Modeling and Optimization of Fiber Optic
Chemical Vapor Sensor
Budi Mulyanti1, Harry Ramza2, Roer Eka Pawinanto3, Faizar Abdurrahman4, Latifah Sarah Supian5
Norhana Arsad6, Mohammad Syuhaimi Ab-Rahman6
1
Department of Electrical Engineering, Faculty of Engineering and Vocational Education,
Universitas Pendidikan Indonesia.
2
Department of Electrical Engineering, Faculty of Engineering,
Universitas Muhammadiyah Prof. Dr. HAMKA, Indonesia.
3
Department of Informatics Engineering, Politeknik Negeri Indramayu, Indonesia.
4
Department of Electrical Engineering, Politeknik Negeri Lhokseumawe, Indonesia.
5
Department of Electrical Engineering, Faculty of Engineering, Universiti Pertahanan Nasional Malaysia.
6
Department of Electrical, Electronic and System Engineering, Faculty of Engineering and Built Environment,
Universiti Kebangsaan Malaysia.
roer.eka@gmail.com
Abstract—This paper discusses the application of Box–
Behnken Design (BBD) to get a mathematical model for
chemical vapor liquid detection with the objective of
optimizing the optical fiber optic sensor probe.
The
parameters of input process were considered as variables to
create the output parameters (response) using Response
Surface Methodology (RSM). Input parameters such as length
of probe, diameter of probe, photo-initiator liquid, vacuum
pressure of chamber and purity of liquid detector were
processed with Box – Behnken design approach for making
POF (plastic optical fiber) probe of chemical sensor. Design
Expert software was used to design the experiments with
randomized runs. The main aim is to create an equation model
as a platform for the probe design of POF chemical vapors
detection similar to acetone, ethanol and methanol liquid. The
experimental data were processed by considering the input
parameters. The contribution of this research is the
mathematic equation model that applies the polynomial
equation. The final result of the wavelength application was
between five to be three wavelengths, 434.05 nm, 486.13 nm
and 656.03 nm. These wavelengths are the significant result of
optimization measured using three chemical vapors. The
optimization process uses the analysis of variables (ANOVA) to
produce the quadratic model equation.
Index
Terms—Box–Behnken
Design;
Design–Expert
Software; Fiber Optic Chemical Vapor Sensor; MathOptimization Model.
I. INTRODUCTION
Plastic optical fiber probe for chemical vapor detection has
been widely used in experimental and industrial scale [1, 2].
Most of POF (plastic optical fiber) probe usually were
created with the many of custom variables. One of the main
variables is cladding modification of POF by substituting
Zinc Oxide (ZnO) layer. The method of measurement uses
fabry–perot interferometer and LED super bright as a light
source to get sufficient reflection of light from the end of
probe tip.
In order to function as a sensor, chemical vapor detection
is conducted in a chamber that can regulate air pressure.
Three chemical liquid such as Acetone, Ethanol and
Methanol were chosen in this experiment to get the chemical
vapor that drives the changes of refractive index from the
sensor probe. To produce the optimum optical probe, it must
be considered other variables, such as the length of probe as
𝑥1 , diameter of probe as 𝑥2 , doping of photoinitiator liquid
as 𝑥3 , vacuum pressure in chamber as 𝑥4 and purity of
chemical vapor detection as 𝑥5 , as the independent variables.
The reflection intensity from five particular wavelengths
was used as dependent variables.
Box–Behnken experimental design or BBD, which is a
well known and most common multi-factorial design of
response surface methodology (RSM) in various
experiments[3] has been applied in the optimization of
probe sensing design. The second-order model has always
been used in RSM due to its many advantages: It consists of
less number of experiments, suitable with multi-variables
and able to explain correlation of each variable [4-6].
The final result determined the optimum values generated
from the mathematical model platform. In this study, the
authors investigated the chemical vapor detection using
plastic optical fiber probe [7] created by the modification of
the cladding site with ZnO nano-powder.
II. EXPERIMENTAL PROCEDURE
In this experiment, the vacuum chamber was used as a
place of POF probe for detection of chemical vapor from the
liquid chamber. The chamber circumstances were set to lowpressure using the vacuum pump that was intended to take
up the chemical vapor so that it can change the refractive
index of POF probe. The air pressure was regulated by two
air valves positioned at the top of the testing chamber.
Figure 1 shows the set-up experiment of chemical vapor
detection at the vacuum chamber. Air pressure was sucked
using the oil vacuum pump. Light source was injected into
the POF Y-coupler to allow it to transmit until the end of
probe. When the chemical vapor affected the probe surface
area, the POF probe changed the refractive index, which
then caused the reflecting light to move the spectrometer.
