Open Journal of Civil Engineering, 2013, 3, 1-7 Published Online September 2013 (
Copyright © 2013 SciRes. OJCE
Significance Analysi s of Fl e x ur al Be h aviour
of Hybrid Sandwich Panels
Jauhar Fajrin, Yan Zhuge, Frank Bullen, Hao Wang
Centre of Excellence in Engineering Composite (CEEFC), University of Southern Queensland Toowoomba, Toowoomba, Australia
Received June 2013
This paper presents the significance analysis of a new type of hybrid composite sandwich wall panel which can be
manufactured as modular panelised system. Two different types of natural fibers reinforced plastics (NFRP) laminate
were incorporated into the new sandwich panel as an intermediate layer. The significance analysis in this research has
been carried out using analysis of variance (ANOVA). As the aim of the analysis is to select the most appropriate natu -
ral fiber composites for the intermediate layer, the experiments were arranged as a single factor experiment in which 3
levels of a factor have been examined. The factor refers to the type of intermediate layer used in the sandwich panel.
The result of this study shows that the incorporation of intermediate layer has significantly enhanced the load carrying
capacity of the sandwich panels.
Keywords: Hybrid Structure; Sandwich Panels; Significance Analysis; Natural F iber Composites; Buil ding
1. Introduction
Providing accommodation on a large scale at relatively
affordable price has always been a challenging task not
only for government but also for housing industry. There
are two key factors in order to solve these two challenges.
The first factor is to reduce the weight of the stru cture to
maintain affordable prices. In building construction, the
self-weight of a structure represents a large proportion of
the total load on a structure. The reduction of self-weight
of structure by adopting appropriate material results in
the reduction of element cross section, size of foundation
and supporting elements thereby reduce the overall cost
of the housing construction. The second factor is to util-
ize the panelized housing system to encourage the mass
production of houses. A panelized housing system is a
form of construction in which all housing components
are pre-fabricated at factory and shipped to the site for
erection. With this construction system, a house can be
built faster than stick-built homes. In most cases, pane-
lized homes can be assembled in a matter of days which
means that lesser labor is needed and more homes can be
built. Other advantages of panelized homes are such as
the system can eliminate costing delays, less weather
damage during construction and also precision engi-
neered to the highest quality.
In order to meet these challenges, a research project
has been carried out aimed at developing a new type of
hybrid composite sandwich wall panel which can be
manufactured as modular panelised system. The typical
sandwich panel used in building application commonly
consists of metal skins and soft core. Although oriented
strand board (OSB) is commonly employed for the skin
of sandwich structure in structural insulated panels (SIPs),
the observed shortcomings of this typical skin such as
mould build-up and disintegration in the presence of
flood water [1] have reduced their usage. In this study,
metal based skins such as aluminium or steel are adopted.
Metal skins are actually preeminent choice for their
many advantages, but the price is always a concern.
Consequently, reducing the thickness of the skin as much
as possible is the only way to keep a competitive and
reasonable overall cost. However, using thinner skins
may result in the early failure of sandwich structure, such
as face wrinkling or inundation. The sustainable hybrid
concept offered in this research has been considered as a
practical solution where an intermediate layer made from
natural fibers reinforced plastics (NFRP) laminate was
introduced. Natural fibers are a major renewable resource
material throughout the world especially in the tropics.
Statistical significance is a mathematical tool that is
commonly used to determine whether the outcome of an
experiment is the result of a relationship between specific
factors or merely the result of chance [2]. Commonly,
such concept is used in the fields in which research is
conducted through experimentation. It is frequent to
Copyright © 2013 SciRes. OJCE
summarize statistical comparisons by declarations of
statistical significance or non -significance [2]. In a scien-
tific research, a hypothesis is proposed prior to data col-
lection and analyses. The statistical analysis of the data
will produce a number that is statistically significant if it
falls below a certain percentage called the confidence
level or level of significance. Further, Reference [2] ex-
plains that statistical significance is used to reject or ac-
cept what is called the null hypothesis, which usually
states that there is no relationsh ip between two variables.
In a simple expression, statistical sign ificance means that
there is a good re lat ionshi p exis ti ng bet ween two variables.
This paper focuses on the significance analysis of the
experimental results on the flexural behaviour of the sand-
wich pane l s develope d by the authors.
