12.2 Time-Integration
Parameters
TIME_INTEGRATION
TIME_INTEGRATION integration_type = etc...
Specify the parameters for the time stepping procedure to use for all equations in all subdomains/element groups belonging to any given solution stagger.
Note |
Variable Name |
Type |
Default |
Description |
|
|
|
|
|
(1) |
Integration_type |
list |
[*] |
Implicit/explicit time integration type |
|
implicit / explicit |
|
|
|
|
|
|
|
|
(2) |
Equation_type |
list |
[*] |
Equation type (see Chapter 9 for details) |
|
elliptic |
|
|
Elliptic Boundary Value problem |
|
parabolic |
|
|
Parabolic Initial Boundary Value problem |
|
hyperbolic |
|
|
Hyperbolic Initial Boundary Value problem |
|
|
|
|
|
|
Analysis_type |
list |
[*] |
Analysis type |
|
direct |
|
|
Direct one-step time integration |
|
modal |
|
|
Modal integration |
|
spectral |
|
|
Spectral integration |
|
Runge_Kutta_Cash_Karp |
Runge_Kutta fifth-order Cash_Karp integration with adaptive stepsize control |
||
|
Runge_Kutta_Bulirsh_Stoer |
|
Runge_Kutta integration with Bulirsh_Stoer steps and adaptive stepsize control |
|
|
|
|
|
|
|
Alpha |
real |
[1.0] |
Algorithmic parameter |
|
|
|
|
= 1.0 for Elliptic BVP |
|
|
|
|
0.0 for Parabolic Initial BVP |
|
|
|
|
0.5 for Hyperbolic Initial BVP |
|
|
|
|
|
(3) |
Beta |
real |
[0.0] |
Algorithmic parameter |
|
|
|
|
|
• Modal and Spectral Analysis Options |
||||
|
Number_of_modes |
integer |
[0] |
Number of modes |
|
Modal_damping_ratio |
real |
[0.0] |
Modal damping ratio |
|
|
|
|
|
• Spectral Analysis Option |
||||
|
Spectrum_load_time |
integer |
[0] |
Spectrum load-time function number 1 |
|
|
|
|
|
EXAMPLE
Time_integration /
Integration_type = implicit / # implicit time integration
Equation_type = hyperbolic / # hyperbolic initial BVP
Alpha = 0.5, Beta = 0.25 # Select Trapezoidal rule
Notes/
(1) Explicit time integration is performed using a diagonal mass matrix.
Notes / (cont'd)
(2) The application of the finite element discretization to the governing balance equation(s) of a field theory generates a matrix system of equations. These equations are either zero-, first- or second order in the time variable, and are referred to in the following as elliptic, parabolic and hyperbolic, respectively. One-step algorithms are used to integrate the finite element semidiscrete equations of motion as follows (for simplicity in the presentation linear systems are used in the following):
A. Hyperbolic and Parabolic-Hyperbolic Initial Boundary Value Problems: the Newmark [3] family of finite difference time stepping algorithms is used which consist of the following equations:
where is a parameter taken to be in the interval [1/2, 3/2] and . Unconditional stability requires that be taken to be in the interval . Maximal high-frequency numerical dissipation is provided by selecting [1] for a given . Some well-known integrators are identified as follows:
Explicit central difference
Trapezoidal
A particularly convenient form of C is the Rayleigh damping matrix:
where a0 and a1 are parameters (see e.g., Section 10.1) referred to as mass and stiffness damping, respectively. Then, the resulting viscous damping can be computed as:
for
each modal frequency i (i = 1, neq) see Fig. 12.2.1. The parameters a0 and a1
may be selected to produce desired damping characteristics (e.g., by adjusting
a0 and a1 for two eigenfrequencies).
overall damping
damping for
damping for
angular frequency
Fig. 12.2.1 Effect of Viscous Damping
As a result of the numerical integration, artificial damping and period distortion are introduced. The following results have been obtained:
Where = algorithmic damping ratio, = algorithmic frequency. Note that first-order errors resulting from manifest themselves only in the form of excess numerical dissipation, and not in period discrepancies.
B. Parabolic Initial Boundary Value Problems: the generalized trapezoidal family of finite difference time stepping algorithms is used which consist of the following equations:
where is a parameter taken to be in the interval . Unconditional stability requires that be taken to be in the interval . Maximal high-frequency numerical dissipation is provided by selecting . Some well-known integrators are identified as follows:
Explicit forward Euler
Crank Nicolson/midpoint rule
Implicit backward Euler
C. Elliptic Boundary Value Problems: a backward finite difference (backward Euler) time stepping algorithms is used which consist of the following equations:
(3) For implicit time integration of hyperbolic IBVP, if , set internally to:
References / Bibliography
1. Hilber, H.M.,
"Analysis and Design of Numerical Integration Methods in Structural
Dynamics," Rep. No. EERC 76-29,
2. Hughes, T.J.R., Pister,
K.S., and
3.
Notes . .
Notes . .