Autodesk Moldflow 2019 Technology Preview

Autodesk and Moldflow are registered trademarks or trademarks of Autodesk, Inc., and/or its subsidiaries and/or affiliates in the USA and/or
other countries. All other brand names, product names, or trademarks belong to their respective holders. Autodesk reserves the right to alter
product and services offerings, and specifications and pricing at any time without notice, and is not responsible for typographical or graphical
errors that may appear in this document.

© 2018 Autodesk, Inc. All rights reserved.

Contents

Autodesk Moldflow 2019 technology preview Features ............................................. 3

Prerequisite ........................................................................................................................... 3
Assembly analysis ................................................................................................................ 4
Automatic mesh type classifier ....................................................................................... 17
Advection Solver application programming interface (API) ....................................18
Reactive Microcellular Injection Molding (with PU Foaming) ................................ 29
Anisotropic Thermo-Viscoelastic Residual Stress Model ........................................ 37
Tiger Stripe Prediction ...................................................................................................... 45
Underflow diagnostic ........................................................................................................ 50
Retractable core-pin support .......................................................................................... 55
Mesh aggregation for buckling analysis ...................................................................... 58
Fiber concentration ............................................................................................................ 60
Annealing analysis with viscoelastic material properties for midplane and dual-
domain .................................................................................................................................. 67
Shrinkage calibrated coefficient of thermal expansion for unfilled semi-
crystalline materials ........................................................................................................... 71
Thickness shrinkage scheme for residual stress accuracy improvement ............. 72
Autodesk Moldflow 2019 technology preview Features
The following new features are supported in Technology Preview 2019 release:

Assembly analysis
Automatic mesh type classifier
Advection Solver application programming interface (API)
Reactive Microcellular Injection Molding (with PU Foaming)
Anisotropic Thermo-Viscoelastic Residual Stress Model
Tiger Stripe Prediction

The following features were carried over from the last technology preview
release:

Underflow diagnostic
Retractable core-pin support
Mesh aggregation for buckling analysis
Fiber concentration
Annealing analysis with viscoelastic material properties for midplane and dual-
domain
Shrinkage calibrated coefficient of thermal expansion for unfilled semi-
crystalline materials
Thickness shrinkage scheme for residual stress accuracy improvement

Prerequisite

Autodesk Moldflow 2019 Synergy and Insight need to be installed for Autodesk
Moldflow 2019 technology preview to work.
Assembly analysis

Deformation prediction of injection molded components mounted into

designed position

Introduction
In some industrial applications such as automobiles, aircraft and home
appliances, it is essential to meet tight dimensional tolerances after the
injection molded components are mounted into the designed position [6].
Injection molding process-induced shrinkage and warpage may cause the
manufactured part to deviate from the designed nominal shape and size. For
the flexible plastics part, forces could be applied to bring them back to the
nominal position during the assembly. Sometimes injection molded
components’ warpage problem can be eliminated once they are mounted by
over-constrained assemblies. On the other hand, the flat components could
become distorted during the compliant assembly due to the unexpected
shrinkage magnitude during injection molding process. For example, when the
front door trim panel in the prototype stage assembled with sheet metal (as
shown in Figure 1), the desirable uniform clearance is not achieved, and the
gap is beyond the tolerance requirement of geometric dimensioning and tolerancing
(GD&T). Predicting the deformation and stress of injection molded components after
being mounted into designed position can be a valuable design tool for identifying the
problem and optimizing the design.
Figure 1. Assembled automotive door at prototype stage (Courtesy of ChangAn
Automotive Group)

Simulation Theory

Final deformation of injection molded components after assembly includes

process-induced and assembly-induced deformation. The prediction can
normally be done with a combination of warpage analysis, an interface
between warpage analysis and structural analysis and a separate structural
analysis. When the flexible thin-walled plastic components are mounted into
high-modulus metal or rigid parts, only deformation analysis on the plastic
part is required. Based on the superposition principal of the linear system, the
deformation can be regarded as a linear combination of warpage analysis and a
structural analysis with forced displacement. Automatically calculated imposed
displacement at mounting nodes can be treated as a special loading case. The
analysis can be run in one go as long as the deformation is in the linear range
and the forced displacement is set up properly.

Depending on the type of material to be injection molded, mold makers must

take its shrink factor into consideration when producing the cavity set so that
the process-induced shrinkage can be compensated. In practice, the material’s
nominal shrinkage is normally used to scale up the cavity.

With the cavity mesh scaled up using the material’s nominal shrinkage S, the
imposed displacement U at mounting nodes can be derived as:
 U= (X’-C) S/(1+S)

where, X’ represents new coordinate after scaling-up at mounting nodes. C

represents the coordinates of the reference origin for cavity mesh scaling-up.

The algorithms have been implemented for the midplane solution, dual-domain
solution and three-dimensional solution [1]. Both the small deflection analysis
and buckling analysis are valid for the application.

Workflow for Assembly Analysis

The workflow for calculating the deformation and stress after the components
are mounted into the designed positions can be summarized as:

Import the designed component, and mesh it.

Scale up the mesh by the material nominal shrinkage.
It can be done by clicking Utilities =>Move=>Scale in Synergy.
Set appropriate boundary conditions for mounting points. A local coordinate system
can be used when necessary. As a minimum, sufficient constraint must be applied to
constrain rigid body motion.

Switching on “Calculate final deformation after Assembly” and input “Nominal

shrinkage for mold making”.
Run Cool+Flow+Warp or Flow+Warp analysis

The Deflection result will include the effects of the assembly constraints. In addition, a
new “Deviation from part design” in the warpage results shows the geometric
deviation between the final mounted part and the original part design (geometry
before scaling up to compensate for molding shrinkage).

Numerical Examples

Two examples are given to illustrate the effectiveness of the approach

suggested in this paper. The thermo-viscous-elastic residual stress model is
used in both examples.
Case 1. Molded Square Plate Fastened to Working Position at Four Corners

This case is used to illustrate the various scenarios in the assembly

simulation. The designed part size is 40mmX40mmX2mm, and modelled using a
midplane shell mesh. The material is PA6, Tarnamid T-27 natural from Azoty
Tarnow. The processing conditions are shown in Table 1.

Mold Temperature 80C

Melt Temperature 270C
Injection Time 1 second
Packing Time 10 Seconds
Packing Pressure 10MPa
Table 1. Processing conditions

Figure 2. Deformation after mounting with 1.34% cavity scaling-up

To compensate the shrinkage in the injection molding process, the mold cavity
is scaled up by the material nominal shrinkage. From the shrinkage simulation,
the plate will shrink by about 1.34%. If we use 1.34% as the material nominal
shrinkage for cavity scaling-up, we can get a molded part very close to the
designed size. Using the proposed approach, the predicted deformation after
mounting the plate at four corners are shown in Figure 2. It shows that the
geometric deviation from the original design is nearly zero. However, if 1% is
used for material nominal shrinkage for cavity scaling-up, the molded part will
be too small. When it is forcefully mounted at four corners, the molded plate
has been in tension. The predicted deformation after the mounting is shown in
Figure 3. The edges of the plate are slightly bowed in.

Figure 3. Deformation after mounting with 1.0% cavity scaling-up (midplane

solution)

On the other hand, if 2% is used for material nominal shrinkage for cavity
scaling-up, the molded part will be too large. When it is forcefully mounted at
four corners, the molded plate has been in compression. The predicted
deformation after the mounting is shown in Figure 4. The edges of the plate are
slightly bowed out. Figure 5. shows the predicted compressive stress level.
Figure 4. Deformation after mounting with 2.0% cavity scaling-up

The assembly-induced compressive stress could cause part buckling during

the assembly. The buckling analysis[7] from the compressive stress level shown
in Figure 5 shows that the eigenvalue is 0.5935, and it means that the plate
will buckle if the molded part is forcefully mounted at the four corners, and the
buckling mode is shown in Figure 6.
Figure 5. Compressive stress after mounting with 2.0% cavity scaling-up

Figure 6. Buckling analysis after mounting with 2.0% cavity scaling-up

Though these analyses are done on a midplane model, the approach is also
implemented in the well-known dual-domain simulation solution and true
three-dimensional simulation solution for injection molding. Figure 7 shows
the deformation results using dual-domain solution after mounting with 1.0%
cavity scaling up, which similar to the results from midplane solution in Figure
3.

