講演番号 320 文献番号 20235320

3-D Bulk Metal Forging Simulation with Adaptive Finite Elements

Li Zhang 1) Yong Guo 2) Wei Hu 2) Philip Ho 2)

1) ANSYS, Inc.
Suite 100, 1740 W Big Beaver Road, Troy, MI 48084, USA
2) ANSYS, Inc.
7374 Las Positas Road, Livermore, CA 94551, USA

ABSTRACT: Bulk metal forging has been used to make everything from weapons and various tools etc. in the ancient time to shafts,
gears, household screws/nails, to precision weapons, electronic components, and high-end consumer products in the modern-day. Forged
metals have finer grain structures, are stronger and the manufacturing process is very efficient. In this paper, we present features and
capabilities of utilizing 3-D r-adaptive remeshing technology within Ansys LS-DYNA for cold and hot forging simulations and
applications. This technology has been applied on a daily basis to improve product/process development efficiency and reduce/eliminate
physical Design Of Experiments (DOE). Features within LS-PrePost related to forging applications are also presented.

KEY WORDS:cold and hot forging, 3D mesh r-adaptivity, Cockcroft-Latham damage.

(1),
1. INTRODUCTION are captured by controlling the variable RMIN the minimum
Ansys LS-DYNA is a powerful nonlinear finite element TETRA size for the high tool curvature region. The variable
analysis software widely used for simulating complex engineering RMAX (1) defines the maximum element size for the smaller tool
problems involving structural mechanics, fluid dynamics,
thermomechanical problems, and various manufacturing
processes. It is used in automotive, aerospace, defense, precision
consumer product and electronics manufacturing industries to
simulate and optimize product designs and manufacturing
processes, analyzing crash and safety scenarios and predicting the
behavior of materials under extreme conditions. During forging,
metal is shaped and compressed through a die (tool), resulting in
large plastic deformation, changing its microstructure and
mechanical properties. By simulating forging processes using LS- Fig. 1 A Simple Forging Model.
DYNA, engineers can create virtual models of the forging process
and analyze materials under different temperatures, strain rates
and tool geometries. The resulting information can be used to
optimize the forging process (initial workpiece size and shape,
tool shape, passes, etc.), predict forging tonnage (press
capacity/allocation), assess manufacturing feasibility, determine
and minimize defects, and reduce or eliminate physical DOE. The
Fig. 2 A Final Deformed State.
virtual forging simulation has the potential to drastically shorten
curvature region, which controls, to a certain degree, the
product development and engineering cycles, and maximally
total element size for the refinement. Another variable, ADPENE
increase manufacturing efficiency and productivity. (1), is used to control the size of the refinement transition region
from RMAX to RMIN. The bigger the ADPENE value the larger
2. 3-D ADAPTIVE REMESHING FEATURES
or more spread/bigger the transition area for the refinement
region. In Fig. 1, a simple forging model is created to demonstrate
2.1. Remeshing Controls
the various features discussed in this paper. A rectangular
3-D r-adaptive remeshing refines (or coarsens) tetrahedron
workpiece of 24x10x4 mm is situated with an overhang (0.875mm)
(TETRA) meshes of the workpiece based on the tool curvature.
on a flat and rigid plate, to be compressed 50% in vertical
Tools can be rigid or deformable body, which can be constructed
direction. The workpiece is pressed down with the punch (blue)
with hexahedra elements. Deformable tools can be used to assess
against the rigid flat plate (green). The average initial TETRA
tools' structure and durability analysis. The smallest tool features