2. Significance Analysis
Although significance or statistical analysis is rarely
found as a primary approach in composite sandwich pan-
el research, it has actually been extensively used in the
field of composite material research. A number of re-
searches reported in the literature provide the idea of how
it was applied to composite material research. A study on
the significance effect of microwave curing on tensile
strength of carbon fibre composites was reported by [3].
The statistical analysis employed was two-way analysis
of variance (ANOVA) using statistical software SPSS
14.0. The results showed that the curing time and mi-
crowave process had significant effect on the tensile
strength of the carbon fibre composites. Reference [4]
reported their work on the optimization of processing
variables in wood-rubber composite panels manufactur-
ing process. The results of experiments were statistically
analysed using response surface method (RSM). The
Design-Expert statistical software was employed to de-
termine the significant factors that affected the properties
of the composite panels. It was concluded that the den-
sity and the interaction of different variables were the
significant factors affecting the final properties of the
Reference [5] used design of experiments (DoE) to
study the significance of low energy impact on modal
parameters for composite beams. The experiment was
designed as a 5 × 2 full factorial design. The results
showed that damping ratio is more sensitive parameter
for the damage detection than the natural frequency.
Reference [6] reported their work on the mechanical and
absorption properties of woven jute/banana hybrid com-
posites. Statistical analysis using one-way ANOVA was
employed to analyze the results of tensile, flexural and
impact tests of various composite configurations. The
results suggested that th e layering pattern had significant
effect on the mechanical properties of the composites.
A response surface methodology (RSM), which is a
statistical design of experiment method, was employed
by [7] to analyze the factors influencing deflection in
sandwich panels subjected to low-velocity impact. The
results revealed that the deflection increased with the
increasing of height of fall the mass of impactor. The
deflection was only slightly increased with the incr easing
in the core thickness.
It is worth to note that when a statistical test res ulted in
a significant outcome it means that a finding has a
chance of being true due to the relation between variables,
not just really a chance occurrence. Statistical signific-
ance does not always mean that the finding is important
or that it has any decision-making utility, so that the re-
searcher must always examine both the statistical and the
practical significance of any research finding.
3. Experimental Program
Before the significance analysis is performed, the flex-
ural testing program will be briefly reviewed in the fol-
lowing sections.
3.1. Testing Specimens
The sandwich samples were cut and shaped into a span
length of 450 mm and the size of 550 × 50 × 22 mm for
length, width and thickness, respectively. An aluminium
5005 H34 sheet with the thickness of 0.5 mm was used
as the skins for all samples. An expanded polystyrene
(EPS) is used for the core of this hybrid sandwich panel.
The thickness of EPS core for control level was 21 mm
and 15 mm for the other two levels to maintain a constant
overall thickness of 22 mm. Several types of natural fiber
composites have been investigated and the two best per-
formed natural fibers were employed as the intermediate
layer; jute and hemp fiber composites with a thickness of
3 mm. Each level was replicated 5 times; hence the total
of samples tested was 15 samples. The arrangements of
the flexural test specimens are shown in Table 1.
Sandwich panel specimens were manually prepared
using a pressing system. All constituent parts were cut
into the same length and width and glued together using
structural grade adhesive. The NFCs intermediate layers
were sanded-up using sanding machine to obtain uniform
thickness while alu minium sheet were roughed manually
using sandpaper. The EPS core was sliced using hot
knife foam cutter to obtain the required thickness. When
all constituents ready, they were glued and placed in the
pressure system. The process of sample preparation is
shown in Figure 1.
3.2. Testing Results
The static flexural test was conducted in accordance with
Copyright © 2013 SciRes. OJCE
the ASTM C 393-00 standard which is a standard test
method for flexural properties of sandwich constructions.
The testing set up is shown in Figure 2.
The failure loads and deflections of specimen tested
under flexural testing scheme in this experiment are
listed in Table 2. For a simplicity reason, the signific-
ance analysis in this paper is only made for the data of
load. As it can be seen from Table 2 that the coefficient
variation (CV) of the actual data ranges from 9.39 to
16.05. The CV values in this range are considered as
fairly acceptable. The ratio of mean to standard deviation
or CV should be of the order of 3 or more, but the value
of 33% has often been stated as the permissible upper
limit of CV (Patel et al., 2001).