Figure 7. Deformation after mounting with 1.0% cavity scaling-up (dual-

domain solution)

Case 2. Automotive Component

An automotive component is shown in Figure 8. The material used for the

analysis is a 33% glass fiber filled PA66 manufactured by DuPont Performance
Polymers under the name Zytel 70G33L NC010. The process conditions are
given in Table 2.
 Melt 295.5C
Temperature
Injection Time 3.5second
Packing Time 10 Seconds
Packing 21MPa
Pressure
Cooling time 6.5seconds

Table 2. Processing conditions

Figure 8. Model and cooling configuration details

The cooling, filling. packing, fiber orientation and residual stress and
warpage analyses were performed on three-dimensional model. The average
fiber orientation results are shown in Figure 9. For the normal warpage analysis
without considering assembly constraints, the predicted deformed shape is
shown in Figure 10, which showed quite significant distortion on the left side of
the edge.
Figure 9. Predicted average fiber orientation

Figure 10. Warped shape without considering assembly

 The component is mounted into the designed position by six mounting areas.
The mounting area at the top is the anchor position where the constraints in all
directions are applied. In the other 5 mounting positions, only the out-of-plane
displacement is restricted. It is assumed that 1% nominal shrinkage is used for
the cavity scaling-up. To calculate the final deformation after the assembly, the
imposed displacement is calculated by the formulation in the previous section,
and set up as a special boundary condition automatically. A warpage analysis
with considering assembly constraint is run, and final deviation from the
originally designed component shape and size is predicted, as shown in Figure
11. It can be seen that the distortion on the left side of edge is actually reduced
after the assembly.

Figure 11. Final geometric deviation from designed shape and size after
assembly

References

Zhiliang Fan, Russel Speight. Deformation and stress prediction of injection molded
components after being mounted into designed position, SPE ANTEC Proceedings,
Orlando, Florida (2018)
Automatic mesh type classifier
An automatic mesh type classifier has been included in Moldflow 2019 TP.
Instead of choosing mesh type only based on match ratio as before, the new
algorithm uses a combination of new measures to capture 3D features. It was
developed based on hundreds of surface meshes classified manually by
Moldflow experts. The accuracy of automatic classification has been improved
to 93%.
Mesh type recommendation can be found by Mesh Statistics and Mesh Match
Diagnostics in Synergy.

.
Advection Solver application programming interface (API)

In 2019 Technical Preview release we have added a solver API for applying an
arbitrary advection equation. The api can be used for research in different
areas including crystallization, fiber orientation, bending breakage and
migration, multiple component co-injection, etc. Currently the new api is not
fully supported for beam elements.

A few examples of its usage are presented here.

Residence time example

As a simple example we use just the transport equation for the residence time
that can be written as:
𝐷
R=1
𝐷𝑡
R(t=0) =0
R (Inj Nodes)=0,
where R is the residence time.

The solver API code that implements the described equation is simply looks
like:

//A simple example of using advection solver API: calculating residence time
//At the injection node the residence time is zero
//The evolution equation is dx/dt=1

DLLEXPORT int STDCALL SolverUserHb3dUserNodeFieldsDimensions()

{
// Single dimension user field: The current residence time
return 1;
}
//Initialized to zero at the start
DLLEXPORT void STDCALL SolverUserHb3dUserNodeFieldsInitialize(double* X, size_t nodeID,
size_t Lam) {
int nDim = SolverUserHb3dUserNodeFieldsDimensions();
for (int i = 0; i < nDim; i++)
X[i] = 0.0;
}

DLLEXPORT void STDCALL SolverUserHb3dInitialize()

{
// Write a message to the screen at the start of the analysis.
std::wostringstream Message1;
Message1 << L" Example of the Advection Api: Calculating Residence time. ";
SolverUtilityHb3dInfo(Message1.str().c_str());
}

DLLEXPORT void STDCALL SolverUserHb3dUserNodeFieldsEvolution(double* dXdt, double const*

X, size_t nodeID, size_t Lam) {
// Residence time dXdt=1
 dXdt[0] = 1.0;
}

DLLEXPORT void STDCALL SolverUserHb3dUserNodeFieldsCheckAndFix(double* X, size_t nodeID,

size_t Lam) {
//if injection node then the residence time is zero, this is a boundary condition
if (SolverUtilityHb3dIsInjNode(nodeID)) {
X[0] = 0.0;
}
}

For details of the implementation load Program Files\Autodesk\Moldflow

Insight Scandium2
019\data\solverapi\MoldflowUserFunctions_UserfieldExample1.vcxproj project.

We have tested the new api on a plate. The predicted residence time “Solver
API User nodal plot 1” at the end of fill looks like:
Figure 1 Predicted residence time at the end of fill obtained using solver api
UserFieldExample1.

As expected the residence time is the highest on the surface of the part while
in the core zone closed to the injection node the residence time is the lowest.

Thermal degradation example

Another example will be a thermal degradation, where we predict degradation

of plastic during the injection molding.
One model for thermal degradation of PVC that can be described as a first order
kinetic equation:

𝐷 −𝐸𝑎𝑐𝑡
𝐷𝑡
𝛼 = 𝐾(1 − 𝛼)exp⁡( 𝑅𝑇
),
where 𝛼 is the degree of temperature degradation, K and 𝐸𝑎𝑐𝑡 are parameters.
Initially the degradation is zero. The heart of the solver api code is:

Only one-dimensional user field:

extern "C" int STDCALL SolverUserHb3dUserNodeFieldsDimensions()
{
return 1;
}

Initialized the field to zero:

extern "C" void STDCALL SolverUserHb3dUserNodeFieldsInitialize(double* X, size_t nodeID,
size_t Lam) {
int nDim = SolverUserHb3dUserNodeFieldsDimensions();
for (int i = 0; i < nDim; i++)
X[i] = 0;
}

Kinetic equation:
//First order Arrhenius equation
extern "C" void STDCALL SolverUserHb3dUserNodeFieldsEvolution(double* dXdt, double const*
X, size_t nodeID, size_t Lam) {
// Residence time dXdt=1
double T; //Temperature
if (SolverUtilityHb3dGetNodeScalarResult(nodeID, "Temperature", &T)>0) {
T += 273.0;
dXdt[0] = k0*(1.0 - X[0])*exp(-Eact / R / T);
}
else {
//Report error
std::wostringstream Message1;
Message1 << L"Cannot find node with ID " << nodeID;
SolverUtilityHb3dFail(Message1.str().c_str());
}
}
There is also a simple code for reading the kinetic parameters from the
SOLVER_API string. For details load the provided Program
Files\Autodesk\Moldflow Insight Scandium
2019\data\solverapi\MoldflowUserFunctions_UserfieldExample2.vcxproj
project.

A user can specify the kinetic parameters 𝐸𝑎𝑐𝑡 and K for thermal destruction via
the “General text” string. E.g. 𝐸𝑎𝑐𝑡 = 160⁡𝑘𝐽/𝑚𝑜𝑙 and K=8.8E15 can be set as:
Figure 2 Entering kinetic parameters for thermal destruction using solver api
UserfieldExample2

For a square plate the api would predict final thermal destruction showing
“burn marks” near injectors:

Figure 3 Predicted final thermal destruction using solver api

UserfieldExample2

Polystyrene degradation and its effect on the filling pattern

Another example is thermal degradation of polystyrenes. Boon proposed a

model that assumes polystyrene thermal degradation as a zero-order reaction:
𝐷 𝐸𝑎𝑐𝑡
𝛼 = 𝐾exp⁡(− ),
𝐷𝑡 𝑅𝑇
where the activation energy 𝐸𝑎𝑐𝑡 ~170E3 J/mol

Assuming α is the number of breaks per polymer chain we can write:

𝑀𝑤 ~𝜂1/3.4 ~𝑀𝑤0 (1 + 𝛼)−1,
where 𝑀𝑤 is the polymer molecular weight and 𝑀𝑤0 is the molecular weight
prior to thermal destruction.
We can rewrite the last equation as:
𝜂(𝑇, 𝛾̇ , 𝛼) = ⁡𝜂(𝑇, 𝛾̇ )(1 + 𝛼)−3.4,
that we use as the viscosity equation. For the viscosity function without
thermal degradation effect we use Cross-WLF equation. The code is
implemented in the supplied Program Files\Autodesk\Moldflow Insight
Scandium
2019\data\solverapi\MoldflowUserFunctions_UserfieldExample3a.vcxproj
project. It is very simple and most of it is just reading the parameters from the
user-supplied strings.

