Research Article

Optimization of strain measurement procedure based on fuzzy quality

evaluation and Taguchi experimental design
Mehdi Moayyedian1 · Javad Farrokhi Derakhshandeh1 · Sang Heon Lee2

© Springer Nature Switzerland AG 2019

Abstract
The reliability of the measurement system, accuracy, and precision are significant criteria at any measurement proce-
dure. Strain gage sensors are one of the tools to measure the strain of structures in engineering applications. Different
controllable and extraneous parameters will change the accuracy and precision of the strain gauges. As a result, it is
valuable to investigate the most significant parameters and the percentage of their contributions affecting the strain
measurement. This paper presents the multi-objective optimization in strain gauge measurements to determine the Reli-
ability and Noise indices. The best set of parameters with a high-reliability index (which means low noise index) for strain
measurement based on two significant objectives of accuracy and precision are evaluated. The significant parameters,
their optimal levels and percentage of contributions are identified according to the Taguchi and Fuzzy Analytic Hierarchy
Process (FAHP) with Technique for the Order Performance by its Similarity to the Ideal Solution (TOPSIS) methods. To
obtain this evaluation, three levels (L1, L2, and L3) of temperature (P1), the length of wires (P2), and the point of applied
force (P3) are considered. This configuration leads to L9 orthogonal array of Taguchi approach based on the number of
parameters and their levels. The triangular membership function for the weightiness of accuracy and precision is used
via AHP. By using the Taguchi method, the optimum set of parameters are identified as P1 at L1, P2 at L2, and P3 at L3 to
have a higher reliability index compared to trial 8. By applying the optimum design of experiment, it is found that the
reliability index is 95%. The results also show that the highest reliability index is related to trial 8 with 87.65% reliability.
Finally, the combination of FAHP and TOPSIS with Taguchi method for multi-objective optimization is introduced and
proposed as a practical and useful model to determine the best set of parameters with the highest reliability index in
strain measurement.

Keywords Strain measurement · Accuracy · Precision · TOPSIS · Fuzzy · FAHP · Taguchi · Reliability index

1 Introduction is required for modeling calculation to a constitutive equa-

tion for the stress–strain relationship [22, 33]. The basic
Piezo-resistivity or change in the electrical resistivity can function of the strain gauge is based on transforming the
be considered as one of the widely used physical phe- strain in a certain direction to change its electric resist-
nomena at different kind of sensory applications. Strain ance [20, 38]. While the strain gauges have been employed
gauges are useful and valuable devices to measure the for a few decades, still there is little guidance on how to
time-dependent magnitude of deformation and can be evaluate mechanical properties and the aspects of strain
associated with determining the stresses of the structure measurement have been covered in brief [22, 31, 33].
in a wide range of applications [5, 11–13, 19, 21, 23, 34]. To There are several uncertainties in measuring the strain
achieve this, a deep knowledge of mechanical engineering of the object, which can be related to the geometry,

* Mehdi Moayyedian, [email protected] | 1College of Engineering and Technology, American University of the Middle
East, Kuwait City, Kuwait. 2School of Engineering, University of South Australia, Mawson Lakes Campus, Mawson Lakes, SA 5095, Australia.

SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x

Received: 25 June 2019 / Accepted: 9 October 2019 / Published online: 14 October 2019

Vol.:(0123456789)
Research Article SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x

