Difference between revisions of "Transversely Isotropic Viscoelastic Material"
Line 58: | Line 58: | ||
=== Isotropic with Time-Dependent Bulk Modulus === | === Isotropic with Time-Dependent Bulk Modulus === | ||
The available [[Viscoelastic Material|isotropic viscoelastic]] material is limited to materials with time-independent bulk modulus because that is a good approximation for most isotropic, viscoelastic materials. | The available [[Viscoelastic Material|isotropic viscoelastic]] material is limited to materials with time-independent bulk modulus because that is a good approximation for most isotropic, viscoelastic materials. These transversely-isotropic materials, however, do not place any restrictions on which properties are time dependent. As result, it can model an isotropic material with a time-dependent bulk modulus as a special case. Imagine an isotropic material with ''K(t)'' and ''G(t)'' as time-dependent bulk and shear moduli, respectively. To model using a transversely isotropic material, choose ''G<sub>A</sub>(t) = G<sub>T</sub>(t) = G(t)'' along with: | ||
| |
Revision as of 12:17, 16 February 2021
Constitutive Law
(This material is available only in OSParticulas because it is still in development)
This anisotropic MPM material is a small strain, linear viscoelastic material that extends the Viscoelastic Material to model anisotropic viscoelasticity. The stress (σ) and strain (ε) are related by:
[math]\displaystyle{ \sigma(t) = \mathbf{C}(t) * \varepsilon(t) }[/math]
Here [math]\displaystyle{ * }[/math] indicates convolution (or Boltzman's superposition) between time-dependent stiffness tensor ([math]\displaystyle{ \mathbf{C}(t) }[/math]) and strain tensor. In Voight-notation with unique axis in the z direction, the time-dependent stiffness tensor is
[math]\displaystyle{ \mathbf{C}(t) = \left[\begin{array}{cccccc} K_T(t)+G_T(t) & K_T(t)-G_T(t) & \ell(t) & 0 & 0 & 0 \\ K_T(t)-G_T(t) & K_T(t)+G_T(t) & \ell(t) & 0 & 0 & 0 \\ \ell(t) & \ell(t) & n(t) & 0 & 0 & 0 \\ 0 & 0 & 0 & G_A(t) & 0 & 0 \\ 0 & 0 & 0 & 0 & G_A(t) & 0 \\ 0 & 0 & 0 & 0 & 0 & G_T(t) \end{array}\right] }[/math]
Here [math]\displaystyle{ K_T(t) }[/math] is the plane strain, bulk modulus, [math]\displaystyle{ G_T(t) }[/math] is the transverse shear modulus, [math]\displaystyle{ G_A(t) }[/math] is the axial shear modulus, and [math]\displaystyle{ n(t) }[/math] and [math]\displaystyle{ \ell(t) }[/math] give time-dependence of the C33 and C13=C23 elements of the stiffness tensor (as defined here). The time dependence of each property is modeled with a sum of exponentials:
[math]\displaystyle{ K_T(t) = K_{T0} + \sum_{k=1}^{N_{KT}} K_{Tk} e^{-t/\tau_{KT,k}} \qquad G_T(t) = G_{T0} + \sum_{k=1}^{N_{GT}} G_{Tk} e^{-t/\tau_{GT,k}} \qquad G_A(t) = G_{A0} + \sum_{k=1}^{N_{GA}} G_{Ak} e^{-t/\tau_{GA,k}} }[/math]
[math]\displaystyle{ n(t) = n_0 + \sum_{k=1}^{N_n} n_k e^{-t/\tau_{n,k}} \qquad\qquad \ell(t) = \ell_0 + \sum_{k=1}^{N_\ell} \ell_k e^{-t/\tau_{\ell,k}} }[/math]
In terms of axial modulus [math]\displaystyle{ E_A }[/math], and Poisson's ratio, [math]\displaystyle{ \nu_A }[/math], we can write:
[math]\displaystyle{ n(t) = E_A(t) + 4K_T(t)\nu_A(t)^2 \qquad\qquad \ell(t) = 4K_T(t)\nu_A(t) }[/math]
This material lumps all these time dependencies into [math]\displaystyle{ n(t) }[/math] and [math]\displaystyle{ \ell(t) }[/math], but not that selection of those properties will determine time dependencies of [math]\displaystyle{ E_A(t) }[/math] and [math]\displaystyle{ \nu_A(t) }[/math]
[math]\displaystyle{ \nu_A(t) = {\ell(t)\over 2K_T(t)} \qquad {\rm and} \qquad E_A(t) = n(t) - \frac{\ell(t)^2}{K_T(t)} }[/math]
TIViscoelastic 1 and 2