2023 JSAE Annual Congress (Spring) May 24 to 26 ， 2023

Issued on May 19 ， 2023

2023 JSAE Annual Congress Proceedings (Spring) 1

 either caused by ADPFREQ or by DTMIN. Unlike for most of the
explicit dynamic calculations such as sheet metal forming or
crash/safety application, mass scaling in terms of setting the
minimum time step DT2MS (1) is not used in 3D adaptive forging
simulation, as it interferes with the remeshing set by DTMIN,
which is required for most forging and extrusion applications.
Additional remeshing controls also include volume fraction loss
Fig. 3 Effects of ADPENE. Initial Element Number: 7680. and mass ratio gain, which are not often used for explicit but more
useful for implicit calculations.
elements’ length used on the workpiece is 1.15mm, with RMAX
In the explicit dynamic regime, the absence of mass scaling in
and RMIN set as 1.5 and 0.2mm, respectively. In Fig. 2, a final
terms of setting a bigger DT2MS is beneficial for an accurate
deformed state of the workpiece is shown. Effects of ADPENE on
the mesh distribution and the total number of elements are shown
in Fig. 3. As ADPENE increases, the adaptive region expands, the
final number of elements becomes greater, and the maximum
effective plastic strain gets slightly higher. Note that the selection
of ADPENE can affect the computation time, shown in Fig. 4.
For explicit dynamic calculations, care should be taken when
setting the RMIN value. Since the RMIN determines the time step
size, it could greatly affect the overall calculation time required to
complete the simulation. The small value of RMIN will also Fig. 5 Tool Velocity Effect on EPS and Tonnage.

prediction of tonnage required to forge a workpiece to its final

shape. Artificial changes to workpiece density or Young's
modulus to speed up the calculation time will affect the tonnage
calculation and deformation mode, and are not advisable. A more
common method to speed up the calculation time is to scale up the
tool velocity from the actual tool speed in the press, which is
around ~100 mm/s. Studies show for most forging applications,
Fig. 4 Effect of ADPENE on Computing Time. this speed can be scaled up to 1000 mm/s without too much effects
on the tonnage. Some variations can be seen in the Effective
increase the total number of elements in a model, slowing down
Plastic Strain (EPS) and deformation mode if the speed goes
the simulation. Remeshing algorithm can be activated with the
(1) above 500 mm/s, as shown in Fig. 5. For a more conservative
variable ADPFREQ to trigger a remeshing during the
approach, speed of 500 mm/s can be used routinely. It is important
simulation. This is done either with a fixed and equal time interval
to note that the scaled-up tool velocity will have an effect on
or through a curve which defines the interval as a function of time
tonnage estimate and deformation modes if strain rate dependent
(LCADP). The curve option offers more control for some
hardening is considered.
applications, especially in implicit calculation. Generally speaking,
Strain hardening, strain rate dependent hardening,
when setting this variable, fewer adaptive remeshings are
temperature dependent hardening, and friction are all critical in
preferred over more frequent remeshing. An additional, “on-
estimating tonnages in forging. In addition, compression test data
demand” remeshing control is used by setting the minimum value
(1), should be used for forging processes instead of tensile test data.
for time step size DTMIN which starts the remeshing/
During an adaptive step, the state variables including effective
remapping process if the current time step size falls below the set
plastic strain (EPS) at Gauss points will be averaged to the nodes
value. Falling time step size is an indication of the deteriorating
prior to remeshing and stored. After remeshing, a restart will read
element deformation, distortion, or size. After remeshing, if it is
the old mesh data, and the state variables at new Gauss points of
successful, the time step size will become larger again than the set
the new mesh are interpolated from the nodal values of the old
value, implying the remeshed elements’ qualities are better for the
mesh. The values of the state variables will drop (cannot jump)
next round of forging until the next remeshing happens again,
due to the averaging effect and the percentage of the drop depends

