Optimization and Numerical Simulation Analysis for Molded

Thin-Walled Parts Fabricated Using Wood-Filled

Polypropylene Composites Via Plastic Injection Molding

M.D. Azaman,1,2 S.M. Sapuan,1,3 S. Sulaiman,1 E.S. Zainudin,1,3,4 A. Khalina5,6

1
Department of Mechanical and Manufacturing Engineering, Universiti Putra Malaysia, 43400 UPM Serdang,
Selangor, Malaysia
2
School of Manufacturing Engineering, Universiti Malaysia Perlis, 02600 Arau, Perlis, Malaysia
3
Laboratory of Biocomposite Technology, Institute of Tropical Forestry and Forest Products (INTROP), Universiti
Putra Malaysia, 43400 UPM Serdang, Selangor, Malaysia
4
School of Engineering and Design, Brunel University, Uxbridge Middlesex, UB8 3PH, UK
5
Department of Biological and Agricultural Engineering, Universiti Putra Malaysia, 43400 UPM Serdang, Selangor,
Malaysia
6
Aerospace Malaysia Innovation Centre, Universiti Putra Malaysia, 43400 UPM Serdang, Selangor, Malaysia

Plastic injection molding is discontinuous and a compli-
cated process involving the interaction of several variables
for control the quality of the molded parts. The goal of this
research was to investigate the optimal parameter selec-
tion, the significant parameters, and the effect of the
injection-molding parameters during the post-filling stage
(packing pressure, packing time, mold temperature, and
cooling time) with respect to in-cavity residual stresses,
volumetric shrinkage and warpage properties. The PP 1 60
wt% wood material is not suitable for molded thin-walled
parts. In contrast, the PP 1 50 wt% material was found to
be the preferred type of lignocellulosic polymer composite
for molded thin-walled parts. The results showed the lower
residual stresses approximately at 20.10 MPa and have
minimum overpacking in the ranges of 20.709% to
20.174% with the volumetric shrinkage spread better over
the part surface. The research found that the packing pres-
sure and mold temperature are important parameters for
the reduction of residual stresses and volumetric shrink-
age, while for the reduction of warpage, the important
processing parameters are the packing pressure, packing
time, and cooling time for molded thin-walled parts that
are fabricated using lignocellulosic polymer composites.
POLYM. ENG. SCI., 00:000–000, 2014. V
C 2014 Society of Plastics
Engineers
INTRODUCTION
Injection molding in serial production is mostly a very stable
process. It is discontinuos and complicated process, which
involves the interaction of several variables for the achievement
of a good-quality part. These variables can be classified in terms
of molding parameters, materials, product design, and mold
design. The process needs appropriate setting parameters. The
selection of appropriate machining parameters for the injection-
molding process becomes more difficult for applications that
concern thin-walled parts and use lignocellulosic polymer com-
posites (i.e., wood-filled polypropylene composites). Further-
more, the processing of lignocellulosic polymer composites is
Correspondence to: M.D. Azaman; e-mail: azaman@unimap.edu.my
DOI 10.1002/pen.23979
Published online in Wiley Online Library (wileyonlinelibrary.com).
C 2014 Society of Plastics Engineers
V method, the signal-to-noise and analysis of variance (ANOVA)
usually limited to temperatures below 230 C to minimize fiber
degradation as reported by Sanadi et al. [1]. Processing at low
temperatures makes it difficult for the polymer melt to flow into
the mold cavities and often leads to an inconsistent distribution
of residual stresses, volumetric shrinkage and warpage in
molded products, particularly in thin-walled parts.
Residual stresses, shrinkage, and warpage are the three major
challenges in injection molding. They are inevitable in many
cases, especially for parts with complex geometries, thin-walled
parts, micro-parts, certain materials, etc. According to Jacques
[2], the warpage of a molded part results from an asymmetrical
stress distribution over the thickness of the part that is caused
by unbalanced cooling during the cooling stage but also by ani-
sotropic fiber or filler orientations. Therefore, the thinnest region
is the most sensitive to warpage. Furthermore, Wang et al. [3]
have also noted that residual stresses in turn result from nonuni-
form shrinkage of molded parts. Unfortunately, a nonuniform
shrinkage distribution on molded parts is inevitable because of
several factors related to inconsistent cooling processes, nonuni-
form pressure distributions, the orientation of molecules and/or
additives, part design, mold design, processing conditions, and
the interactions among all these factors.
Therefore, a researcher or process engineer should have
knowledge and understanding of methods for the statistical
design of experiments (DOE) to identify the optimum interac-
tions among the variables in the injection-molding process. Pre-
viously, Azaman et al. [4] have reported that it is preferable to
design a product that uses shallow thin-walled parts rather than
flat thin-walled parts when molding lignocellulosic polymer
composites in terms of quality, rigidity of structure, and econ-
omy. On the basis of this premise, the next stage of optimization
concerns the selection of variables in the molding process
parameters (filling time, packing pressure, packing time, cooling
time, mold temperature, injection pressure, etc.) and the deter-
mination of their effects on part quality. According to Giboz
et al. [5], the level of warpage and shrinkage is highly depend-
ent on the molding process parameters. They propose that
efforts to reduce warpage and shrinkage to an acceptable level
should be focused on the careful control of the molding process
parameters.
Because of these considerations, application of the Taguchi

