Viscoelastic materials are characterized by a combination of elastic behavior, which stores energy during deformation, and viscous behavior, which dissipates energy during deformation.
The elastic behavior is rate-independent and represents the recoverable deformation due to mechanical loading. The viscous behavior is rate-dependent and represents dissipative mechanisms within the material.
A wide range of materials (such as polymers, glassy materials, soils, biologic tissue, and textiles) exhibit viscoelastic behavior.
Following are descriptions of the viscoelastic constitutive models, which include both small- and large-deformation formulations. Also presented is time-temperature superposition for thermorheologically simple materials and a harmonic domain viscoelastic model.
For additional information, see Viscoelasticity in the Structural Analysis Guide.
The following formulation topics for viscoelasticity are available:
The following figure shows a one dimensional representation of a generalized Maxwell solid. It consists of a spring element in parallel with a number of spring and dashpot Maxwell elements.
The spring stiffnesses are μi, the dashpot viscosities are ηi, and the relaxation time is defined as the ratio of viscosity to stiffness, τi = ηi / μi.
In three dimensions, the constitutive model for a generalized Maxwell model is given by:
(4–9) |
where:
| σ = Cauchy stress |
| e = deviatoric strain |
| Δ = volumetric strain |
| τ = past time |
| I = identity tensor |
and G(t) and K(t) are the Prony series shear and bulk-relaxation moduli, respectively:
(4–10) |
(4–11) |
where:
| G0, K0 = relaxation moduli at t = 0 |
| nG, nK = number of Prony terms |
| α i G, α i K = relative moduli |
| τi G, τi K = relaxation time |
For use in the incremental finite element procedure, the solution for Equation 4–9 at t1 = t0 + Δt is:
(4–12) |
(4–13) |
where si and pi are the deviatoric and pressure components, respectively, of the Cauchy stress for each Maxwell element.
By default, the midpoint rule is used to approximate the integrals:
(4–14) |
(4–15) |
An alternative stress integration method is to assume a constant strain rate over the time increment. Then the stress update is:
(4–16) |
(4–17) |
The model requires input of the parameters in Equation 4–10 and Equation 4–11. The relaxation moduli at t = 0 are obtained from the elasticity parameters input using the MP command or via an elastic data table (TB,ELASTIC). The Prony series relative moduli and relaxation times are input via a Prony data table (TB,PRONY), and separate data tables are necessary for specifying the bulk and shear Prony parameters.
For the shear Prony data table, TBOPT = SHEAR,
NPTS = nG, and the constants in
the data table follow this pattern:
| Table Location | Constant |
|---|---|
| 1 | α 1 G |
| 2 | τ1 G |
| ... | ... |
2(NPTS) - 1 | α nG G |
2(NPTS) | τnG G |
For the bulk Prony data table, TBOPT = BULK,
NPTS = nK, and the constants in
the data table follow this pattern:
| Table Location | Constant |
|---|---|
| 1 | α 1 K |
| 2 | τ1 K |
| ... | ... |
2(NPTS) - 1 | α nK K |
2(NPTS) | τnK K |
Use TBOPT = INTEGRATION with the Prony table
(TB,PRONY) to select the stress update algorithm. If the table is
not defined, or the value of the first table location is equal to 1, then the
default midpoint formula from Equation 4–14 and Equation 4–15 are used. If the value of the first table location is equal
to 2, then the constant strain rate formula from Equation 4–16 and
Equation 4–17 are used.
This model is used when the large-deflection effects are active (NLGEOM,ON).
To account for large displacement, the model is formulated in the co-rotated configuration using the co-rotated deviatoric stress Σ = RTsR, where R is the rotation obtained from the polar decomposition of the deformation gradient. The pressure component of the Cauchy stress does not need to account for the material rotation and uses the same formulation as the small-deformation model.
The deviatoric stress update is then expressed as:
(4–18) |
where ΔR = R(t1)RT(t0) is the incremental rotation.
Parameter input for this model resembles the input requirements for the small-deformation viscoelastic model.
The large-strain viscoelastic constitutive model is a modification of the model proposed by Simo. Modifications are included for viscoelastic volumetric response and the use of time-temperature superposition.