It includes the kinetic model:

//Zero order Arrhenius equation
extern "C" void STDCALL SolverUserHb3dUserNodeFieldsEvolution(double* dXdt, double
const* X, size_t nodeID, size_t Lam) {
double T; //Temperature
if (SolverUtilityHb3dGetNodeScalarResult(nodeID, "Temperature", &T)>0) {
T += 273.0;
dXdt[0] = k0*exp(-Eact / R / T);
}
else {
//Report error
std::wostringstream Message1;
Message1 << L"Cannot find node with ID " << nodeID;
SolverUtilityHb3dFail(Message1.str().c_str());
}
}
It also includes the viscosity model that should be coded separately for the
nodal and elemental viscosities. The Cross-WLF viscosity is calculated using a
ViscosityCalculator class.

//Using viscosity calculaor to get CrossWLF viscosity and mulipying it on the thermal
destruction term
//The thermal destruction TG is obtaine by using utility function
SolverUtilityHb3dGetUserNodeField
//Implementation for 3D nodes
extern "C" double STDCALL SolverUserHb3dViscosityAtNode(double temperature, double
shearRate, double pressure, size_t nodeId)
{
// Compute viscosity at given temperature, shear rate and pressure.
assert(viscosityCalculator);
assert(temperature > -273.15);
 assert(shearRate >= 0.0);
assert(pressure > -101325.0);

// In power law example, pressure is not used.

double TG = SolverUtilityHb3dGetUserNodeField(nodeId, 0, 0);
return viscosityCalculator->CalculateViscosity(shearRate, temperature,
0.0)*pow((1.0+TG),-3.4);
}

// Using viscosity calculaor to get CrossWLF viscosity and mulipying it on the thermal
destruction term
//The thermal destruction TG is obtaine by using utility function
SolverUtilityHb3dGetUserNodeField
//Implementation of laminated nodes
extern "C" double STDCALL SolverUserHb3dViscosityAtLaminateOfNode(double temperature,
double shearRate, double pressure, size_t nodeId, unsigned int gridId)
{
// Compute viscosity at given temperature, shear rate and pressure.
assert(viscosityCalculator);
assert(temperature > -273.15);
assert(shearRate >= 0.0);
assert(pressure > -101325.0);

double TG = SolverUtilityHb3dGetUserNodeField(nodeId, gridId, 0);

return viscosityCalculator->CalculateViscosity(shearRate, temperature, 0.0)*pow((1.0 +
TG), -3.4);
}

// Using viscosity calculaor to get CrossWLF viscosity and mulipying it on the thermal
destruction term
//The thermal destruction TG is obtaine by using utility function
SolverUtilityHb3dGetUserNodeField
//Implementation for 3D elements
extern "C" double STDCALL SolverUserHb3dViscosityAtElement(double temperature, double
shearRate, double pressure, size_t elementId)
{
// Compute viscosity at given temperature, shear rate and pressure.
assert(viscosityCalculator);
assert(temperature > -273.15);
assert(shearRate >= 0.0);
assert(pressure > -101325.0);
//First lets calculate average TG
double lTG = 0.0;
size_t lCount = SolverUtilityHb3dGetConnectedNodeCount(elementId);
if (lCount > 0) {
size_t* lNode=new size_t [lCount];
SolverUtilityHb3dGetConnectedNodeIds(elementId, lNode);
for (int i = 0; i < lCount; i++) {
double lTG_node = SolverUtilityHb3dGetUserNodeField(lNode[i], 0, 0);
lTG += lTG_node / (double)lCount;
}
delete[] lNode;
}
// In power law example, pressure is not used.
return viscosityCalculator->CalculateViscosity(shearRate, temperature, 0.0)*pow((1.0 +
lTG), -3.4);
}
 A user is supposed to provide kinetic parameters K and Eact in the general
configuration string and the parameters for the WLF equation in the viscosity
configuration string:

Figure 4 Entering kinetic parameters for thermal destruction and Cross-WLF

parameters for viscosity using solver api UserfieldExample3a

We run simulation for a custom double-T intersection mold there the shear
heating in the runners creates flow imbalance in otherwise naturally balanced
mold. Without thermal destruction simulation significantly under-predicts the
disbalance and the hypothesis is that taking into account thermal degradation
could improve the prediction.
We did simulation with different K_0 and E=170E3 J/mol.
If we estimate the uneven filling as the maximal difference in filling time
between branches we have the following results:

K_0 Max filling time, s Min filling time, s Difference, s

0 0.207 0.201 0.006
1e15 0.207 0.201 0.006
3E17 0.207 0.198 0.009
3E18 0.204 0.196 0.008
As the results show using thermal degradation moderately increases predicted
filling unbalance making results closer to the experiment. Further investigation
of the phenomena requires more accurate kinetic equations of thermal
degradation.

Multi-component co injection example

Another example we provide is co-injection of more than two materials (not

supported by the current version AMI). The userfield is concentration of the i-
th component, 𝑋𝑖 (the dimension is number of the components). The evolution
equation is trivial:
𝐷
𝑋 =0
𝐷𝑡 𝑖
The boundary condition is that near the i-th injector the i-th concentration is
1.0 and the other concentrations are zeroes. To avoid excessive smear, we use
“renormalization”. Each component has its own Cross-WLF equation for
viscosity (we can also implement individual PVTs for each component, etc.).
The code is in the Program Files\Autodesk\Moldflow Insight Scandium
2019\data\solverapi\MoldflowUserFunctions_UserfieldExample5.vcxproj.
To make the plot more informative instead of the default UserField plots
showing the i-th component of the nodal field (concentration of the i-th
polymer material) we plot the most prevalent material (the material with the
highest concentration near a particular node). This is done by coding a custom
SolverUserHb3dUserPlotResultAtNode routine that in turn uses a new utility function
SolverUtilityHb3dGetUserNodeField for accessing components of the user field:

double SolverUserHb3dUserPlotResultAtNode(int plotNumber,

double time,
size_t nodeId
) {
// return the component that has the maximal concentration:
int maxComp = 0;
double maxval = 0.0;
for (int i = 0; i < InjectorNodes.size(); i++) {
double frac = SolverUtilityHb3dGetUserNodeField(nodeId, 0, i);
if (frac > maxval) {
maxComp = i;
maxval = frac;
}
}
return (double)maxComp;
}
As an example, we simulated co-injection of five components. The first
component has 10 times as much viscosity as the 2nd,3d and 4th while 5th
component has 1/10 of viscosity of 2nd,3d and 4th. In the “General text” string we
give the node numbers for the injection of those components and in the
viscosity configuration we give Cross-WLF parameters of each component:
Figure 5 Entering nodal numbers for injection of each component and Cross-
WLF parameters for their viscosities using solver api UserfieldExample5

Here are the final distributions of the components:

Figure 6 Final distribution of the components in the five-component co-

injection. Sover api example UserfieldExample5.