POF probe was inserted into the hole of the detection
chamber. Here, the chemical vapor moved into the chamber
when the vacuum pump sucked the air vapor. This situation
was controlled until the air pressure was positioned at low
level by regulating the outlet valve. In this set-up, white
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
73
Journal of Telecommunication, Electronic and Computer Engineering
light source was used to read the changing of the light
reflection intensity. Spectrometer USB4000 VIS-NIR used
to detect the light was connected to the data recorder.
Figure 1: Set-up experiment of chemical vapor
Figure 2 exhibits all parts used in this experiment,
together with three chemical vapors detector, namely the
Acetone, Ethanol and Methanol that were tested at separate
places and different time. Data were collected from the
recorder created by SPECTRASUITE application from
Ocean Optic. The number of experiment was adjusted
according to the methodology of the experiment.
Figure 2: Setup Experiment with three chemical vapors; Acetone,
Ethanol and Methanol
III. COATING MATERIAL
ZnO nano powder from SIGMA – ALDRICH was mixed
with the methanol liquid. Additional adhesive liquid called
photo-initiator was granted with the amount of 0.05 ml, 0.10
ml and 0.20 ml. It was given based on the code of -1, 0 and
1, where -1 as the lowest and +1 as the highest value. The
coating liquid was mixed following the process of probes
coating as shown in Table 1. After this process, it was dried
for six hours until the powder can be attached to the probe
head. The mixed liquid was used in the coating process of
the optical fiber as shown in Table 1. After the coating
process, the optical fiber was left to dry for about 24 hours,
so that the ZnO powder can stick into the end of the probe
tip.
Figure 3(a) shows ZnO nano powder as coating material.
The mixing process consists of 30 ml methanol that is mixed
74
with 0.4884 gram of ZnO material. This process is shown in
Figure 3(b) that used the hotplate and stirrer.
(a)
(b)
Figure 3: ZnO material;(a). Intake of ZnO nanopowder adjusted with
mixing ratio. (b). Mixing process of ZnO with Methanol liquid using
hotplate and stirrer
IV. METHODOLOGY
It can be shown from the references [8-10] the
development and current applications can be improved using
RSM for explaining the parameters of output in many
variables input. The response surface methodology promotes
the relations between the two or more response variables.
The main idea of RSM is to apply a sequence of designed
experiments to find an optimal response. The design
procedure of RSM is as follows; (1) creating a series of
experiments for adequate and reliable measurement, (2)
extending a mathematical model of the second order
response, (3) searching for the optimum experimental
parameters that generate a maximum or minimum value of
response, (4) explaining direct and interactive effects of the
parameter procedure using graphs. Figure 4 shows the
flowchart of the design experiment that aims for efficiency
to get more information from fewer experiments by focusing
on collecting required information only
RSM design recommends us to calculate interaction and
even quadratic effects. It gives us an idea of the local shape
of response surface under investigation. Box-Behnken
design is one of design experiments of RSM. It is an
efficient design for fitting second-order polynomials to
response surface because it applies relatively small number
of observation to calculate the parameters. The detection
process of POF probe,
V. BOX BEHNKEN DESIGN
Box Behnken is an experimental design for response
surface methodology (RSM) to achieve the following aims:
i. Each factor or independent variable is placed at one
of three equally spaced values, usually coded as -1, 0
and 1.
ii. It should be sufficient to fit a quadratic model that is
one containing squared terms and products of two
factors.
iii. The ratio of experimental number points the number
of coefficients in the quadratic model that should be
reasonable in the range 1.5 to 2.6.
iv. The estimation variance should more or less depend
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
Modeling and Optimization of Fiber Optic Chemical Vapor Sensor
on the distance from the center only.
The Box–Behnken proposed three level designs for fitting
response surface. These designs were created by combining
2k factorials with incomplete block design [11]. Figure 5
shows the three variables of Box–Behnken design. It can be
remarked that Box–Behnken design is a spherical scheme
with all points lying on radius sphere. Box–Behnken design
does not contain any point at the nodes of the cubic region
created by the upper and lower limits. BBD needs fewer
treatment combinations than a CCD and rotatable in
problem with 3 and 5 factors.
CCD design. The application of Box–Behnken can reduce
the sum of constant number without the lack of optimization
constant in comparison to the traditional factorial design
[11-14].
Figure 5: Three variables of Box – Behnken design[8].
Table 1
Minimum and maximum level from four variables factors at coded and
un-coded symbol.