material, and optical of the test case. Besides, other uncer- significant parameter was feed rate on the surface rough-
tainties which can be considered in analyzing the system ness [35]. Later, Sarikaya et al. [36] showed that DOE is able
are related to the methods of measurement such as con- to reduce the amount of oil used for machining purposes
figuration, the number, and distribution of measuring with an improvement in machinability of material. Also, in
points, the number of sampling, and the operator [8, 40]. drilling process, the most significant parameters affecting
Davis [8] conducted an excellent study using different the dimensional accuracy is cutting speed for both coated
techniques of strain measurement during tensile testing. and uncoated tools using DOE [24].
It was found that the size of the sample, environmental Two conventional methods which determine the practi-
conditions, measurement requirement for accuracy and cal and possible factors in increasing the quality in differ-
precision of anticipated strain levels are important factors ent applications are Fuzzy logic and TOPSIS. It was shown
and can significantly affect the deformation of the struc- that the combination of these methods can be used in
ture [8]. Therefore, it was found that a precise optimization injection molding to find the optimum set of param-
method is required to characterize the tensile testing of eters and reduce the number of internal and external
materials properly. defects [17, 29]. Consequently, choosing the wrong set
To have a product with high quality and low-cost, of geometrical and process parameters will decrease the
manufacturing companies typically attempt to produce mechanical properties of the injected part [26, 27]. The
valuable products in a short time to compete with others. Taguchi method and DOE can be employed and used as
For a complex product, the design of the product requires tools for single and not for multi-objective problems. To
optimization tasks comprising multiple objectives. This guarantee that different internal and external defects can
optimization can be considered both in numerical simu- be controlled appropriately, the fuzzy logic theory can be
lations and experimental measurements. In computational proposed [6]. It was shown that in injection molding, in
simulations, for instance, it will take more than 100 h to run order to find the best set of process parameters, the com-
one crash simulation for a non-very complicated model bination of fuzzy and TOPSIS can be employed, which has
[16, 39]. For a two-variable optimization problem, assum- the high level of moldability for producing the product
ing on average 50 iterations are needed by optimization [7, 29, 30, 32].
and assuming each iteration needs one crash simulation, There is an extensive investigations which implemented
the total computation time would be more than 2 months. the TOPSIS concept for optimization process. For instance,
This type of study is mostly associated with trial and error. in micromachining procedure, to have an accurate result,
Trial and error has been identified as the main methodol- TOPSIS was applied [15]. To determine the effect of differ-
ogy in problem solving and has been employed exten- ent input parameters on the machining accuracies, two
sively, in particular, in numerical modeling [1, 4, 9, 10, 16]. types of fluids were employed. Hence, sodium chloride
Therefore, to decrease the number of test case studies, the was selected as a fluid with higher material removal rate
design of experiments and optimization of measurement using TOPSIS [15]. Another application of TOPSIS method is
are valuable skills, which can be employed to enhance pro- related to the surface performance measurements in elec-
ductivity and efficiency. trical discharge machining process. Huo et al. [18] found
One of the optimization method, which is implemented that magnitude of the current has the most influence on
for quality evaluation in industry is the Design of Experi- the performance measurement of the process.
ment (DOE) and using relevant tools such as Taguchi The above review intimates that that although the
method, Signal to noise (S/N) ratio and Analysis of Vari- integration of FAHP with TOPSIS and AHP with FTOPSIS
ance (ANOVA) [2, 25, 28]. One of the most advantage of are useful methods for quality evaluation at different
DOE is to find the significant parameters and their opti- engineering applications, they are unable to cover all dif-
mum level. This assists to find the best quality evaluation ferent possibilities to determine the optimum level and
of the products. Amer et al. [3] showed that DOE can be the contribution rate of significant controllable factors. To
considered as a valuable tool in order to define the right the best of the author’s knowledge, there is no published
parameters and levels in injection moulding with less study employing Taguchi with FAHP and TOPSIS in strain
internal and external defects. The effect of cutting fluid measurement to evaluate the optimum parameters of
in manufacturing process also can be optimized by DOE. DOE. Consequently, the objective of this paper is to evalu-
Sarikaya and Güllü [35] found to have a lower rate of tool ate and determine the most significant parameters affect-
wear pattern and surface roughness, the most significant ing the reliability of strain gauges for different industrial
parameters and their levels are vegetable based cutting applications. This paper investigates and proposed a new
fluid with 180 mL/h fluid flow rate and 30 m/min cutting approach for the multi-objective optimization method in
speed. Besides, the surface roughness was investigated for strain measurement based on two significant objectives
turning machine and the results indicated that the most comprising accuracy and precision. Thus, a new approach

Vol:.(1234567890)
SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x Research Article

based on Taguchi with FAHP and TOPSIS for quality evalu- a single objective (Reliability index), Taguchi method is
ation of strain measurement process is developed and implemented to find the most significant parameters, their
introduced. There are controllable and extraneous param- optimum levels, and their contribution to find the highest
eters to evaluate the final quality of any process. One of reliability index.
the main application of Taguchi method is to enhance
the control of the parameters. In this paper, three control-
lable parameters namely the temperature, the length of 2 Proposed methodology to enhance
wire and the point of application are considered. Besides, the quality of measurement
uncontrollable parameters such as operator which leads
to produce noises is noted. A comprehensive methodology is developed and intro-
In general, the Taguchi method is applied for a single duced herein by applying FAHP along with TOPSIS and
objective. So, the novelty of the paper is to determine Taguchi that aims to optimize the multi-objective process
the reliability index in strain measurement based on two in strain measurement. The proposed approach determines
objectives (accuracy and precision) via Fuzzy Analytic the best alternatives with higher reliability, as shown in Fig. 1.
Hierarchy Process (FAHP) and TOPSIS. Finally, by having

Fig. 1  Flowchart of optimiza-

tion of strain measurement
based on fuzzy approach and
the Taguchi method

Vol.:(0123456789)
Research Article SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x

The objective of this paper is to determine the significant Table 1  Triplet description of Linguistic variables for the evaluation
parameters affecting the reliability of strain gauges. of accuracy and precision weightiness associated with Fig. 1

