You are on page 1of 9

Journal of Manufacturing Processes 62 (2021) 28–36

Contents lists available at ScienceDirect

Journal of Manufacturing Processes


journal homepage: www.elsevier.com/locate/manpro

A new approach to the evaluation of ejection friction in micro


injection molding
Davide Masato a, *, Marco Sorgato b, Giovanni Lucchetta b
a
Department of Plastics Engineering, University of Massachusetts Lowell, One University Avenue, Lowell, MA, United States
b
Department of Industrial Engineering, University of Padova, via Venezia 1, Padova, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: The ejection phase influences the quality and integrity of micro injection molded parts. The successful design of
Ejection force robust micro mold ejection systems requires studying the tribological interactions at the mold/polymer interface.
Ejection friction At the micro-scale, the tooling topography can have a significant impact on the ejection friction. Here we propose
Modeling
a novel approach to the evaluation of ejection friction in micro injection molding. The two main contributors to
Shrinkage calibration
Micro injection molding
the ejection force are the normal force due to shrinkage and the static friction coefficient. The former is addressed
by developing a procedure for shrinkage characterization at the micro-scale. The latter is studied using exper­
imental measurements of the ejection force in micro injection molding. Comparing the numerical and the
experimental results allows identifying the friction coefficient as a function of polymer, process parameters, and
mold surface roughness.

1. Introduction on the thermo-mechanical analysis of stresses generated during the


process. Shrinkage models base their calculation by correlating the
In micro injection molding (μIM), the ejection phase is carried out by simulated pressure and temperature history to the pvT curves, measured
applying an ejection force in some designed locations of the part [1]. for a specific resin. The selection of different process parameters and the
Ejector pins can apply high local stresses to the part, causing possible part’s geometry are significant variables that could affect shrinkage and
distortion, stress marks, and fractures [2]. increase the interaction with mold core geometries. Shrinkage predic­
Approaches to the modeling of the injection molding ejection force tion accounts for the thermally induced stress, which arises from
are not systematic and well-established yet. In particular, the analysis of freezing and subsequent shrinkage of the material and the pressure-
the ejection phase is complicated for micro parts [3]. The development induced stress. The use of material shrinkage data allows for signifi­
of predictive models for the ejection force is required to improve mold cant improvements in shrinkage predictions [8]. The Corrected Residual
design, especially for smaller features that are more prone to damage In-Mold Stress (CRIMS) model can improve the prediction using exper­
[4]. However, accurate and calibrated models for predicting the imental data from shrinkage measurements [9]. The model improves the
demolding force in μIM are currently not available in the literature. prediction that could be obtained with material data obtained under
Amonton`s law of friction has been widely used in the literature to laboratory conditions. Indeed, test parts are molded under different
estimate the ejection force [5,6]. Following this approach, the force processing conditions, allowing the inclusion in the model of processing
required to eject a plastic molding from a core is a function of the normal phenomena that are significant for shrinkage behavior. CRIMS param­
force and the static coefficient of friction developed at the polymer/­ eters are usually determined for conventional parts with wall thick­
mold interface [7]. Modeling these parameters required focusing on two nesses that are larger than 1 mm. When considering micro-scale plastic
aspects: the polymer shrinkage’s characterization and the study of the parts, the shrinkage behavior drastically changes for small wall thick­
friction coefficients. Both aspects have been studied in the literature: ness [10]. This creates a need for calibrating the shrinkage behavior
several researchers have developed models to improve the predictions’ using micro-scale parts. The ejection force prediction for micro parts
accuracy. requires accurate shrinkage models that consider micro-molding pro­
The prediction of polymer shrinkage for injection molding is based cessing conditions.

* Corresponding author.
E-mail address: Davide_Masato@uml.edu (D. Masato).

https://doi.org/10.1016/j.jmapro.2020.12.032
Received 3 June 2020; Received in revised form 26 October 2020; Accepted 15 December 2020
Available online 22 December 2020
1526-6125/© 2020 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Fig. 1. Flow chart of the approach developed to calibrate a shrinkage model and determine ejection friction coefficients.

