Comparison of implicit and explicit procedures

Abaqus/Standard is more efficient for solving smooth nonlinear

problems; on the other hand, Abaqus/Explicit is the clear choice
for a wave propagation analysis. There are, however, certain
static or quasi-static problems that can be simulated well with
either
e ther program. Typ
Typically,
cally, these are problems that usually would
be solved with Abaqus/Standard but may have difficulty
converging because of contact or material complexities, resulting
i a large
in l number
b of
f iterations.
it ti S h analyses
Such l are expensive
i i
in
Abaqus/Standard because each iteration requires a large set of
linear equations
q to be solved.
 Comparison of implicit and explicit procedures
Whereas Abaqus/Standard must iterate to determine the
solution to a nonlinear problem, Abaqus/Explicit determines the
solution without iterating by explicitly advancing the kinematic
state from the previous increment. Even though a given analysis
may requ
require
re a large number of ttime
me increments
ncrements us
using
ng the expl
explicit
ct
method, the analysis can be more efficient in Abaqus/Explicit if
the same analysis in Abaqus/Standard requires many iterations.

Another advantage of Abaqus/Explicit is that it requires much

less disk space
p and memoryy than Abaqus/Standard
q for the same
simulation. For problems in which the computational cost of the
two programs may be comparable, the substantial disk space and
memory savings
savin s of Abaqus/Explicit make it attractive.
attractive
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods
Comparison of Implicit and Explicit Methods

STATIC ‘QUASI’ STATIC DYNAMIC

PUNCH

BLANK

DIE

Structural Problems Metal Forming Impact Problems

ΣF = 0 ΣF≈ 0 Σ F = ma

IMPLICIT METHOD
EXPLICIT METHOD
 Comparison of Implicit and Explicit Methods
Implicit Time Integration:
•Inertia effects ([C] and [M]) are typically not included
•Average acceleration - displacements evaluated at time t+Dt:

{u t + Δt } = [K ]−1 {Fta+ Δt }
Linear Problems:
¾Unconditionally stable when [K] is linear
¾Large time steps can be taken
Nonlinear problems:
¾Solution obtained using a series of linear approximations
(Newton Raphson)
(Newton-Raphson)
¾Requires inversion of nonlinear stiffness matrix [K]
¾Small iterative time steps are required to achieve convergence
¾Convergence is not guaranteed for highly nonlinear problems
 Comparison of Implicit and Explicit Methods
Explicit Time Integration:
Central difference method used - accelerations evaluated at time t:

Wh
Where {Ftext} is
i the
h applied
li d externall and
dbbody
d fforce vector,

{Ftint} is the internal force vector which is given by:

{a t } = [M ]−1 ([Ftext ]− [Ftint ])
F int = Σ⎜⎛ ∫ B Tσ n dΩ + F hg ⎞⎟ + F contact
⎝Ω ⎠
• Fhg is the hourglass resistance force (see ELEMENTS Chapter)
and Fcont is the contact force.
• The velocities and displacements are then evaluated:
{vt + Δt / 2 } = {vt −Δt / 2 } + {at }Δt t
{ut + Δt } = {ut } + {vt + Δt / 2 }Δt t + Δt / 2
where Δtt+Δt/2=.5(Δtt+ Δtt+ Δt) and Δtt- Δt/2=.5(Δtt- Δtt+ Δt)
 Comparison of Implicit and Explicit Methods
Explicit Time Integration:
The geometry is updated by adding the displacement increments to
the initial geometry {xo}:

{xt + Δt } = {xo } + {ut + Δt }

• Nonlinear problems:
¾ Lumped mass matrix required for simple inversion
¾ Equations become uncoupled and can be solved for directly
(explicitly)
¾ No inversion of stiffness matrix is required. All nonlinearities
(including contact) are included in the internal force vector.
¾ Major computational expense is in calculating the internal forces.
¾ No convergence
g checks are needed
¾ Very small time steps are required to maintain stability limit
 Stability Limit
Implicit Time Integration: Explicit Time Integration:
For linear problems, the time Only stable if time step size
step
t can b
be arbitrarily
bit il llarge i smaller
is ll than
th critical
iti l time
ti
(always stable) step size
For nonlinear problems
problems, time 2
Δt ≤ Δt crit
=
step size may become small ωmax
due to convergence difficulties

Where wmax = largest natural

circular frequency
Due to this very small time
step size, explicit is useful
only for very short transients
Critical Time Step Size

Critical time step size of a rod

- Natural frequency:
q y
c E
ωmax=2
l with c= (wave propagation velocity)
ρ

Critical time step:

l
Δt=
c

- Courant-Friedrichs-Levy-criterion
y
- ∆t is the time needed of the wave to propagate through the
rod of length l
Note: The critical time step size for explicit time integration
depends on element length and material properties (sonic speed).
ABAQUS/EXPLICIT Time Step Size

ABAQUS/EXPLICIT checks all elements when calculating the

required time step.