POLYMER ENGINEERING AND SCIENCE—2014

seems to be a more practical approach to the statistical design
of experiments than other methods, which appear to be more
complicated. The more complex approaches of other researchers
[6, 7], such as Liao and Hsieh [6], who have applied back-
propagation artificial neural networks (BPANN) to this problem,
do yield more optimal process conditions than does Taguchi’s
method of minimizing the shrinkage and warpage of injected
thin-walled parts. However, in the author’s view, the selection
of a method for the statistical design of experiments should be
based on multiple criteria, including practicality, efficacy, ease
of construction, and adequate accuracy. These criteria, when
considered collectively, favor the Taguchi method.
Many industries concur with this assessment and have
employed the Taguchi method over the years to improve their
products or manufacturing processes. The Taguchi approach
seems to provide practical and effective tools for solving chal-
lenging quality problems. This method has been used quite suc-
cessfully in several industrial applications such as the
optimization of manufacturing processes and the design [3, 8].
According to Chin [9], the Taguchi method was developed by
systematically allocating factors and levels to suitable orthogo-
nal arrays, then performing an analysis of the signal-to-noise
ratio (S/N) and an analysis of variance (ANOVA) to determine
the optimal combination of parameters, validate the results and
identify the significant parameters that affect the part quality.
This article presents a detailed method for such a calculation
that can be used as a reference for researchers/engineers to build
their calculations using available software, such as Microsoft
Office Excel.
Chen et al. [10] have analyzed a cover component used in a
power supply that was molded using a novel heat-resistant poly-
amide (PA9T). The application of numerical simulations (Mold-
Flow) and the integration of experimental and statistical
techniques were used to determine the optimal parameter selec-
tion and the most significant parameters that affected the warp-
age results. The melt temperature and the packing pressure were
found to be the most significant factors that affected warpage in
both the simulation and the experimental analysis.
Altan and Yurci [11] have focused on the optimization of the
injection-molding process parameters (melt temperature, mold
temperature, and cooling time) to minimize thermal residual
stresses in the surface regions of polystyrene (PS) and high-
density polyethylene (HDPE) parts. The melt temperature and
mold temperature were found to be the most significant parame-
ters that affected residual stresses in the surface regions of the
PS and HDPE parts, respectively. Altan [12] has also investi-
gated the optimal molding parameters to ensure the minimum
shrinkage value for a rectangular-shaped molded part using the
Taguchi method and ANOVA. The results demonstrated that the
packing pressure and melting temperature were most significant
factors that affected the shrinkage of the polypropylene (PP) and
polystyrene (PS), respectively, while the injection pressure was
the least significant factor for both materials.
Wang et al. [3] have investigated the effects of processing
parameters, including melt temperature, injection time, packing
pressure, packing time, and cooling time, on the warpage prob-
lem for a molded LCD TV front shell. The results indicated that
the part warpage can be effectively reduced based on the opti-
mal combination of parameters determined using signal-to-noise
analysis, while the packing pressure is the most significant fac-
2 POLYMER ENGINEERING AND SCIENCE—2014
tor in controlling the warpage. Chang and Faison [13] have
studied the shrinkage and optimization of general-purpose poly-
styrene (GPS), high-density polyethylene (HDPE), and
acrylonitrile-butadiene-styrene (ABS) parts using the Taguchi
method and ANOVA. The optimization results indicated that the
mold and melt temperatures as well as the holding pressure and
holding time were the most significant factors that affected the
shrinkage.
Erzurumlu and Ozcelik [14] have constructed a study apply-
ing three-level factorial designs, an L18 orthogonal array, the
S/N ratio, ANOVA and numerical simulation to determine the
effects of rib design and molding parameters on warpage and
sink indices. The packing pressure was found to be the most sig-
nificant factor affecting warpage and sink indices in PC/ABS,
while the melting temperature and rib layout angle had signifi-
cant effects on the warpage and sink index, respectively, for
polyamide (PA66) material. Ozcelik and Sonat [15] via ANOVA
analysis, have shown that the packing pressure has the greatest
effect on the warpage of the final product and that the casing
structure is strongest when filled with 15 wt% carbon fiber-
reinforced PC/ABS, compared with casing with no filler.
Huang and Tai [16] have used computer simulations
TM
(C-MOLD ) with the assistance of the Taguchi method to
determine the effective molding parameters that affect the warp-
age behavior of a molded, box-shaped geometric part. The
results indicated that packing pressure is the most important fac-
tor that affects warpage. Of the other factors, gate dimension
and filling time were found to be the least important factors
affecting warpage. In addition, Wang et al. [17] have also
reported that packing pressure is the most important factor that
affects the shrinkage and warpage of injected thin-walled parts.
Tang et al. [18] have conducted experiments based on the Tagu-
chi method, using an L9 orthogonal array and optimizing via
ANOVA. This research investigated the effects of molding
parameters such as melt temperature, filling time, packing pres-
sure and packing time on the warpage of thin-walled parts. The
results indicated that melting temperature was the most signifi-
cant factor, while filling time was least significant factor that
affected the minimum warpage.
Oktem et al. [19] have used the Taguchi method to investi-
gate the warpage problem related to shrinkage variation, which
can be reduced by modulating the molding parameters (injection
time, packing pressure, packing time, and cooling time) for
molded thin-shell features. The results indicated that the packing
pressure and the packing time were the most significant factors
affecting the warpage and shrinkage behaviors. Lio et al. [20]
using the Taguchi method and an F-test, have found that the
packing pressure is the most important factor that affects the
warpage and shrinkage of a molded thin-walled part, while other
parameters such as mold temperature, melt temperature, and
injection speed were less significant.
The limitation of all this research is the lack of direct experi-
mental validation of the results of numerical simulations. How-
ever, previous researchers have nevertheless used results from
numerical simulations for the purpose of optimization. Hakimian
and Sulong [8] have also used numerical simulations to determine
the optimal molding parameters for molded microgears fabricated
using polymer composites. A similar approach has been used by
Erzurumlu and Ozcelik [14] to investigate the minimization of the
warpage and sink index of molded thermoplastic parts.
DOI 10.1002/pen
 TABLE 1. Specifications of injection-molding machine.

Parameters Units Value

Maximum machine injection stroke mm 176.8

Maximum machine injection rate cm3/s 112.1
Machine screw diameter mm 30
Maximum machine injection pressure MPa 247
Maximum machine clamp force MPa 79.80

FIG. 1. A shallow, thin-walled part.
A review of previous studies indicates that investigations of
molded thin-walled parts fabricated using lignocellulosic poly-
mer composites in the injection-molding process have been
rarely reported. Most researchers have focused on the use of
nonreinforced polymers and reinforced polymers in the molding
of thin-walled parts to investigate the relations between the
molding parameters and their effects on part quality. Hence, the
aim of this study is to investigate the effects of injection-
molding parameters during the post-filling stage (packing pres-
sure, packing time, mold temperature, and cooling time) and to
determine the optimal selection of parameters for three different
types of lignocellulosic polymer composite materials with
respect to in-cavity residual stresses, volumetric shrinkage, and
warpage properties. Numerical simulations with the assistance
of the Taguchi method, the S/N ratio and ANOVA are used in
this research. In addition, this study should permit the reduction
of trial molding times and the improvement of part quality, and
importantly, it can also serve as a reference in the further inves-
tigation of the molding defects of thin-walled parts fabricated
using lignocellulosic polymer composite.
METHODOLOGY
Part Design
R
Autodesk Inventor ProfessionalV was used to model the thin-
walled molded parts, as shown in Fig. 1. A shallow, thin-walled
part was created as a 3D design. The general dimensions of the
part were 55 mm 3 50 mm 3 0.7 mm.
Numerical Simulation
R
Autodesk MoldFlow InsightV was used to simulate and ana-
lyze the injection-molding process. A mesh model was devel-
oped, as shown in Fig. 2. The injection-molding machine and
the materials used in the simulation adhered to the following
specifications: Arburg Allrounder 370c 88-ton injection-molding
machine (screw diameter of 30 mm); PP 1 40 wt% wood com-
posite from NCell 40, GreenCore Composites; PP 1 50 wt%
wood composite from Isoform Lip CPCW50, Isokon; and
PP 1 60 wt% wood composite from WPC-2-mv, Fraunhofer
Institut. Table 1 shows the specifications of the injection-
molding machine. Table 2 shows the material properties of the
lignocellulosic polymer composites. The simulation was per-
formed using a set analysis (Fill 1 Cool 1 Fill 1 Pack 1 Warp)
for these models. The post-filling processing parameters for the
simulations are shown in Table 3. The fixed parameters used for
further detailed analysis of the post-filling parameters are setting
the injection time to 1 s; melt temperature to 185 C; ejection
temperature to 104 C and mold open time to 5 s.
Measurement of In-Cavity Residual Stresses, Volumetric Shrinkage,
and Warpage
The results indicate that the in-cavity residual stresses lie
along the first principle direction (plotted at the centre of the
surface). According to the suggestion of Altan and Yurci [11],
the stress values near the surface region were taken into account
during the optimization analysis. High stresses near the surface
regions of molded parts cause the parts to be more susceptible
to cracking caused by environmental stresses and to be sensitive
to chemical diffusion. Similarly, the volumetric shrinkage and
warpage were measured at the center of the surface.

TABLE 2. Material properties of lignocellulosic polymer composites.

