16th International LS-DYNA® Users Conference NVH

Recent Developments in Time Domain Fatigue Analysis

with LS-DYNA®
Zhe Cui, Yun Huang
Livermore Software Technology, an ANSYS Company

Abstract
A series of new options were implemented to the time domain fatigue analysis features since the last international LS-DYNA User’s
Conference 2018. They include:
⋅ Fatigue mean stress correction methods
⋅ Load steps definition
⋅ Fatigue damage evolution
⋅ Fatigue failure simulation
⋅ Multiaxial fatigue analysis
⋅ Fatigue summation
This paper gives a brief review of these new options for time domain fatigue analysis with LS-DYNA. Some examples are provided to
demonstrate the new feature of LS-DYNA and show how to use this feature towards different loading cases.

Keywords: LS-DYNA, time domain, fatigue analysis

Fatigue mean stress correction methods

Mean stress has important effect on fatigue behavior of metal structures. Mean stress correction is necessary for
accurate prediction of fatigue life of those metal structures. Under different mean stress, the SN curve of the
same material can change quite a lot.
In LS-DYNA, two categories of mean stress correction methods are available.
Use equations to perform mean stress correction, based on the SN curves obtained by fully reversed testing (R =
-1, or mean stress = 0). Following mean stress correction equations are available
 Goodman equation
 Soderberg equation
 Gerber equation
 Goodman tension only equation
 Gerber tension only equation
 Morrow equation (for fatigue analysis based on EN curve)
 Smith-Watson-Topper equation (for fatigue analysis based on EN curve)
Use *DEFINE_TABLE to define a family of SN curves. Each curve corresponds to a unique mean stress. In
*MAT_ADD_FATIGUE keyword, use the table ID for the SN curve. When a mean stress is not represented by
the existing SN curves, interpolation is performed to find the corresponding number of cycles for failure N, for
the given stress range or stress amplitude S, under current mean stress.
Figure1 and 2 show a pipe model cumulative damage ratio comparison with and without mean stress correction.
One can see that the original damage ratio is 0.002853 and the damage ratio is 0.002917 with mean stress
correction.

June 10-11, 2020 1

16th International LS-DYNA® Users Conference NVH

Figure1. Damage ratio without mean stress correction

Figure2. Damage ratio with mean stress correction

Load steps definition

A new keyword *FATIGUE_LOADSTEP was implemented to define load steps in fatigue analysis.
One can choose which segments of loading history are needed in fatigue analysis. Sometimes user may want to
skip the starting transient response in fatigue analysis and use only the steady state cyclic response.
One can compute fatigue cumulative damage ratio for a long-term load, based on representation on a shorter
load step. The cumulative damage ratio, computed on the shorter load step, is multiplied by a scale factor
(which is the ratio between the duration of real load and the duration of the representative load step), to provide
estimation of the cumulative damage ratio for the real load, which could be much longer and be prohibitive to
compute otherwise. Of course, it is assumed that stress / strain response in the shorter load step is a good
representation of the behavior in the real load step. And the material properties don’t change with the number of
load cycles, or with the load sequence. In other words, the fatigue behavior of the structure is linear.
The example pipe is modelled by *MAT_ELASTIC_PLASTIC_THERMAL. The thermal loading is defined
by *LOAD_THERMAL_LOAD_CURVE. The keyword cards for *FATIGUE_LOADSTEP and other
keywords for the load can be found in Figure 3. The thermal loading time history is in Figure4.

June 10-11, 2020 2

16th International LS-DYNA® Users Conference NVH

Figure3. Keyword setting for running Figure4. Thermal loading time history
fatigue time step

The pipe is subjected to two steps of cyclic thermal loading. For the first load step, the temperature varies
between 0°F and 200°F and this last for 10000 seconds. For the second load step, the temperature varies
between 0°F and 400°F and this last for 20000 seconds. It is very time consuming to run finite element
simulation for the whole thermal loading history of 30000 seconds. To get a quick estimation of the cumulative
damage ratio, we can reduce the duration for each load step to only 50 seconds, and multiply the cumulative
damage ratio generated in each step by a scale factor which is the ratio between the real loading period and the
reduced loading period.
Figure 5 shows the distribution of effective stress near the end of simulation. Figure 6 shows the cumulative
damage ratio of the pipe, after the 30000 seconds thermal loading. One can see that the maximum values of the
effective stress and the cumulative damage ratio appear near the bottom of the pipe, probably due to the stress
concentration at the constraints.

June 10-11, 2020 3

16th International LS-DYNA® Users Conference NVH

Figure5. Effective stress at the end of Figure6. Cumulative damage ratio

simulation

Fatigue damage evolution

With a nonzero DT in *DATABASE_D3FTG, LS-DYNA can perform fatigue analysis and dump out d3ftg
database every DT time. Multiple states are saved in d3ftg and can be plotted using LS-PrePost® 4.7 or newer
versions. Each state saves cumulative damage ratios for the whole structure at one time point. With this
database, user can track the fatigue damage ratio evolution for the structure.
Figure7 shows an L-beam fixed to a bottom plate by four bolts. The plate is constrained to ground. Prescribed
harmonic motion (displacement) is applied on the edge of the hole on the L-beam, in the vertical direction. The
prescribed displacement time history is shown in Figure8.
The cumulative damage ratio fringe plots at time 0.01s, 0.02s and 0.03s are shown in Figure9. Constant color
scale from 0 to 1.0 is used for all the plots so that one can easily compare the magnitude of the cumulative
damage ratio and trace the development of the damage. It is clear that the area at the lower edge of the hole
experiences higher fatigue damage. The damage ratio increases with time and the damage area expands with
time.