2023 JSAE Annual Congress Proceedings (Spring) 2

 be applied in a global coordinate system. Both
(1)
*CONSTRAINED_LOCAL and *BOUNDARY_SPC
(1)
_SYMMETRY_ PLANE allow global and local constraints.
All three options provide users with a tolerance within the
symmetric plane to constrain the existing and newly generated
Fig. 6 EPS before (left), and after remeshing (right).
nodes that fall within the tolerance.
on the mesh size. The finer the mesh, the smaller of the
drop percentage. From the example shown in Fig. 6, we can see 3. ELEMENTS
one good thing brought by the averaging procedure. Before Solid elements are used in 3-D forging simulations for both
adaptivity, some elements become distorted and have higher EPS the workpiece and tools.
(may not be physical) and after adaptivity, those with Element type 13 adaptive tetrahedron is used for the
higher EPS are smoothed out. With the mesh being refined workpiece. It includes volume averaged 1-point nodal pressure to
(1,2).
enough, the drop of EPS should become very small. As the state reduce volumetric locking It averages the volumetric strain
variables are used to calculate the damage accumulation, it will over neighboring elements to smooth the pressure response. The
also drop after remeshing. assumed strain displacement is not restricted to a single element
Another important feature in 3D adaptive remeshing is the
(1),
"monotonic resizing" variable MM meaning the adaptive
remeshing cannot coarsen a mesh during the simulation. This
feature is critical in maintaining the feature lines of the forged
parts, and it is available for forging in both explicit and implicit
methods. In Fig. 7, an initial rectangular-shaped block is extruded

Fig. 8 Pancake Crushing Using adaptive TETRA 13.

nodal connectivity but is spread over a region of adjacent elements.

It has no kinematic modes so no hourglass control is needed. This
element is the workhorse for many applications with
incompressible or nearly incompressible materials including
rubber and ductile elastoplastic metals in bulk forming. It is a very
stable element, able to handle severe and extreme mesh distortion.
Figure 7 Monotonic resizing. Left: MM=0. Right: MM=1.
As an example, we show in Fig. 8 (left) a small block of rubber
through a slot into a bread shaped part, the difference is shown
(dimension: 14.8x6.6x5mm) defined using *MAT_077H
with and without this feature activated. When MM is set to 0, the
(hyperelastic rubber) being crushed into a pancake shape with a
mesh coarsens as soon as the workpiece leaves the die slot opening,
final thickness of only 0.18mm, Fig. 8 (right).
enlarging the radius that was previously forged as it goes through
Another frequently used element for the workpiece is type 42,
the die corners. In the case of MM=1, the sharp corner radius is
adaptive mesh-free (Element Free Galerkin – EFG) tetrahedron.
maintained even after the workpiece has left the slot opening. Care
EFG 42 tetrahedron uses meshfree approximation function, which
should be taken though, as the mesh will not be coarsened
has high order continuity and smoothness compared to standard
throughout the simulation duration. To avoid having excessively
FEM 13. EFG 42 has a better element distortion tolerance for large
large number of elements during and at the end of the simulation
deformation. The EFG approximation function has been
resulting in long simulation time, the initial workpiece mesh is
implemented as a more accurate alternative to the mapping
best to have a variable sized TETRA elements (see Section 7 LS-
function to transfer variables from old mesh to the new one during
PrePost Features), with finer mesh in fine featured tooling area,
the adaptive process (3~5).
and coarser mesh elsewhere.
Hexahedral solid elements are often used for the tools. The
elements on the tool mesh should be close to a square shape,
2.2. Boundary Conditions
uniform, and small enough to capture the tools’ features. The tool
Symmetric boundary conditions can be applied to reduce
surfaces may need to be offset back to allow for a small clearance
simulation time. There are three options to apply symmetric
to be defined in sliding contact for contact stability. Tools can be
conditions. *CONSTRAINED_GLOBAL (1) allows constraints to

2023 JSAE Annual Congress Proceedings (Spring) 3

either rigid (element type 1 (1)) or deformable (element types -1 or
-2 (1)) for tool stress/structure/durability analysis.

4. CONTACT
For explicit, *CONTACT_AUTOMATIC_SURFACE_TO_
(1)
SURFACE is used with soft constraint penalty (SOFT=1)
formulation (1). An optional solid element thickness defined by the Fig. 9 Strain rate and Temperature Dependent Strain
Hardening.
variable SLDTHK (1) is often used for a robust contact. The use of
defined with these material models, or using *MAT_ADD_
the variable essentially covers the tools’ solid elements with a
THERMAL_EXPANSION keyword. Thermal properties such as
layer of null shells for contact treatment. This contact type is
heat capacity and thermal conductivity, etc. can be defined using
sometimes also used in implicit calculations. In the contact
various *MAT_THERMAL cards.
member definition, the workpiece can be defined as part ID, while
For modified Johnson-Cook model (*MAT_107), the yield
the tools are defined as segments; only the tool surface segments
stress takes the following form (1):
expected to be in contact with the workpiece need to be included. 2
𝑟̇ 𝐶
For implicit analysis, the most often used contact type is 𝜎𝑦 = {𝐴 + 𝐵𝑟 𝑛 + ∑ 𝑄𝑖 [1 − exp(−𝐶𝑖 𝑟)]} (1 + )
𝜀0̇
*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE_MO 𝑖=1