Solver API for user-provided fiber orientation

2019 TP provides also an ability for the user to calculate their own custom fiber
distribution. This feature is illustrated by provided FiberExample1 where a user
can set a constant fiber orientation in every node:
// A simple example of using user-calculated fiber orientation:
//Just setting harcoded fiber orientation for all nodes
extern "C" void STDCALL SolverUserHb3dFiberOrientation(size_t nodeID, size_t Lam, double*
axx, double* ayy, double* axy, double* axz, double* ayz)
{
//Just setting fixed fiber orientation
*axx = 0.75;
*ayy = 0.2;
*axy = 0.02;
*axz = 0.02;
*axz = 0.02;
}
 This user-provided fiber orientation will be used instead of those predicted by
the solver in calculations of mechanical properties, CTE, shrinkage and
warpage. They also be used in all fiber-related plots.

The functionality is intended to be used together with the tools for advection
solver api described above and can be used to implement custom fiber
orientation models. To facilitate creation of custom fiber models the api can
access a new fiber orientation configuration string with the parameters of the
model and the evolution function can access the deformation rate tensor: the
parameter often used in fiber orientation calculations.

Most fiber-orientation models including the classical Folgar-Tucker model use

the deformation gradient tensor as an input to the model. In order to help
users in creation custom fiber orientations models via solver api we added an
additional keywords "DeformGradientAxx" … "DeformGradientAzz" to the
SolverUtilityHb3dGetNodeScalarResult utility function. These keywords are
used for the nine components of the deformation gradient tensor.

As an example, the following solver api code would plot the nine components
of the tensor gradients:

#include "MoldflowUserFunctions.h"
#include <string>
#include <cassert>
#include <sstream>
#include <vector>
#include "MoldflowSolverUtility.h"
#ifdef _WIN32
#define STDCALL __stdcall
#else
#define STDCALL
#endif

//custom nodal plot, one plot

int SolverUserHb3dNumberOfUserNodalPlots() {
return 9;
}

// outputted at each timestep

int SolverUserHb3dUserNodalPlotOutputStagesOption(int plotNumber) {
return 1;
}

// output results for filled nodes

int SolverUserHb3dUserNodalPlotOutputDomainOption(int plotNumber) {
return 2;
}

double SolverUserHb3dUserPlotResultAtNode(int plotNumber,

double time,
size_t nodeId
) {
double lRes=0.0;
int lStatus=0;
switch (plotNumber) {
case 1:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAxx", &lRes);
break;
case 2:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAxy", &lRes);
break;
case 3:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAxz", &lRes);
break;
case 4:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAyx", &lRes);
break;
case 5:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAyy", &lRes);
break;
case 6:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAyz", &lRes);
break;
case 7:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAzx", &lRes);
break;
case 8:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAzy", &lRes);
break;
case 9:
lStatus = SolverUtilityHb3dGetNodeScalarResult(nodeId, "DeformGradientAzz", &lRes);
break;
default:
lStatus = 0;
}
if (lStatus) return lRes;
return 0.0; //if something went wrong
}
extern "C" void STDCALL SolverUserHb3dInitialize() {
// Write a message to the screen at the start of the analysis.
std::wostringstream Message1;
Message1 << L" Example of the Utility function: Plotting components of the deform
gradient tensor";
SolverUtilityHb3dInfo(Message1.str().c_str());
}

Reactive Microcellular Injection Molding (with PU

Foaming)

In 2019 TP release, the feature for the simulation of Reactive Microcellular

Injection Molding is added for 3D. This can be used to simulate PU foaming
process.
Workflow

To use this process, “Reactive Microcellular Injection Molding” needs to

be selected as in Figure 1 with mesh type of 3D.

Figure 1: Selecting “Reactive Microcellular Injection Molding” process.

There are several analysis sequences for reactive microcellular process

as shown in Figure 2. There are 2 groups for analysis sequences. One group is
for a regular reactive microcellular process. The analysis sequences in this
group don’t have “(CFM”) in the sequence name. Another group is for CFM
(Chemical Foaming Molding) and the sequences in this group have “(CFM)” in
the sequence name. For PU foaming analysis, an analysis sequence with
“(CFM)” needs to be selected.

Figure 2: Selecting analysis sequence

For the analysis sequences for “regular microcellular injection molding”,

the process setting is similar to that for “thermoplastics microcellular injection
molding” as shown in Figure 3. This is for the simulation of microcellular
process for reactive materials with foaming gas dissolved in the melt at the
barrel at the start of injection.
(a)

(b)
Figure 3: For regular reactive microcellular injection molding: (a) page 1 and
(2) page 2.

For the analysis sequences for CFM (Chemical Foaming Molding), the
process settings are shown in Figure 4. This is for the case where the foaming
gas is generated during molding at the cavity by chemical reaction. For CFM,
only constant number of cells per volume is supported (no bubble nucleation
model is supported).

(a)
(b)
Figure 4: For CFM (Chemical Foaming Molding) analysis sequence: (a) page 1
and (b) page 2.

For CFM analysis, the initial filling method can be selected as shown in
Figure 5(a). This is how the initial filling of the cavity will be done. For CFM
process (such as PU foaming), initially, a small portion of the cavity is filled.
And the initially filled region is expanded to fill the whole cavity by foaming
gas generation, bubble nucleation and growth. The initial filling can be done by
injection or by initial charge. When the initial filling is done by injection,
process settings (such as filling method, V/P switch-over and packing/holding
control) can be assigned as shown in Figure 5(b). Typically, V/P switch-over will
occur at low volume filled for CFM cases.

When the initial filling is done by initial charge, the initially filled region
can be specified by clicking the 3D elements, and select “Initially filled” for the
elements that are filled initially as shown in Figure 5(c).

(a)

(b)
(c)

Figure 5: Selecting the “initial filling method” for CFM cases: (a) Selection of
initial filling method, (b) Filling method setting for initial filling “By injection”
case and (c) assigning the initially filled status for 3D elements.

For CFM, the “Foaming gas generation” method can be selected as shown
in Figure 6. It can be by “PU Foaming” or “General chemical blowing agent
reaction.

Figure 6: Foaming gas generation method.

The specific reaction for “PU foaming” is describe below.

This case involves 2 reactions. One is the gelling reaction as described by the
following reaction:

(1)

Another is the blowing reaction (which generates foaming gas (CO 2)) as
described by the following reaction.

(2)

For any other reaction cases, “General chemical blowing agent” reaction
method can be selected.
 For “PU foaming” case, PU foaming data can be specified as in Figure 7.
Figure 7(a) represent the data needed for PU foaming analysis. The data
needed include “initial water concentration”, “initial OH concentration”, “initial
NCO concentration”. These are by weight %. It also requires “initial dissolved
CO2 concentration”. This is the amount of CO2 gas that can be dissolved in the
melt before nucleation. “Equivalent weight of OH” and “Equivalent weight of
NCO” also need to be assigned.

Equivalent weight (also known as gram equivalent) is a term which has

been used in several contexts in chemistry (Wikipedia). In its most general
usage, it is the mass of one equivalent, that is the mass of a given substance
which will
1. combine or displace directly or indirectly with 1.008 parts by mass of
hydrogen or 8 parts by mass of oxygen or 35.5 parts by mass of chlorine;
or
2. supply or react with one mole of hydrogen cations (H+) in an acid–base
reaction; or
3. supply or react with one mole of electrons (e−) in a redox reaction.
Equivalent weight has the dimensions and units of mass. The equivalent
weight can be obtained or can be calculated from the data provided by the
material supplier.

Optionally, phase changing physical blowing agent (PBA) can also be added
for the blowing reaction. In this case, the “PBA data if using PBA” box can be
checked, and the appropriate data can be specified as in Figure 7(b). The
required data for PBA include “MB” (molecular weight of the blowing agent gas
generated from PBA), “Mno” (initial average molecular weight of the
polymerizing mixture), “Density, BL” (the density of the physical blowing agent
in liquid or solid form), “Heat capacity, BL” (heat capacity of the physical
blowing agent in liquid or solid form), and “Heat capacity, BL” (heat capacity of
the foaming gas generated from physical blowing agent). It also requires
“TB,A” and “TB,B”. These are for the following relation:

XBL = TB,A TB + TB,B (3)

Where, XBL is for the mole fraction of the blowing agent in liquid phase, and T B
is for boiling temperature of blowing agent. It also requires “Heat of reaction”
and “Initial liquid PBA mass ratio”.
(a)

(b)

Figure 7: For PU foaming case: (a) PU foaming data, (b) PBA data.