The prediction of plastic parts’ friction on the injection mold surface representative geometry, were compared to the simulated results to
has been the focus of several researchers [11–14]. Ejection friction has optimize the model’s material data shrinkage. The calibrated model was
been most often studied using offline tribological setups designed to then applied to the simulation of the contact pressure generated on
mimic the polymer’s ejection behavior in contact with a mold surface micro-scale deep cores’ surface. The friction of behavior of different
[12]. However, correlation with injection molding experiments is still thermoplastic resins over mold surfaces with different roughness was
limited [15]. Most of these studies focus on plastic parts of a conven­ studied. The coefficient of friction was modeled using the shrinkage
tional scale. Minimal knowledge of micro-scale polymer/mold interac­ prediction and experimental measures of the ejection force.
tion is available in the literature [4]. Studies of ejection friction for
micro-molding applications focus mainly on surface treatments and 2. Approach to ejection force modeling
coatings that can reduce stiction [16]. Sorgato et al. studied the effect of
different mold coatings on the ejection force for a part with micron-sized In injection molding, for simple mold geometries (e.g., deep cores),
holes [17]. Models for predicting the static coefficient of friction at the the ejection force (Fe) can be predicted by using the following equation:
micro-scale need to be developed to promote modeling of ejection forces
Fe = FN ∙μ (1)
for micro molding.
In this work, a model for predicting the ejection force in micro in­ where FN the normal force acting on mold cores during the demolding
jection molding was developed and applied to a case study geometry. phase, and μ is the friction coefficient [18]. The procedure developed to
The calibration of a numerical shrinkage model was performed, calibrate the ejection friction model is shown in Fig. 1. The coefficient of
following an inverse analysis approach. Experimental shrinkage values friction controlling the mold/polymer interactions was determined
obtained from injection molding experiments, carried out with a using a hybrid approach based on numerical shrinkage prediction and

29
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Fig. 2. Schematics of the model used to calculate the normal force generated by a ring of material acting on a core.

experimental micro injection molding. The procedure follows two main


Table 1
directions: (i) the calibration of a shrinkage model and (ii) the experi­
Main properties of the polymers selected for the study of the ejection force
mental measure of ejection forces with different mold topographies.
model.
Two different parts were selected for shrinkage calibration and ejection
Material Crystal 1540 Ultraform H2320 Topas 5013L-10
force measurements. The part selected for shrinkage calibration was
adapted from conventional injection molding standards (cf. Section Type Polystyrene Polyoxymethylene Cyclic Olefin
3.2.1). The ejection force measurements were carried out with a (PS) (POM) Copolymer (COC)
Structure Amorphous Crystalline Amorphous
different part to evaluate mold geometries critical for ejection (cf. Sec­
Transition 100 145 134
tion 3.3.1). temperature (◦ C)
When modeling the ejection phase, polymer shrinkage is crucial as it Ejection 86 110 100
determines the normal force FN acting on the mold surface. Here we temperature (◦ C)
Melt temperature 235 200 300
propose to calculate FN using the deformation (i.e., shrinkage) of a
(◦ C)
polymer ring onto a mold core (Fig. 1 – Shrink. A). For simple ring-core Elastic modulus 3200 3200 3200
geometry (Fig. 2), the following equation can be used [19]: (MPa)
Linear Mold 0. 4− 0. 7 2.1 0.4− 0.7
δrg Eg ( )
FN = ( )∙ 2πrf L (2) Shrinkage (cm/
2 +r2
rog cm)
rf f
2 − r2
rog
+ νg
f

friction coefficient that characterizes the ejection was determined by


where δrg is the radial displacement for a ring of polymer around a mold monitoring the ejection force (Fig. 1 – Eject. D) and using the calibrated
core, rf and L are the dimensions of the polymer ring shrinking around shrinkage models. The resulting information on mold/polymer friction
the core, rog is the deformed ring’s external radius and Eg, νg are the properties demonstrates the procedure’s efficacy and can be used to
mechanical properties of the ring. model the ejection of more complex geometries.
The fundamental cause for shrinkage in injection molding is the
thermodynamic behavior of polymers, which is characterized by
3. Experimental
determining the relationship between pressure (p), specific volume (v),
and temperature (T) in the pvT curve. During the injection molding
3.1. Polymers selection
process, polymer shrinkage causes residual stresses that affect the con­
tact pressure at the interface with the mold. Upon ejection, the polymer/
Three different thermoplastic resins were selected for the calibration
mold interface stresses control the force required to eject the plastic part.
and validation of the ejection force model, specifically a polystyrene (PS
In this work, we develop and validate a model to predict the radial
– Total, Crystal 1540), a cyclic olefin copolymer (COC – Topas, 5013 L-
displacement from a shrinkage simulation (Fig. 1 – Shrink. D), which is
10) and a polyoxymethylene (POM – Basf, Ultraform H2320). These
previously calibrated for a simple geometry (Fig. 1 – Shrink. C) using
were selected to study the model for different resins that are widely used
dimensional measurements of micro molded parts (Fig. 1 – Shrink. B).
in μIM applications. PS and COC are commonly used for medical ap­
The calibrated simulation allows accurate prediction of the normal force
plications because of their biocompatibility and transparency, while
considering the residual stresses generated upon cooling of the polymer
POM is generally used for micromechanical injection molded parts.
melt due to its processing temperature and pressure history.
Table 1 reports the main properties of the selected polymers.
According to Eq. 1, the ejection force prediction is also controlled by
the static coefficient of friction between the solidified polymer and the
core surface (μ) (Fig. 1 – Eject. A). This work aims to evaluate the friction 3.2. Shrinkage calibration
behavior of different thermoplastic resins when molded over different
mold topographies (Fig. 1 – Eject. E). Indeed, the micromachining pro­ 3.2.1. Part and mold design
cess leaves a footprint on the mold topography, crucial for ejection The part considered in this study is a square plaque with a side length
friction [20]. Moreover, as Masato et al. reported, polymer selection of 10 mm and a thickness of 350 μm, as shown in Fig. 3 (a). The part’s
significantly affects the interlocking of the mold/polymer interface [15]. design was selected based on conventional injection molding standards
The procedure adopted here included (i) generation of different mold [21] and methodologies proposed in the literature for the μIM process
topographies using μEDM (Fig. 1 – Eject. B), and (ii) injection molding [22]. The square cavity was positioned on the mold’s B-side at the end of
experiments using different polymers (Fig. 1 – Eject. C). The static a fan gate (thickness: 0.2 mm), which ensured a linear flow front and a