RTAR. The variable SLDTHK also applies. 𝑇 − 𝑇𝑟 𝑚

∗ [1 − ( ) ]
𝑇𝑚 − 𝑇𝑟
In situations where the workpiece is expected to fold onto itself,
where r is the damage-equivalent plastic strain defined as
*CONTACT_AUTOMATIC_SINGLE_SURFACE can be used
𝑟
in both explicit and implicit situations. 𝜀𝑝 =
1 − 𝛽𝐷
Thermal option can be added to the above *CONTACT cards and the parameter β is equal to one for coupled damage and equal
when working with thermomechanical forging. Heat transfer to zero for uncoupled damage. All other variables and parameters
conductance and other related thermal variables can be defined. are explained in the User’s Manual (1). On the right-hand side of
THERMAL_FRICTION option allows definition of temperature the yield stress equation, the first term represents the strain
dependent friction coefficients, and temperature and pressure hardening, the second term the strain-rate hardening and the third
dependent thermal contact conductance. term the thermal softening.
For the tabulated Johnson-Cook model (*MAT_224), the
5. MATERIAL MODELS flow stress 𝜎𝑦 (1) is expressed as a function of plastic strain 𝜀𝑝 ,
Ansys LS-DYNA has over more than three hundred different plastic strain rate 𝜀̇𝑝 and temperature 𝑇 via the following formula:
(1)
material models that cover a wide range of different 𝑘𝑡 (𝜀𝑝 , 𝑇)
𝜎𝑦 = 𝑘1 (𝜀𝑝 , 𝜀̇𝑝 )
engineering materials. The most often used material models for 𝑘𝑡 (𝜀𝑝 , 𝑇𝑅 )
forging and extrusion applications are *MAT_024 (piecewise where 𝑘1 (𝜀𝑝 , 𝜀̇𝑝 ) is strain rate dependent hardening curves (Fig. 9
linear plasticity, strain rate dependent or independent, non- left), 𝑘𝑡 (𝜀𝑝 , 𝑇) is the temperature dependent hardening curves
thermal), *MAT_106 (elastic-viscoplastic thermal), (Fig. 9 right), and 𝑇𝑅 is the room temperature. We can see the
*MAT_107 (modified Johnson-Cook, non-thermal) and similarities of the yield stress functions between *MAT_107 and
*MAT_224 (tabulated Johnson-Cook thermal). *MAT_024 is *MAT_224.
mostly used in cold forging. *MAT_106 is a comprehensive Recently, a new damage accumulation feature was
thermal enabled model and is most often used for implemented in *MAT_224 based on Cockcroft-Latham’s work
thermomechanical forging. This material model allows for inputs (8). In order to assess the damage accumulated in the material
of a strain hardening curve, or strain hardening curves for each during the forging process, the Cockcroft-Latham damage value
temperature, or hardening curves for each strain rate under each is calculated during the forging simulation. The Cockcroft-Latham
temperature. The material model *MAT_224 (thermal enabled) is (C-L) damage value is defined as following:
an elasto-viscoplastic material which accounts for plastic work 𝐷𝐶 𝜀𝑝
𝐷= ∫ 〈𝜎1 〉𝑑𝜀𝑝
caused adiabatic heating therefore material softening. This 𝑊𝐶 0
material model allows direct, separate inputs of the strain-rate and where, 𝜎1 is the first principal stress, Dc and Wc are material
temperature dependent strain hardening curves, making it much parameters. The user input of the critical Cockcroft-Latham
easier to use than *MAT_107. Thermal expansion can also be damage parameter 𝑊𝐶 (critical plastic work) is required, and 𝐷𝐶 is