If “General chemical blowing agent reaction” method is chosen, the

required data are as shown in Figure 8. This is for the case where the CBA
reaction is not that in PU foaming. The required data include “Foaming gas
conversion % of CBA by weight”, “initial concentration of CBA”, “Density of
solid/liquid CBA”, “Specific heat of solid/liquid CBA”, and “Specific heat of
foaming gas”.
Figure 8: General reaction data for “General chemical blowing agent reaction”
case.

The kinetics of blowing agent to generate foaming gas can be specified

as shown in Figure 9. For PU foaming, this is for the CO2 gas generation
reaction shown in equation (2). The data use the same format as that for
thermoset curing kinetics (Kamal’s model).

Figure 9: Chemical blowing agent kinetics data.

An Example Case
One example case is a PU foaming case in a rectangular geometry as
shown in Figure 10(a). This case is from Mitani et al (Polymer Engineering and
Science, v. 43, 2003). The length of the plate is 1100 mm, and the width is 300
mm. The thickness is either 35 or 65 mm. The mold temperature is 30C, initial
melt temperature is 30C. 9.1% of the cavity is filled initially by injection. The
initial water concentration is 0.906%, initial OH concentration is 45.31%, and
the initial NCO concentration is 53.78%. The equivalent weight of OH is 153.7,
and the equivalent weight of NCO is 135. No PBA is used in this case. The
number of cells was assumed to be 1.0x1012 /m3.
The simulation result (fill pattern) is shown in Figure 10(b).
 (a) (b)
Figure 10: (a) The mesh used for PU foaming simulation, and (b) the filling
pattern obtained from simulation.

Anisotropic Thermo-Viscoelastic Residual Stress Model

Introduction

The thermally and pressure-induced residual stresses in the injection molding

process are built up due to the restriction of thermal contraction during the
process, coupled with frozen layer growth with varying pressure history. The
residual stresses are affected by different process conditions such as melt
temperature, mold temperature, injection velocity, packing pressure, packing
time and cooling time. The stress relaxation occurs as the hot polymer is cooled
down to room temperature, which also complicates the stress field.

To simulate the effect of stress relaxation on molded-in residual stress,

shrinkage and warpage of injection molded plastics parts, a three-dimensional
linear anisotropic thermo-viscoelastic residual stress model is developed. The
thermo-rheological simplicity is assumed for the material, and the viscoelastic
master curve is fitted with a generalized Maxwell model. Tabulated time-
temperature shift factors are preferred over the WLF equation and Arrhenius
equation due to its general applicability.

Simulation Technology
The in-mold residual stress will relax a certain amount depending on the
temperature history during cooling, and it can only be predicted by the thermo-
viscoelastic model. The constitutive equation can be written as

 rs T
 ij t    Cijrs  (t )   ( ) ) (
t
  rs )d
0  

where Cijrs is the viscoelastic relaxation modulus,  ij t  is the stress tensor,  ij is

the strain tensor, t is time, T is temperature, and (t ) is a pseudo-time scale
defined as where aT is the time temperature shift factor that
accounts for the effect of temperature on material response.  ij is the tensor of
thermal coefficients of expansion.

The viscoelastic characterization of a material is required for the thermo-viscoelastic model.

It is to determine the relaxation spectrum and the temperature dependence of
a material. The relaxation spectrum over a wide range of time and for the range
of temperatures which occur during the injection molding process. The actual
experiments are done within a narrow range of time with various temperatures.
The time-temperature superposition is used to obtain a single master curve,
which represents the relaxation spectrum at reference temperature. The master
curve can be fitted to a Prony series representing the generalized Maxwell
Model (Figure 1). The general equation can be expressed:

where G(t) is relaxation modulus at time t , which is evaluated at reference

temperature. G(0) is the instant modulus. N is the number of spring-dashpot
Maxwell elements used, represents the relaxation times, and is the
relative dimensionless modulus with .
Figure 1. Generalized Maxwell model

The time-temperature shift factor represents the temperature dependence of

a material, which is obtained at forming the master curve. An example of a
typical shift factor-temperature data is plotted in Figure 2.

Figure 2. Time-Temperature shift factor

For polymer materials, two commonly-used models for the shift factor are the
WLF model and the Arrhenius model. As we can see from Figure 2, it is not
appropriate to fit the shift factor data to a single model over a complete
temperature range, though it is quite common that only one set of shift factor
parameters is given in literature or material database. A more rational
approach is to use two different models for fitting the shift factors over two
different temperature ranges. For example, for amorphous material, the WLF
equation can be used to describe the time-temperature shift factor with the
temperature ranging from Tg to Tg+100o C. For temperatures outside the above
range, an Arrhenius expression can be used. To avoid introducing errors during
the model fitting, a tabulated form for describing the temperature-dependence
of the shift factors is preferred as the input in our implementation.

An incremental constitutive equation can be derived from the linear thermo-

viscoelastic equation, in which the stress state at time can be evaluated
from the known stress state at the previous time step[1]. It is used to calculate
the initial stresses as the driving force for the shrinkage and warpage of an
injection-molded plastics part in a 3D solution.

GUI for Viscoelastic Analysis Option

Viscoelastic material data is required for using the anisotropic thermo-

viscoelastic residual stress model. The setting steps are as follows:

In the process setting wizard -fill+Pack Setting, select “Advanced options”,

then select “Edit” for “Molding material” to Mechanical Properties”
Select “Unified viscoelastic” on “viscoelastic properties data” option.
Click “Edit Unified viscoelastic properties”, and edit the “Viscoelastic
Relaxation Spectrum” and “Temperature shift model” data. “Table data” is a
preferred option for the temperature shift model.

If these data are not available in the database for the chosen material but you
have the data, you can input these data manually. The relaxation spectrum
values should be the relative dimensionless modulus such that .
With these options on, the viscoelastic residual stress model will be used in the
flow and warpage analyses.
Figure 3 . GUI for viscoelastic residual stress model

Numerical example

A numerical case is presented to evaluate the effect of stress relaxation on

molded-in residual stresses and shrinkage by comparing the results between
the thermo-viscous-elastic solution and the thermo-viscoelastic solution.
Effect of pressure and thermal history and stress build-up and relaxation are
taken into account through the simulation of the filling, packing and cooling
stages of injection molding process.

The case is a rectangular strip part, molded using an ABS material, Novodur
P2X of Bayer (Figure 4). The processing conditions and measurement of the
residual stresses with the layer removal method were reported in detail in
reference [2]. The mold cooling was slightly asymmetrical, with wall
temperatures of 51.85°C/47.85°C.

There are two time-temperature shift equations for shifting the relaxation data
to the appropriate temperature in [2]: the WLF-shift is used when T is not less
than Tg, and a special exponential relationship is used when T is less than Tg.
In our implementation, only the tabulated form, WLF equation and Arrhenius
equation are implemented, and therefore these two shift equations of the
material are replaced with an equivalent tabulated temperature-time shift
relationship in Table 1, which can be readily used for our algorithm.

Figure 4. Rectangular strip mold (dimensions are in mm)

Temperature (C) Shift Factor

30 9.625e9
50 1.333e7
70 18465.0
80 687.207
90 25.576
99.85 1.000
105 0.0394
115 0.000342
Table 1. Temperature-time shift factor table
4.706e-9 6.058e-2
4.410e-6 1.037e-1
2.082e-3 3.467e-1
6.198e-1 4.884e-1
3.305e6 4.266e-4
2.746e8 1.415e-4
Table 2. Dimensionless relaxation spectrum data

The dimensionless relaxation spectrum data for equation (3) is listed in table 2.
For better capturing the pressure and temperature history and frozen layer
growth in a gap-wise direction, 20 layers of tetrahedral elements are used.
There are 520311 tetrahedral elements in the mesh. Both the thermo-viscous-
elastic residual stress model and the thermo-viscoelastic residual stress model
were run.