The fluctuation in the distribution of exp erimental data
can be easily observed in Figure 3. It is clearly shown in
the figure that each level of samples has at least one out-
lier data. Removing these values, by conducting norma-
lization process, will produce more consistent or less
fluctuated data. For example, the failure loads of speci-
men 2 in CTR and JFC samples, and specimen 4 for HFC
samples are considered as an outlier, with the value of
415 N, 473 N and 734 N, respectively.
The experimental data resulting from the normaliza-
tion process are presented in Table 3.
As it can be noticed in Table 3, the coefficient varia-
tion of CTR, JFC and HFC for loads has now reduced to
3.83%, 5.17% and 8.7%, respectively. The previous CV
value for each level prior to the normalization process
was 15.23%, 9.39% and 11.68% for CTR, JFC and H FC,
respectively. The following significance analysis will use
data provided from the normalization process as listed in
Table 3.
Table 1. Experimental arrangements for flexural testing.
Skin Intermediate Layer Core
0.5mm 3 mm 15 mm
CTR Aluminium None EPS (21 mm)
JFC Aluminium Jute EPS
HFC Aluminium Hemp EPS
Figure 1. Sandwich panel f ab rication pr oc ess.
Figure 2. Testing set up.
Table 2. Failure loads deflections of specimens un der flexural test.
Treatments based upon intermediate layer used
P (N) δ (mm) P (N) δ (mm) P (N) δ (mm)
1 321 11.92 414 56 628 42.24
2 415 15.4 473 62.18 579 36.37
3 307 12.18 379 45.22 524 47.04
4 293 9.89 378 56.53 481 37.86
5 302 13.2 414 64 635 35.09
Average 327.60 12.52 411.60 56.79 569.40 39.72
Stdev 49.90 2.01 38.64 7.34 66.53 4.90
CV 15.23 16.05 9.39 12.93 11.68 12.34
Copyright © 2013 SciRes. OJCE
Figure 3. Dot-plot diagram for medium scale samples under flexural test.
Table 3. The results of normalization process for medium scale samples.
Samples number
Treatments based upon intermediate layer used
P (N) δ (mm) P (N) δ (mm) P (N) δ (mm)
1 321 11.92 414 56 628 42.24
2 307 12.18 379 45.22 579 36.37
3 293 9.89 378 56.53 524 47.04
4 302 13.2 414 64 635 35.09
Average 305.75 11.80 396.25 55.44 591.50 40.19
Stdev 11.70 1.39 20.50 7.73 51.44 5.53
CV 3.83 11.75 5.17 13.94 8.70 13.76
4. Analysis of Results and Discussion
The significance analysis in this research is carried out
using analysis of variance (ANOVA) as described in our
previous paper [8]. As the aim of the analysis is to select
the most appropriate NFCs for the intermediate layer,
either jute fibre composite or hemp fibre composite, the
experiments were arranged as a single factor experiment
in which 3 levels of a factor have been examined. The
factor refers to the type of intermediate layer used in the
sandwich panel and such factor was leveled as 0, 1 and 2
as required by Minitab 15 software. For the specimens
discussed in this paper, level 0 was the conventional
sandwich panel without intermediate layer or control
level (CTR) while level 1 and 2 refer to as jute fiber
composite (JFC) and hemp fiber composite (HFC), re-
spectively. For the analysis purpose, the data for ANOVA
are tabulated as follows in Table 4.
From Table 4, some important parameters for theoret-
ical calculations can be determined such as replications
(n = 5), total number of samples (N = 12), and number of
levels or treatments (a = 3). The results of the theoretical
calculations are presented in Table 5, contains all im-
portant parameter for further use to make a significance
judgment. Such analysis obtained by statistical software
Minitab 15 is presented in Table 6.
As presented in Tables 5 and 6, the theoretical calcu-
lations were in good agreement with the ANOVA results
obtained by statistical software Minitab 15. It can be
noted here that the mean square between treatments
(85311) was few times larger than the mean square
within the treatments or error mean square which was
only 1068. This indicates that the treatments differ. The
other way to make a significance decision is by using the
F value (F0). The value of F0, as seen in Tables 5 and 6,
was 79.91. Instead, the F value obtained from the
F-distribution table for F(0.05;2,9) was 4.26. This value,
which called as F table, was obtained by using the signi-
ficance level of 95% (α = 0.05), 3 levels (a = 3) and 12
Copyright © 2013 SciRes. OJCE
Table 4. Tabulated data for analysis of variance (ANOVA).