PP 1 40 PP 1 50 PP 1 60
wt% wood wt% wood wt% wood

Trade name NCell 40 ISOFORM LIP WPC-2-mv

CPCW 50
Filler content (wt%) 40 50 60
Material structure Semicrystalline Semicrystalline Semicrystalline
Melt flow rate 3 5 3.81
(g/10 min)
Melt temperature ( C) 190 185 190
Mold temperature ( C) 50 45 50
Aspect ratio (L/D) 1 1 1
of fillers
FIG. 2. A mesh model for simulation.

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—2014 3

 TABLE 3. Post-filling processing parameters.
Parameters Values
Packing pressure 0.8Pinject20.95Pinject MPa
Packing time 10–40 s
Mold surface temperature 40–50 C
Cooling time 10–40 s
Experimental Design
Structure of Taguchi Orthogonal Arrays. Taguchi orthogonal
arrays consisting of 16 experiments with four levels were applied,
as four processing parameters were used in the analysis. Table 4
lists the processing parameters and levels, while Table 5 shows
the detailed arrangement of the orthogonal arrays that was used
for the analysis of the three investigated materials.
Statistical Analysis. In determination of the signal to noise
(S/N), the smaller-is-better quality characteristic has been
selected for each of three responses: in-cavity residual stresses,
volumetric shrinkage, and warpage. The optimum set of
parameters was determined from the S/N ratio for each poly-
mer composite. The penultimate stage of the Taguchi method
is the verification of the predicted results via confirmation on
the optimum set of parameters. In addition, the collected data
were also analyzed using ANOVA to determine which signifi-
cant factors would affect all responses. An example of a statis-
tical calculation is presented in Appendix A.
RESULTS AND DISCUSSION
Analysis of the S/N Ratio For Optimization of Molding Parameters
and Materials
The average results for in-cavity residual stresses, volumetric
shrinkage and warpage for different types of lignocellulosic
polymer composites (PP 1 40 wt% wood, PP 1 50 wt% wood
and PP 1 60 wt% wood) were determined and are shown in
Table 6. In the Taguchi method, the (S/N) ratio is used to ana-
lyze the collected data to determine the optimal combination of
parameters and also to predict the optimum value. A smaller-is-
better S/N ratio was defined following Eq. 1 in Appendix A.
This selection is appropriate to the study’s objectives regarding
the reduction of residual stresses, volumetric shrinkage and
warpage on molded thin-walled parts using different types of
lignocellulosic polymer composites.
Based on Table 7, the highest S/N ratio for each processing
parameter gives the optimum combination of processing condi-
tions for the minimization of the three selected responses of the
investigated materials, while the difference between the maxi-
mum and minimum values of the S/N ratio for each mean
parameter can be used to make an initial prediction regarding
the parameters that significantly influence the responses. This
claim is proven and explained by a statistical analysis using
analysis of variance (ANOVA) in a discussion that appears later
in the paper. The study found that the packing pressure and
mold temperature are the most important parameters for the
reduction of residual stresses and volumetric shrinkage on thin-
walled parts. This is contrary to the results for the reduction of
warpage, for which the most important processing parameters
for all three materials are the packing pressure, packing time
and cooling time.
As listed in Table 8, for the PP 1 40 wt% wood material, the
optimum processing conditions for the minimization of in-cavity
residual stresses, volumetric shrinkage and warpage are a pack-
ing pressure of 0.95Pinject, a packing time of 30 s, a mold tem-
perature of 40 C and a cooling time of 40 s; a packing pressure
of 0.95Pinject, a packing time of 40 s, a mold temperature of
40 C and a cooling time of 10 s; and a packing pressure of
0.85Pinject, a packing time of 40 s, a mold temperature of 50 C
and a cooling time of 10 s, respectively.
For PP 1 50 wt% wood, the optimum processing conditions
for the minimization of in-cavity residual stresses, volumetric
shrinkage, and warpage are a packing pressure of 0.95Pinject, a
packing time of 40 s, a mold temperature of 40 C and a cool-
ing time of 30 s; a packing pressure of 0.95Pinject, a packing
time of 40 s, a mold temperature of 40 C and a cooling time
of 30 s; and a packing pressure of 0.95Pinject, a packing time
of 10 s, a mold temperature of 55 C and a cooling time of
40 s, respectively.
For PP 1 60 wt% wood, the optimum processing conditions
for the minimization of in-cavity residual stresses, volumetric
shrinkage and warpage are a packing pressure of 0.95Pinject, a
packing time of 10 s, a mold temperature of 45 C and a cooling
time of 30 s; a packing pressure of 0.95Pinject, a packing time of
10 s, a mold temperature of 45 C and a cooling time of 30 s; and
a packing pressure of 0.95Pinject, a packing time of 10 s, a mold
temperature of 50 C and a cooling time of 30 s, respectively.
The optimum combination of parameters for each material to
affect each response (in-cavity residual stresses, volumetric shrink-
age, and warpage) were varied, although the values used for the
processing parameters were the same, as shown in Table 3. This
difference can be attributed to the different filler loadings and also
to the different viscosity behaviors of each lignocellulosic polymer
composite, as shown in Table 2. These differences affect the flow
behavior along thin-walled parts.

TABLE 4. Process parameters and levels.

PP 1 40 wt% wood PP 1 50 wt% wood PP 1 60 wt% wood

Levels Levels Levels

Factor Description (unit) 1 2 3 4 1 2 3 4 1 2 3 4

A Packing pressure (%) 80 85 90 95 80 85 90 95 80 85 90 95

B Packing time (sec) 10 20 30 40 10 20 30 40 10 20 30 40
C Mold temperature ( C) 40 45 50 55 40 45 50 55 40 45 50 55
D Cooling time (sec) 10 20 30 40 10 20 30 40 10 20 30 40

4 POLYMER ENGINEERING AND SCIENCE—2014 DOI 10.1002/pen

 TABLE 5. The combination of parameters in the orthogonal array L1644. to provide precise values for the determination of the S/N ratio
values compared to the other types of lignocellulosic polymer
Parameters composites. Although the melt flow rate was in the appropriate
range (greater than 3.0 g/10 min) for the injection-molding pro-
No. trial A B C D cess, the conditions were not suitable for molded thin-walled
1 80 10 40 10
parts. The high filler content made it difficult for the melt to flow
2 80 20 45 20 easily, and rapid solidification occurred. This finding conflicts
3 80 30 50 30 with some statements made by Hakimian and Sulong [8]; they
4 80 40 55 40 have claimed that the short-shot problem occurs during the mold-
5 85 10 45 30 ing of microparts because of insufficient injection speeds, injec-
6 85 20 40 40 tion pressure, and molding temperatures. Hence, a material with a
7 85 30 55 10
wood-filler content of greater than 50 wt% is not suitable for
8 85 40 50 20
9 90 10 50 40
molding thin-walled parts using lignocellulosic-filler-reinforced
10 90 20 55 30 thermoplastic composites.
11 90 30 40 20 From the optimum processing conditions setting for the minimi-
12 90 40 45 10 zation of volumetric shrinkage, the Fig. 3 illustrates the overpack-
13 95 10 55 20 ing that occurred in the gate regions for all three lignocellulosic
14 95 20 50 10 polymer composites. However, the values of overpacking were in
15 95 30 45 40
the ranges of 21.980% to 21.025% and 21.182% to 20.115%
16 95 40 40 30
for PP 1 40 wt% wood and PP 1 60 wt% wood, respectively,
while the value for PP 1 50 wt% wood was in the ranges of
20.709% to 20.174%. These value differences can be explained
Simulation Verification Run in terms of the viscosity and shear rate behaviors rather than the
By substituting the optimum S/N ratio into Eq. 30 for the opti- filler contents. The melt flow index for PP 1 50 wt% wood was
mum combination of processing parameters, the predicted mini- higher than the others, as shown in Table 2. A high value of the
mum responses for the three lignocellulosic polymer composites melt flow index indicates low-viscosity behavior.
were obtained. All values are given in Table 8. However, it was In addition, Fig. 4 shown the viscosity behavior for PP 1 40
necessary to perform verification tests to confirm the efficacy of wt% wood shows Newtonian flow at low shear rates and power
the optimal parameter combinations in producing the desired law behavior until high shear rates whereas the PP 1 50 wt%
responses. The results demonstrated that the simulated results and PP 1 60 wt% wood show nearly ideal power law behavior
agreed well with the predicted results, and most relative deviation within the measured range of shear rates. Although the viscosity
errors were less than 10%. However, relative deviation errors of behavior of PP 1 60 wt% wood is approximately similar to
more than 10% were obtained for PP 1 60 wt% wood. This is PP 1 50 wt% wood, the higher amount of filler increases the
because the short-shot problem occurred in these molded thin- viscosity by approximately a factor of 5 and becomes the limit-
walled parts. This problem caused the data obtained to be unable ing factor in molding. The low viscosity at high shear rates