Variables
Symbol
UnCoded
X1
X2
Level
Coded
-1
0
1
x1
x2
2.5
0.51
6.25
0.69
10
0.87
Probe length (cm)
Probe Diameter (mm)
Doping
Photo-initiator
(ml)
Pressure of Vacuum
Chamber (mBar)
Purity of Liquid detector
(%)
x3
X3
0.05
0.125
0.2
x4
X4
0
50
100
x5
X5
10
55
100
The minimum and maximum interval of five variables
represented by coded and un-coded symbols are shown in
Table 2. The correlation of coded and un-coded variables
are explained by the following equations [15-17]:
(𝑋1 − 6.25)
3.75
(𝑋2 − 0.69)
𝑥2 =
0.18
(𝑋3 − 0.125)
𝑥3 =
0.075
(𝑋4 − 50)
𝑥4 =
50
(𝑋5 − 55)
𝑥5 =
45
𝑥1 =
Figure 4: Design of experiments flowchart
The design of RSM allows the calculation of variable
interaction and even-quadratic effect. It also gives the ideas
from RSM form that was being investigated. The BoxBehnken design has a maximum efficiency for RSM
problems involving five factors with three-level factorial.
The process number lower than the center composite design
(CCD) is required. The RSM is an optimum way to assess
the relationship between the experiment output (response)
and any factors called as X1, X2, X3, and others. This
method is always used in the combination form with
factorial design method like the Box–Behnken design and
(1)
(2)
(3)
(4)
(5)
where X1, X2, X3, X4, X5 are the un-coded variables and x1,
x2, x3, x4,x5 are the coded variables. Two types of variable
that have particular unit and effects of variables on detection
efficiency can be approached by using a second order of
polynomial model that is written in equation [14, 18-20],
𝑘
𝑘
𝑖=1
𝑖=1
𝑦 = 𝛽0 + ∑ 𝛽𝑖 𝑥𝑖 + ∑ 𝛽𝑖 𝑥𝑖 2 + ∑ ∑ 𝛽𝑖𝑗 𝑥𝑖 𝑥𝑗 + 𝜀
𝑖
(6)
𝑗
where 𝜀 is random error, 𝛽0 is defined as the intercept
coefficient, 𝛽𝑖 is the linear and quadratic interaction
coefficient, 𝛽𝑖𝑗 is the second order of interaction coefficient
and 𝑘 is the number of independent parameter. Equation (6)
can be rewritten into the matrix form as in (7):
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
75
Journal of Telecommunication, Electronic and Computer Engineering
𝑦 = 𝛽𝑥 + 𝜀
(7)
Detail solution of Equation (6) and (7) can be explained
with the matrix approach with reference to [10]. The
preparation of chemical vapor experiment using optical fiber
is shown in Figure 2. This experiment consists of chemical
liquid chamber, chemical vapor detection chamber, vacuum
suction pump, pressure meter, white light source and
spectrometers Ocean Optic (USB4000VIS-NIR). In the
internal part of the detection chamber, it is a place of probe
position that performs the experiment of fiber optic with
variety of sizes which are detectable in each of the probes.
Vacuum suction pump was used to attract the chemical
vapor in the liquid chamber. Chemical vapor escaped
through the output line from the suction pump. The air
pressure on the detection chamber was controlled by the
alignment of outlet valve with the suction power by the
pump. Spectrometer was used as a detector of spectrum
from the light source transmitted by the fiber optic.
VI. RESULT AND DISCUSSION
The proposed Box–Behnken design requires 46 processes
for response surface model [4, 21, 22]. The parameter
process for experimental runs was selected based on the
standard design shown in Figure 5. Detail of the
experimental runs with input data set that has been carried
out is shown in Table 2.
Design expert application software was used to design the
experiment and randomized process [4]. Randomized
process is more useful to ensure that the conditions in the
term do not depend on the previously run process and
predict the situation in the next run. The randomized process
is important in the experiment as it helps to follow the right
track and can be defended, depending on the result of the
run.
The experiment result requires meaningful analysis of
each variable. Table 3 shows that F-Value is 1.87 implying
that the model was not significant for the surrounding
disturbance. There is a 12.20% chance that a large F-value
could occur due to noise and the values of “prob > F”. Pvalue less than 0.05 indicates that the model is significant.
In this case, A-Probe Length is a significant model. Values
greater than 0.1 indicate the model terms are not-significant
[3, 4, 23].
If many parts of the model terms are not significant (not
including models that is required to support the hierarchical
modeling), the model reduction can improve the proposed
model. The lack of fit from F-value is 0.80, implying that
the relative value is not significant to the Pure Error. There
is 69.45% chance that the lack of fit from F-value can cause
disturbance. No significant value due to lack of fit is a good
value because the model expectancy is suitable to be
applied.
