Bahador Bahmani*, Katherine A. Acton**, Justin M. Garrard*, Connor Sherod**, Philip L. Clarke*, Reza Abedi*
*University of Tennessee Space Institute / Knoxville (UTK), ** University of St. Thomas (UST)
The use of statistical volume elements for homogenization
and fracture analysis of quasi-brittle materials
Acknowledgments: This material is based upon work supported by the National Science Foundation under Grant No. 1538332 and two supplementary ROA and RUI awards (UTK / UST)
Motivation
➢ Response of brittle materials is highly sensitive to their
microstructure due to the lack of energy dissipative mechanismsAl-Ostaz et. al. (1997)Kozicki and J Tejchman (2007)
Nonlinear response
Different fracture patterns Variations in strength
and toughness
Approach: Statistical Volume Elements (SVEs) for brittle fracture
Statistical Volume Element (SVE)
12 Random field statistics3
1. By using SVEs material inhomogeneities & sample to sample variations (randomness) are preserved.
2. Still no need to resolve all microscale details!A very efficient and accurate model for fracture modeling
Macroscopic
domainRandom field realization4 Macroscopic fracture
simulation5
Ostoja-Starzewski 1998
RVE limit: 𝑙 is very large
PDF of a material property, e.g. stiffness
SVEs appropriate for fracture modeling(still reduce problem size by homogenization)
• RVEs are good for many elastic regime problems• For fracture modeling (particularly quasi-brittle material):
• Need to preserve spatial inhomogeneity• Randomness (sample to sample variations)
Spatial inhomogeneity: more realistic fracture patterns
1x1 SVEs 2x2 SVEs 4x4 SVEs Uniform strength
Homogenized by:
Corresponding fracture patterns under horizontal tensile loading
Smaller SVEs are used for homogenization: More realistic dynamic fracture patterns are observed
Voronoi tessellation vs. square SVEs: Effect of shape / size of SVEs
Original Volume Element Square SVEs:
Concern: Intersection of inclusions
Voronoi Tessellation based SVEs:
- SVE is generated by Voronoi cells around inclusions.- Better homogenized properties are expected.
No inclusion intersection problem
Sample SVEs at 2, 5, 10, 25% of VE size
Square SVEs:- Large variations especially for small SVE sizes.
Voronoi SVEs:- Much smaller variation of
homogenized properties.- Fracture strength does not
decrease significantly asSVE size increases.
Large variation and degradation of strength as SVE size increases
PDE of bulk modulus Size effect plot for bulk modulus Size effect plot for normal strength
Anisotropy of homogenized properties
Interesting question: How does fracture strength for a given angle (for example 10 degrees) behave as a random field?
Mean strength෩𝑺𝒏
Angle-dependent strength𝒔𝒏(𝜽)
Measures of anisotropy:- Span / mean- COV = s.div / mean
span
Objectives:- Define measures of anisotropy for fracture and
elasticity properties.- How does the shape of SVE affect anisotropy?- By what rate anisotropy decreases as SVE size
increases?
Small SVE: d = 3.125 Large SVE: d = 12.5
Square SVEs: normal strength
- High strength- Low strength
Small SVE: d = 3.125 Large SVE: d = 12.5
- High strength- Low strength
Square SVEs: shear strength
Small SVE: d = 3.125 Large SVE: d = 12.5
Voronoi SVEs: normal strength
Isotropic!
𝛿 = 2
𝛿 = 10
Small SVE: d = 3.125 Large SVE: d = 12.5
Voronoi SVEs: shear strength
- High strength- Low strength
Isotropic!
𝛿 = 2
𝛿 = 10
- Shape of SVE result in artificially weaker and stronger angles. - Voronoi-based SVEs result in an isotropic response (for a macroscopically isotropic material) when SVE size is small.
Effect of SVE size on measures of anisotropy
- X axis: SVE size - Y axis: Measure of anisotropy (coefficient of variance) for different elastic and fracture properties.
As SVE size increases -> homogenized properties become more isotropic.
Voronoi-based SVEs (blue) result in more isotropic properties than square SVEs (black)
Fracture properties (red) are more anisotropic than elastic properties (black)
Another application: Microcracked rock + circular VEs
PDFs of microcrack length:
- Weibull model (m =1
2, 1, 2, 4)
- vs. Power law- Same mean length for all
m = 1
2m = 1 m = 2 m = 4
- Yuen-Long marbles all with crack density 24.3% and same mean microcrack length.
- Weibull model with m =1
2has a distribution closest to power law. It has a wider range of crack lengths.
2a (mm)
PD
F
Strength (MPa)
PD
F
PDFs of strengths obtained by size 1 SVEs for different distributions
m = 1
2m = 1
m = 2 m = 4m = 4 results in higher strength and more uniform fracture strength field.
Macroscopic fracture simulation under dynamic tension in vertical direction
Strain
Stre
ss
Macroscopic strain-stress:
m =1
2: weakest
m = 4: strongest
Concluding remarks:
- Homogenization by SVEs maintains sample-to-sample variation and material inhomogeneity which are very important in brittle fracture – Shape and size of SVE affect the inhomogeneity and anisotropy of homogenized properties – The distribution of defects affect macroscopic response.
Fracture strength fields