TIViscoelastic 1 and TIViscoelastic 2 give identical materials but with different initial orientations. TIViscoelastic 1 has the unrotated axial direction along the z (or θ if axisymmetric) axis (see above tensors) while TIViscoelastic 2 has unrotated axial direction along the y (or Z if axisymmetric) axis (exchange y and z directions in above tensors). You can change the unrotated direction to any other orientation when defining material points by selecting rotation angles. For 2D analyses, the two options are needed to allo for axial direction in the x-y (or R-Z if axisymmetric) analysis plane (TIViscoelastic 2) or normal to that plane (TIViscoelastic 1). For 3D analyses, only TIViscoelastic 1 is allowed (and it in the only one needed).
Isotropic with Time-Dependent Bulk Modulus
The available isotropic viscoelastic material is limited to materials with time-independent bulk modulus because that is a good approximation for most isotropic, viscoelastic materials. These transversely-isotropic materials, however, do not place any restrictions on which properties are time dependent. As result, it can model an isotropic material with a time-dependent bulk modulus as a special case. Imagine an isotropic material with K(t) and G(t) as time-dependent bulk and shear moduli, respectively. To model using a transversely isotropic material, choose GA(t) = GT(t) = G(t) along with:
[math]\displaystyle{ K_T(t) = K(t) + \frac{G_T(t)}{3},\quad n(t)= K(t) + \frac{4G_T(t)}{3},\quad {\rm and}\quad \ell(t) = K(t) - \frac{2G_T(t)}{3} }[/math]
Finally, any other material properties (such as thermal expansion coefficients) should be set to the special cases for an isotropic material.
Fibrous Materials
A potential use for a transversely isotropic viscoelastic material is to model fiber-reinforced composites including wood (although transversely isotropic is not a great model for wood because it cannot represent low transverse shear modulus compared to radial and tangential transverse tensile moduli). If the axial direction is the fiber direction, one might expect that time dependence will be much weaker in the axial direction than in the transverse direction. This behavior could be approximated by setting [math]\displaystyle{ n }[/math] and [math]\displaystyle{ \ell }[/math] independent of time. Because an elastic material has [math]\displaystyle{ n=E_A+4K_T\nu_A^2 }[/math] and [math]\displaystyle{ \ell = 2K_t\nu_A }[/math], this approximation implies [math]\displaystyle{ E_A }[/math], [math]\displaystyle{ K_T }[/math], and [math]\displaystyle{ \nu_A }[/math] are all independent of time. The final result is that only [math]\displaystyle{ G_T(t) }[/math] and [math]\displaystyle{ G_A(t) }[/math] are time dependent. This model is one approximation, but would have zero creep in the axial direction and therefore no use in modeling that response (unless experiments also show zero creep).
A second option might be to let [math]\displaystyle{ K_T(t) }[/math] depend on time while [math]\displaystyle{ E_A }[/math] and [math]\displaystyle{ \nu_A }[/math] remain independent of time. This approach does not work and results in non-physical response to axial loading. The problem appears that it predicts that [math]\displaystyle{ n(t) }[/math] decreases in time. For isotropic materials, [math]\displaystyle{ n(t) }[/math] increases in time and it also likely increases for fibrous materials as well. In brief, to model viscoelastic properties of fibrous material that might include effects in the axial direction, the modeling will need to input all five time-dependent properties allowed by this material. Those properties should be based on experimental observations.
Effective Time Implementation
This material handles variations in temperature and solvent concentration by the same methods used for isotropic viscoelastic materials.