Table 4 shows that F-value is 3.07, implying that the
model is significant. There is 0.44% chance that one of the
F-values can occur due to disturbances. P-value less than
0.05 means that the part of the model term is significant.
In the case of E-Purity of Detection Liquid, E^2 is a
significant part of the model. If there are many parts that are
not significant (not including model that is required to
support hierarchical modeling), model reduction can
improve the proposed model. The lack of fit value is 0.90,
76
implying that it is not obvious to a Pure Error. There is a
61.13% chance that a lack of fit that indicates good
conditions because it is considered as suitability of the
model.
In addition, the model is validated using analysis of
variable (ANOVA). As shown in Table 4, the model is
validated by experiments with new parameter to measure the
value of ranged reactions and compare with the prediction of
equation model. The details of the experiments and
calculated from the output variables are given in Table 6, 7
and 8.
Table 2
Box – Behnken Design for the Experiment
Run
Probe
Length
(cm)
Probe
Diameter
(mm)
Doping
Photoinitiator (ml)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
10
10
6.25
10
6.25
6.25
2.5
2.5
6.25
6.25
6.25
6.25
6.25
6.25
10
6.25
6.25
6.25
2.5
6.25
6.25
2.5
10
6.25
10
6.25
6.25
2.5
6.25
2.5
6.25
6.25
6.25
2.5
6.25
6.25
6.25
6.25
2.5
6.25
6.25
6.25
10
6.25
10
6.25
0.87
0.69
0.87
0.69
0.69
0.69
0.69
0.69
0.51
0.69
0.51
0.69
0.51
0.69
0.69
0.87
0.69
0.51
0.69
0.87
0.69
0.69
0.69
0.69
0.51
0.87
0.51
0.69
0.69
0.87
0.69
0.69
0.69
0.69
0.69
0.69
0.69
0.51
0.51
0.69
0.87
0.87
0.69
0.69
0.69
0.69
0.125
0.125
0.05
0.125
0.125
0.05
0.125
0.125
0.2
0.125
0.125
0.05
0.125
0.2
0.125
0.125
0.125
0.125
0.125
0.125
0.125
0.2
0.05
0.05
0.125
0.125
0.05
0.125
0.2
0.125
0.125
0.125
0.125
0.05
0.2
0.125
0.2
0.125
0.125
0.05
0.125
0.2
0.2
0.125
0.125
0.125
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
Pressure
of
Vacuum
Chamber
(mBar)
50
50
50
50
50
50
50
100
50
50
50
0
100
50
0
0
100
0
0
50
50
50
50
100
50
100
50
50
50
50
0
50
0
50
100
50
0
50
50
50
50
50
50
50
100
100
Purity of
detection
liquid
(ml)
55
100
55
10
55
10
10
55
55
55
10
55
55
100
55
55
100
55
55
10
55
55
55
55
55
55
55
100
10
55
100
55
10
55
55
55
55
100
55
100
100
55
55
55
55
10
Modeling and Optimization of Fiber Optic Chemical Vapor Sensor
VII. RESEARCH CONTRIBUTION
The experimental result discussed in the design of
chemical detection is a mathematic equation as a platform
for the probe design by considering five variables, namely
the probe length, probe diameter, doping photo-initiator,
pressure of vacuum chamber and purity of detection liquid.
Mathematic equation model can be applied with the
calculated significant value. For chemical vapor detection,
the formation of polynomial equation is drawn from
Equation (8).