30
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Fig. 3. Plastic part designed to calibrate a shrinkage model: (a) 2D drawing with nominal dimensions indicated in millimeters; (b) isometric view.

experiments. These were set considering datasheet information and


Table 2
optimized to guarantee proper filling and part quality.
Processing parameters used for manufacturing of the micro parts for shrinkage.
Parameter PS POM COC 3.2.3. Shrinkage measurements
Melt temperature [◦ C] 240 235 305 The linear shrinkage was evaluated by measuring the molded
Mold temperature [◦ C] 60 80 50 microparts through a coordinate measurement machine (Werth, Video-
Injection speed [mm/s] 200 200 200
Check-IP 400). The measurements were performed at room temperature
Packing pressure [bar] 150 150 150
Packing time [s] 6 6 6 48 h after the ejection [23]. The following procedure was defined and
Cooling time [s] 6 6 6 followed for the measurements (Fig. 4):

i the coordinates of five equally spaced points for each edge of the
balanced filling flow. square molded parts were acquired, and a straight line was inter­
The mold cavity was machined with a 5-axis micro-milling machine polated (Fig. 4 (a)),
(Kugler, Micromaster 5X), and its dimensions were characterized using a ii lines were fitted on the four edges of the part using the acquired
multi-sensor coordinate measurement machine (Werth, Video-Check-IP points (Fig. 4 (b)),
400). The cavity’s real dimensions were 9.899 mm in the flow direction iii two symmetry lines were created in the flow and transverse di­
and 9.906 mm in the perpendicular direction. rections and used to define a midpoint for each edge of the part
(Fig. 4 (c)),
3.2.2. Manufacturing system iv calculation of the distance between the edge midpoints on the
A state-of-the-art μIM machine (Wittmann Battenfeld, MicroPower opposite edges of the parts (Fig. 4 (d)).
15) with a maximum injection speed of 750 mm/s and a maximum
clamping force of 150 kN was used in the experiments. The shrinkage was calculated in the flow (Sf) and the perpendicular
The mold temperature was set and controlled using four electrical directions (St) as the percentage of reduction from the mold dimensions:
cartridges, two for each mold side, and two thermocouples.
Table 2 reports the process parameters that were adopted for the μIM

Fig. 4. The procedure adopted for the shrinkage measurements performed on the molded micro plaques.

31
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Table 3 3.3. Ejection force monitoring


EDM discharge parameters used to machine the mold cores to the design 400 μm
diameter. 3.3.1. Mold design and machining
Parameter/Core Set A B C D The ejection force model was applied to a case study geometry used
Incremental Depth [mm] 0.0008 0.001 0.0015 0.002
to obtain experimental values of the frictional force acting on the
Frontal Gap [mm] 0.002 0.0025 0.003 0.005 polymer during the demolding phase. The interaction at the interface
Frequency [Hz] 180 140 130 90 between the mold and the polymer was maximized using the deep core
Current [A] 100 100 50 80 geometry (diameter of 400 μm and a height of 2 mm – aspect ratio of 5)
Voltage [kV] 90 100 130 150
described by Sorgato et al. [20]. The designed plastic part is a disc with a
Gain 230 350 1500 1000
uniform wall thickness of 2 mm characterized by six through holes
(diameter: 400 μm). The mold’s ejection system comprises three ejector
pins (diameter: 2 mm) regularly positioned at 120◦ intervals on a
Table 4
15.5 mm circumference.
Process parameters selection for the three injection molding polymers.
Four different sets (A, B, C, D) of interchangeable cavity cores were
Process parameter PS Crystal POM Ultraform COC 5013L- realized employing micro electro-discharge machining (μEDM) using
1540 H2320 10
different machining parameters [15]. Considering the direct propor­
Melt Temperature (◦ C) 240 235 305 tionality between the friction coefficient (μ in Eq. 1) and the ejection
Mold Temperature (◦ C) 60 80 50 force, the generation of different topographies allows for more in-depth
Injection Speed (mm/s) 200 200 200
Packing pressure (bar) 150 150 150
validation of the proposed model (cf. Section 2).
Packing Time (s) 6 6 6 Table 3 reports the main machining parameters used to machine the
Switch-over v/p (bar) 600 640 670 deep cores using μEDM. The mold surface’s roughness was characterized
Cooling Time (s) 10 10 10 using an optical profiler (Sensofar, Plu Neox), operating in confocal
Ejection Temperature 85 110 115
mode with a 20X objective [24–26].
(◦ C)