The predicted gap-wise molded-in residual stress profiles in parallel and

perpendicular directions at a position between P2 and P3, where the stress
measurement bars were cut out, are shown in Figures 5 and 6. The measured
residual stress results from [2] are also shown in the figures. Both models
obviously predict the correct residual stress pattern. The final stress profile
shows that some tensile stresses have been built up in the surface layer,
followed by a compressive stress region further inward, then a tensile stress
region in the core. This kind of stress profile is typical for injection-molded
parts, which is closely associated with the relatively low pressure under which
the skin and core have frozen, and high pressure when the intermediate region
has frozen. In addition, the model also correctly predicted the slightly
asymmetrical stress distribution due to the slightly asymmetrical cooling.
Figure 5. Molded-in residual stress profile in parallel direction (unit: MPa)

Figure 6. Molded-in residual stress profile in perpendicular direction (unit:

MPa)

The thermo-viscoelastic residual stress model predicts lower tensile and

compressive residual stress than the thermo-viscous-elastic residual stress
model. This is purely due to the stress relaxation effect. The magnitude of
stress relaxation in the surface layer is more significant as these layers are
frozen and in-cavity stress build-up in the mold has occurred earlier than core
layers, and it has undergone a long stress relaxation time at a relatively high
temperature. The predicted shrinkage from the thermo-viscoelastic model is
about 17% smaller than that from the thermo-viscous-elastic model due to the
stress relaxation effect.

References

1. Zhiliang Fan, Russel Speight. Effect of Stress Relaxation on Shrinkage and Warpage
of Injection Molded Parts, SPE ANTEC Proceedings, Orlando, Florida (2018)

2. W. F. Zoetelief, L. F. A. Douven & A.J. Ingen Housz, Residual Thermal Stresses in

Injection Molded Products, Polymer Engineering and Science, 36, 1886-1896(1996)

Tiger Stripe Prediction

Some injection molded parts show visible surface marks which are
characterised as alternating bands of glossy and dull stripes aligned
perpendicular to the flow direction and with the sequence of gloss and dull
bands inverted on the underside of the part. This defect is commonly known as
“Tiger Stripes”.

Bulters and Schepens [1] at DSM showed experimentally that the Tiger Stripe
pattern was caused by oscillations of an unstable flow front. However, the
existence of flow front oscillations is not by itself considered to be a sufficient
condition for the visibility of Tiger Stripes. Hirano et al.[2] established that the
variations in gloss were caused by the stretch of rubber additive particles in
the polymer, with alternating opposing bands of stretch and non-stretch
caused by the flow front oscillations (see Figure 1). Polymer grades which do
not contain neither such rubber particles nor other additive particles which can
be oriented or stretched by the oscillating flow front will not display visible
Tiger Stripes even if a flow front instability exists during filling.
Figure 1: Alternating stretch pattern due to flow front instability. (Hirano et
al.[2])

The predictive capability provided in this Technology Preview is aimed at

predicting the flow front instability but does not consider whether particles are
present in the polymer compound which will cause bands of visible gloss
differences. The user is required to consider the composition of the polymer
grade when deciding whether a predicted flow front instability will cause
visible Tiger Stripes.

The Tiger Stripe prediction feature is available as part of a Fill analysis (or
Fill+Pack analysis) for 3D mesh types. Two new result plots are provided by the
Tiger Stripe feature: A “Flow front stability factor” result and a “Tiger stripe
region” result. Both these results are automatically added to the default list of
displayed results.

The Flow front stability factor result (Figure 2) can be used to identify the
regions where the likelihood of Tiger Stripe is highest. This result scales a
measure of the conditions leading to flow front instability against the stability
threshold parameter. Stability factor results below 1 indicate regions where a
flow front instability is not expected. Regions of the part with the highest flow
front stability factor are the areas where the likelihood of flow front instability
leading to Tiger Stripes is expected to be highest. However, Tigers Stripes are
not expected close to the gate because the amplitude of the osciallation will
still be weak.
Figure 2: Flow front stability factor result

The Tiger stripe region result (Figure 3) will display a banded pattern in those
regions where the flow front stability factor exceeds 1 and the distance from
the gate is great enough to allow the instability to develop into strong
oscillations. Tiger Stripe bands are not expected close to the gate even if the
flow front instability factor is greater than 1 because the initial oscillations are
weak and do not contact the mold surfaces.

The width of banding depends on the thickness and flow front speed. However,
whether a flow front instability causes visible gloss bands will depend on the
filler composition of the polymer grade. For some polymer grades, a flow front
instability may not lead to significant gloss level banding. The prediction of
this visibility characteristic is not considered in this Tiger Stripe technology
preview.
Figure 3: Tiger stripe region result, 5mm node spacing

The prediction of the flow front stability factor and the regions of potential
Tiger Stripes will be performed for all filling analyses. No user selection is
required to enable this calculation.

In the polymer grade’s material properties, a tab is provided which displays the
properties relevant to Tiger Stripe prediction. (See Figure 4). Viscoelastic
properties and the Tiger Stripe instability threshold can be viewed and edited
on this tab.
Figure 4: Tiger Stripe Material data tab.

The viscoelastic relaxation spectrum follows the same model definition as used
by the viscoelastic stress relaxation feature. If the viscoelastic relaxation
spectrum data is not available, a characteristic relaxation time will be
estimated from the non-Newtonian Cross-WLF viscosity model.

The Tiger Stripe instability threshold parameter is the threshold value of the
characteristic describing the tendency for flow front instability. When this
characteristic exceeds the threshold value (i.e. their ratio is greater than 1), a
flow front instability is expected. It is anticipated that the Tiger Stripe
instability threshold will be tuned for each polymer grade based on observed
Tiger Stripe behaviours from molded parts. Such identification of an optimized
threshold parameter will be specific to whether the characteristic relaxation
time is from measured viscoelastic data or estimated from the Cross-WLF
viscosity model. If the Tiger Stripe instability threshold is not specified, a
default value of 2.6 will be used. Analyses run which include the Tiger Stripe
feature will report the used Tiger Stripe instability threshold in the analysis log
(Figure 5).

Figure 5: Analysis log showing Tiger Stripe Instability Threshold

References

1. M.Bulters and A.Schepens. “The Origin of the Surface Defect ‘Slip-Stick’ on

Injection Moulded Products”. PPS-16 (2000) Shanghai, China. pp.144-145.

2. K. Hirano, Y. Suetsugu, T. Kanai, “Morphological Analysis of the Tiger Stripe on

Injection Molding of Polypropylene/Ethylene-Propylene Rubber/Talc Blends
Dependent on Based Polypropylene Design”, J. of App. Polymer Science, vol104, pp
192-199, 2007.

Underflow diagnostic

You can now easily check the results of 3D filling simulation to find whether an
underflow occurs during filling of your part and what region is affected.
Underflow

Underflow is when a flow front reverses direction. It occurs when flow fronts
from two directions meet, pause momentarily, then one of the flows reverses
direction and flows back between the outer frozen layers. When the flow
reverses direction the frozen layer partly re-melts due to shear heating. This
flow reversal gives poor part quality, both from surface appearance and
structural viewpoints.

Underflow often occurs near weld and meld lines but may happen
independently from them.

In previous releases the users were supposed to identify underflow by playing

the fill time animation from beginning to end and watching for any flow fronts
meeting, then considering the geometry surrounding this point. It is tedious,
especially for 3d simulation.

Severity of underflow
In the present release the software calculates the “severity of underflow”
parameter calculated by the following formula:

𝑡
‖∫0 𝑓𝑖𝑙 𝑣⃑(𝑡)𝑑𝑡‖
𝑆𝑢 = 1 − 𝑡
∫0 𝑓𝑖𝑙 ‖𝑣⃑(𝑡)‖𝑑𝑡

Where 𝑆𝑢 is the severity of underflow, 𝑣⃑(𝑡) is melt velocity vector and 𝑡𝑓𝑖𝑙 is the filling time.
Severity of underflow is always between zero and one. The value of zero shows that melt
velocity has been always in the same direction while the large values indicate that velocity
significantly change direction so a significant underflow occurs.