Factor levels Observations Totals Averages
1 2 3 4
Level 0 (CTR) 321 307 293 302 1223 305.75
Level 1 (JFC) 414 379 378 414 1584 396.25
Level 2 (HFC) 628 579 524 635 2366 591.50
Table 5. Analysis of variance table for single-factor experimental design.
Source of variations Sum of square Degrees of freedom Mean square F0
Between treatments
treatments i
SS y
= −
Error (within treatments)
E T treatments
Table 6. The theoretical results of ANOVA for sandwich panels.
Source of variations Sum of square Degrees of freedom Mean square F0
Intermediate lay e r 170621.20 2 85310.58 79.91
Error 9608.50 9 1067.61
Total 180229.70 11
samples (N = 12). Since the value of F0 (79.91) was
much higher than the value of F table (4.26), the null
hypothesis (H0) should be rejected, meaning that there
are a significant difference in the treatment means.
For this experiment, the null hypothesis was actually
trying to state that the load carrying capacity of all types
of tested sandwich panels (CTR, JFC and HFC) were
equal. However, the statistical significance analysis
showed that the average values of the three types of
sandwich panels were significantly different. Since the
average values or means of the JFC and HFC was higher
than CTR, it can be concluded that the load carrying ca-
pacity of hybrid sandwich panels is significantly higher
than the conventional sandwich panels. In addition, as
the confidence level used in the Minitab analysis was
95%, it means that the finding has a 95% chance of being
true. Alternatively, the finding has a 5% of not being true
as it was analyzed with the assumption of the probability
error (α) of 0.05. A value of P, or frequently called as
P-value, could also be used for drawing a conclusion.
The rule is that if the P-value is less than α (0.05) which
is an error tolerance level, it can be concluded that there
has factor levels or treatments that have different means.
It is clearly presented in Table 6 that the P-value of me-
dium scale specimen was very small, which is approx-
imately 0.000.
A pairwise comparison between all factor levels might
be conducted to support the decisions drawn from
ANOVA results. There are several possible test methods
for this purpose such as Dunnet’s test, Tukey’s test and
Fishers test. The three pairwise tests were also con-
ducted using Minitab 15 software and the results are dis-
cussed as follows.
The result of Tukey’s test for medium scale sandwich
panels is presented in Table 7. The rule for making a
decision is that whenever the Tukey’s confidence inter-
vals contain zero number, it means that the means are not
different or in other word if none of the Tukey’s confi-
dent intervals equals to zero, it indicates that all of the
means are different. As can be observed from Table 7,
all the confidence intervals have a positive number. The
comparison of level 0 to level 1 and level 2 has the value
of 25.97 and 221.22 fo r the lower values and 155. 03 and
350.28 for the upper values. While for the comparison of
level 0 to level 2, the lower value was 130.72 and 259.78
for the upper value. Since all the confidence intervals
included only positive numbers, it can be concluded that
all the treatment means differ. This sugges ts that the load
carrying capacity of hybrid sandwich panels with JFC
and HFC intermediate layer is significantly different with
the conventional sandwich panels (CTR). The term of
significantly differ en th as the same meaning with sig-
nificantly higherbecause the load carrying capacity of
hybrid sandwich panels was higher than conventional
sandwich panels.
The second pairwise comparison method was the
Copyright © 2013 SciRes. OJCE
Table 7. Analysis of variance results for sandwich panel obtained by statistical software Minitab 15.
One-way ANOVA: Flexural Load versus Intermediate Layer
Source DF SS MS F P
Intermediate Layer 2 170621 85311 79.91 0.000
Error 9 9609 1068
Total 11 180230
S = 32.67 R-Sq = 94.67% R-Sq(adj) = 93.48%
Individual 95% CIs For Mean Based on Pooled StDev
Level N Mean StDev ---+---------+---------+---------+------
0 4 305.75 11.70 (---*--)
1 4 396.25 20.50 (---*--)
2 4 591.50 51.44 (---*---)
300 400 500 600
Pooled StDev = 32.67
Dunnet’s test that only compares the reference levels or
control with other factor levels. This means that the
Dunnet’s test only compares level 0 to level 1 and level 2,
it is not comparing level 1 to level 2. Likewise the Tu-
key’s method, the approach for making a significance
judgment is by checking whether confidence intervals
contain zero number. The result of Dunnet’s test con-
ducted with Minitab 15 shows that all confidence inter-
vals include only a positive numbers. Another way of
drawing a decision is by looking at the critical value of
each level. The results indicated that the critical value of
reference level was 2.61. This value is much lower than
the critical value of level 1 and level 2 which was 90.50
and 285.75, respectively. Overall, it can be concluded
that the load carrying capacity of level 1(JFC) and level 2
(HFC) is much higher than level 0 (CTR).