3.3.2. Experimental approach


( )
Lf ,part − Lf ,mold Scientific injection molding experiments were carried out to deter­
Sf = ∙100 (3)
Lf ,mold mine the friction force that characterizes ejection when using different
mold topography (i.e., surface roughness from 0.23 μm up to 1.67 μm)
( )
Lt,part − Lt,mold and different resins (i.e., PS, POM, COC). Table 4 reports the processing
St = ∙100 (4)
Lt,mold parameters used to mold the case study parts. A single forward stroke of
the ejector pins (3 mm at a speed of 10 mm/s) was used to separate the
where Lf,part and Lt,part represent the dimensions of the molded part as molded plastic from the mold cavity.
experimentally measured in the flow and the transverse directions, During the ejection phase, the force exerted by all the 2 mm pins on
respectively. Lf,mold, and Lt,mold are the dimensions of the mold measured the plastic part was acquired by a single piezoelectric transducer (Kis­
after machining. tler, 9223A) mounted on the ejector plate. The sensor’s voltage output
was then processed and logged using a data acquisition unit (National

Fig. 5. Modeling of the plaque used for shrinkage calibration – (a) mesh and (b) deflection (all effects) result for one of the runs. Note: the deflection result is scaled
by a factor of 2.

32
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Table 5 experiments.
Selected ranges of variation for the CRIMS parameters for each polymer. The CRIMS model was calibrated using the shrinkage measurements
CRIMS PS POM COC performed on the micro plaque samples, as described in Section 3.2.3.
Due to the simple part geometry, a Dual Domain meshing approach was
a1 0 - 2.49 0 - 3.38 0 - 1.52
a2 − 0.37 - 0.41 − 2.09 - 1.85 − 0.49 - 0.32 adopted to discretize the part (Fig. 5 (a)). Then, simulations were run to
a3 0 - 0.01 − 0.03 - 0.05 0 - 0.01 determine part deflection (Fig. 5 (b)) and compare with experimental
a4 0 - 2.89 0 - 6.57 0 - 1.61 measurements. The values of the CRIMS parameters that minimize the
a5 − 0.56 - 0.42 − 3.2 - 2.74 − 0.65 - 0.37 difference between experimental measures and numerical calculations
a6 0 - 0.01 0 - 0.06 0 - 0.01
were determined using an inverse analysis approach (Section 4.2).