𝑦(𝜆𝑛 , 𝜔𝑚 ) = 𝑎0,𝜆𝑛 ,𝜔𝑚 + 𝑏1,𝜆𝑛,𝜔𝑚 𝑋1 + 𝑏2,𝜆𝑛,𝜔𝑚 𝑋2 + 𝑏3,𝜆𝑛 ,𝜔𝑚 𝑋3
+ 𝑏4,𝜆𝑛,𝜔𝑚 𝑋4 + 𝑏5,𝜆𝑛 ,𝜔𝑚 𝑋5 + 𝑐1,𝜆𝑛,𝜔𝑚 𝑋1 𝑋2
+ 𝑐2,𝜆𝑛 ,𝜔𝑚 𝑋1 𝑋3 + 𝑐3,𝜆𝑛 ,𝜔𝑚 𝑋1 𝑋4
+ 𝑐4,𝜆𝑛 ,𝜔𝑚 𝑋1 𝑋5 + 𝑑1,𝜆𝑛 ,𝜔𝑚 𝑋2 𝑋3
+ 𝑑2,𝜆𝑛,𝜔𝑚 𝑋2 𝑋4 + 𝑑3,𝜆𝑛 ,𝜔𝑚 𝑋2 𝑋5
+ 𝑒1,𝜆𝑛,𝜔𝑚 𝑋3 𝑋4 + 𝑒2,𝜆𝑛 ,𝜔𝑚 𝑋3 𝑋5
+ 𝑓1,𝜆𝑛,𝜔𝑚 𝑋4 𝑋5 + 𝑔1,𝜆𝑛 ,𝜔𝑚 (𝑋1 )2
+ 𝑔2,𝜆𝑛,𝜔𝑚 (𝑋2 )2 + 𝑔3,𝜆𝑛,𝜔𝑚 (𝑋3 )2
+ 𝑔4,𝜆𝑛,𝜔𝑚 (𝑋4 )2 + 𝑔5,𝜆𝑛,𝜔𝑚 (𝑋5 )2
(8)
where, 𝑎, 𝑏, 𝑐, 𝑑, 𝑒, 𝑓 and 𝑔 are defined as constant
equation depending on the wavelength and purity of
detection liquid, 𝜆𝑛 is the wavelength used for vapor
detection and n = 434.05 nm, 486.13 nm, and 656.03 nm,
𝜔𝑚 is the chemical vapor material and 𝑚 is an integer
represented as 1 for Acetone, 2 for Ethanol and 3 for
Methanol.
The application of mathematical model in Equation (8)
must be considered with the environment situation, the
experiment of vacuum chamber at each chemical vapor
detection and the positioning of probe. Table 5 shows the
results of the different stage of responsive based on
Equation (8) to re-test with the variables randomly. The
result of re-testing was then compared with the results of the
existed prediction.
Parameter of 𝑋1 , 𝑋2 , 𝑋3 , 𝑋4 and 𝑋5 was tested to get the
validity of the data that is randomly valued to re-create the
probe. This remake was the objective of testing the model
that has been designed in Equation (8).
Table 6, 7 and 8 do not provide the information of the
wavelength response for 397.14 nm and 410.05 nm. The
reaction at these wavelengths from the beginning was to
provide a non-significant value. Both of these wavelengths
used custom processing methods of models design, namely
the average, linear, two-factor interaction, quadratic, cubic,
order-5 and order-6 that produce a non-discriminant analysis
significantly.
VIII. CONCLUSION
Box–Behnken design and the experimental design were
conducted by selecting five input variables and levels. The
minimum of the experiment process, data collection and
models were explained and developed. Confirmation of the
suitability of each model was conducted using ANOVA
technique (analysis of variance). The results show that the
whole model can be used with a confidence level of 0.95
into the next stage of design. The model validation was
conducted by collecting additional experimental data where
it has a high confident level to adopt the chosen parameters.
Mathematical equation model in equation (8) was
explained as a platform to design optical probe for three
chemical vapor detection, namely the acetone, ethanol and
methanol. The optimization set of input parameters can be
identified by getting into consideration the probe length,
probe diameter, doping photo-initiator, pressure of vacuum
chamber and purity of detection liquid. By reducing the
number of experimental runs, the expected result was very
convincing and logically acceptable. It can be followed to
obtain a solution for planning purposes as well as saving
time and cost.
For future work, we plan that the application of uncladding plastic optical fiber to be included with the
optimization of chemical vapor detection based on the
absorbance rate with the other liquid concentrations. In
addition, the comparison of ANOVA process method, such
as modification process, model design, linearity process and
two-factor interaction process. Finally, the model and
optimization of chemical vapor detection with high pressure
or zero pressure will be conducted.