Linguistic variables Fuzzy rating Triple description

2.1 Problem description
Very high a5 (0.75, 1, 1)
High a4 (0.5, 0.75, 1)
Two common objectives in strain measurement which
Medium a3 (0.25, 0.5, 0.75)
evaluate the reliability of the measurement procedure are
Low a2 (0, 0.25, 0.5)
defined as precision and accuracy. These two objectives are
Very low a1 (0, 0, 0.25)
significantly related to different controllable parameters. The
accuracy of the measurement is the difference between the
measured value and the real value. Any deviation of the
measured value from the real value is the error within the where 𝛼 , 𝛽 , and 𝜆 are positive, negative, and regulative fac-
system [31, 37]. Furthermore, the precision is the closeness tors, respectively.
of two or more measurement to each other. By employing
these two criteria, the reliability of strain measurement is 2.3 Variable weight definition and implementation
determined.
For rating the seriousness of each criterion, a fuzzy eval- The initial weights for accuracy and precision are imple-
uation method is applied. Five different levels are defined, mented via AHP. The variable weight vector W is the nor-
namely 𝛼̃ 1 , 𝛼̃ 2 , 𝛼̃ 3 , 𝛼̃ 4 , and 𝛼̃ 5. The seriousness of each cri- malized product of constant weight factor w and variable
terion is evaluated by applying linguistic terms named as weight state vector s which can be evaluated based on
very low, low, medium, high, and very high. For rating the Eq. 2.
seriousness of precision and accuracy, the triangle member- � �
ship functions with triplet descriptions are used as shown in � � wj sj xj
Wj xj = ∑n � � (2)
Fig. 2 and Table 1, respectively [6, 30]. w s x
k=1 k k j

2.2 Weight calculation of state variables 2.4 Taguchi orthogonal array

For weighting the selected parameters, Analytic Hierarchy Taguchi is a technique to reduce the number of experi-
Process (AHP) is implemented based on the weight profit ments which create robustness for different industrial
factor [7]. applications. Hence, there is a huge reduction of time and
The vector S(xj ) = {s1 (xj ), s2 (xj ), … , sp (xj )} is given as cost throughout the manufacturing process [14]. In this
follow: paper, the interaction of Taguchi with TOPSIS and FAHP is
implemented to investigate the highest reliability index
⎧ 𝜆1 𝛼 xj = 𝛼̃ 1 throughout the strain measurement procedure. Selected
⎪
� � ⎪𝛼 xj = 𝛼̃ 2 parameters and their relevant levels are listed and intro-
sj xj = ⎨ 1 xj = 𝛼̃ 3 (1) duced, as shown in Table 2.
⎪𝛽 xj = 𝛼̃ 4 Based on the number of parameters and their levels, the
⎪
⎩ 𝜆2 𝛼 xj = 𝛼̃ 5 L9 orthogonal array is created as shown in Table 3.

2.5 TOPSIS

In this study, m represent the trial number of experi-

Very low Low Medium High Very High ments number n represent different objectives for quality
Membership Grade

1.0
0.5
0.0
Table 2  Three levels of the temperature (P1 ), the length of wires
(P2 ), and the point of applications (P3 )

0 0.25 0.5 0.75 1.0 P1 (°C) P2 (cm) P3 (cm)

Accuracy and Precision Level
Level 1 23 5 1
Level 2 37 10 2
Fig. 2  Triangular membership function for the weightiness of accu- Level 3 50 15 3
racy and precision

Vol:.(1234567890)
SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x Research Article

Table 3  L9 orthogonal array of Trial number P1 P2 P3 Table 4  Descriptions of the experimental components associated
Taguchi with Fig. 3
1 1 1 1
Part number Description Technical information
2 1 2 2
3 1 3 3 1 Load cell Crow-tail sensor with
HX711 amplifier
4 2 1 2
2 Arduino micro-controller (ATmega328)
5 2 2 3
3 Analog to digital convertor Digital I/O pins 24
6 2 3 1
4 Intermediate software Lab view
7 3 1 3
8 3 2 1
9 3 3 2
di+
Ri = i = 1, 2, … , m. (8)
di+ + di−
evaluation. Initial weight calculation is implemented via
AH. Then, fuzzy relative matrix based on the seriousness
of accuracy and precision is applied and different weight di−
Ni = i = 1, 2, … , m. (9)
of each criterion is considered by Eqs. 1 and 2. Finally, the di+ + di−
varied weighted fuzzy evaluation matrix is based on Eq. 3
as follow: 2.6 Experimental setup
[ ]
Ṽ = ṽ ij m×n i = 1, 2, … , m and j = 1, 2, … , n (3) The experimental components used in strain measure-
ment are described in Table 4, including the technical
here ṽ ij = r̃ij × Wj = (rij1 Wj , rij2 Wj , rij3 Wj ). information for each component.
TOPSIS is applied to rank the order of 9 experiments. In the experimental tests, a Crow-tail load cell sensor
Based on the selected fuzzy decision matrix, it is evident was employed. The HX711 amplifier was embedded in
that the elements ṽ ij of normalized positive triangle num- the kit, with high accuracy of a 24-bit analog-to-digital
bers and their range is related to interval [0, 1]. Hence, the converter (ADC) as shown in Fig. 3a. The Crow-tail load
definition of fuzzy positive ideal solution (FPIS) and fuzzy cell has been designed for a weigh scale and industrial
negative ideal solution (FNIS) is defined as: control applications with the accuracy of the mass sensor
{ } of 0–5 kg as shown in Fig. 3b. The calibration process for
A+ = ṽ 1+ ṽ 2+ … ṽ n+ (4)
the selected sensor is implemented by applying the same
{ }
A− = ṽ 1− ṽ 2− … ṽ n− . (5) mass (40 g) to guarantee that no noise is produced.
In the experimental tests, a C++ program was provided
where ṽ j+ = (vj+ , vj+ , vj+ ) , ṽ j+ = max(vij+ ), ṽ j− = (vj− , vj− , vj− ) ,
and embedded in the Arduino controller. The measured
and vj− = min(vij− ). data as outputs were collected from the controller by
The distance of each alternative or experiment based altering the inputs. In the next section, the details of strain
on the orthogonal array can be calculated by following measurements are explained.
equations:

∑
n ( )
di+ = d ṽ ij , ṽ j+ , ∀i = 1, 2, … , m (6) 3 Result and discussion
j=1
3.1 Strain measurement
∑
n ( )
di− = d ṽ ij , ṽ j− , The strain value was evaluated based on the mass, the
∀i = 1, 2, … , m (7)
j=1 cross-section area of the square aluminum bar and the
module of elasticity employing the following equations:
where d(̃vij , ṽj± ) = [1∕3((vij1 − v11
± 2 ± 2
) + (vij2 − v12 ± 2 0.5
) + (vij3 − v13 ) )] . The total stress was evaluated by Eq. 10:
Finally, the quality index of “n” alternatives can be calcu- mg
F
lated as the reliability index (Ri ) and noise index (Ni ) (Eqs. 8 𝜎= = , (10)
A A
and 9). The higher reliability index, and lower noise index
means that there is less noise within the measurement where F (N) stands for the applied force, A ­(m2) represents
procedure. the cross-section area of the bar, and m (kg) is the mass.
Equation 11 is used to calculate the strain value as follow:

Vol.:(0123456789)
Research Article SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x

Fig. 3  a Crow-tail series sensor

with HX711 amplifier, front and
back view, b Square aluminum
beam used in the tests with
the attached load cell

𝜎
s=
E
. (11)

here s and 𝜎 are the strain (unitless) and the stress (N/m2),
respectively. E (N/m2) represents the module of elasticity
of the aluminum. Three points of the applied forces with
a gap of 0.03 (m) are considered as shown in Fig. 4. Both
the S/N ratio and the analysis of variance were applied to
determine the significance and percentage of contribution
of each parameter.
Table 5 summarizes the evaluation of outcomes, includ-
ing the stresses and strains of all experimental tests. The
tests were conducted for three measurements known
as m1, m2 , and m3 , for the same mass (40 g), which are
obtained by Arduino micro-controller. This repetition
enhances the accuracy of the strain measurements. The
Fig. 4  Point of applied force on the strain gauge (L1 , L2 , and L3 ), mass was evaluated according to the employed material,
including the gaps between the loads which is aluminum (EAl = 69 × 109 MPa).

Table 5  Strain measurement of the test cases based on the L9 orthogonal array, using three types of masses
Trial number m1 (g) m2 (g) m3 (g) Stress 1 (Pa) Stress 2 (Pa) Stress 3 (Pa) Strain 1 Strain 2 Strain 3

1 39.5 39.6 39.5 40.40 40.62 40.40 5.84E−08 5.86E−08 5.84E−08

2 38.9 38.8 38.8 39.09 39.87 39.87 5.76E−08 5.74E−08 5.74E−08
3 38.9 38.9 38.9 39.09 39.09 39.09 5.76E−08 5.76E−08 5.76E−08
4 40.9 40.6 41 41.46 41.81 41.68 6.05E−08 6.01E−08 6.07E−08
5 42.2 42 41.6 43.31 42.87 42.75 6.24E−08 6.22E−08 6.16E−08
6 42 41.7 41.5 42.80 42.22 42.78 6.22E−08 6.17E−08 6.14E−08
7 38 38.3 38.3 38.10 39.785 39.78 5.62E−08 5.67E−08 5.67E−08
8 40 40 40.1 40.50 40.50 40.71 5.92E−08 5.92E−08 5.93E−08
9 39.5 38.8 41.7 40.40 39.87 42.21 5.84E−08 5.74E−08 6.17E−08

Vol:.(1234567890)
SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x Research Article