The third pairwise method is a Fisher’s test which is
pretty similar to the Tukey’s test. Th e result of Fisher test
indicated that none of the confident intervals contain zero
number meaning that all levels differ. It is also noticeable
that the critical values or th e center confident levels were
comparable to the critical values obtained on Tukey’s
test. The difference is only for their lower and upper
5. Conclusion
The significance analysis has been carried out to the
testing data of ultimate load from flexural test of hybrid
and conventional sandwich panels. The experiments were
designed following the principle of statistical design of
experiments and the works reported in this paper was
specifically designed as a single factor experiments. Two
types of hybrid sa ndwich panels were compared with the
conventional sandwich panels without intermediate layer.
The primary conclusion drawn was that the incorporation
of intermediate layer has significantly enhanced the load
carrying capacity of sandwich panels. The results also
indicated that the value of F0 was much higher than the F
value obtained from the F-distribution (F(0.05;2,9)). The F0
was 79.91 while the F-table was only 4.26. Therefore, the
null hypothesis (H0) should be rejected, meaning that
there are significant differences in the treatment means.
All pairwise tests Tukey’s, Dunnet’s and Fisher’s tests
obtained positive confident intervals which suggest the
means of treatments differ. The inference statements
suggested that the load carrying capacity of hybrid sand-
wich panels with JFC and MDF intermediate layer was
significantly higher than the conventional sandwich pa-
[1] A. S. A. Vaidya, N. Uddin and U. Vaidya, “Structural
Characterization of Composite Structural Insulated Panels
for Exterior Wall Applications,” ASCE Journal of Com-
posites for Construction, Vol. 14, 2010, pp. 464-469.
[2] A. Gelman and H. Stern, “The Difference between Sig-
nificant and Not Significant Is Not Itself Statistically Sig-
finicant,” The American Statistician, Vol. 60 , No. 4, 2006,
pp. 328-331.
[3] B. Balzer and J. McNabb, “Significant Effect of Micro-
wave Curing on Tensile Strength of Carbon Fiber Com-
Copyright © 2013 SciRes. OJCE
posites,” Journal of Industrial Technology, Vol. 24, No. 3,
2008, pp. 2-8.
[4] Z. Jun, X. Wang, J. Chang and K. Zheng, “Optimization
of Processing Variables in Wood-Rubber Composite Pa-
nel Manufacturing Technology,” Journal of Bioresource
Technology, Elsevier, Vol. 99, No. 7, 2008, pp. 2384-
[5] A. Shahdin, L. Mezeix, C. Bouver, J. Morlier and Y.
Gourinat, “Fabrication and Mechanical Testing of Glass
Fiber Entangled Sandwich Beams: A Comparison with
Honeycomb and Foam Sandwich Beams,” Composite
Structures, Elsevier, Vol. 90, No. 4, 2009, pp. 404-412.
[6] N. Venkateshwarah, A. Elayaperumal and G. K. Sathiya,
“Prediction of Tensile Properties of Hybrid-Natural Fiber
Composites,” Composites Part B: Engineering, Elsevier,
Vol. 43, No. 2, 2012, pp. 793-796.
[7] M. Periasamy, B. Manickam and K. Hariharasubrama-
niam, “Impact Properties of Aluminium—Glass Fiber
Reinforced Plastics Sandwich Panels,” Materials Re-
search, Vol. 15, No. 3, 2012.
[8] J. Fajrin, Y. Zhuge, F. Bullen and H. Wang, “Flexural
Strength of Sandwich Panel with Lignocellulosic Compo-
sites Intermediate Layer—A Statistic Approach,” Inter-
national Journal of Protective Structures, Vol. 2, No. 4,
2011, pp. 453-464.