2023 JSAE Annual Congress Proceedings (Spring) 4

the critical damage value. For most materials 𝐷𝐶 is 1.0, which criterion used in ductile failure analysis. Further examination of
means that the material failure occurs when 𝐷 reaches 1.0. No the first principal stress contour (Fig. 11 left) in the area reveals a
coupling is assumed between ductile damage and the constitutive negative compression state. As materials are more unlikely to fail
relation. under compression than in tension in a forging scenario, the C-L
The critical Cockcroft-Latham damage parameter 𝑊𝐶 can be damage criterion may be a better failure indicator than the ratio of
considered as the total plastic work required for the material to fail EPS over plastic failure strain.
and result in fracture. It is calculated by the following integral:
𝜀𝑓
6. SOLVERS
𝑊𝑐 = ∫ 〈𝜎1 〉𝑑𝜀𝑝
0 Ansys LS-DYNA is a general purpose, nonlinear finite
where 〈 〉 is the Macauley bracket which is defined as element solver, and forging and extrusion simulations benefit
𝑆 if 𝑆 > 0 from its powerful capabilities in both explicit dynamics and
〈𝑆〉 = {
0 if 𝑆 ≤ 0
implicit (statics and dynamics) solvers. While LS-DYNA explicit
where 𝑆 is a scalar variable. 𝜀𝑓 is the critical failure strain. The
dynamics is the most well-known method used in crash, safety,
values of 𝑊𝐶 and 𝜀𝑓 can be calibrated by experiment and
medical, consumer product, ballistic and explosive simulation, etc.
(6).
numerical simulation of a simple tension test The Cockcroft-
it is also the most widely applied method in forging
Latham damage accumulation can be activated with the variable
simulation. Compared with the explicit dynamic capability within
ADPDAM set to 2 in *CONTROL_ADAPTIVE.
LS-DYNA, its less known counterpart, yet the very powerful
Adiabatic heating is calculated as:
implicit method can sometimes offer the same high quality results
σ: d𝑝 𝜎̃𝑒𝑞 𝑟̇
𝑇̇ = 𝜒 =𝜒 with much faster computational speed. These capabilities can all
𝜌𝐶𝑝 𝜌𝐶𝑝
be employed to solve the most challenging forging problems. In
where 𝜌 is the density and 𝐶𝑝 is the specific heat, 𝜒 is the Taylor-
Quinney parameter, which gives the portion of plastic work
converted into heat therefore temperature, causing metal stress
softening. In a mechanical only simulation, under room
temperature of 300 °𝐾, using the strain hardening curves (Fig. 9
left) for each strain rate ranging from 0 to 12000 1/s, and for Fig. 10 EPS and Cockcroft-Latham Damage.
temperatures (Fig. 9 right) ranging from 300 to 775 °𝐾, the same
workpiece (Fig. 1) is formed with a scaled-up punch velocity of
500mm/s, and with no overhang for the bottom flat support plate.
The resulting EPS contours are shown in Fig. 10 (left), and C-L
damage accumulation contours are shown on the right (assuming
Fig. 11 First Prin. Stress (MPa) and Temperature (°𝑲).
𝑊𝐶 = 300 𝑀𝑃𝑎) . A white solid dot indicates the maximum
damage value location. The maximum C-L damage accumulation
value (stored as history variable #28) is predicted at 0.65, well
below the failure value of 1.0. Notice the maximum C-L damage
value that occurs near the top right corner/side wall, where a
combination of high levels of first principal stress in tension (Fig. Fig. 12 Plastic Failure Strain vs. Stress Triaxiality.
11 left) and EPS are found. The plastic work (assuming 100%
conversion to heat) induced temperature during the forging is
predicted with the maximum value of 557 °𝐾 (Fig. 11 right).
Temperature contours closely follow the EPS contour pattern and
are also elevated in the maximum C-L damage location. In
addition, with the plastic failure strain vs. stress triaxiality defined Fig. 13 Plastic Failure Strain and the Ratio of EPS over
Plastic Failure Strain.
in Fig. 12, plastic failure strain contour is shown in Fig. 13 (left),
addition, LS-DYNA’s MPP (Message Passing Parallel)
and the ratio of EPS over EPS failure strain contour in Fig. 13
technology works with both explicit and implicit methods, and
(right), with the area of the maximum value marked with a white
offers unparalleled computing experience.
solid dot. The ratio of EPS over EPS failure strain is often a failure