Table 6  Calculation of initial weights Table 9  Fuzzy evaluating results of nine experiments for accuracy
and precision
S1 S2 Initial weight
Trial number Accuracy Precision
Step weight 0.5 0.5 –
Accuracy 1 – 0.5 1 (0, 0.125, 0.25) (0, 0.125, 0.25)
Precision – 1 0.5 2 (0.11, 0.22, 0.33) (0, 0.14, 0.28)
3 (0.0875, 0.175, 0.2625) (0, 0, 0.162)
4 (0.125, 0.25, 0.375) (0125, 0.25, 0.375)
Table 7  Fuzzy rating of 9 trial numbers based on accuracy and pre- 5 (0.45, 0.6, 0.6) (0, 0.1, 0.2)
cision 6 (0.45, 0.6, 0.6) (0.2, 0.3, 0.4)
Trial number Accuracy Precision 7 (0.45, 0.6, 0.6) (0.2, 0.3, 0.4)
8 (0, 0, 0.0875) (0, 0, 0.162)
1 ᾶ2 (0,0.25,0.5) ᾶ2 (0,0.25,0.5)
9 (0, 0, 0.125) (0.375, 0.5, 0.5)
2 ᾶ3 (0.25,0.5,0.75) ᾶ2 (0,0.25,0.5)
3 ᾶ3 (0.25,0.5,0.75) ᾶ1 (0, 0, 0.25)
4 ᾶ3 (0.25,0.5,0.75) ᾶ3 (0.25,0.5,0.75)
5 ᾶ5 (0.75,1,1) ᾶ2 (0,0.25,0.5) Table 10  Reliability and Noise indices of nine experiments based
on accuracy and precision
6 ᾶ5 (0.75,1,1) ᾶ4 (0.5, 0.75, 1)
7 ᾶ5 (0.75,1,1) ᾶ4 (0.5, 0.75, 1) Trial number di+ di− di+ + di− Ri Ni
8 ᾶ1 (0, 0, 0.25) ᾶ1 (0, 0, 0.25)
1 0.8744 0.3227 1.1972 0.7304 0.2695
9 ᾶ1 (0, 0, 0.25) ᾶ5 (0.75,1,1)
2 0.7681 0.4183 1.1865 0.6474 0.3525
3 0.8833 0.2828 1.1661 0.7574 0.2425
4 0.6346 0.5400 1.1746 0.5402 0.4597
3.2 Optimization process 5 0.4948 0.6836 1.1784 0.4199 0.5800
6 0.3026 0.8654 1.1680 0.2590 0.7409
The initial weight of each criterion, based on their classifi- 7 0.3026 0.8654 1.1680 0.2590 0.7409
cations (Fig. 2), is calculated using AHP, and the results are 8 1.0246 0.1443 1.1690 0.8765 0.1234
tabulated in Table 6. 9 0.6336 0.5342 1.1678 0.5425 0.4574
Fuzzy rating for accuracy and precision of nine experi-
ments is implemented. Based on their seriousness, different
level of fuzzy rating is considered as triangular fuzzy num-
bers, and the results are tabulated in Table 7. that trial number 8 has the highest reliability index based
Based on the literature, the regulative factors are selected on accuracy and precision as shown in Table 10 after all cal-
as α = β = 1.25, λ1 = λ2 = 1.5 [7]. The varied weight of culations (Fig. 5).
each criterion was determined in Table 7 via Eqs. 1 and 2. A sample of calculation is performed here for trial test 1,
The reliability index was also evaluated for all trial numbers for evaluation of accuracy employing Eq. 3. The result can be
(Eq. 8) as shown in Table 8. Then, fuzzy evaluation was used evaluated as follow.
to determine the results of nine experiments for accuracy Sample of calculation:
and precision as shown in Table 9 using Eq. 3. It is found (0, 0.25, 0.5) × 0.5 = (0, 0.125, 0.25)

Table 8  Calculation of varied Trial number Accuracy Precision

weight based on variable
weight profit factor xj sj(xj) wj(xj) xj sj(xj) wj(xj)

1 ᾶ2 1.25 0.50 ᾶ2 1.25 0.50

2 ᾶ3 1.00 0.44 ᾶ2 1.25 0.56
3 ᾶ3 1.00 0.35 ᾶ1 1.87 0.65
4 ᾶ3 1.00 0.50 ᾶ3 1.00 0.50
5 ᾶ5 1.87 0.60 ᾶ2 1.25 0.40
6 ᾶ5 1.87 0.60 ᾶ4 1.25 0.40
7 ᾶ5 1.87 0.60 ᾶ4 1.25 0.40
8 ᾶ1 1.00 0.35 ᾶ1 1.87 0.65
9 ᾶ1 1.87 0.50 ᾶ5 1.87 0.50

Vol.:(0123456789)
Research Article SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x

0.9 Table 11  S/N ratio for all test cases

Reliability
0.8 Noise Trial number NI MSD S/N ratio

0.7 1 0.2695 0.06 11.70

2 0.3525 0.09 10.45
0.6
3 0.2425 0.04 13.97
0.5
4 0.4597 0.16 7.95
5 0.5800 0.25 6.02
0.4 6 0.7409 0.49 3.09
7 0.7409 0.49 3.09
0.3
8 0.1234 0.01 20
0.2 9 0.4574 0.16 7.95