Instruments, NI cDAQ-9172) and the National Instruments Labview 4.2. Inverse analysis
2013 software (sampling rate: 90 kHz, i.e., time step of 0.01 ms).
The values of the CRIMS parameters that force the numerical results
4. Shrinkage modeling to fit the measured shrinkage values were determined using an inverse
analysis approach. The differences between numerical and experimental
The ejection force model presented in Section 2 was calibrated, dimensions of a representative micro part were minimized using an
considering the variables that characterize Eq. 1. The normal force iterative optimization algorithm. To reduce the computational time, an
acting on the cores (FN) was calculated using shrinkage-calibrated in­ artificial neural network (ANN) was used as a meta-model to approxi­
jection molding simulations. The static friction coefficient (μ), which mate the simulation results locally [27]. The artificial neural networks
controls polymer/mold surface interactions, was obtained from injec­ were first trained to reproduce the simulation of the micro plaque
tion molding experiments in which the ejection force was acquired using described in Section 3.2.1. The difference between numerical and
a piezoelectric force cell (cf. Section 3.3.2). experimental results was minimized using an optimization procedure
based on a MOGA-II genetic optimization algorithm [28]. For each
iteration (set of CRIMS parameters), the meta-model calculates the
4.1. CRIMS shrinkage model
shrinkage and iteratively compares it to the experimental values.
The initial range of variation of the CRIMS parameters was defined
In Autodesk Moldflow, different models are proposed to calculate the
for each polymer, considering similar polymers in the Autodesk Mold­
polymer’s shrinkage because of the processing pressure and temperature
flow database. The material database contains CRIMS parameters for
distributions [9]. The most accurate model is the ‘Corrected Residual
several materials obtained by experimental characterization using a
In-Mold Stress` (CRIMS) model, which efficacy lies in the calibration of
large (40 × 200 × 2 mm) rectangular plaque. Table 5 reports the range
its parameters through experimental shrinkage measurements [8].
of each CRIMS parameter that was selected for each polymer.
Shrinkage calculation in the CRIMS model is based on the following
equations:
4.3. Normal force modeling
ε(c) (p) 0
11 = a1 ε11 + a2 ε + a3 (5)
The calibrated shrinkage model was used to determine the normal
ε(c) (p) 0
22 = a4 ε22 + a5 ε + a6 (6)
force due to shrinkage acting on the mold surface upon part ejection.
where ε(c) (c)
11 , ε22 are the corrected principal strain values in the parallel, The model is used to calculate shrinkage around 0.4 mm mold cores that
and perpendicular directions to the flow, ε(p) (p)
11 and ε22 are the predicted characterize the plastic part’s geometry designed in Section 3.3.1. Fig. 6
principal strain values in the two directions, and ε0 is the value of the (a) shows the dual-domain mesh created to analyze the case study
strain related to orientation effects. The CRIMS parameters, a1, a2, a3, a4, plastic part.
a5, a6, are data-fitted coefficient determined by inverse analysis to fit the A Fill + Pack + Warp analysis sequence was run, and the Isotropic
residual stresses to shrinkage values obtained from injection molding Shrinkage result was considered to evaluate the displacement of the

Fig. 6. Mesh of the case study part (a) and isotropic shrinkage contour plot for one simulation run.

33
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Table 6 Table 7
Optimized CRIMS parameters for each resin. Shrinkage for a ring of material around a mold core for the different polymers.
CRIMS PS POM COC Shrinkage [%]
Position
a1 1.2275028 1.2384333 0.0610429 PS POM COC
a2 − 0.0103715 − 0.1714228 0.0086159
Inner 0.992 2.417 0.291
a3 − 0.0016706 − 0.0045474 0.0023909
Outer 1.001 2.436 0.293
a4 0.0072468 0.0000001 0.0968220
a5 − 0.5539396 − 0.3147111 − 0.4649388
a6 0.0061820 0.0204584 0.0029365
Eq. 2 as the radial displacement δrg.

5. Results and discussion

5.1. Shrinkage calibration results

For each polymer, the inverse analysis approach was used to deter­
mine the set of CRIMS parameters that minimize the difference between
numerical deflection predictions and experimental dimensional mea­
surements. Table 6 reports the optimized values of CRIMS parameters
obtained from the ANN used in the inverse analysis calibration.
Fig. 7 displays the comparison between experimental shrinkage
values and those obtained with the different simulation models. The
CRIMS values obtained from the optimization, which used ANN, were
input into Moldflow to run calibrated simulations. The results indicate
that the shrinkage values obtained from calibrated simulations are very
close to those calculated by the ANN. A prediction error smaller than 4%
confirms the accuracy of using the ANN in place of the numerical
simulation.
Considering experimental shrinkage measurements, the errors
committed by the CRIMS-calibrated model were 4%, 0.2 %, and 2% for
PS, POM, and COC, respectively. Such small errors confirm the accuracy
of the proposed approach to shrinkage calibration for the micro molded
part. Moreover, for all the polymers, the CRIMS model yielded the
closest results to the experimental shrinkage measurements.
Compared with the Uncorrected Residual Mold Stress (URMS), the
CRIMS calibration significantly improved the numerical prediction. The
URMS model calculates shrinkage considering the polymer’s linear
thermo-viscoelastic behavior, which is included in Moldflow through
the pvT curve for each resin. Using the URMS model, the difference
between predicted and experimental shrinkage values of 25 %, 10 %,
and 260 % for PS, POM, and COC, respectively. This indicates the
importance of calibrating shrinkage models using injection molding
information (i.e., dimensional plastic part measurements) and not only
data from material testing (i.e., pvT curve).