June 10-11, 2020 4

16th International LS-DYNA® Users Conference NVH

Figure7. A L-BEAM constrained to a Figure8. Prescribed harmonic

bottom plate displacement on the hole

Time = 0.01s Time = 0.02s Time = 0.03s

Figure9. Cumulative damage ratio at different time points

Fatigue failure simulation

A new keyword *FATIGUE_FAILURE was implemented to introduce a mechanism to model the failure of
elements due to fatigue. With this keyword, user can define a threshold cumulative damage ratio (the default
value is 1.0) and all the elements with cumulative damage ratio larger or equal to this value can be removed
from the structure for subsequent simulation. For increased safety factor, the threshold cumulative damage ratio
can be defined as a number smaller than 1.0.
This is a simple way to show the local failure of structures due to fatigue, and it provides an opportunity to
study the effect of local fatigue failure on the overall behavior of structures in a long term. An approximate
fatigue crack propagation trajectory can be obtained by this approach.

June 10-11, 2020 5

16th International LS-DYNA® Users Conference NVH

A more accurate simulation of the fatigue crack propagation can be achieved by using the approach by fracture
mechanics or using the cohesive zone modelling.

The max cumulative damage ratio at time 0.03 second is 1.76651 (see Figure9). It is obvious that several
elements have failed (including element 5622, which exhibits the max cumulative damage ratio 1.76651). With
*FATIGUE_FAILURE and IFAILURE= 1 and DRATIO=1.0, LS-DYNA automatically removes those
elements whose cumulative damage ratio ≥ 1.0 from the structure. The remaining elements and their cumulative
damage ratio fringe plot are shown in Figure10. Then the cumulative damage ratio of the remaining elements
continues to grow with the loading. Figure11 shows the cumulative damage ratio at 0.04 second. One can see
that the cumulative damage ratio of several other elements goes beyond 1.0 at 0.04 second (e.g. element 5587),
and this results in failure of those elements too. Those failed elements are removed too, as shown in Figure 12.
It is expected that with the loading cycles going on, more and more elements will have cumulative damage ratio
≥ 1.0 and will fail and be removed from the structure. Figure13 shows the keyword setting for modelling fatigue
damage evolution and fatigue failure.

Figure10. Cumulative Figure12. Cumulative

damage ratio at 0.03s Figure11. Cumulative damage ratio at 0.04s
(failed elements are damage ratio at 0.04s (failed element are
removed) removed)

Figure13. Keyword setting for modelling fatigue damage evolution and fatigue failure.

June 10-11, 2020 6

16th International LS-DYNA® Users Conference NVH

Multiaxial fatigue analysis

Many mechanical components experience multiaxial cyclic loadings during their service life. Compared with
the uniaxial fatigue problem, the multiaxial fatigue problem is more complex due to the complex stress / strain
states and loading histories. Stress / strain state is always three dimensional.

Three multiaxial fatigue methods are provided in LS-DYNA time domain fatigue analysis. Figure14 shows
keyword setting for modelling multiaxial fatigue.

• A scalar index (e.g. Von-Mises stress, 1st principal stress) can be used
• Fatigue damage is computed on multiple planes and the max value is picked
• A critical plane is located, and fatigue analysis is performed on the critical plane

Figure14. Keyword setting for modelling multiaxial fatigue

Figure15 and 16 show the cumulative damage ratio of a simple plate with different multiaxial approaches. One
can see that the damage ratio is 1.2655 with MAXIAL=0 and the damage ratio is 1.3045 with MAXIAL=2.

Figure15. Cumulative damage ratio for MAXIAL=0

Figure16. Cumulative damage ratio for MAXIAL=2

June 10-11, 2020 7

16th International LS-DYNA® Users Conference NVH

Fatigue summation
This keyword reads in existing fatigue databases defined by
*INITIAL_FATIGUE_DAMAGE_RATIO and sum up the damage ratio results from them to obtain the final
cumulative damage ratio. The final cumulative damage ratio results are dumped to a new d3ftg database. The
Figure17 and 18 show a comparison of a simple plate cumulative damage ratio with and without damage from
transient preload. One can see that the damage ratio is 0.3440 from fatigue load and the damage ratio is 0.3443
from fatigue load plus transient preload.

Figure17. Damage ratio from fatigue load

Figure18. Cumulative damage ratio from transient preload + fatigue load

June 10-11, 2020 8

16th International LS-DYNA® Users Conference NVH

Summary
This paper reviews recent updates in time domain fatigue analysis in LS-DYNA and introduces several new
keywords and options for running these features. These new options and enhancements enable users to solve
more comprehensive problems in NVH and durability analysis.

June 10-11, 2020 9