0.1

0
1 2 3 4 5 6 7 8 9 Table 12  The response table of S/N ratio
Trail Number
P1: Temperature P2: length of wire P3: point of
application
Fig. 5  Comparison of Reliability and Noise indices at 9 levels apply-
ing the data of Table 10 Level 1 12.04 7.58 11.59
Level 2 5.69 12.15 8.79
Reliability index (Ri) and Noise index (Ni) are calculated Level 3 10.35 8.34 7.69
for nine experimental tests and shown in Table 10. The Difference |ΔT| 6.35 4.57 3.90
maximum reliability index is 0.8765 and the minimum noise
index is 0.1234 which are related to experiment number 8.
Although, the best set of parameters are P1 at level 3, P2 at
level 2, and P3 at level 1 It, the calculation of Signal to Noise S∕N = −10log(MSD) (13)
ratio will determine the optimum level of individual param- From the collecting data in Table 11, the average
eters with higher reliability index and lower noise index. The S/N ratio for the response table can be calculated. The
distance of each alternative or experiment based on the responses are summarized in Table 12 to determine the
orthogonal array can also be calculated using Eq. 8: optimal levels of the selected parameters.
[ ]0.5 The difference between the levels is defined as |ΔT|.
+
dAccuracy = 1∕3((0 − 0.6)2 + (0.125 − 0.6)2 + (0.25 − 0.6)2 ) The larger value of |ΔT| demonstrates the significance
= 0.4858 of the parameter. It is clear that P1 is the most signifi-
cant parameter followed by P3 , and P2 , respectively.
+
[ ]0.5 The optimum set of parameters can be evaluated from
dprecision = 1∕3((0.5 − 0)2 + (0.5 − 0.125)2 + (0.5 − 0.25)2 )
Table 12 by selecting the highest level of S/N for each
= 0.3886 parameter. The best results are P1 at level 1, P2 at level 2,
So the distance of trial 1 is calculated as: and P3 at level 1. This is clearly highlighted in a bar chart
d1+ = 0.4858 + 0.3886 ≅ 0.8744 as shown in Fig. 6. The results reveal that by setting the
temperature at 23 °C, length of wires at 10 cm, and point
After determining the Reliability and Noise Indexes,
of application at 1 cm, the highest reliability is achiev-
the next step is to apply S/N ratios to find the significant
able with minimum noise. After setting the controllable
parameters and their optimum level in order to find the
parameters to their optimal levels, the final result for the
minimum noise (high accuracy and precision). To reduce
selected mass was 40 g.
the noise, “The smaller the better” is chosen for the Noise
Once the optimum level of individual parameters was
index. The calculated results for temperature measure-
determined, the degrees of freedom, the sum of squares,
ment and S/N ratio have been determined in Table 11.
the variance, F ratio, the pure sum of a square and their con-
The value for the S/N was calculated based on Eqs. 12 and
tribution are calculated via ANOVA. The individual sum of a
13 for each experiment. MSD stands for mean squared
square and the total sum of the square are calculated based
deviation.
on Eqs. 14–16:
1∑ 2
n
MSD = y (12)
N i=1 i

Vol:.(1234567890)
SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x Research Article

14 The variance of the individual parameter (VPi ) and the

Temperature
variance of error (Ve ) are calculated as follow:
length of wire
12
point of application
SSPi
10
VPi = , (20)
Fi

8
SSe
Ve = (21)
6
Fe

Finally, the percentage of contribution of each param-

4 eter (PPi ) can be determined by Eq. 22.

2 SSPi
PPi = , (22)
SST
0
1 2 3
Level Se = ST − SP1 − SP2 − SP3 (23)
Fig. 6  Comparison of Temperature, the length of wire and the By using Eqs. 10–18, the weight or percentage of con-
point of application at three levels based on Table 12 tribution of each parameter is calculated, as shown in
Table 13 which is associated with Fig. 7. The results indi-
cate that the temperature with 67.36% has the highest
contribution, followed by the length of wire with 17.15%
�∑ �2
� � N and the point of application with 15.48%. According to
�
KA
A2i x
i=1 i
Eqs. 18 and 19, the amount of error for degree of free-
SSPi = − , (14)
i=1
nAi N dom is 2. Also, based on Eq. 23 the error for both sum of
square and the percentage of contribution is zero. The
�∑ �2
N
�
KA x
i=1 i Table 13  Analysis of variance (ANOVA) for the selected parameters
SST = xi2 − , (15)
i=1
N F SS P (%)

( ) P1: Temperature 2 0.1398 67.36

∑
KA
P2: length of wire 2 0.03561 17.15
SSE = SST − SSPi , (16) P3: point of application 2 0.03214 15.48
i=1
Error 2 0 0
where Ai is the average temperature for each level, nAi Total 8 0.2076 100
stands for the number of levels, xi represents the tempera-
ture value in each experiment, N is the number of experi-
Point of application
ments and kA represents the number of parameters. For
the calculation of the degree of freedom of each level (Fi)
and the total degree of freedom (Ft) the following equa-
tions were used, respectively:
Fi = xi − 1, (17) Length of wire