5.2. Radial displacement calculations

The calibrated shrinkage model was applied to evaluate the poly­


mer’s radial displacement around mold cores (cf. Section 2), allowing
calculation of the normal force for each resin. The average shrinkage did
not vary significantly through the part (i.e., variations smaller than 5%)
considering the reduced dimensions of the part and the cores. Thus, for
Fig. 7. Comparison of average shrinkage values for experimental ANN, CRIMS, each polymer, a single representative shrinkage value was considered by
and URSM. calculating each polymer’s average deformation around the six cores.
Table 7 reports the values of shrinkage determined for the different
polymer around the core (Fig. 6 (b)), according to the definition pro­ polymers. Among the different resins, POM has the highest shrinkage
vided in Section 2. Eq. 2 was applied considering the shrinkage of a ring due to its semi-crystalline nature. Moreover, higher shrinkage was
of polymer around each one of the through-holes. The ring had an inner observed for PS compared to COC due to the different chemical structure
diameter of 0.4 mm, the same as the mold cores, and an outer diameter and processing parameters.
that was fixed at 0.8 mm, which was defined considering the shrinkage
distribution around the hole. The shrinkage for a ring of material was 5.3. Ejection force experimental results
calculated considering (i) the triangular elements that create the inner
surface around the through-hole and (ii) elements located at a radial The topography of the mold cores, machined with different selec­
distance of 0.8 mm from the center of the through-hole. tions of μEDM parameters, was evaluated considering the roughness
The shrinkage calculation was used to determine the elastic strain amplitude of the topography Sa, allowing the evaluation of the differ­
that loads the plastic around each mold core, which eventually enters ences between the core surfaces. Different sets of mold cores were used

34
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

Table 8 Table 10
Results of experimental ejection force measurements with different sets of mold Mechanical properties (at room temperature) of the polymers used to model
cores. elastic strain due to shrinkage.
Fpeak [N] Polymer properties Symbol Unit PS POM COC

Cores Set Sa [μm] PS POM COC Elastic Modulus Eg MPa 3200 2450 2850
Poisson Coefficient vg – 0.35 0.42 0.41
Avg. 0.23 65.8 34.6 115.3
A
Std. Dev. 0.02 0.9 0.8 1.1
Avg. 0.60 88.3 45.3 158.9
B
Std. Dev. 0.05 1.8 0.8 1.3
Table 11
Avg. 1.01 98.7 49.7 187.1
C Ejection friction coefficients for the different polymers with different mold
Std. Dev. 0.03 0.7 0.7 1.5
Avg. 1.67 113.5 68.2 278.4 roughness.
D
Std. Dev. 0.02 0.4 0.9 1.0 Friction Coefficient

Sa PS POM COC

0.23 0.38 0.12 2.20


Table 9 0.60 0.51 0.15 3.04
Parameters used for the calculation of the normal force due to shrinkage. 1.01 0.57 0.17 3.58
1.67 0.66 0.23 5.32
Symbol Value Unit

Core geometry
Length L 2 mm which yielded radial displacements for each polymer (Table 7). Table 9
Radius rc 0.2 mm
summarizes the parameters that were used to calculate the normal force
Center Interspace D 3.5 mm
Number n 6 due to shrinkage using the model of Eq. 2. From the radial displacement
Dimensions of a ring molded over a core predicted for each polymer using the calibrated shrinkage model, the
Internal radius rig 0.2 mm elastic strain around the mold cores was calculated, and thus the normal
External radius rog 0.4 mm force was obtained. Table 10 reports the mechanical properties of the
Mold properties
Elastic modulus Ec 210,000 MPa
polymers that were used for the calculations.
Poisson coefficient vc 0.33 The calculation yielded a normal force due to shrinkage of 172 N,
293 N, and 52 N for PS, POM, and COC, respectively. These results
correlate with the polymers’ different shrinkage behavior, indicating
for an experimental campaign in which the ejection force peak was that the semi-crystalline structure of POM leads to higher shrinkage
measured. Table 8 reports the average and standard deviation values for compared to amorphous PS and COC. The effect of mold surface
the different mold topographies and the measured ejection force peak roughness was not considered when modeling the normal force due to
for each polymer. shrinkage. Indeed, mold topography has a negligible effect on polymer
The topography characterization of the different sets of mold cores solidification, and thus on its shrinkage during injection molding.
indicates that surface roughness increases by 161 %, 339 %, and 626 % An iterative procedure was implemented to determine the coefficient
when changing μEDM combinations of parameters from A to B, to C, and of friction using the calculated normal force values. The ejection force
to D. This confirms the importance of discharge parameters on the was calculated using the model, and the difference with the experi­
footprint left by the machining process onto the mold surface [20]. mental values was minimized. The inverse analysis minimization target
The different sets of mold cores resulted in significantly different was defined as the sum of square differences between predicted and
ejection behavior. In particular, the high surface roughness of set D experimental ejection force values at the different values of surface
resulted in the highest ejection force peak. The rough mold topography roughness.
generates the highest stiction at the polymer/mold interface. Reducing Table 11 shows the values of the friction coefficients determined
surface roughness by changing the set of mold cores from D to C, from C using the iterative procedure. The results indicate that the polymers
to B, and from B to A resulted in an average reduction of Fpeak by 23 %, generate diverse interaction at the polymer/mold interface, which leads
10 %, and 22 %, respectively. to different stresses during ejection. The lowest friction was observed
The effect of mold surface roughness on the ejection force was when molding acetal, which is well-known for its low friction coefficient
consistent for the different resins, indicating a linear correlation be­ and high stiffness. Conversely, high friction was measured with COC,
tween surface roughness and Fpeak. However, the rate of change was which has low shear viscosity and high flowability into the mold
steeper for some polymers compared to others. In particular, COC was topography. The stiction created with this resin leads to a frictional force
more sensitive to roughness variations than PS and POM. This indicates (i.e., topographical effect) higher than the normal force (i.e., shrinkage
the importance of modeling the mold/polymer interactions, such as effect). The calculated values are higher than the coefficient of friction
polymer melts’ ability to replicate the mold topography, and of studying numbers available in the literature with offline friction testing [12].
the effect of polymer shrinkage. Indeed, the online ejection force measurement accounts for the topog­
raphy replication and the effect of the injection molding processing
5.4. Ejection friction conditions.