TABLE 6. Results for residual stresses, volumetric shrinkage and warpage and the corresponding S/N values for the investigated materials.

PP 1 40 wt% wood PP 1 50 wt% wood PP 1 60 wt% wood

Residual Residual
stresses Shrinkage Warpage stresses Shrinkage Warpage Residual Shrinkage Warpage
(MPa) (%) (mm) (MPa) (%) (mm) stresses (MPa) (%) (mm)

Run Average S/N Average S/N Average S/N Average S/N Average S/N Average S/N Average S/N Average S/N Average S/N

1 46.39 233.33 2.047 26.22 0.0140 37.08 29.12 229.28 3.181 210.05 0.0390 28.18 57.26 235.16 2.486 27.91 0.0761 22.37
2 45.00 233.06 1.968 25.88 0.0742 22.59 31.30 229.91 3.560 211.03 0.0335 29.50 62.08 235.86 2.747 28.78 0.0685 23.29
3 52.42 234.39 2.414 27.65 0.0222 33.07 24.48 227.78 2.528 28.06 0.0222 33.07 65.86 236.37 3.001 29.55 0.0939 20.55
4 60.21 235.59 2.991 29.52 0.0235 32.58 27.04 228.64 2.934 29.35 0.0177 35.04 152.0 243.64 9.933 219.94 0.4415 7.10
5 47.64 233.56 2.009 26.06 0.0157 36.08 23.01 227.24 2.178 26.76 0.0189 34.47 59.77 235.53 2.501 27.96 0.0425 27.43
6 40.77 232.21 1.667 24.44 0.0539 25.37 21.46 226.63 2.021 26.11 0.0179 34.94 55.60 234.90 2.723 28.70 0.2188 13.20
7 41.02 232.26 1.591 24.03 0.0081 41.83 29.64 229.44 3.025 29.61 0.0425 27.43 65.93 236.38 2.843 29.08 0.0438 27.17
8 47.88 233.60 1.967 25.88 0.0081 41.83 25.96 228.29 2.576 28.22 0.0442 27.09 57.86 235.25 2.582 28.24 0.0633 23.97
9 31.44 229.95 1.056 20.47 0.0242 32.32 26.17 228.36 2.392 27.58 0.0557 25.08 57.86 235.25 2.403 27.62 0.2279 31.09
10 35.07 230.90 1.317 22.39 0.0352 29.07 29.32 229.34 2.884 29.20 0.0381 28.38 61.64 235.80 2.516 28.01 0.0351 29.09
11 37.91 231.58 1.334 22.50 0.0181 34.85 21.02 226.45 1.867 25.42 0.0482 26.34 138.5 242.83 7.694 217.72 0.2355 12.56
12 32.40 230.21 1.042 20.36 0.0142 36.95 23.88 227.56 2.137 26.60 0.0577 24.78 55.66 234.91 2.351 27.43 0.0887 21.04
13 28.76 229.18 0.8243 1.68 0.0585 24.66 31.75 230.03 2.901 29.25 0.0128 37.86 49.23 233.84 1.888 25.52 0.0351 29.09
14 27.15 228.68 0.8038 1.90 0.0545 25.27 24.01 227.61 2.009 26.06 0.0266 31.50 53.02 234.49 2.041 26.20 0.0441 27.11
15 20.49 226.23 0.6299 4.01 0.0317 29.98 23.31 227.35 1.948 25.79 0.0294 30.63 41.19 232.30 1.561 23.87 0.0472 26.52
16 18.31 225.25 0.5105 5.84 0.0276 31.18 20.98 226.44 1.708 24.65 0.0445 27.03 38.43 231.69 1.424 23.07 0.0749 22.51

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—2014 5

 TABLE 7. The responses of S/N ratios for in-cavity residual stress, volumetric shrinkage and warpage.

Packing Packing Mold Cooling Packing Packing Mold Cooling

pressure time temperature time pressure time temperature time

2
In-cavity PP 1 40 Level 1 34.09 231.50 230.59 231.12 Volumetric PP 1 40 Level 1 27.32 22.77 21.83 22.18
residual wt% wood Level 2 232.91 231.21 230.77 231.85 shrinkage wt% wood Level 2 25.1 22.7 22.07 23.15
stress Level 3 230.66 231.11 231.65 231.03 Level 3 21.43 22.54 23.03 22.57
Level 4 227.33 231.17 231.98 231.00 Level 4 3.36 22.48 23.57 22.6
Variance 6.76 0.39 1.39 0.85 Variance 10.68 0.29 1.74 0.97
PP 1 50 Level 1 228.9 228.73 227.2 228.47 PP 1 50 Level 1 29.62 28.41 26.56 28.08
wt% wood Level 2 227.9 228.37 228.02 228.67 wt% wood Level 2 27.68 28.10 27.54 28.48

6 POLYMER ENGINEERING AND SCIENCE—2014

Level 3 227.93 227.75 228.01 227.70 Level 3 27.20 27.22 27.48 27.17
Level 4 227.86 227.73 229.36 227.74 Level 4 26.44 27.20 29.35 27.21
Variance 1.04 1 2.16 0.97 Variance 3.18 1.21 2.79 1.31
PP 1 60 Level 1 237.76 234.94 236.15 235.23 PP 1 60 Level 1 211.54 27.25 29.35 27.65
wt% wood Level 2 235.52 235.26 234.65 236.95 wt% wood Level 2 28.49 27.92 27.01 210.06
Level 3 237.2 236.97 235.34 234.85 Level 3 210.19 210.05 27.9 27.15
Level 4 233.08 236.37 237.42 236.52 Level 4 24.66 29.67 210.64 210.03
Variance 4.68 2.03 2.77 0.38 Variance 6.88 2.8 3.63 2.91
Packing Packing Mold Cooling
pressure time temperature time
Warpage PP 1 40 wt% Level 1 31.33 32.54 32.12 35.28
wood Level 2 36.28 25.58 31.4 30.98
Level 3 33.30 34.93 33.12 32.35
Level 4 27.77 35.64 32.03 30.06
Variance 8.51 10.06 1.72 5.22
PP 1 50 Level 1 31.45 31.40 29.12 27.97
wt% wood Level 2 30.98 31.08 29.84 30.20
Level 3 26.14 29.37 29.19 30.74
Level 4 31.76 28.49 32.18 31.42
Variance 5.62 2.91 3.06 3.45
PP 1 60 Level 1 18.33 27.50 17.66 24.42
wt% wood Level 2 22.94 23.17 24.57 22.23
Level 3 23.45 21.7 25.68 24.90
Level 4 26.31 18.66 23.11 19.48
Variance 7.98 8.84 8.02 5.42

The significance of bold indicate the highest S/N ratio.This is to indicate the optimum combination of processing parameter conditions for the minimisation of the three selected responses (residual stress,
shrinkage and warpages).