The linear structure of the formulation is provided by the generalized Maxwell model. Extension to large-deformation requires only a hyperelastic model for the springs in the Maxwell elements. Hyperelasticity is defined by a strain energy potential where, for isotropic materials:
(4–19) |
where:
|
|
|
|
|
|
The second Piola-Kirchhoff stress in the Maxwell element springs is then:
(4–20) |
and the large-deformation stress update for the Maxwell element stresses is given by:
(4–21) |
(4–22) |
where:
|
|
|
|
An anisotropic hyperelastic model can also be used for Equation 4–19, in which case the form of the Maxwell element stress updates are unchanged.
This model requires the Prony series parameters to be input via the Prony data table (as described in Small Deformation). The hyperelastic parameters for this model are input via a hyperelastic data table (TB,HYPER). For more information, see Hyperelasticity.
For a physical interpretation of the Prony series formulation, the dissipated energy in the viscoelastic material is the energy used to deform the dashpots in the Maxwell elements. The increment of energy used by the dashpots over a time increment is:
where, for dashpot i, is the deviatoric stress,
is the increment in deviatoric strain,
is the pressure, and
is the increment in volumetric strain.
For an additive decomposition of the Maxwell element strains, the dashpot strain increment is given by:
where is the strain increment,
is the Maxwell spring compliance, and
is the stress increment. Defining
as the instantaneous compliance, the Maxwell spring compliance
is:
The dashpot strain increment and compliance can be applied to both deviatoric and
volumetric components, and then corresponds to either the deviatoric or volumetric relative
modulus for Maxwell element
i.
A scalar equivalent viscous strain increment can be defined as:
where:
|
|
For large deformation, the incremental dissipation uses strain increments calculated from the rate of deformation tensor and a Maxwell compliance tensor calculated from the instantaneous hyperelastic stiffness and the relative moduli.
For thermorheologically simple materials, the influence on the material behavior due to changing temperature is the same as that due to changing time. For these materials, a rate-dependent material response, P (a function of temperature and time), can be reduced to:
(4–23) |
where:
| T = current absolute temperature |
| Tr = constant absolute reference temperature |
| ξ = shifted time given by ξ = t / A(T), where A(T) = shift function. |
The constitutive equations are solved in the shifted time scale. This method has the potential to reduce the experimental effort required to determine the material parameters but requires the determination of the shift function.
The shift functions, A(T), are evaluated in an absolute temperature scale determined by adding the temperature offset value (TOFFST) to the current temperature, reference temperature, and fictive temperature in the shift functions.
The following forms of the shift function are available:
The Williams-Landel-Ferry shift function has the form:
(4–24) |
where C1 and C2 are material parameters. (The shift function is often given in the literature with the opposite sign.)
The parameters are input via a shift function data table (TB,SHIFT).
For the Williams-Landel-Ferry shift function, TBOPT =
WLF, and the required input constants are:
| Table Location | Constant |
|---|---|
| 1 | Tr |
| 2 | C1 |
| 3 | C2 |
Two forms of the Tool-Narayanaswamy shift function are available, one of which includes a fictive temperature.
The first form is given by:
(4–25) |
where is the scaled activation energy.
The parameters are input in a shift function data table (TB,SHIFT).
For the Tool-Narayanaswamy shift function, TBOPT = TN,
and the required input constants are:
| Table Location | Constant |
|---|---|
| 1 | Tr |
| 2 |
|
The second form of the Tool-Narayanaswamy shift function includes an evolving fictive temperature. The fictive temperature is used to model material processes that contain an intrinsic equilibrium temperature that is different from the ambient temperature of the material. The shift function is given by:
(4–26) |
where:
| X = weight parameter |
| TF = absolute fictive temperature. |
The partial fictive temperatures are calculated in the relative temperature scale defined by the input material parameters. The evolving fictive temperature is given by:
(4–27) |
where:
| nf = number of partial fictive temperatures |
| Cfi = fictive temperature relaxation coefficient |
| Tfi = partial fictive temperature |
The evolution of the partial fictive temperature is given by:
(4–28) |
where:
|
|
| 0 (superscript) = values from the previous time step |
The fictive temperature model modifies the volumetric thermal strain model and gives an incremental thermal strain as:
(4–29) |
where:
| ΔT = temperature increment over the time step. The temperature increment in the first increment is the body temperature at the end of the increment minus the fictive thermal strain reference temperature, Tref, defined in the shift function table. If Tref is 0 or undefined in the shift function table, the shift function reference temperature, Tr, is used to calculate the temperature increment in the first time step. |
| α g and α l = glass and liquid coefficients, respectively, of thermal expansion given by: |
(4–30) |
(4–31) |
Equation 4–30 and Equation 4–31 are evaluated at the relative current and fictive temperatures.