2023 JSAE Annual Congress Proceedings (Spring) 5

 6.2. Implicit Statics and Dynamics
Implicit method (1) offers advantage over explicit method for
forging small, tiny and miniature geometry (radius and length)
features and with relatively large forging stroke, as step
size (therefore total simulation time) is not dependent on the

Table 1. CFL time step size of a trirectangular tetrahedron of an smallest characteristic element length in the model. Currently,
aluminum alloy. DTMIN used to trigger adaptive remeshing in explicit is not
6.1. Explicit Dynamics available for implicit, but will be made available in the future.
During 3D mesh adaptivity, the characteristic element length Remeshing is mostly done with the use of an even adaptive
becomes smaller and smaller as a simulation progresses, posing a interval, or with a curve specifying the adaptive intervals in any
huge challenge for explicit dynamic method (1) to finish in a given step segment. The number of remeshings during the stroke
reasonable amount of computational time. Table 1 lists different should be determined by the stroke distance and the tool geometry,
stable time step sizes as required by Courant-Friedrich-Levy (CFL) and it may require more remeshing steps than that of explicit
Criterion for a few characteristic element lengths of a dynamics, especially in the final stage of a forging process. The
trirectangular tetrahedron element. The smaller the geometric implicit step size is in the range of 0.1% to 1% of the total punch
feature in the model, the smaller the element size is required to stroke, dependent on the stroke distance and tool geometry. The
better capture the features (in terms of radius and length). For
some of the tiny electronic components, the smallest
element lengths will need to go down to 0.01~0.02 mm range,
corresponding to characteristic element lengths in the range of
0.00577~0.01154 mm in the most ideal situation, which dictates
the time step size to be in the range of 8.6E-10 ~ 1.72E-9 sec. This Fig. 14 Extrusion of an Aluminum Block (mm).
range of time step size is compared with sheet metal forming
application (where LS-DYNA’s explicit dynamics is also widely
used), where, in some cases, the smallest shell element size needs
to be around 1mm, which corresponds to a time step size of around
1.82E-7 sec. With mass scaling this value can increase to 5.0E-7,
Fig. 15 EPS comparison between Explicit Dynamics and
which is considered very small for a typical metal forming; but Implicit Statics.
still, it is over 100 times larger than those for TETRAs in forging
applications. Fortunately, most of these electronic components
also have an equally small forging stroke making the total
computation time still feasible. For situations which require tiny
element length and longer forging stroke, implicit method for both
static and dynamic analysis can become very efficient.
Another key factor that controls the explicit dynamic
simulation, especially in 3D forging simulation, is the stable
surface contact time step. In certain situation this time step size Fig. 16 Extruding Force Comparison between Explicit
Dynamics and Implicit Statics.
may be much smaller than required by CFL criterion; therefore,
reducing the time step size is essential to avoid contact instability. step size should also be determined by workpiece and tools
Although the minimum stable surface contact time step is contact condition, so sliding contact can be accurately
provided in the beginning of a simulation by LS-DYNA, it is captured. Convergence has not been a major problem (with
not always obvious especially when adaptivity remeshing is adequate remeshing) in 3D forging application, although the
involved. It is helpful to identify issues in the contact areas model size (total number of deformable TETRAs) could make the
between a workpiece and tools so the time step size can be simulation slow as stiffness matrix inversion can take more time
adjusted to avoid contact instability. with large models. In addition, as the model size gets bigger the
physical memory required for in-core matrix inversion

