 96 × 96 pixel with 1 pixel = 0.65 μm.
High-resolution CT imaging, despite its robustness, results in measured fibre diameters slightly smaller than those measured by a laser diffraction system (LDS). This difference can affect the calculated fibre volume fraction (V_f) and hence the mechanical response of image-based models [3].
To correct fibre diameters measured by CT images, a correction factor can be applied via fitting with LDS measurements [3]. Additionally, segmentation can create errors regarding the measured position of fibre centres. These two sources of error may lead to unrealistic instances of fibres touching each other, or too small inter-fibre distances, resulting in difficulties in creating image-based finite element models.
The process of creating periodic RVEs from images involves moving and shrinking the fibres close to the image border, which demands a careful balance between the two: maintaining the original fibre position as much as possible while minimizing the deviation from the original V_f. This balance becomes challenging when processing images with dense fibre packings near the image border.
Our new proposed algorithm uses an iterative process, which can automatically adjust the inter-fibre distance for those cases of touching- or too close fibres without losing representativity for the fibre diameter and V_f. Furthermore, it creates periodicity at the boundaries of the microstructures for high V_f (as shown in Fig. 1). This work is highly useful for the microstructural analysis and RVE representation of composites, helping to gain a deeper understanding of the impact of their microstructure on mechanical properties. The validity of our algorithm will be demonstrated by measuring homogeneous elastic properties under the periodic boundary condition and comparing them with analytical solutions.
Fig. 1. The overlay of fibre centre and greyscale images: (a) Original non-periodic microstructures and (b) periodic microstructures: the image size is 96 × 96 pixel with 1 pixel = 0.65 μm.