The parameters are input in a shift function data table (TB,SHIFT).
For the Tool-Narayanaswamy with fictive temperature shift function,
TBOPT = FICT, NPTS =
nf, and the required input constants are:
| Table Location | Constant |
|---|---|
| 1 | Tr |
| 2 | H/R |
| 3 | X |
4 to 3(NPTS +
1) | Tf1, Cf1, τf1, Tf2, Cf2, τf2, ..., Tfn, Cfn, τfn |
3(NPTS + 1) + 1
to
3(NPTS + 1) + 5 | α g0, α g1, α g2, α g3, α g4 |
3(NPTS + 1) + 6
to
3(NPTS + 1) + 10 | α l0, α l1, α l2, α l3, α l4 |
3(NPTS + 1) + 11 | Tref |
Other shift functions can be accommodated via the user-provided subroutine UsrShift ,
described in the Programmer's Reference.
Given the input parameters, the routine must evolve the internal state variables, then return the current and half-step shifted time.
For use in harmonic analyses, the generalized Maxwell model can be used to provide a constitutive model in the harmonic domain.
Assuming that the strain varies harmonically and that all transient effects have subsided, Equation 4–9 has the form:
(4–32) |
where:
|
|
|
|
|
|
|
|
Comparing Equation 4–32 to the harmonic equation of motion, the material stiffness is due to the storage moduli and the material damping matrix is due to the loss moduli divided by the frequency.
The following additional topics for harmonic viscoelasticity are available:
The storage and loss moduli are related to the Prony parameters by:
(4–33) |
(4–34) |
Input of the Prony series parameters for a viscoelastic material in harmonic analyses follows the input method for viscoelasticity in the time domain detailed above.
Storage and loss moduli can also be input as piecewise linear functions of
frequency on a data table for experimental data. Isotropic elastic moduli can be input for the complex
shear, bulk and tensile modulus as well as the complex Poisson's ratio.
The points for the experimental data table (input via the TBPT command) have frequency as the independent variable, and the dependent variables are the real component, imaginary component, and tan(δ). If the imaginary component is empty or zero for the data point, the tan(δ) value is used to determine it; otherwise tan(δ) is not used.
Input complex shear modulus on an experimental data table
(TB,EXPE) with TBOPT = GMODULUS. The
data points are defined by:
| Position | Value |
|---|---|
| 1 | f |
| 2 | |
| 3 | |
| 4 | |
Input complex bulk modulus on an experimental data table
(TB,EXPE) with TBOPT = KMODULUS. The
data points are defined by:
| Position | Value |
|---|---|
| 1 | f |
| 2 | |
| 3 | |
| 4 | |
Input complex tensile modulus on an experimental data table
(TB,EXPE) with TBOPT = EMODULUS. The
data points are defined by:
| Position | Value |
|---|---|
| 1 | f |
| 2 | |
| 3 | |
| 4 | |
Input complex Poisson's ratio on an experimental data table
(TB,EXPE) with TBOPT = NUXY. The data
points are defined by:
| Position | Value |
|---|---|
| 1 | f |
| 2 | |
| 3 | |
| 4 | |
Using experimental data to define the complex constitutive model requires elastic
constants (defined via MP or by an elastic data table
[TB,ELASTIC]). The elastic constants are unused if two sets of
complex modulus experimental data are defined. This model also requires an empty
Prony data table (TB,PRONY) with TBOPT
= EXPERIMENTAL.
Two elastic constants are required to define the complex constitutive model. If only one set of experimental data for a complex modulus is defined, the Poisson's ratio (defined via MP or by elastic data table) is used as the second elastic constant.
For thermorheologically simple materials in the frequency domain, frequency-temperature superposition is analogous to using time-temperature superposition to shift inverse frequency. The Williams-Landel-Ferry and Tool-Narayanaswamy (without fictive temperature) shift functions can be used in the frequency domain, and the material parameter input follows the shift table input described in Time-Temperature Superposition.
Frequency-temperature superposition can be used with either the Prony series complex modulus or any of the experimental data for complex moduli or Poisson's ratio.
The magnitude of the real and imaginary stress components are obtained from expanding Equation 4–32 and using the storage and loss moduli from either the Prony series parameters or the experimental data:
(4–35) |
(4–36) |
where:
| Re(σ) = real stress magnitude |
| Im(σ) = imaginary stress magnitude |