Ft = N − 1. (18) Temperature
Thus, to determine the error of the degree of freedom,
Eq. 19 is applied as follow:

∑
KA
Fe = Ft − Fi (19)
i=1
Fig. 7  Pie chart of contribution of Variance (ANOVA) for the three
selected parameters, including Temperature, the length of wire and
the point of application

Vol.:(0123456789)
Research Article SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x

combination of Taguchi and TOPSIS was implemented. design. Doctoral dissertation, WASET-World Academy of Sci-
The determination of an experiment with the highest ence Engineering
4. Arasaratnam P, Sivakumaran KS, Tait MJ (2011) True Stress-True
reliability of strain gauge was considered and experi- Strain models for structural steel elements. Int Sch Res Netw
ment number 8 is nominated with the highest reliability ISRN Civ Eng 2011:11. https​://doi.org/10.5402/2011/65640​1
index. In addition, the contribution of Taguchi led to the 5. Ashutosh Kumar A, Chaturvedi SK, Chaturvedi V, Yadaw RC
determination of significant parameters and their per- (2012) Design studies and optimization of position of strain
gauge. Int J Sci Eng Res 3(10):1–4
centage of contribution affecting the reliability of strain 6. Chen MY, Tzeng HW, Chen YC, Chen SC (2008) The applica-
gauges for different applications. tion of fuzzy theory for the control of weld line positions in
injection-molded part. ISA Trans 47:119–126
7. Cheng J, Feng Tan, Wei JW (2008) Optimization of injection
mold based on fuzzy moldability evaluation. J Mater Process
4 Conclusion Technol 208:222–228
8. Davis JR (2004) Tensile testing, 2nd edn. ASM International,
Combination of Taguchi and FAHP with TOPSIS offers a Materials Park
well-organized use approach for multi-objective prob- 9. Derakhshandeh JF, Arjomandi M, Dally B, Cazzolato B (2014)
The effect of arrangements of two circular cylinders on the
lems. Determining the highest reliability index in strain maximum efficiency of vortex- induced vibration power using
measurement is the objective of this study. By using a scale-adaptive simulation model. J Fluids Struct 49:654–666
the Taguchi method, the significant controllable factors 10. Derakhshandeh JF, Arjomandi M, Dally B, Cazzolato B (2014)
and their optimum levels were identified. Two common Effect of a rigid wall on the vortex induced vibration of two
staggered cylinders. J Renew Sustain Energy Rev 6:033114
concepts in measurement were selected as a different 11. Derakhshandeh JF, Arjomandi M, Cazzolato B, Dally B (2014)
criterion to evaluate the reliability and noise index. The Experimental and computational investigation of wake induced
proposed method was applied for three different control- vibration. In: 19th Australasian fluid mechanics conference
lable parameters, namely the temperature, the length of (AFMC), Melbourne, Australia
12. Derakhshandeh JF, Arjomandi M, Dally B, Cazzolato B (2015)
wire, and the point of application. For rating the serious- Harnessing hydro-kinetic energy from wake induced vibra-
ness of each criterion, a fuzzy evaluation is considered by tion using virtual mass spring damper system. J Ocean Eng
defining five different levels. TOPSIS was chosen to rank 108:115–128
the order of nine experiments. According to the TOPSIS 13. Derakhshandeh JF, Arjomandi M, Dally B, Cazzolato B (2016)
Flow-induced vibration of an elastically mounted airfoil under
approach, it is found that the highest reliability index is the influence of oncoming vortices. J Exp Therm Fluid Sci
related to trial number 8. The best set of parameters are 74:58–72
P1 at level 3, P2 at level 2, and P3 at level 1 with 88% reli- 14. Figliola RS, Beasley D (2015) Theory and design for mechanical
ability index and 12% noise index. The optimum levels are measurements. Wiley, New York
15. Geethapriyan T, Muthuramalingam T, Kalaichelvan K (2019)
P1 at level 1, P2 at level 2, and P3 at level 1 and the tem- Influence of process parameters on machinability of Inconel
perature is the most significant parameters with 67.36% 718 by electrochemical micromachining process using TOPSIS
contribution, which alters the accuracy and precision in technique. Arab J Sci Eng 44(9):7945–7955
strain measurement. 16. Gu L (2001) A comparison of polynomial based regression mod-
els in vehicle safety analysis, in Diaz. In: ASME design engineer-
ing technical conferences—design automation conference,
ASME, Pittsburgh, PA, USA, September 9–12, DAC-21063
Compliance withy ethical standards 17. He W, Zhang YF, Lee KS, Fuh JYH, Nee AYC (1998) Automated pro-
cess parameter resetting for injection moulding: a fuzzy-neuro
Conflict of interest The authors declare that they have no conflict of approach. J Intell Manuf 9:17–27
interest. 18. Huo J, Liu S, Wang Y, Muthuramalingam T, Pi VN (2019) Influ-
ence of process factors on surface measures on electrical dis-
charge machined stainless steel using TOPSIS. Mater Res Exp
6(8):086507
19. Li X, Zhang R, Yu W, Wang K, Wei J, Wu D, Cao A, Li Z, Cheng
References Y, Zheng Q, Ruoff RS (2012) Stretchable and highly sensitive
graphene-on-polymer strain sensors. Sci Rep 2:870
20. Lin MW, Rogers CA (1994) Bonding layer effects on the actuation
1. Alam MM, Zheng Q, Derakhshandeh JF, Rehman S, Ji C, Zafar mechanism of an induced strain actuator/substructure system.
F (2018) On forces and phase lags between vortex sheddings SPIE 2190:658–670
from three tandem cylinders. Int J Heat Fluid Flow 60:117–135 21. Lindsey LH, Kirstin P, Metin S (2017) Soft actuators for small-scale
2. Amer Y, Moayyedian M, Hajiabolhasani Z, Moayyedian L (2012) robotics. Adv Mater 29(13):1603483
Reducing warpage in injection moulding processes using 22. Lord DJ, Morrel RM (2010) Elastic modules measurement-obtain-
Taguchi method approach: ANOVA. Doctoral dissertation, ing reliable data from the tensile test. Metrologia 47:502–515
ACTA Press, Calgary 23. Lu N, Lu C, Yang S, Rogers J (2012) Highly sensitive skin-mounta-
3. Amer Y, Moayyedian M, Hajiabolhasani Z, Moayyedian L (2013) ble strain gauges based entirely on elastomers. Adv Funct Mater
Improving injection moulding processes using experimental 22:4044–4050