Table 3
The Result Analysis of ANOVA for the Linear Model of Acetone Vapor
Detection ( = 410.05 nm)
Source
Model
A-Probe
Length
B-Probe
diameter
C-Doping
photoinitiator
D-Pressure
of Vacuum
chamber
E-Purity of
detection
liquid
Residual
Lack of fit
Pure Error
Cor Total
Sum of
Squares
1.741E+5
df
Mean square
F-Value
p-value
“Prob > F”
0.1220 Not significant
5
34810.78
1.87
96064.35
1
96064.35
5.15
0.0287
significant
51548.30
1
51548.30
2.76
0.1042
Not significant
12.83
1
12.83
6.881E-4
0.9792
Not significant
15017.89
1
15017.89
0.81
0.3749
Not significant
11410.51
1
11410.51
0.61
0.4387
Not significant
7.461E+5
6.327E+5
1.133E+5
9.201E+5
40
35
5
45
18651.85
18077.87
22669.74
0.80
0.6945
Not significant
Table 4
The Result Analysis of ANOVA for the Quadratic Model of Acetone Vapor
Detection (=434.05nm)
Source
Model
A-Probe
Length
B-Probe
diameter
C-Doping
photo-initiator
D-Vacuum
pressure of
chamber
E-Purity of
detection
liquid
AB
AC
AD
AE
BC
BD
BE
CD
CE
DE
A^2
B^2
C^2
D^2
E^2
Residual
Lack of fit
Pure error
Cor Total
Sum of
Squares
1.058E+6
df
Mean square
F-Value
20
52876.05
3.07
p-value “Prob
> F”
0.0044
3477.76
1
3477.76
0.20
0.6569
51972.60
1
51972.60
3.02
0.0945
109.67
1
109.67
6.375E-3
0.9370
2630.66
1
2630.66
0.15
0.6991
83706.06
1
83706.06
4.87
0.0368
28170.27
25937.10
805.99
49375.06
10315.45
6771.64
12729.48
16961.16
20049.14
8840.70
7668.68
26676.32
266.85
21856.38
4.891E+5
4.301E+5
3.370E+5
93120.77
1.488E+6
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
25
20
5
45
28170.27
25937.10
805.99
49375.06
10315.45
6771.64
12729.48
16961.16
20049.14
8840.70
7668.68
26676.32
266.85
21856.38
4.891E+5
17204.89
16850.07
18624.15
1.64
1.51
0.047
2.87
0.60
0.39
0.74
0.99
1.17
0.51
0.45
1.55
0.016
1.27
28.43
0.2124
0.2309
0.8304
0.1027
0.4460
0.5361
0.3979
0.3303
0.2907
0.4801
0.5105
0.2246
0.9019
0.2704
< 0.0001
0.90
0.6113
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
Significant
Not significant
77
Journal of Telecommunication, Electronic and Computer Engineering
Table 5
The Constant Value from Mathematic Equation Model for Three Chemical Vapors
Constant
𝑎0
𝑏1
𝑏2
𝑏3
𝑏4
𝑏5
𝑐1
𝑐2
𝑐3
𝑐4
𝑑1
𝑑2
𝑑3
𝑒1
𝑒2
𝑓1
𝑔1
𝑔2
𝑔3
𝑔4
𝑔5
X1
𝝀𝟒𝟑𝟒
462.80908
111.85802
3439.89815
6687.63241
1.95036
-6.89487
-124.32593
-286.31111
0.075707
0.65839
-3761.66667
4.57167
-6.96451
-17.36467
-20.97704
-0.020894
-2.10794
-1706.39146
-983.03704
-0.020017
0.11691
Acetone Liquid
𝝀𝟒𝟖𝟔
755.38509
123.79383
5031.91204
8080.87685
3.41617
-12.85265
-160.04444
-338.59556
0.057600
0.32464
-5637.22222
5.47667
-8.60216
-27.74000
-9.06074
-0.027512
-0.43567
-2522.96811
-843.55556
-0.020745
0.19073
𝝀𝟔𝟓𝟔
376.33258
109.07676
4630.22492
7813.96944
2.53363
-9.44871
-129.22222
-360.60444
0.068520
0.32960
-4916.48148
5.71472
-6.86111
-29.06933
-15.48741
-0.022776
-0.42230
-2495.71116
886.18519
-0.014564
0.14974
𝝀𝟒𝟑𝟒
3098.4
8.44198
1330.18
-1418.42
-3.18158
-9.69591
-23.5148
-72.6489
-0.0935867
0.332919
-2489.26
6.76944
-6.92438
-8.688
-1.87481
-0.0194856
1.4917
1270.41
14351.3
0.00421983
0.136087
Ethanol Liquid
𝝀𝟒𝟖𝟔
𝝀𝟔𝟓𝟔
3442.58
3372.01
-20.4043
-30.6251
43.0143
-856.992
635.625
-559.467
-3.46232
-6.29731
-12.6787
-10.0643
-16.4111
-25.8556
-142.658
-68.6667
0.00893333
-0.0148933
-0.0394519
0.0090963
-4171.85
-3210
7.4625
10.2139
-10.5157
-10.0562
-18.1587
-14.6713
0.937778
-0.386667
-0.0265833
-0.02532
3.02308
3.66424
494.2
976.44
13978.4
14326.7
0.0129458
0.018253
0.20626
0.173746