Our observations show that if severity of underflow is lower than an empirical value of 0.05
no signs of underflow occur.

If somewhere in the part severity of underflow exceed this empirical value, then WARNING
313412 is written to the log file showing the maximal and average severity of
underflow, an additional result is created showing the distribution of
underflow severity in the molding.

What to do

If 3D flow analysis show WARNING 313412 please inspect the underflow

severity plot. If underflow occurs in critical regions of the molding and the
severity is high you should balance the flow by changing the locations of
injection points, balancing the runners or adding valve gates.

Example

Let’s consider a cavity filled from two injection points:

The filling time and velocity show that melt changes velocity when the flow
fronts meet in the middle between the two injection points:
With the new underflow diagnostic results, we immediately see that underflow
occurs from the log files:

The underflow severity plot immediately identifies the area there underflow
occurs:
Since the maximal underflow is quite large (0.83) and underflow occurs in a
large and prominent region the injection mold should be rebalanced by moving
injection points or by balancing melt flow between those points.

Retractable core-pin support

You now can simulate injection molding with retractable core pins.

Injection molding with retractable core pins

Injection molds with retractable pin supporting inserts are used for moldings
that completely cover inserts. Common examples of such moldings are golf
balls where injection molded plastic covers rubber inserts.

Polymer pressure on the rubber inserts causes their deformation while even
the slightest imbalance in the melt flow - e.g. originated from different flow
paths from the sprue towards the gates causes a shift of the inserts that in
turn causes a stronger imbalance, causing higher shift etc. The progressively
increasing imbalance of flow may cause a significant shift of the rubber inserts,
warpage of the balls, variations of the plastic thickness, etc. Those factors are
critically important for the quality of the balls, thus, prediction of the
deformations and shifts of the inserts is an important part of the mold filling
analysis.

The retractable supports are also quite usable for limiting the effect of the core
shift if the cores are not rigid enough to sustain molding pressure. Since
limiting the core shift effects is the main motivation for putting the retractable
pins the prediction of the core shift effects is very important for such molding
processes too.

We take into account retractable core pins as retractable constraints (A) that
are excluded from the analysis after some retraction time. The approach does
not require the user to model the geometry of the support pins themself, while
the second approach is more accurate, allowing to account for the deformation
of the retractable pins themself.

Figure 1 Retractable constraints (A) applied to an insert (B)

Case Study

As a case study we would consider a thermoplastic polyurethane ball filled

from multiple gates in the equatorial plane. The spherical rubber insert is
supported by two steel retractable pins at the poles as shown on Figure 2. The
pins are retracted at 0.6 seconds while the filling time is 0.664 seconds.
Figure 2 Case study: rubber insert is supported by two retractable steel pins
modelled as fixed retractable supports.

The simulation shows that displacements of the rubber insert approach 0.55
mm prior to retraction of the supports (see Figure 3)

Figure 3 Displacements of the rubber insert prior to retraction of supports.

Magnification factor x10

After retraction of the support displacements of the insert redistribute to a

more uniform shift of approximately 0.2 mm. The displacements of the insert
do not change much during packing and are fixed at the end of cool by frozen
thermoplastic polyurethane (see Figure 4) leaving a significantly unbalanced
wall thickness. Those golf balls would probably not be acceptable and a change
in design of the mold or molding conditions is required.

Figure 4 Final displacements of the rubber insert at the end of cooling.

Magnification factor x10

The case study shows the importance of the core shift analysis for the molds
with retractable insert supports.

What to do?

Core shift analysis with retractable core pins predicts the displacements of the
inserts after the injection molding. If the displacement values are not
acceptable you have to redesign the injection mold probably using more
retractable pins and better balanced injection, sometimes modification of
injection molding process parameters may help.

Mesh aggregation for buckling analysis

The buckling analysis type for a 3D mesh model is an eigenvalue buckling

solution which predicts the theoretical buckling strength of an elastic part. The
analysis results suggest the possible deformed shapes of the part as well as
the corresponding eigenvalue which by multiplying the actual load gives the
critical buckling load. The buckling analysis type in Moldflow warpage
simulation predicts the polymer part’s buckling mode as the result of the
residual stresses from the polymer manufacturing process, e.g. injection
molding.

Generally, a buckling analysis is computationally expensive and requires a lot

of computer resources. For a typical 3D model, there are a large number of
degrees of freedom and the computing performance becomes more critical for
a buckling analysis. In this technology preview release, the mesh aggregation
feature is enhanced and extended to eigenvalue buckling analysis. The mesh
aggregation feature aggregates the original mesh and generates an equivalent
coarser mesh with second order elements for the structure analysis, in this
case, the buckling analysis. After structural analysis, the results are mapped
back to the original mesh for result inspection. This way, the required computer
resources are reduced and run speed increased.

The user can check the mesh aggregation option when the buckling analysis
type is chosen in the Warp Settings page of the Process Settings Wizard as
shown below.

Below is an example of applying mesh aggregation to a buckling analysis. This

example is an injection molded part of a simple geometry, the polymer is a
filled PP material. As the result, the difference in the eigenvalue result between
before and after mesh aggregation is 1.93%, while the mesh aggregation option
is 5.02 times faster than the original analysis.
 Original Mesh aggregation
Eigenvalue = 2.164 Eigenvalue = 2.123
Runtime = 532 s Runtime = 106 s

Fiber concentration

Introduction

The fiber segregation effect is known when injecting fiber reinforced plastic. It
is commonly observed that there are more fibers in the core region while less
fibers in the shear region. Such phenomenon is believed to be caused by shear
induced migration.
The fiber concentration is solved in conjunction with 3D flow solution. Once
complete the fiber concentration field is used for composites mechanical
property calculation. The fiber concentration together with fiber orientation
and fiber breakage determine the mechanical properties and thermal
expansion coefficients of the composites, which are important factors to the
final part warpage and subsequent structure analysis

Fiber concentration theory

The fiber concentration module employs the suspension balance model (SBM)
to predict the fiber segregation effect. In SBM the particle normal stress
gradient is believed to be the driven force for particles to migrate, hence the
shear induced fiber segregation and convection of fiber concentration from
upstream to downstream can both be considered.

The SBM model was first proposed by Nott et al, and further improvements
were made by Morris. SBM models the fluid and particle as two phases, but the
full Naiver-Stokes equations are only solved on the suspension (fluid and
particle mixture), and one extra particle-phase mass balance equation, which is
similar to a transport equation, is needed [1] [2] [3].

In Moldflow the SBM model is defined by the following governing equations

∂ρ
⁡ + ∇ ∙ (ρV) = 0
∂t (1)

DV
ρ = −∇P + ∇ ∙ τ⁡ + ⁡ρg
Dt (2)

DT
ρCp = ∇ ∙ (k∇T) + τ: ∇V
Dt
DP (3)
+ ⁡βT
Dt
∂∅
∂t
⁡ + V ∙ ∇∅⁡ = −∇ ∙ j⊥ (4)

ρ is the suspension density; t is time; V is the velocity vector; P is the pressure;

τ is the viscous stress tensor; g is the gravitational acceleration vector; Cp is
the specific heat capacity; k is the thermal conductivity; β is the expansivity; ∅
is the particle volume fraction; j⊥ is the particle migration flux relative to the
bulk motion which is defined in (5)
 2a2
j⊥ = ⁡ f(∅)∇ ∙ ⁡ Σp
9η0 (5)

f(∅) is the sedimentation hindrance function; η0 is the suspension fluid

viscosity; a is the particle radius; Σp is the particle normal stress which is
defined in (6) [3]

Σp = ⁡ −η0 ηn γ̇ Q (6)

ηn ⁡is the particle normal stress viscosity; Tensor Q defines the anisotropy of the
normal stress; γ̇ is shear rate;
(7)
ηn = K n (∅⁄∅m )2 (1 − ∅⁄∅m )−2
λ1 0 0
Q =⁡[ 0 λ2 0]
0 0 λ3 (8)

2
γ̇ = ⁡ √2E: E (9)

∅m is the maximum packing particle volume fraction; K n is a fitted parameter to

match the experimental result; λ1 , λ2 , and⁡λ3 correspond to the principal
directions of flow, gradient, and vorticity;⁡E is the shear rate tensor.