DOI 10.1002/pen
 TABLE 8. The optimal combinations of parameters and the verification of the simulation results.

Factors

A B C D Predicted Simulated Deviation

Residual PP 1 40 wt% wood

stresses Values 95 30 40 40 20.63 MPa 19.33 MPa 6.30%
Levels 4 3 1 4
PP 1 50 wt% wood
Values 95 40 40 30 20.06 MPa 20.97 MPa 4.54%
Levels 4 4 1 3
PP 1 60 wt% wood
Values 95 10 45 30 31.12 MPa 46.51 MPa 49.45%
Levels 4 1 2 3
Shrinkage PP 1 40 wt% wood
Values 95 40 40 10 0.58% 0.56% 3.45%
Levels 4 4 1 1
PP 1 50 wt% wood
Values 95 40 40 30 1.62% 1.755% 8.33%
Levels 4 4 1 3
PP 1 60 wt% wood
Values 95 10 45 30 0.99% 1.723% 74.04%
Levels 4 1 2 3
Warpage PP 1 40 wt% wood
Values 85 40 50 10 0.0064 mm 0.0059 mm 7.81%
Levels 2 4 3 1
PP 1 50 wt% wood
Values 95 10 55 40 0.0150 mm 0.0137 mm 8.67%
Levels 4 1 4 4
PP 1 60 wt% wood
Values 95 10 50 30 0.0156 mm 0.0233 mm 49.35%
Levels 4 1 3 3

permits the melt to flow more easily when molding thin-walled
parts. This is caused by orientation of the long polymer chains,
a phenomenon often referred to as shear thinning. This behavior
allows processors to move polymer composite melts long distan-
ces through thin-walled part and minimizes the occurrence of
overpacking in various regions of the molded parts. Therefore,
the differences in volumetric shrinkage on the part surface
become smaller and more uniformly distributed.
Based on the predicted results, the in-cavity residual stresses
for PP 1 50 wt% wood should be 20.06 MPa, which is lower
than the 20.63 MPa and 31.12 MPa stresses expected for
PP 1 40 wt% wood and PP 1 60 wt% wood, respectively. This
may be attributable to the homogeneity of the filler-to-matrix-
polymer ratio, which causes the thermal stresses to be uniformly
distributed along thin-walled parts. Meanwhile, the predicted
volumetric-shrinkage value for PP 1 50 wt% wood is 1.62%,
which is higher than the values of 0.58% and 0.99% predicted
for PP 1 40 wt% wood and PP 1 60 wt% wood, respectively.
However, visual inspection of Fig. 3 shows that the differences
in the values of the contour-pattern distribution for PP 1 50
wt% wood are small compared with the other types of lignocel-
lulosic polymer composites. Smaller variations in value in the
contour-pattern distribution indicate a better distribution of the
volumetric shrinkage on the part surface. A molded part may
become distorted because of a lack of homogeneity in its shrink-
age. A negative value of volumetric shrinkage in the contour
pattern indicates that overpacking occurred. Greater overpacking
was observed for the PP 1 40 wt% wood polymer composite
than for PP 1 50 wt% wood. Any non-homogeneity in volumet-
ric shrinkage will also affect the warpage results. Therefore,
PP 1 50 wt% wood is considered to be suitable for applications
that involve molding thin-walled parts using lignocellulosic
polymer composites.
Based on the obtained predictions, the expected warpage is
0.01 mm for all types of lignocellulosic polymer composites.
The value calculated for PP 1 40 wt% wood is 0.0064 mm,
which is lower than the values of 0.0150 mm and 0.0156 mm
calculated for PP 1 50 wt% wood and PP 1 60 wt% wood,
respectively. However, among the three investigated materials,
PP 1 50 wt% wood is the preferred type of lignocellulosic poly-
mer composite for the manufacture of molded thin-walled parts.
This material, which was selected after evaluation in terms of
economy in manufacturing, offers good results with respect to
the quality of the final part without significant risk of the short-
shot problem and represents the optimal or maximum filler con-
tent that should be used in a polymer composite.
Analysis of Variance (ANOVA)
Tables 9–11 show the results of the ANOVA for in-cavity
residual stresses, volumetric shrinkage, and warpage determined
for thin-walled parts fabricated using the three investigated types
of lignocellulosic polymer composites. The following discussion
focuses on the ANOVA results obtained for the thin-walled
parts molded using PP 1 40 wt% wood and PP 1 50 wt% wood.
The ANOVA results for PP 1 60 wt% wood are neglected. This
material was found to be unsuitable for molded thin-walled parts
because of the problem of early solidification (short-shot

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—2014 7

 FIG. 3. Volumetric-shrinkage distributions of (a) PP 1 40 wt% wood, (b) PP 1 50 wt% wood, and (c) PP 1 60 wt%
wood. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
problem), and the analysis yielded statistical residual errors that
were higher than the molding parameters. Wang et al. [3] have
also reported similar results concerning the percentage contribu-
tions to the residual errors. This can most likely be attributed to
some interaction between parameters that was not included in
the statistical design of the experimental set-up.
On the basis of the ANOVA results summarised in Table 9,
the packing pressure and mold temperature are the most signifi-
cant factors that affect the in-cavity residual stresses for molded
parts fabricated using PP 1 40 wt% wood and PP 1 50 wt%
wood, respectively. The packing pressure serves to ensure the
replenishment of polymer melt into the molded part, which is
necessary because of the loss caused by the volume shrinkage
of the cooled part during the solidification phase. Zhou and Li
[21] have noted that during the packing stage, the frozen-in
stress caused by the packing pressure should be taken into
account when measuring residual stresses. Wang and Young
[22] have reported that the effect of packing pressure on resid-
ual stresses is usually more significant at low packing pressures
than at high packing pressures, resulting in a nearly identical
8 POLYMER ENGINEERING AND SCIENCE—2014
distribution of residual stresses. In contrast to the results of Aza-
man et al. [23], a lower packing pressure can lead to early solid-
ification phenomena during the packing stage, causing the inner
stress on the part to increase. The optimal results also indicate
that the optimum packing pressures for all three types of ligno-
cellulosic polymer composites are almost all at the same high
level, as shown in Table 8. In contrast, Altan and Yurci [11]
have indicated that the most important parameter for the distri-
bution of residual stresses on the surface regions of high-density
polyethylene (HDPE) parts is the mold temperature. The mold
temperature was determined to be the most effective parameter
for the reduction of residual stresses because when the mold is
hotter, the cooling rate is slower [24]. As a consequence, a
higher mold temperature is able to increase the crystallization
time, which consequently induces slower crystallization with
lower stresses that become frozen with sufficient relaxation.
The results summarized in Table 10 show that the packing
pressure is the most significant parameter that contributes to the
volumetric shrinkage for PP 1 40 wt% wood, while the packing
pressure and mold temperature are the most effective parameters
DOI 10.1002/pen
 FIG. 4. Viscosity behavior of (a) PP 1 40 wt% wood, (b) PP 1 50 wt% wood, and (c) PP 1 60 wt% wood [31].

for the control of the volumetric shrinkage of molded parts fabri- ence the distribution of shrinkage along molded parts. It has been
cated using PP 1 50 wt% wood. Similarly, Lotti et al. [25] have shown that the volumetric shrinkage is more strongly dependent
found that the holding pressure and mold temperature most influ- on the packing pressure and mold temperature for molded thin-

TABLE 9. ANOVA for in-cavity residual stresses.

