Professional Documents
Culture Documents
Autostabilization in Finite Element Method
Autostabilization in Finite Element Method
For the case of static stabilization the mass matrix for Timoshenko beams is
always calculated assuming isotropic rotary inertia, regardless of the type of
rotary inertia specified for the beam section definition (“Rotary inertia for
Timoshenko beams” in “Beam section behavior,” Section 29.3.5).
It is assumed that the problem is stable at the beginning of the step and that
instabilities may develop in the course of the step. While the model is stable,
viscous forces and, therefore, the viscous energy dissipated are very small.
Thus, the additional artificial damping has no effect. If a local region goes
unstable, the local velocities increase and, consequently, part of the strain
energy then released is dissipated by the applied damping. Abaqus/Standard
can, if necessary, reduce the time increment to permit the process to occur
without the unstable response causing very large displacements.
Abaqus/Standard calculates and prints to the message file the damping
factor, c, based on the solution of the first increment of a step. In most
applications the first increment of the step is stable without the need to apply
damping. The damping factor is then determined in such a way that the
dissipated energy for a given increment with characteristics similar to the first
increment is a small fraction of the extrapolated strain energy. The fraction is
called the dissipated energy fraction and has a default value of 2.0 × 10–4. If
the default value for the dissipated energy fraction is used, the adaptive
automatic stabilization scheme discussed in the next section will be activated
automatically by default in the step.
Alternatively, you can specify the non-default dissipated energy fraction for
automatic stabilization directly.
Abaqus/CAE Usage: Step module: Create Step: General: any valid step
type: Basic: select Specify dissipated energy
fraction from the Automatic stabilization field
There are cases where the first increment is either unstable or singular (due to
a rigid body mode). In such cases it is not possible to obtain a solution to the
first increment without applying some damping. Therefore, some damping is
already applied during the first increment. The damping factor used for the
initial increment is chosen such that the average element damping matrix
component, divided by the step time, is equal to the average element stiffness
matrix component multiplied by the dissipated energy fraction. If the calculated
strain energy change in this increment indicates that the solution without
damping is stable, the damping factor is recalculated based upon the energy
method described previously. However, if the strain energy change indicates
that the solution is unstable or singular, the initially calculated damping factor
is maintained, and a warning message is issued indicating that the amount of
damping applied may not be appropriate. In many cases the amount of
damping may actually be rather large, which can affect the solution in ways
that are not desirable. Therefore, if the above mentioned warning message is
issued, check the viscous forces (VF) and compare them with the expected
nodal forces to make sure that the viscous forces do not dominate the solution.
If necessary, follow the stabilized step with another step in which stabilization
is not used or with a step in which a much smaller damping factor is used.
You can also specify the damping factor directly. Unfortunately, it is generally
quite difficult to make a reasonable estimate for the damping factor unless a
value is known from the output of previous runs; the damping factor depends
not only on the amount of damping but also on mesh size and material
behavior.
Input File Usage: Use any of the following options to specify the
damping factor directly:
*COUPLED TEMPERATURE-DISPLACEMENT,
STABILIZE, FACTOR=damping factor
*SOILS, STABILIZE, FACTOR=damping factor
*STATIC, STABILIZE, FACTOR=damping factor
*STEADY STATE TRANSPORT, STABILIZE,
FACTOR=damping factor
*VISCO, STABILIZE, FACTOR=damping factor
The adaptive automatic stabilization scheme, in which the damping factor can
vary spatially and with time, provides an effective alternative approach. In this
case the damping factor is controlled by the convergence history and the ratio
of the energy dissipated by viscous damping to the total strain energy. If the
convergence behavior is problematic because of instabilities or rigid body
modes, Abaqus/Standard automatically increases the damping factor. For
example, the damping factor may increase if an analysis takes extra severe
discontinuity or equilibrium iterations per increment or requires time increment
cutbacks. On the other hand, Abaqus/Standard may reduce the damping
factor automatically if instabilities and rigid body modes subside.