Vol:.(1234567890)
SN Applied Sciences (2019) 1:1392 | https://doi.org/10.1007/s42452-019-1428-x Research Article

24. Meral G, Sarıkaya M, Dilipak H, Şeker U (2015) Multi-response 33. Motra HB, Hildebrandb J, Dimmig-Osburg A (2014) Assessment
optimization of cutting parameters for hole quality in drilling of strain measurement techniques to characterize mechanical
of AISI 1050 steel. Arab J Sci Eng 40(12):3709–3722 properties of structural steel. Eng Sci Technol Int J 17:260–269
25. Moayyedian M, Abhary K, Marian R (2015) Improved gate system 34. Ponnusamy P, Masood SH, Ruan D, Palanisamy S, Rashid R
for scrap reduction in injection molding processes. Proc Manuf (2018) High strain rate dynamic behaviour of AlSi12 alloy
2:246–250 processed by selective laser melting. Int J Adv Manuf Technol
26. Moayyedian M, Abhary K, Marian R (2016) The analysis of defects 97(1–4):1023–1035
prediction in injection molding. World Acad Sci Eng Technol Int 35. Sarıkaya M, Güllü A (2015) Multi-response optimization of mini-
J Mech Aerosp Ind Mechatron Manuf Eng 10(12):1876–1879 mum quantity lubrication parameters using Taguchi-based grey
27. Moayyedian M, Abhary K, Marian R (2016) Gate design and filling relational analysis in turning of difficult-to-cut alloy Haynes 25.
process analysis of the cavity in injection molding process. Adv J Clean Prod 91:347–357
Manuf 4(2):123–133 36. Sarıkaya M, Yılmaz V, Güllü A (2016) Analysis of cutting parame-
28. Moayyedian M, Abhary K, Marian R (2017) The analysis of short ters and cooling/lubrication methods for sustainable machining
shot possibility in injection molding process. Int J Adv Manuf in turning of Haynes 25 superalloy. J Clean Prod 133:172–181
Technol 191(9–12):3977–3989 37. Singh J, Kumar A, Sharma ND, Bandyopadhyay AK (2011) Reli-
29. Moayyedian M, Derakhshandeh JF, Sherif S (2018) Experimental ability and long term stability of a digital pressure gauge (DPG)
investigations of significant parameters of strain measurement used as a standard-a case study. Mapan 26(2):115–124
employing Taguchi method. SN Appl Sci 1–9 38. Sirohi J, Chopra I (2000) Fundamental understanding of piezo-
30. Moayyedian M, Abhary K, Marian R (2018) Optimization of electric strain sensors. J Intell Mater Syst Struct 11:246–257
injection molding process based on fuzzy quality evaluation 39. Wang GG, Shan S (2007) Review of metamodeling techniques
and Taguchi experimental design. CIRP J Manuf Sci Technol in support of engineering design optimization. Trans ASME
21:150–160 129:370–380
31. Moayyedian M (2018) Intelligent optimization of mold design 40. Willink R (2007) On the uncertainty of the mean of digitized
and process parameters in injection molding. Springer, New measurement. Metrologia 44:73–81
York
32. Moayyedian M, Mamedov A (2019) Multi-objective optimiza- Publisher’s Note Springer Nature remains neutral with regard to
tion of injection molding process for determination of feasible jurisdictional claims in published maps and institutional affiliations.
moldability index. Proc CIRP 84:769–773

Vol.:(0123456789)