The characteristic length l and the wave propagation velocity c are

dependent
p on element type:
yp
E
Beam elements: l = length of the element c=
ρ

Shell elements:

l= A , for triangular shells: l= 2A

max ((L 1 , L 2 , L 3 , L 4 ) L3 max ((L 1 , L 2 , L 3 )
E L4
c= A
ρ( 1 - ν 2 ) L2
L1
ABAQUS/EXPLICIT Time Step Size

– The concept of a stable time increment is explained

easily by considering a one
one-dimensional
dimensional problem.

One-dimensional problem

– The stable time increment is the minimum time that a dilatational

wave takes to move across any element in the model.
model
• A dilatational wave consists of volume expansion and contraction.
 ABAQUS/EXPLICIT Time Step Size

– Thus, the stable time increment can be expressed as

l
Δt=
c
– Decreasing L and/or increasing c will reduce the size of the
stable time increment.
• Decreasing element dimensions reduces L. L
• Increasing material stiffness increases c.
• Decreasing material compressibility increases c.
• Decreasing material density increases c.
– ABAQUS/Explicit monitors the finite element model
throughout the analysis to determine a stable time increment.
Summary
Summary

Implicit Time Integration (used by ANSYS) -

•Finite Element
El method used
•Average acceleration calculated
•Displacements evaluated
•Always stable – but small time steps needed to capture
transient response
•Non-linear materials can be used to solve static problems
•Can solve non-linear (transient) problems…
•…but only for linear material properties
•Best for static or ‘quasi’ static problems
Summary
Summary

Explicit Time Integration (used by LS Dyna)

•Central Difference method used

•Accelerations (and stresses) evaluated
• Accelerations -> velocities -> displacements
•Small time steps required to maintain stability
•Can solve non-linear problems for non-linear materials
•Best for dynamic
y problems
p
 Overview of the Explicit
p Dynamics
y Procedure
• Stress wave propagation
• This stress wave propagation
p p g
example illustrates how the
explicit dynamics solution
procedure works without
iterating or solving sets of
linear equations. Initial configuration of a rod
with a concentrated load, P,
• We consider the ppropagation
p g
at the free end
of a stress wave along a rod
modeled with three elements.
We study the state of the rod
as we increment through time.
• Mass is lumped at the
nodes.
Overview of the Explicit
p Dynamics
y Procedure

P − u&1
u&&1 =
M1 ∫
⇒ u&1 = u&&1dt ⇒ ε&el1 =
l ∫
⇒ dε el1 = ε&el1dt

⇒ ε el1 = ε 0 + dε el1 ⇒ σ el1 = Eε el1

Configuration at the end of Increment 1

Overview of the Explicit
p Dynamics
y Procedure

P − Fel1 u& 2 − u&1

⇒ u&1 = u&1old ∫ ∫
ld
u&&1 = + u&&1dt ε&el1 = ⇒ dε el1 = ε&el1dt
M1 l
Fel1 ⇒ ε el1 = ε1 + dε el1
u&&2 =
M2 ∫
⇒ u& 2 = u&&2 dt
⇒ σ el1 = Eε el1
Configuration of the rod at the beginning of Increment 2

Configuration of the rod at the beginning of Increment 3

 Explicit Dynamics method
M .U&& + K .U = F
& U (t n +1 ) −U (t n )
U (t n + (1/ 2) ) = ,
Δt
U & (t ) − U& (t
&& n + (1/ 2) n − (1/ 2) ) U (t n +1 ) − 2U (t n ) + U (t n −1 )
U (t n ) = = ,
Δt ( Δt ) 2

Errors are of the order O ( (∆t) 2) for time steps ∆t → 0,

M .U&& (t n ) = F (t n ) − K .U (t n )
U& (t n + (1/ 2) ) = U& (t n −(1/ 2) ) + Δt U&& (t n ) U (t 0 ) = U 0
U (t n +1 ) = U (t n ) + Δt U& (t n + (1/ 2) ), U& (t ) = U&
0 − (1/ 2) 0
 Implicit Dynamics method

M .U&& (t n +1 ) + (1 + α )K .U (t n +1 ) − α K .U (t n ) = F (t n +1 + αΔt ) (*)

1
U (t n +1 ) = U (t n ) + ΔtU (t n ) + (Δt ) 2 [(1 − 2 β )U&& (t n ) + 2 βU&& (t n +1 )],
& )]
2
U& (t n +1 ) = U& (t n ) + Δt [(1 − γ )U&& (t n ) + γU&& (t n +1 )],

α ∈ [−1/ 3, 0], β = (1 − α ) 2 / 4, γ = (1 − 2α ) / 2

U (t 0 ) = U 0
U& (t ) = U&
0 0

U&& (t 0 ) from eq . * with α = 0,

0 n = −1,
1