Factors Degrees of freedom, DOF Sum of squares, SOS Mean squares, MS F-ratio % contribution

PP 1 40 wt% wood
Packing pressure 3 106.08 35.36 10.29 85.40
Packing time 3 0.37 0.12 0.04 0.29
Mold temperature 3 5.47 1.82 0.53 4.40
Cooling time 3 1.99 0.66 0.19 1.60
Error 3 10.31 3.44 8.30
Total 15 124.22 100
PP 1 50 wt% wood
Packing pressure 3 3.06 1.02 0.97 14.10
Packing time 3 2.87 0.96 0.91 13.22
Mold temperature 3 9.65 3.22 3.07 44.48
Cooling time 3 2.97 0.99 0.95 13.71
Error 3 3.14 1.05 14.49
Total 15 21.69 100
PP 1 60 wt% wood
Packing pressure 3 52.89 17.63 0.94 35.55
Packing time 3 10.75 3.58 0.19 7.22
Mold temperature 3 16.94 5.65 0.30 11.39
Cooling time 3 12.10 4.03 0.22 8.14
Error 3 56.07 18.69 37.70
Total 15 148.75 100

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—2014 9

 TABLE 10. ANOVA for volumetric shrinkage.

Factors Degrees of freedom, DOF Sum of squares, SOS Mean squares, MS F-ratio % contribution

PP 1 40 wt% wood
Packing pressure 3 261.52 87.17 15.43 90.64
Packing time 3 0.22 0.07 0.01 0.08
Mold temperature 3 7.94 2.65 0.47 2.75
Cooling time 3 1.89 0.63 0.11 0.66
Error 3 16.95 5.65 5.87
Total 15 288.52 100
PP 1 50 wt% wood
Packing pressure 3 22.13 7.38 4.18 41.36
Packing time 3 4.54 1.51 0.86 8.49
Mold temperature 3 16.42 5.47 3.10 30.71
Cooling time 3 5.11 1.70 0.97 9.55
Error 3 5.29 1.76 9.89
Total 15 53.49 100
PP 1 60 wt% wood
Packing pressure 3 106.60 35.53 1.10 37.44
Packing time 3 21.88 7.29 0.23 7.68
Mold temperature 3 30.72 10.24 0.32 10.79
Cooling time 3 28.56 9.52 0.29 10.03
Error 3 96.96 32.32 34.05
Total 15 284.72 100

walled parts that are fabricated using lignocellulosic-filler-
reinforced thermoplastic composites. The packing pressure assists
in filling additional polymer melt into the mold for compensation
during the solidification phase. Meanwhile, the mold temperature
helps to ensure consistent solidification rates during the solidifica-
tion process between the cavity and the core side of the mold for
thin-walled parts. Bushko and Stokes [26] have reported that the
packing pressure has a significant effect on shrinkage. As
expected, a higher packing pressure results in lower shrinkage in
both the in-plane and through-thickness directions. A low packing
pressure results in high volumetric shrinkage. Thus, the magni-
tude of the packing pressure has an effect on the shrinkage distri-
bution of the molded part. According to Azaman et al. [27], an
optimum mold temperature in the range of 40–45  C provides an
adequately low orientation between the fillers and the matrix
polymer, resulting in the minimal volumetric shrinkage for a
molded thin-walled part. However, Jansen et al. [28] have found
the mold temperature does not have a great effect on shrinkage
for molded parts fabricated using semi-crystalline materials that
are unfilled and or filled with glass fibers.

TABLE 11. ANOVA for warpage.

Factors Degrees of freedom, DOF Sum of squares, SOS Mean squares, MS F-ratio % contribution

PP 1 40 wt% wood
Packing pressure 3 152.76 50.92 3.79 29.69
Packing time 3 253.07 84.36 6.29 49.19
Mold temperature 3 6.09 2.03 0.15 1.18
Cooling time 3 62.33 20.78 1.55 12.11
Error 3 40.26 13.42 7.83
Total 15 514.51 100
PP 1 50 wt% wood
Factors Degrees of freedom, DOF Sum of squares, SOS Mean squares, MS F-ratio % contribution
Packing pressure 3 83.93 27.98 1.05 35.20
Packing time 3 23.14 7.71 0.29 9.71
Mold temperature 3 24.67 8.22 0.31 10.34
Cooling time 3 26.80 8.93 0.34 11.24
Error 3 79.91 26.64 33.51
Total 15 238.45 100
PP 1 60 wt% wood
Factors Degrees of freedom, DOF Sum of squares, SOS Mean squares, MS F-ratio % contribution
Packing pressure 3 131.02 43.67 0.84 19.43
Packing time 3 162.28 54.09 1.04 24.06
Mold temperature 3 151.73 50.58 0.97 22.50
Cooling time 3 73.55 24.52 0.47 10.91
Error 15 155.82 51.94 23.11
Total 674.4 100