The fiber segregation can be modelled by the SBM model, but problem is that
SBM requires the particles to be tiny hard spheres but fibers are not. The way
to resolve this problem is to use an equivalent sphere to represent the fiber.
The radius of the equivalent sphere 𝑅𝑒 ⁡is calculated based on the volume of the
fiber.

4 1
𝜋𝑅𝑒 3 = 4 𝜋𝑑 2 𝑙 (10)
3

𝑑 is the diameter of the fiber cross section; 𝑙 is the length of fiber.

The fiber concentration does not impact the suspension viscosity, fiber
orientation or fiber breakage, but the mechanical property calculation takes the
non-uniform fiber concentration into account.

Fiber concentration workflow

The fiber concentration feature can be enabled from the fiber solver
parameters entry which is shown in figure 1
Figure 1 Fiber concentration entry

The fiber property page has been extended to include the fiber concentration
related parameters which are shown in figures 2 and 3.

Figure 2 Fiber concentration parameters entry

Figure 3 Fiber concentration parameters

Initial fiber volume fraction is calculated via fiber weight fraction, material
solid density, and fiber density. Using material solid density for calculating the
initial volume concentration makes the volume concentration result match the
mechanical property calculation without fiber concentration feature being
turned on.

∅𝑖𝑛𝑖𝑡 = 𝑤 ∗ 𝜌𝑠 /𝜌𝑓 (11)

∅𝑖𝑛𝑖𝑡 is the initial fiber volume fraction; 𝑤 is the fiber weight fraction;⁡𝜌𝑠 is the
material density of solid state; 𝜌𝑓 is the fiber density;

The fiber equivalent radius can be determined from equation (10). The default
value is 0.02mm.

The maximum fiber volume fraction, which is ∅𝑚 in equation (7), is the

maximum packing volume fraction in the suspension. This value cannot be
smaller than the initial volume fraction. Its default value for fiber
concentration is 50%.

Particle normal stress coefficients can be found in equations (6) to (8). The
particle normal stress can be determined from experiments and in Moldflow K n
is given 3.0 by default for fiber concentration calculation; 𝜆1 , 𝜆2 , 𝑎𝑛𝑑⁡𝜆3 are 1.0,
0.8 and 0.5 by default.

Run the flow analysis with the fiber concentration turned on. Once completed
there is a new result named “Fiber volume concentration” in the default result
list. One demonstrative plot is given in fig 4 to show the fiber segregation
through the thickness of the plate. It can be seen that the lowest fiber
concentration is the shear region while the highest fiber concentration is at the
center of the plate.

Figure 4 Fiber volume concentration

Given the fiber concentration is no longer uniform through the part the
mechanical properties and thermal expansion coefficients prediction of the
composites have also been modified to take the non-uniformed concentration
into account. Figures 5, 6 and 7 show the plots through thickness of the fiber
concentration, the tensile modulus and thermal expansion coefficient (CTE) in
the first principle direction comparison between the case without fiber
concentration and the case which has the fiber concentration on. It can be seen
that previously as of the fiber alignment at the shear region hence the tensile
modulus in the first principle is maximum and the CTE is minimal, while the
core region shows the opposite result as it is least aligned there; However, with
the non-uniformed fiber concentration being taken into account it relieves
such effect as there is less fiber in the shear region now but more fiber in the
core region.

Figure 5 Fiber concentration

Figure 6 Tensile modulus comparison

Figure 7 CTE comparison

References

1. P. R. Nott and J. F. Brady, “Pressure-driven flow of suspensions: simulation and

theory”, J. Fluid Mech. 275, 157-199, (1994).

2. J. F. Morris and F. Boulay, "Curvilinear flows of noncolloidal suspensions: The role

of normal stresses", J. Rheol.43 (5), 1213-1237, (1999).

3. JR Clausen, "Using the suspension balance model in a finite-element flow solver",

Computers & Fluids 87, 67-78
Limitations:

1. Only support thermoplastic injection molding process.

2. Only support 3D mesh.

Annealing analysis with viscoelastic material properties

for midplane and dual-domain

A new capability of annealing analysis for midplane and dual-domain

thermoplastics injection molding processes has been added. Viscoelastic
material data are needed for this analysis. At present, uniform temperature
change in the part is assumed at the specified value during the reflow stage.

Workflow

To run annealing analysis, “Add reflow profile for annealing analysis” option
needs to be checked as in Figure 1(a). The reflow profile can be specified as in
Figure 1(b).

(a)
(b)
Figure 1: (a) “Add reflow profile for annealing analysis” option and (b) an
example of reflow temperature profile.

For the “Mechanical Properties”, “Unified viscoelastic” as shown in figure 2

needs to be provided with appropriate data for the viscoelastic properties.

Figure 2: Using “Unified viscoelastic” data for the viscoelastic analysis.

Example Case

The example is for a simple rectangular plate case shown in figure 3. The
material is PS. The mold temperature is 35°C, and the melt temperature is
218°C. Mid-plane mesh is used for the analysis. For this case, the reflow profile
shown in figure 1(b) is used.
Figure 3: Mesh used in the example case study.

The warpage calculation results (without and with reflow) are shown in Figure
4. For this case, the warpage has decreased after reflow.

(a)
(b)

Figure 4: (a) Calculated warpage result without reflow and (b) with reflow.
Shrinkage calibrated coefficient of thermal expansion for
unfilled semi-crystalline materials

The lack of good material data in anisotropic thermo-viscoelastic residual

stress calculations and crystallization simulation necessitates the use of a
simplified thermo-viscous-elastic residual stress model. This model has a
constant solidification temperature with available material data obtained
under laboratory conditions. It can generally provide acceptable results
helping users optimize the processing conditions for mold and part design. The
model calibration and correction can improve the predicted accuracy further.

Measured shrinkage data for a material complements measured mechanical

values quite well, containing extra information such as temperature-dependent
material properties and stress relaxation effects from testing samples. Using
measured shrinkage data for calibrating the coefficient of thermal expansion
(CTE) values in the 3D residual stress model can improve the residual stress
and shrinkage prediction. This scheme is available in the 3D residual stress
calculation on unfilled semi-crystalline materials with available data in the
Shrinkage Molding Summary field of the Shrinkage Properties tab in the
Thermoplastics material dialogue.

The steps for activating the option is process Setting Wizard->Advanced

options-> Warpage analysis->tick “Calibrate CTE from shrinkage measurement
for 3D residual stress calculation” (as shown in the figure below).
Below is the validation result for Moldflow’s lab tray. The left figure below is
the buckling mode shape result using the 3D residual stress model with the
shrinkage-calibrated CTE scheme, which is consistent with the experimental
result (right picture). Without the shrinkage-calibrated CTE scheme, the
predicted buckling mode is a dome shape which does not agree with the
experimental result.

Thickness shrinkage scheme for residual stress accuracy

improvement

This scheme uses the part thickness information to assist the shrinkage
calculation in the 3D warpage simulation when using the 3D residual stress
shrinkage model. It is particularly useful for better warpage prediction for
parts with corner-like geometrical features. This alternative scheme is
available if the Uncorrected residual stress is selected as the 3D shrinkage
model option. To activate this scheme, check the UI option in the Warp Analysis
tab of the Thermoplastics injection molding solver parameters (3D) dialogue as
shown below.

Below is an example of how the alternative scheme helps improve the warpage
prediction of a box part, for which the shrinkages between the centres of edges
are measured and compared. The example is a box part with a fiber-filled Poly
Propylene material injection molded with conformal cooling. Four different
mold temperatures are tested in the example. The simulation results are
obtained with the 3D mesh type.
In this technology preview release, the thickness shrinkage scheme is
improved and is extended to fiber-filled materials.

Limitation:

The scheme of thickness shrinkage calculation is not available for the Generic
shrinkage model.