𝝀𝟒𝟑𝟒
1823.51
132.3
-436.562
7238.74
-6.82407
-5.84429
-25.0296
-458.738
0.146973
0.201081
-2358.33
8.19194
-4.85
9.43267
0.372593
0.000248889
-5.65467
429.874
-14475.8
-0.00613917
0.0986574
Methanol Liquid
𝝀𝟒𝟖𝟔
𝝀𝟔𝟓𝟔
1725.73
1180.4
158.367
150.794
1449.14
1231.6
9356.86
9732.82
-2.34541
-1.15765
-8.14692
-3.18591
-47.0259
-46.9593
-513.031
-518.818
0.08044
0.0767067
-0.314696
-0.307259
-4475.19
-3606.48
4.54611
3.30944
-9.06759
-9.95247
-0.521333
-4.518
-2.17852
-0.8
-0.00835778 -0.00591667
-4.42421
-3.60622
-397.492
-239.712
-12690.6
-16369
-0.0034505
-0.00350433
0.182593
0.136242
Table 6
Prediction and Experiment Results for Validation Data in Acetone Vapors
X2
X3
X4
X5
2.5
0.83
0.2
100
2.5
0.8
0.15
100
5
0.79
0.05
50
7.5
0.81
0.2
50
R* = Prediction; U* = Experiment
100
100
100
100
(434.05nm)
R*
U*
2165.12
2160.01
2245.22
2250.12
2495.46
2490.02
2326.64
2329.72
(486.13nm)
R*
U*
3227.04
3225.01
3317.19
3320.93
3428.77
3410.40
3203.82
3210.72
(656.03nm)
R*
U*
2649.58
2640.97
2733.65
2740.27
2854.04
2840.66
2636.61
2639.18
Table 7
Prediction and Experiment Results for Validation Data in Ethanol Vapors
X1
X2
X3
X4
X5
2.5
0.83
0.2
100
2.5
0.8
0.15
100
5
0.79
0.05
50
7.5
0.81
0.2
50
R* = Prediction; U* = Experiment
100
100
100
100
(434.05nm)
R*
U*
4614.06
4611.33
4510.44
4530.21
4670.43
4611.89
4614.06
4599.98
(486.13nm)
R*
U*
3381.631
3370.67
3386.003
3400.03
3486.888
3450.96
3297.87
3210.78
(656.03nm)
R*
U*
2861.15
2840.16
2849.37
2860.18
2911.03
2920.63
2757.09
2760.27
Tabel 8
Prediction and Experiment Results for Validation Data in Methanol Vapors
X1
X2
X3
X4
X5
2.5
0.83
0.2
100
2.5
0.8
0.15
100
5
0.79
0.05
50
7.5
0.81
0.2
50
R* = Prediction; U* = Experiment
100
100
100
100
(434.05nm)
R*
U*
2469.97
2470.01
2462.09
2440.12
2436.08
2450.32
2366.08
2390.97
IX. REFERENCES
[1]
[2]
[3]
[4]
[5]
78
L. Wang, M. Cao, Q. Liu, G. Wei, B. Li, and C. Lin,. "Modeling and
experimental verification of polarization errors in Sagnac fiber optic
current sensor". Optik - International Journal for Light and Electron
Optics, 2005. vol. 126, pp. 2743-2746.
M. Islam, M. Ali, M.-H. Lai, K.-S. Lim, and H. Ahmad, Chronology
of Fabry-Perot Interferometer Fiber-Optic Sensors and Their
Applications: A Review, Sensors, 2014. vol. 14, pp. 7451-7488.
J. Prakash Maran, S. Manikandan, K. Thirugnanasambandham, C.
Vigna Nivetha, and R. Dinesh,. Box-Behnken design based statistical
modeling for ultrasound-assisted extraction of corn silk
polysaccharide, Carbohydrate Polymers, 2013. vol. 92, pp. 604-611.
D. Montgomery C, Design and Analysis of Experiments, 5 ed. Tempe,
Arizona: John Wiley & Sons, Inc, 2007.
G. Srinu Babu, R. R. Shiva Kiran, N. Lokeswari, and K. Jaya Raju,.
Optimization of protease production from Aspergillus oryzae sp.
[6]
[7]
[8]
[9]
(486.13nm)
R*
U*
3537.78
3540.39
3582.20
3550.88
3454.13
3425.74
3242.91
3290.94
(656.03nm)
R*
U*
2846.70
2810.93
2902.08
2920.00
2797.97
2783.80
2605.50
2620.59
using Box-Behnken experimental design. E-Journal of Chemistry,
2007. vol. 4, pp. 145-153.
B. Wang and C. Q. Lan,. Optimising the lipid production of the green
alga Neochloris oleoabundans using box-behnken experimental
design, Canadian Journal of Chemical Engineering, 2011. vol. 89,
pp. 932-939.
M. Islam, M. Ali, M.-H. Lai, K.-S. Lim, and H. Ahmad,. Chronology
of Fabry-Perot Interferometer Fiber-Optic Sensors and Their
Applications: A Review, Sensors, 2014. vol. 14, p. 7451.