The ejection force measurements from the experimental injection 6. Discussion


molding campaign were used to study ejection friction using the model
presented in Section 2. An inverse analysis approach was used to The modeling of the ejection force carried out in this work allows
determine the friction coefficient μ that characterizes the ejection force considerations on the combined effect of shrinkage and mold surface
model (Eq. 1). The ejection force was model was calibrated by calcu­ roughness. It can be observed that when designing an ejection system for
lating the values of the friction coefficient μ that minimize the difference micro injection molding, it is essential to consider both polymer
between the predicted ejection force values and those obtained from shrinkage and mold surface topography.
injection molding experiments. The efficacy of the procedure developed to calibrate polymer
The normal load (FN) acting on mold cores during ejection was shrinkage was confirmed using a case study geometry. The CRIMS
predicted using the shrinkage-calibrated model described in Section 4.3, model’s use allowed the evaluation of micro injection molding

35
D. Masato et al. Journal of Manufacturing Processes 62 (2021) 28–36

We anticipate modeling of ejection forces for micro injection mold­


ing will require the creation of large databases for friction coefficients.
This could be accomplished using online ejection force measurement,
rather than offline friction testing with sample surfaces, as proposed in
this work. The ejection system’s successful design for plastic micro parts
would have to rely on experimental data that represent different mold/
polymer interfaces.

Declaration of Competing Interest

The authors report no declarations of interest.