The ratio of the energy dissipated by viscous damping to the total strain energy
is limited by an accuracy tolerance that you specify. Such an accuracy
tolerance is imposed on the global level for the whole model. If the ratio of the
energy dissipated by viscous damping to the total strain energy for the whole
model exceeds the accuracy tolerance, the damping factor at each individual
element is adjusted to ensure that the ratio of the stabilization energy to the
strain energy is less than the accuracy tolerance on both the global and local
element level. The stabilization energy always increases, while the strain
energy may decrease. Therefore, Abaqus/Standard restricts the ratio of the
incremental value of the stabilization energy to the incremental value of the
strain energy for each increment to ensure that this value has not exceeded
the accuracy tolerance if the ratio of the total stabilization energy to the total
strain energy exceeds the accuracy tolerance. The accuracy tolerance is a
targeted value and can be exceeded in some situations, such as when there is
rigid body motion or when significant non-local instability occurs.
If the accuracy tolerance is not specified but the dissipated energy fraction with
the default value of 2.0 × 10–4 is used, the adaptive automatic damping
algorithm will be activated automatically with an accuracy tolerance of 0.05.
Input File Usage: Use any of the following options to activate adaptive
automatic stabilization with the default stabilization
energy tolerance:
*COUPLED TEMPERATURE-DISPLACEMENT,
STABILIZE
*SOILS, STABILIZE
*STATIC, STABILIZE
*STEADY STATE TRANSPORT, STABILIZE
*VISCO, STABILIZE
*COUPLED TEMPERATURE-DISPLACEMENT,
STABILIZE,
ALLSDTOL=accuracy tolerance
*SOILS, STABILIZE, ALLSDTOL=accuracy tolerance
*STATIC, STABILIZE, ALLSDTOL=accuracy
tolerance
*STEADY STATE TRANSPORT, STABILIZE,
ALLSDTOL=accuracy tolerance
*VISCO, STABILIZE, ALLSDTOL=accuracy tolerance
By default, the initial value of the damping factor is typically equal to the value
that would be used for automatic stabilization with a constant damping factor
(see “Calculating the damping factor based on the dissipated energy fraction”
above). In some cases additional factors that are considered with adaptive
automatic stabilization cause some differences in the initial damping factor.
Alternatively, you can specify the initial damping factor directly. The damping
factor is adjusted based on the convergence history and the accuracy
tolerance through the step.
Input File Usage: Use any of the following options to specify the initial
damping factor directly with the default stabilization
energy tolerance:
*COUPLED TEMPERATURE-DISPLACEMENT,
STABILIZE, FACTOR=damping factor, ALLSDTOL
*SOILS, STABILIZE, FACTOR=damping factor,
ALLSDTOL
*STATIC, STABILIZE, FACTOR=damping factor,
ALLSDTOL
*STEADY STATE TRANSPORT, STABILIZE,
FACTOR=damping factor,
ALLSDTOL
*VISCO, STABILIZE, FACTOR=damping factor,
ALLSDTOL
Propagating the damping factors from the immediately preceding general step
into the current step
Input File Usage: Use any of the following options to indicate that the
damping factors in the current step are propagated
from the immediately preceding general step:
*COUPLED TEMPERATURE-DISPLACEMENT,
STABILIZE, ALLSDTOL, CONTINUE=YES
*SOILS, STABILIZE, ALLSDTOL, CONTINUE=YES
*STATIC, STABILIZE, ALLSDTOL, CONTINUE=YES
*STEADY STATE TRANSPORT, STABILIZE,
ALLSDTOL,
CONTINUE=YES
*VISCO, STABILIZE, ALLSDTOL, CONTINUE=YES
The automated procedure of computing damping factors works well for many
applications. However, there are cases where the computed damping factor is
either too small, thus not controlling the instability, or too high, thus leading to
inaccurate results. These problems are more likely to occur when using a
constant damping factor—the damping factor is computed in the first
increment, which may not be representative of behavior in the rest of the step.
For example, consider a sequentially coupled thermal-stress analysis in which
a mechanical analysis reads temperatures from a previous transient thermal
analysis. Typically the thermal analysis exhibits a diffusive process, where
rapid changes in temperature occurs early in the analysis and minor changes
in temperature occur once steady state is reached. In such a case Abaqus will
compute the extrapolated strain energy based on the temperatures
corresponding to the time of the first increment (in this case there may be a
significant change in temperature for the first increment), thus yielding a larger
then expected extrapolated strain energy. This in turn leads to a damping
factor that is too large, resulting in inaccurate results.