M. Manohar, J. Joseph, T. Selvaraj, and D. Sivakumar,. Application
Box-Behnken design to optimize the parameters for turning Inconel
718 using coated carbide tools. International Journal of Scientific &
Engineering Research, 2013. vol. 4, pp. 620-642.
P. Qiu, M. Cui, K. Kang, B. Park, Y. Son, E. Khim, M. Jang, and J.
Khim,. Application of Box-Behnken design with response surface
methodology for modeling and optimizing ultrasonic oxidation of
arsenite with H2O2, Central European Journal of Chemistry, 2014.
vol. 12, pp. 164-172.
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
Modeling and Optimization of Fiber Optic Chemical Vapor Sensor
[10] M. A. Tekindal, H. Bayrak, B. Ozkaya, and Y. Genic,. Box-Behnken
experimental design in factorial experiments: The importance of
bread for nutrition and health, Turkish Journal of Field Crops, 2012.
vol. 17, pp. 115-123.
[11] Z. He, X. T. Zhang, and G. Q. Xie, "Product quality improvement
through response surface methodology: A case study," Phuket,
Thailand, 2010. pp. S4,120-S4,130.
[12] D. Das, R. Thakur, and A. K. Pradhan,. Optimization of corona
discharge process using Box–Behnken design of experiments, Journal
of Electrostatics, 2012. vol. 70, pp. 469-473.
[13] S. L. C. Ferreira, R. E. Bruns, H. S. Ferreira, G. D. Matos, J. M.
David, G. C. Brandão, E. G. P. da Silva, L. A. Portugal, P. S. dos
Reis, A. S. Souza, and W. N. L. dos Santos, Box-Behnken design: An
alternative for the optimization of analytical methods. Analytica
Chimica Acta, 2007. vol. 597, pp. 179-186.
[14] B. Ladu Kishor, R. Swain, and R. B. Rao,. Response Surface
Methodology to Assess an Ecofriendly Approach for Utilization of
Berhampur City Municipal Solid Waste, International Journal of
Engineering and Innovative Technology, 2013. vol. 3, pp. 470-478.
[15] G. Shu, C. Dai, H. Chen, and X. Wang, "Application of Box Behnken design in optimization for crude polysaccharides from fruits
of Tribulus terristris L,. Journal of Chemical and Pharmaceutical
Research Research, 2013. vol. 5, pp. 342-350.
[16] P. Qiu, M. Cui, K. Kang, B. Park, Y. Son, E. Khim, M. Jang, and J.
Kim,. Application of Box - Behnken design with response surface
methodology for modelling and optimizing ultrasonic oxidation of
arsenite with H2O2, Central European Journal of Chemistry, 2014.
vol. 12, pp. 164-172.
[17] M. J. Anderson and P. J. Whitcomb, DOE Simplified practical tools
for effective experimentation, 1 ed. New York: Productivity Press,
2007.
[18] N. Aslan and Y. Cebeci,. Application of Box-Behnken design and
response surface methodology for modeling of some Turkish coals,
Fuel, 2007. vol. 86, pp. 90-97.
[19] A. Kumar, B. Prasad, and I. M. Mishra,. Optimization of process
parameters for acrylonitrile removal by a low-cost adsorbent using
Box-Behnken design, Journal of Hazardous Materials, 2008. vol.
150, pp. 174-182.
[20] P. Qiu, M. Cui, K. Kang, B. Park, Y. Son, E. Khim, M. Jang, and J.
Khim,. Application of Box–Behnken design with response surface
methodology for modeling and optimizing ultrasonic oxidation of
arsenite with H2O2," Central European Journal of Chemistry, 2014.
vol. 12, pp. 164-172.
[21] M. Manohar, T. Selvaraj, D. Sivakumar, R. Jeyapaul, and J. Jomy,.
Application of Experimental Design and Analysis of Mathematical
Models for Turning Inconel 718 Using Coated Carbide Tools,
Experimental Techniques, 2012. pp. 1-11.
[22] T. F. Zhang, J. F. Yang, and D. K. J. Lin,. Small Box-Behnken
design, Statistics and Probability Letters, 2011. vol. 81, pp. 10271033.
[23] L. Wang, M. Cao, Q. Liu, G. Wei, B. Li, and C. Lin,. Modeling and
experimental verification of polarization errors in Sagnac fiber optic
current sensor, Optik - International Journal for Light and Electron
Optics, 2015. vol. 126, pp. 2743-2746.
ISSN: 2180-1843 e-ISSN: 2289-8131 Vol. 9 No. 2
79