References
Fig. 8. Correlation between mold surface roughness and friction coefficients
determined using inverse analysis. [1] Araújo BJ, Pouzada AS. Design of ejection systems for injection moulds. O Molde
2002;54:36–41.
[2] Griffiths CA. Investigation of surface treatment effects in micro-injection-moulding.
phenomena that significantly affect polymer shrinkage. In particular, Int. J. Adv. Manuf. Tech. 2010;47:99–110.
the reduced wall thickness, the high shear rates, and the extreme cooling [3] Delaney KD. A structured review and classification of demolding issues and proven
solutions. Int Polym Process J Polym Process Soc 2012;27:77–90.
ratio lead to different part morphology, indicating the necessity of [4] Delaney KD. Demoulding force prediction for micro polymer replication: a review
calibration at the microscale [10]. of relevant literature. 4M Conf. 2010.
The values of the coefficient of friction, calculated by inverse anal­ [5] Menges G, Mohren P. How to make injection molds. 2nd edn. Munich: Hanser
Publishers; 1993.
ysis to the experimental ejection forces, include both deformation and [6] Pontes AJ, Pouzada AS, Pantani R, Titomanlio G. Ejection force of tubular injection
adhesion contributions [29]. However, the resins showed a diverse moldings. Part II: a prediction model. Polym Eng Sci 2005;45(3):325–32. https://
response to an increase in surface roughness. Fig. 8 displays the linear doi.org/10.1002/pen.20275.
[7] Kazmer DO. Injection mold design engineering. Carl Hanser Verlag GmbH Co KG;
correlation between surface roughness and the predicted friction co­
2016.
efficients. Polymers characterized by higher shear viscosity, such as PS [8] Cellere A, Lucchetta G. Identification of CRIMS model parameters for warpage
and POM, are less sensitive to variations of mold surface roughness than prediction in injection moulding simulation. Int J Mater Form 2010;3(1):37–40.
COC, which can completely replicate the mold surface roughness. This [9] Kennedy PK, Zheng R. Shrinkage of injection molded material. In: Greener J,
Wimberger-Friedl R, editors. Precision injection molding: process, materials and
generates strong interlocking at the mold/polymer interface, and ejec­ applications. Munich: Hanser Gardner; 2006.
tion stiction has to be overcome by high ejection force. [10] Masato D, Rathore J, Sorgato M, Carmignato S, Lucchetta G. Analysis of the
It should be noted that the contribution of mechanical interlocking at shrinkage of injection-molded fiber-reinforced thin-wall parts. Mater Des 2017;
132:496–504.
the mold/polymer interface is much higher than that of polymer [11] Correia MS, Miranda AS, Oliveira MC, Capela CA, Pouzada AS. Analysis of friction
shrinkage. The semi-crystalline POM, which is characterized by high in the ejection of thermoplastic mouldings. Int J Adv Manuf Technol 2012;59
shrinkage, yielded the lowest ejection friction because of its reduced (9–12):977–86.
[12] Pouzada AS, Ferreira EC, Pontes AJ. Friction properties of moulding
interaction with the mold surface. Hence, a complete description of the thermoplastics. Polym Test 2006;25(8):1017–23.
machining footprint is fundamental to understand friction at the part/ [13] Masato D, Sorgato M, Parenti P, Annoni M, Lucchetta G. Impact of deep cores
mold interface and develop further ejection force models [20]. surface topography generated by micro milling on the demolding force in micro
injection molding. J Mater Process Technol 2017;246:211–23.
[14] Sasaki T, Koga N, Shirai K, Kobayashi Y, Toyoshima A. An experimental study on
7. Conclusions ejection forces of injection molding. Precis Eng 2000;24(3):270–3.
[15] Masato D, Sorgato M, Lucchetta G. Effect of ultrasound vibration on the ejection
friction in microinjection molding. Int J Adv Manuf Technol 2018;96(1-4):345–58.
In this work, a novel approach to modeling ejection friction was
[16] Saha B, Toh WQ, Liu E, Tor SB, Hardt DE, Lee J. A review on the importance of
developed and characterized using injection molding experiments and surface coating of micro/nano-mold in micro/nano-molding processes.
numerical simulation. The ejection force is modeled from a shrinkage J Micromech Microeng 2015;26(1):013002.
prediction that is obtained from calibrated numerical simulations. Then, [17] Sorgato M, Masato D, Lucchetta G. Tribological effects of mold surface coatings
during ejection in micro injection molding. J Manuf Process 2018;36:51–9.
the ejection force’s online monitoring allowed the analysis of the friction [18] Menges G, Michaeli W, Mohren P. How to make injection moulds. New York:
at the polymer/mold interface for different resins and mold topography. Hanser; 1993.
The following conclusions can be drawn based on the results obtained in [19] Hamrock B. J, fundamentals of machine elements. Singapore: WCB McGrawHill.
International Edition; 1999.
this research: [20] Sorgato M, Masato D, Lucchetta G. Effects of machined cavity texture on ejection
force in micro injection molding. Precis Eng 2017;50:440–8.
• The use of experimental measurements to calibrate the CRIMS model [21] BS EN ISO 294-3. Plastics – injection moulding of test specimens of thermoplastic
materials - part 3: small plates. 2003.
allows significant improvement in numerical simulation of micro [22] Annicchiarico D, Attia UM, Alcock JR. A methodology for shrinkage measurement
parts shrinkage. in micro-injection moulding. Polym Test 2013;32(4):769–77.
• Ejection friction is strongly affected by the selection of different [23] Fischer J. Handbook of molded part shrinkage and warpage. William Andrew;
2012.
resins and by the mold topography. The interface interlocking con­ [24] ISO 4287. Geometrical Product Specifications (GPS) - Surface texture: profile
trols the force required to eject the part. method - Terms, definitions and surface texture parameters. 1997.
• It is possible to calculate the normal force acting on mold cores upon [25] ISO 4288. Geometrical Product Specifications (GPS) - Surface texture: profile
method - Rules and procedures for the assessment of surface texture. 1996.
ejection using an analytical model. This allows the separation of the
[26] ISO 5436-1. Geometrical product specifications (GPS) - Surface texture: profile
shrinkage and friction contribution to the ejection force. method - Part 1: material measures. 2000.
• It is possible to determine the static coefficient of friction for a spe­ [27] Sadeghi BHM. A BP-neural network predictor model for plastic injection molding
cific polymer/mold interface using an inverse analysis approach. process. J Mater Process Technol 2000;103(3):411–6.
[28] Quagliarella D, Périaux J, Poloni C, Winter G. Genetic algorithms and evolution
These resulted significantly higher than those reported in the liter­ strategies in engineering and computer science. John Wiley & Sons; 1997.
ature and measured with offline tribometers. [29] Suh NP, Mosleh M, Howard PS. Control of friction. Wear 1994;175(1–2):151–8.

36

You might also like