2023 JSAE Annual Congress Proceedings (Spring) 6

will become large, requiring appropriate computing resources to workpiece as shown in Fig. 1, but with no overhang for the lower
be used. fixed rigid plate. The temperature dependent hardening curves
In Fig. 14, an aluminum block of 14.5x5x1.8mm is extruded shown in Fig. 9 (right) are used as input to *MAT_106. The
through an opening, until the thickness of the flange area of the workpiece’s initial temperature is set at 523 °𝐾 and the tools have
extruded part reaches 0.18mm. Comparison of the EPS contours a fixed temperature of 298 °𝐾 . The upper tool is moving
are shown in Fig. 15 between explicit dynamics (left) and implicit downward at a scaled-up speed of 500mm/s. We also assume 100
statics (right), and they compare well in both patterns and percent of mechanical work is converted into heat. Convection and
magnitude. In Fig. 16, the extrusion forces also match well radiation are not considered. Fig. 17 shows effective plastic strain
between the two analyses. contours of the workpiece at the finished state. The workpiece
While implicit statics can be used in simulating forging in a surface temperature is shown in Fig. 18 (left), where areas close
quasi-static fashion which is not time dependent, implicit to the tools are cooler and become hotter further away. The hottest
(1)
dynamics method can be used to simulate small (or large) temperature reached 821°𝐾 in the center middle of the workpiece
features with long duration and is time, speed (strain rate) and (Fig. 18 right), where the mechanical work is the greatest and
temperature sensitive. Implicit dynamics is especially useful in distance is the furthest away from the tools. The tools’
simulating warm/hot forging with long stroke/duration, while temperatures rise through the thermal contact interface is also
explicit dynamics is more suitable in high velocity impact with visible. Noticeable is the highest temperature found in the middle
thermal effects. section of the highly compressed area where material softens the
most; this differentiates from the mechanical only simulation
6.3. Thermal Solver where temperature patterns (Fig. 11 right) closely follow those of
Thermal and mechanical effects can be captured in the EPS (Fig. 10 left ).
simulation such as forging. This type of simulation is important
for understanding how a metal part will behave during hot forging, 6.4. MPP Solver
and for optimizing the forging process to produce parts with Ansys LS-DYNA is a pioneer and a leader in MPP
desired properties and dimensions. The two widely used thermal technology, dating back to 1993. MPP is enabled with both
solvers in forging applications are solvers 11 (1) (direct solver) and explicit and implicit solvers within LS-DYNA. As MPP is used
(1)
12 (conjugate gradient iterative). Thermal time integration extensively in automotive crash/safety, metal forming, and all
typically uses the fully implicit method, with thermal time steps possible application fields, it is also indispensable in
taken between every 10~100 of the mechanical time steps. The forging application. MPP divides the model into multiple pieces
thermal time step must be small enough to capture mechanical (number of cores) using domain decomposition method and using
motion or otherwise the temperature field will not be realistic. MPI (Message Passing Interface) to communicate between sub-
Initial temperatures for workpiece and tools can be set individually, domains. Compared with SMP (Shared Memory Parallel) which
and tools’ temperature can be set as either fixed boundary is scalable up to eight cores, MPP offers much higher scalability
(7),
conditions or varied throughout the simulation. As an example, we with hundreds of cores simultaneously and routinely used for
forge in a thermal-mechanical coupled simulation the same large and small forging and extrusion problems. It also has the
flexibility to be run on either a SMP machine or a cluster. With
MPP, the maximum number of deformable TETRAs can be in a
few millions for explicit dynamics, essential to capture all the
details in complex forging/extrusion situations. All of the
simulation examples presented in this paper are run with MPP.

Figure 17 Thermomechanical Forging: Effective Plastic Strain.

6.5. Multi-stage Forging
Multi-stage forging such as cold heading, where the
workpiece is moved from one forging station to others with
different tools, can be simulated with LS-DYNA. The workpiece
for the current station can be output as the "dynain" (or lsda) file,
which contains all state variables, at the end of the forging. The
Figure 18 Thermomechanical Forging: Temperature (°𝐾).