Material Properties
The unusual task for this material is to use multiple terms to define the exponential series used for up to five material properties.
Property | Description | Units | Default |
---|---|---|---|
GT0 | The long term (or fully-relaxed) transverse shear modulus | pressure units | none |
GA0 | The long term (or fully-relaxed) axial shear modulus | pressure units | none |
KT0 | The long term (or fully-relaxed) plane-strain bulk modulus | pressure units | none |
en0 | The long term (or fully-relaxed) C33 element of the stiffness tensor | pressure units | none |
ell0 | The long term (or fully-relaxed) C13=C23 elements of the stiffness tensor | pressure units | none |
ntaus | The number of relaxation times of the previous long-term property that was entered. This property is only needed in XML files and must come before any subsequent Pk or tauk properties. In scripted files, the number is automatically determined from the number of relaxation times you provide. | none | none |
Pk | The next property in the series for the previous long-term property that was entered. Use multiple Pk values for each term on the series. | pressure units | none |
tauk | The next relaxation time in the series for the previous long-term property that was entered. Enter multiple tauk values for each term on the series. | time units | none |
Tref | Reference temperature to shift relaxation times using the WLF equation. If Tref<0, then no shifting is done and relaxation times will be independent of temperature. | K | -1 |
C1 | Coefficient in WLF equation used when Tref≥0 to shift relaxation times | none | 17.44 |
C2 | Coefficient in WLF equation used when Tref≥0 to shift relaxation times | none | 51.6 |
mref | Reference concentration to shift relaxation times using the WLF-style equation. If mref<0, then no shifting is done and relaxation times will be independent of concentration. | K | -1 |
Cm1 | Coefficient in WLF-style equation used when mref≥0 to shift relaxation times | none | 10 |
Cm2 | Coefficient in WLF-style equation used when mref≥0 to shift relaxation times (must be positive) | none | 0.0625 |
TI Properties | Enter thermal expansion, solvent expansion, and poroelasticity properties as usual for transversely isotropic materials. | varies | varies |
(other) | Properties common to all materials | varies | varies |
The material properties need to define the time dependence of 3 properties (when fibrous is 1) or 5 properties (when fibrous is 0). The process for each one is to enter the long-term value first (GT0, GA0, KT0, en0, ell0) and then to follow each one by ntaus (only needed in XML files) and by one Pk and tauk value for each term in the series.
When fibrous is 1, you instead enter time-independent values for EA and nuA. When fibrous is 0, their initial values calculated from:
[math]\displaystyle{ \nu_A(0) = {\ell(0)\over 2K_T(0)} \qquad {\rm and} \qquad E_A(0) = n(0) - 4K_T(0)\nu_A^2 = n(0) - \frac{\ell(0)^2}{K_T(0)} }[/math]
The remaining, initial transverse properties are:
[math]\displaystyle{ \frac{1}{E_T(0)} = \frac{1}{4K_T(0)} +\frac{1}{4G_T(0)} + {\nu_A(0)^2\over E_A(0)} \qquad {\rm and} \qquad \nu_T(0) = \frac{E_T(0)}{2G_T(0)}-1 }[/math]
The above initial values, or the properties at time zero, are sums of all Pk terms for that property. For example
[math]\displaystyle{ G_T(0) = G_{T0} + \sum_{k=1}^{N_{GT}} G_{Tk} }[/math]
For valid modeling, the initial Poisson's ratios must satisfy
[math]\displaystyle{ -1\lt \nu_T(0)\lt 1 \qquad -\sqrt{E_A(0)\over E_T(0)} \lt \nu_A(0) \lt \sqrt{E_A(0)\over E_T(0)} \qquad {E_T(0)\nu_A(0)^2\over E_A(0)} \lt {1-\nu_T(0)\over 2} }[/math]
These relations must apply for all time, but only the initial values are validated before starting a simulation.
History Variables
This material tracks internal history variables (one for each relaxation time and each component of stress) for implementation of linear viscoelastic properties, but currently none of these internal variables are available for archiving.