10 POLYMER ENGINEERING AND SCIENCE—2014 DOI 10.1002/pen

 For PP 1 40 wt% wood, the most significant factor that
affects warpage is the packing time (49.19%), and the packing
pressure (29.69%) is the second-most effective parameter, as
shown in Table 11. Meanwhile, for PP 1 50 wt% wood, the
most significant factor that affects the warpage of a molded
thin-walled part is the packing pressure (35.20%), followed by
the cooling time (11.24%). Several studies have reported a rela-
tion between packing time and packing pressure. If the time of
compaction is too long, excess polymer melt will be packed into
the cavity during the packing stage. Inconsistency in the com-
pressive stress between the frozen layers of the part surface and
the mold wall causes increased warpage [18, 29, 30]. Azaman
et al. [27] have found that further increasing the cooling time
does not significantly affect the distribution of warpage on a
molded thin-walled part. However, their findings agree that the
cooling time plays an important role in providing more time for
the relaxation of the orientation between the filler and the
matrix polymer during the cooling stage. In contrast, Zhou and
Li [30] have reported that a longer cooling time can lead to a
significant decrease in warpage because of the decreased effects
of thermal strain and deformation after demolding.
CONCLUSIONS
In conclusion, Taguchi optimization is sufficient to optimise
with respect to in-cavity residual stresses, volumetric shrinkage
and warpage for molded thin-walled parts fabricated using ligno-
cellulosic polymer composites. Of the lignocellulosic polymer
composites investigated in this study, the PP 1 50 wt% wood
material is preferred over PP 1 40 wt% wood and PP 1 60 wt%
wood for the manufacture of molded thin-walled parts. This
determination was based on the following considerations: (1) The
predicted in-cavity residual stresses for PP 1 50 wt% wood are
20.10 MPa, which is lower than the values of 20.60 MPa and
31.10 MPa predicted for PP 1 40 wt% wood and PP 1 60 wt%
wood, respectively. (2) The differences in value of the contour-
pattern distribution for PP 1 50 wt% wood are small compared to
those for the other types of lignocellulosic polymer composites.
(3) The predicted warpage is 0.01 mm for all types of lignocel-
lulosic polymer composites. This material, which was selected
after evaluation in terms of economy in manufacturing, yields
good results in terms of part quality without a significant risk of
encountering the short-shot problem and represents the optimal or
maximum filler content that should be used in preparing polymer
composites. The ANOVA analysis found that the packing pres-
sure and mold temperature are the most important parameters for
the reduction of residual stresses and volumetric shrinkage, while
the packing pressure, packing time, and cooling time are signifi-
cant to the reduction of warpage in molded thin-walled parts fab-
ricated using various types of lignocellulosic polymer composites.
ACKNOWLEDGMENTS
The authors would like to thank Universiti Putra Malaysia
for financial support through the Research University Grant
Scheme (Project No: RUGS/05-02-12-1917RU).
DOI 10.1002/pen
REFERENCES
1. A. Sanadi, D.F. Caulfield, and R.M. Rowell, Lignocellulosic/
Plastic Composites, The Fibril Angle, Spring Newsletter, 8
(1998).
2. M. St. Jacques, Polym. Eng. Sci., 22, 241 (1982).
3. X. Wang, G. Zhao, and G. Wang, Mater. Des., 47, 779 (2013).
4. M.D. Azaman, S.M. Sapuan, S. Sulaiman, E.S. Zainudin, and
K. Abdan, Mater. Des., 50, 451 (2013).
5. J. Giboz, T. Copponnex, and P. Mele, J. Micromech. Microeng.
17, 96 (2007).
6. S.J. Liao and W.H. Hsieh, Polym Eng Sci., 44, 2029 (2004).
7. K.T. Chiang, and F.P. Chang, Int. J. Adv. Manuf. Technol., 35,
468 (2007).
8. E. Hakimian and A.B. Sulong, Mater. Des., 42, 62 (2012).
9. L. Chin-Win, Comput. Indus. Eng., 13, 386 (1987).
10. C.P. Chen, M.T. Chuang, Y.H. Hsiao, Y.K. Yang, and C.H.
Tsai, Expert Syst. Appl., 36, 10752 (2009).
11. M. Altan and M.E. Yurci, Polym.-Plast. Technol. Eng., 49, 32
(2010).
12. M. Altan, Mater. Des., 31, 599 (2010).
13. T.C. Chang and E. Faison, Polym. Eng. Sci., 41, 703 (2001).
14. T. Erzurumlu and B. Ozcelik, Mater. Des., 27, 853 (2006).
15. B. Ozcelik and I. Sonat, Mater. Des., 30, 367 (2009).
16. M.C. Huang and C.C. Tai, J. Mater. Proc. Technol., 110, 1
(2001).
17. T.H. Wang, W.B. Young, and J.T. Wang, Int. Polym. Proc.
XVII, 2, 146 (2002).
18. S.H. Tang, Y.J. Tan, S.M. Sapuan, S. Sulaiman, N. Ismail, and
R Samin, J. Mater. Proc. Technol., 182, 418 (2007).
19. H. Oktem, T. Erzurumlu, and I. Uzman, Mater. Des., 28, 1271
(2007).
20. S.J. Liao, D.Y. Chang, H.J. Chen, L.S. Tsou, J.R. Ho, H.T.
Yau, W.H. Hsieh, J.T. Wang, and Y.C., Su, Polym. Eng. Sci.,
44, 917 (2004).
21. H. Zhou and D. Li, Int. J. Adv. Manuf. Technol., 25, 700
(2005).
22. T.H. Wang and W.B. Young, Eur. Polym. J., 41, 2511 (2005).
23. M.D. Azaman, S.M. Sapuan, S. Sulaiman, E.S. Zainudin, and
A. Khalina, Mater. Des., 55, 381 (2014).
24. W. Michaeli and G. Po’tsh, Injection Molding: An Introduction,
Hanser Publishers, New York (1995).
25. C. Lotti, M.M. Ueki, and R.E.S. Bretas, J. Injection Molding
Technol., 6, 157 (2002).
26. W.C. Bushko and V.K. Stokes, Polym. Eng. Sci., 36, 322 (1996).
27. M.D. Azaman, S.M. Sapuan, S. Sulaiman, E.S. Zainudin, and
A. Khalina, Mater. Des., 52, 1018 (2013).
28. K.M.B. Jansen, D.J. Van Dijk, and M.H. Husselman, Polym.
Eng. Sci., 38(5), 838 (1998).
29. B. Ozcelik and T. Erzurumlu, J. Mater. Proc Tech., 171, 437 (2006).
30. H. Zhou and D. Li, Polym.-Plast. Technol. Eng., 44(4), 603 (2005).
31. Moldflow Thermoplastic Database, Rheological Properties,
Autodesk Moldflow, Insight, 2011.
POLYMER ENGINEERING AND SCIENCE—2014 11
APPENDIX A: The following is an example of the calculation using the vol-
umetric-shrinkage result for PP 1 50 wt% wood, taken from
Step 1: Selection of factors Table 6:
TABLE A1.
S 1X 1

PP 1 50 wt% wood 5 210log 3:1812 5 210:05 dB (A2)

N 1 i51
Levels

Then, the average S/N ratio can be determined:

Factor Description (unit) 1 2 3 4

A Packing pressure (%) 80 85 90 95 S ð210:05211:0328:06 . . . . . . . . . 25:7924:65Þ

5 527:73 dB
B Packing time (sec) 10 20 30 40 N 16
C Mold temperature ( C) 40 45 50 55 (A3)
D Cooling time (sec) 10 20 30 40
Step 6: Determine the effects of the average S/N ratios of
factors at different levels
Step 2: Orthogonal array For packing pressure:
4
The L164 orthogonal array of the Taguchi method was ð210:05211:0328:0629:35Þ
selected to correspond to the experimental setup. Level 15 529:62 dB (A4)
4
ð26:7626:1129:6128:22Þ
Level 25 527:68 dB (A5)
Parameters 4
ð27:5829:2025:4226:60Þ
No. trial A B C D Level 35 527:20 dB (A6)
4
1 1 1 1 1 ð29:2526:0625:7924:65Þ
2 1 2 2 2 Level 45 526:64 dB (A7)
3 1 3 3 3
4
4 1 4 4 4 For packing time:
5 2 1 2 3
6 2 2 1 4
7 2 3 4 1
ð210:0526:7627:5829:25Þ
Level 15 528:41 dB (A8)
8 2 4 3 2 4
9 3 1 3 4 ð211:0326:1129:2026:06Þ
10 3 2 4 3 Level 25 528:10 dB (A9)
4
11 3 3 1 2
12 3 4 2 1 ð28:0629:6125:4225:79Þ
Level 35 527:22 dB (A10)
13 4 1 4 2 4
14 4 2 3 1
ð29:3528:2226:6024:65Þ
15 4 3 2 4 Level 45 527:20 dB (A11)
16 4 4 1 3 4
For mold temperature:
Step 3: The combination of parameters on the orthogonal
L1644 array ð210:0526:1125:4224:65Þ
Level 15 526:56 dB (A12)
4
The parameters for the effective factors are arranged on the
orthogonal array shown at Table 5 at page 5. ð211:0326:7626:6025:79Þ
Level 25 527:54 dB (A13)
Step 4: Experimental objective 4
ð28:0628:2227:5826:06Þ
The research objective is to determine the optimum Level 35 527:48 dB (A14)
parameters that minimize three responses: in-cavity residual 4
stresses, volumetric shrinkage and warpage. ð29:3529:6129:2029:25Þ
Level 45 529:35 dB (A15)
4
Step 5: Quality characteristics
In determining S/N ratio, the smaller is the better quality For cooling time:
characteristic was selected
ð210:0529:6126:6026:06Þ
S 1X n Level 15 528:08 dB (A16)
5 210log yi2 (A1) 4
N n i51 ð211:0328:2225:4229:25Þ
Level 25 528:48 dB (A17)
Remarks: yi is the value of volumetric shrinkage for the ith 4
test; n is the number of tests. ð28:0626:7629:2024:65Þ
Level 35 527:17 dB (A18)
4