2023 JSAE Annual Congress Proceedings (Spring) 7

dynain file can be combined with the tools for the next station. It
is also possible to do a multi-stage forging in one simulation
without the dynain file. This can be done by combining the
workpiece with tool sets from different stations (“all-in-one”),
establishing necessary contact with each tool set with contact
"birth" and "death" time, and assigning independent tool
kinematics for each tool set. The "dynain" method offers a more
Fig. 19 Creating a Variable Mesh for a Workpiece
"segmented" approach where one could fix forging
issues identified through simulation in the current station before
moving onto the next station, while the "all-in-one" approach is
more seamless and more suitable when most of forging issues are
resolved, albeit it could get complicated as the number of stations
increases. The "dynain" method can be applied in the early
product and process development stage while the "all-in-one" Fig. 20 Flow Curve on a Section Plane of a Workpiece
method can be used as the product and process becomes
In this paper we reviewed aspects of Ansys LS-DYNA’s
more stable and mature.
capabilities in terms of forging and extrusion applications,
including 3D adaptive remeshing, elements, contacts, thermal and
7. LS-PREPOST FEATURES
mechanical material models, MPP, and LS-PrePost features. A
LS-PrePost offers a complete capability related to forging
recently implemented Cockcroft-Latham damage criterion with
simulation. Here we highlight two important features regarding
tabulated Johnson-Cook model *MAT_224 is also presented.
workpiece meshing and flow curve display. As shown in Fig. 19,
With multitude of methods (explicit and implicit) available, and
a workflow is shown to build a variable TETRA mesh for a
MPP technology, LS-DYNA has been proven to solve complex,
cylindrical workpiece, as this type of mesh is important for MM=1
miniature and large forging problems. We are committed to
features in adaptivity for an efficient run (Section 2.1); in addition,
continuing to work with our customers to improve the software.
sometimes a local refined initial mesh is also preferred for
accurate forging dimension in areas where tool contact is not
REFERENCES
present, and severe deformation in the area is expected. First a
(1) LS-DYNA User's Manuals I, II, and Theory Manual III.
uniform bounding surface meshes of 2 mm shells is created on the
(2) T Erhart, Review of Solid Element Formulation in LS-DYNA,
cylindrical surfaces, then the area around where the top cap and
LS-DYNA Forum 2011, Stuttgart, Germany, October 11, 2011.
wall meets are refined to 0.5mm with mixed QUAD and TRI
(3) W Hu, CT Wu, Metal forming analysis using meshfree-
shells, then the entire bounding surface shell is remeshed to TRI
enriched finite element method and mortar contact algorithm,
shells, finally the TETRA solids with size variable from 0.4 ~
Interaction and Multiscale Mechanics, 6 (2013) 237-255.
2mm are created for the cylindrical volume based on the variable
(4) W Hu, CT Wu, Metal forming analysis using meshfree-
TRI shells. For the tool mesh, shells are created first then extruded
enriched finite element method and mortar contact algorithm,
or offset to make solids. Uniform hexahedral elements are
Interaction and Multiscale Mechanics, 6 (2013) 237-255.
preferred. Furthermore, LS-PrePost also offers many mesh quality
(5) CT Wu, W Hu, HP Wang, H Lu, A robust numerical procedure
check features. Related features include check for volume,
for the thermomechanical flow simulation of friction stir welding
minimum element length, internal angle, and Jacobian.
process using an adaptive element-free Galerkin method,
Flow curves can be used to track the metal movement inside
Mathematical Problems in Engineering, 2015.
the workpiece or on the surface during forging. Initially straight
(6) G.H. Majzoobi and F. Rahimi Dehgolan, Determination of
lines can be specified on a section plane and the deformed curves
the constants of damage models, Procedia Engineering, 10, 764-
are displayed (Fig. 20) on the same section plane as the workpiece
773, 2011.
is forged. Noticeable is the initial line on the surface in this case is
(7) J Wang, Recent Development, Deutsch LS-DYNA Forum,
displayed on the deformed workpiece surface outside the section
Bamberg, Germany, 2018.
plane (Fig. 20 left).
(8) M.G. Cockcroft and D.J. Latham, Ductility and the workability
of metals, J Inst Metals, J Inst Metals, 96: 33-9, 1968.
8. CONCLUSION

2023 JSAE Annual Congress Proceedings (Spring) 8