12 POLYMER ENGINEERING AND SCIENCE—2014 DOI 10.1002/pen

 ð29:3526:1127:5825:79Þ 
O524:16 (A28)
Level 45 527:21 dB (A19)
4
Then,
Here, the average S/N ratios for the volumetric shrinkage S=N5210logðMSDÞ (A29)
of PP 1 50 wt% wood polymer composites are summarised.
Remarks: MSD is the mean-square deviation.
2S 
=N
Factor Level 1 Level 2 Level 3 Level 4 MSD510
10
(A30)
52:61 (A31)
A 29.62 27.68 27.20 26.44*
B 28.41 28.10 27.22 27.20* 1X n
C 26.56* 27.54 27.48 29.35 MSD ðQC5SÞ5 yi2 (A32)
n i51
D 28.08 28.48 27.17* 27.21
2:615y2 (A33)
Remarks: * indicates the factor level at the optimal condition pffiffiffiffiffiffiffiffiffi
y5 2:61 (A34)
or the highest value.
Step 7: Identify the best combination of parameters to mini- y51:62% is the predicted value of volumetric shrinkage for
mize volumetric shrinkage PP 1 50 wt% wood.
Step 9: Analysis of variance (ANOVA)

Factor Level Parameter Value i) Degrees of freedom (DOF)

Packing pressure 4 0.95Pinject
Packing time 4 40 sec
fTDOF 5N21 (A35)
Mold temperature 1 40  C
Cooling time 3 30 sec
Remarks: fTDOF is the total number of degrees of freedom, N
is the number of experiments, fx is the number of degrees of
Step 8: Estimate the optimum value determined using the freedom attributable to the considered factors, kx is the num-
best combination of parameters ber of levels of each factor, and fe is the error on the degrees
of freedom.
i) Determine the overall average S/N ratio
X fTDOF 51621515 (A36)
T5 ½FA;L1;L2;L3;L4 1FB;L1;L2;L3;L4 1FC;L1;L2;L3;L4 fA 5kA 21542153 (A37)
(A20)
1FD;L1;L2;L3;L4 4m fB 5kB 21542153 (A38)
fC 5kC 21542153 (A39)
Remarks: m is the total number of average factor effects.
fD 5kD 21542153 (A40)
.
T5½ð29:6227:6827:2026:44Þ1 ............1 ð28:0828:4827:1727:21Þ
(A21) fe 5fTDOF 2fA 2fB 2fC 2fD (A41)
16
51523232323 (A42)
5 27:73 dB (A22)
53 (A43)

ii) Determine the total contribution ii) Sum-of-squares computation

The sum of squares (SS) attributable to the overall experi-
         mental mean:
C5ðF A 2T Þ1ðFB 2T Þ1ðFC 2T Þ1ðFD 2T Þ (A23)
S
SS attributable to the mean5N3 (A44)
Remarks: * indicates the factor level at the optimum N
condition 5163ð27:73Þ2 (A45)

 5956:93 (A46)
C5ð26:4417:73Þ1ð27:2017:73Þ1ð26:5617:73Þ
(A24)
1ð27:1717:73Þ The total sum of squares:
53:57 dB (A25) 16  
X Si S
SSTotal 5 2 pr 2 (A47)
i51
Ni N
iii) Calculate the predicted optimum value
X
16

 T1C SSTotal 5 ð210:0517:73Þ2 1ð211:0317:73Þ2 1 . . .

O5 (A26) i51

O527:7313:57 (A27) 1ð25:7917:73Þ2 1ð24:6517:73Þ2

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—2014 13

 553:49 (A48) iv) The F-test
The F-ratio is given by the equation below. When F is
Sum of squares for factor A: much greater than 1, the effect of the control factor is large
 2  2
compared to the variance that is attributable to experimental
S A1 S S A2 S error and interaction effects.
SSA 5ðNLevel A1 Þ 2 1ðNLevel A2 Þ 2
NA1 N NA2 N
 2  2 (A49) Mean square due to a control factor MSX
S A3 S S A4 S F5 5
Mean square due to experimental error MSError
(A62)
1ðNLevel A3 Þ 2 1ðNLevel A4 Þ 2
NA3 N NA4 N
These are some suggested generalised values for F-ratios:
Remarks: NLevel xi is the number of experiments at level i 5
1, 2, 3, or 4.  F<1: The experimental error outweighs the control effect; the
 2 control factor is insignificant and indistinguishable from the
ð210:05211:0328:0629:35Þ experimental error.
SSA 54 17:73 1 . . . . . . . . .
4  F2: The control factor has only a moderate effect compared
 2 to the experimental error.
ð29:2526:0625:7924:65Þ
14 17:73  F>4: The control factor is strong compared to the experimen-
4 tal error and is clearly significant.
522:13 dB2 (A50) Therefore, for factor A:
MSA 7:38
This procedure is repeated for factors B, C, and D, and the FA 5 5 54:18 (A63)
MSe 1:76
sums of squares are as follows:
This procedure is repeated for factors B, C and D, and the F
SSB 54:54 dB2 (A51) ratios are as follows:
SSC 516:42 dB2 (A52)
FB 50:86 (A64)
2
SSD 55:11 dB (A53) FC 53:10 (A65)
Sum of squares for experimental error: FD 50:97 (A66)

SSe 5SST 2SSA 2SSB 2SSC 2SSD (A54)

v) Calculate the percentage contribution (%P) to determine the most
553:49222:1324:54216:4225:11 (A55) significant or least significant factors
55:29 dB2 (A56)
SSA 22:13
% PA 5 3100%5 3100%541:36% (A67)
SST 53:49
iii) Mean-square variance for a control factor
For factor A: This procedure is repeated for factors B, C, and D and the
experimental error, and the percentage contributions are as
SSA 22:13 follows:
MSA 5 5 57:38 (A57)
fA 3
% PB 58:49% (A68)
For factor B:
% PC 530:71% (A69)
SSB 4:54 % PD 59:55% (A70)
MSB 5 5 51:51 (A58)
fB 3 % Pe 59:89% (A71)
For factor C:
vi) Analysis of variance (ANOVA) table for volumetric shrinkage
SSC 16:42 of PP 1 50 wt% wood
MSC 5 5 55:47 (A59)
fC 3

For factor D: Factors DOF(f) SS MS F-Ratio %P

SSD 5:11 A 3.00 22.13 7.38 4.18 41.36

MSD 5 5 51:70 (A60) B 3.00 4.54 1.51 0.86 8.49
fD 3 C 3.00 16.42 5.47 3.10 30.71
D 3.00 5.11 1.70 0.97 9.55
For mean square (error variance) Error 3.00 5.29 1.76 9.89
SSe 5:29
MSError 5 5 51:76 (A61)
fe 3

14 POLYMER ENGINEERING AND SCIENCE—2014 DOI 10.1002/pen