Study on the Effect of Preparation Parameters of K2CO3/Al2O3 Sorbent on CO2 Capture Capacity at Flue Gas Operating Conditions

doi:10.4236/jeas.2013.32007

Paper Menu >>

Journal Menu >>

Journal of Encapsulation and Adsorption Sciences, 2013, 3, 57-63

http://dx.doi.org/10.4236/jeas.2013.32007 Published Online June 2013 (http://www.scirp.org/journal/jeas)

Study on the Effect of Preparation Parameters of

K2CO3/Al2O3 Sorbent on CO2 Capture Capacity at Flue

Gas Operating Conditions

Javad Esmaili, Mohammad Reza Ehsani

Department of Chemical Engineering, Isfahan University of Technology, Isfahan, Iran

Email: javad [email protected], [email protected]

Received February 6, 2013; revised March 6, 2013; accepted March 14, 2013

Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is

properly cited.

ABSTRACT

In this paper, study on the effect of preparation conditions of K2CO3/Al2O3 sorbent was done. Box-Behnken design was

applied to study the influence of four parameters involve initial solution concentration, impregnation time and calcina-

tion step temperature and time. A quadratic model was used to correlate the sorbent capture capacity. The model was

used to calculate the optimum conditions for preparing sorbent. From the analysis of variance (ANOVA), the most in-

fluential factor on each experimental design response was identified. The potassium-based sorbents used in this study

were prepared by impregnating K2CO3 on Al2O3 support. The CO2 capture capacity was measured in the presence of

H2O in a fixed-bed reactor at CO2 capture temperature of 60˚C using breakthrough curves. The optimum sorbent pre-

pared by this method showed CO2 capture capacity of 77.21 mg CO2/g sorbent. It was observed that the experimental

values obtained were in good agreement with the values predicted by the model, with relatively small errors between

the predicted and the actual values. The results obtained in this study can be used as basic data for study on design and

operating condition optimization of CO2 capture process using these sorbents.

Keywords: CO2 Capture; K2CO3/Al2O3 Sorbent; Response Surface Method; Box-Behnken Design

1. Introduction

Global warming increasingly thought to be associated

with the atmospheric emission of greenhouse gases. Car-

bon Dioxide is a major greenhouse gas that is released

into the atmosphere due to the use of fossil fuels. It can

be removed from flue gas by various methods such as

membrane separation, amine scrubbing, and using mole-

cular sieves [1-3]. These methods however are costly and

energy intensive.

Solid sorbent processes for CO2 capture are also under

study. The use of solid sorbents can be a highly cost-effec-

tive and energy efficient way to remove CO2 [4-7]. Sor-

bents containing alkali and alkaline-earth metals like po-

tassium carbonate and calcium oxide are investigated for

this commercial feasibility [8].

Alkaline-earth method-based sorbents such as MgO

and CaO are applicable at much higher adsorption and

regeneration temperatures [7]. Alkali metal-based sor-

bents were employed in CO2 adsorption at low tempera-

tures (50˚C - 70˚C) with thermal regeneration easily oc-

curring at low temperatures (<200˚C). CO2 capture using

a dry sodium-based sorbent was also reported [9,10]. How-

ever, when CO2 reacted with Na2CO3, the global carbona-

tion reaction rate was rather slow [11]. The CO2 adsorp-

tion and regeneration of potassium-based sorbents with

several supports such as activated carbon [4,5,12], TiO2,

SiO2, MgO, ZrO2, CaO and Al2O3 were studied [6,9,13,

14]. Al2O3 is one of the most proper materials used as

sorbent support.

In most of the previous works, the sorbent prepara-

tion conditions were fixed and the effect of the different

conditions on the capture capacity which is an important

specification of the sorbent for commercial applications,

has not been studied yet. The objective of this work was

to study the influence of preparation condition on sorbent

capture capacity and finding optimum condition for pre-

paration of sorbent at larger scales. Therefore, the scope

of this research was to carry out a statistical optimization

to determine the optimum preparation conditions of K2CO3/

Al2O3 sorbent, in order to achieve high CO2 capture ca-

pacity using the response surface methodology approach

J. ESMAILI, M. R. EHSANI

(RSM). Response surface methodology (RSM) is an em-

pirical statistical technique employed for multiple regres-

sion analysis by using quantitative data. It solves multi-

variate data which is obtained from properly designed

experiments to solve multivariate equation simultaneous-

ly [15]. This method was applied by many of researcher

for study on various multi-variable phenomena [16,17].

2. Experimental

2.1. Preparation of Sorbent

The potassium-based sorbents used in this study were

prepared by impregnating K2CO3 on Al2O3 (Aluminum

Oxide, Merck) as support. Twenty (20.0) g of Al2O3 was

added to an aqueous solution of anhydrous potassium

carbonate (K2CO3, Merck) with particular concentration

in deionized water (Initial solution concentration). Then,

it was mixed with a magnetic stirrer at room temperature

(Impregnation time). After that, the mixture was dried in

a rotary vacuum evaporator at 60˚C. The dried samples

were calcined in a furnace under a N2 flow (100 cc/min)

for a particular time at selected temperature (Calcination

time and temperature).

2.2. Apparatus and Procedures

A fixed-bed Stainless-Steel reactor (diameter of 15 mm),

which was placed in an electric furnace under atmos-

pheric pressure was used for adsorption process. Two

(2.0) g of the sorbent was packed into the reactor. In or-

der to prevent condensation of water vapor injected into

the reactor and GC column the temperatures of the inlet

and outlet lines of the reactor were maintained above

100˚C. The column used in this analysis was a 1/8 in.

stainless tube packed with Porapak Q. When the CO2

concentration of the outlet gas reached the same level as

that of the inlet gas in the CO2 adsorption process, proc-

ess was stopped. The outlet gas from the reactor was au-

tomatically analyzed every 5 min by a thermal conduc-

tivity detector (TCD), which was equipped with an auto

sampler (Valco online valve). Schematic diagram of the

fixed-bed reactor is shown in Figure 1.

The feed stream comprises of Nitrogen, carbon diox-

ide and water. The liquid water flow rate was controlled

using a piston pump and the water was vaporized before

entering the column. Both CO2 and N2 flow rates were

controlled by independent mass flow controllers and

Figure 1. Schematic diagram of the fixed-bed reactor.

J. ESMAILI, M. R. EHSANI 59

these gases were mixed with the vaporized water inside

the oven where experiments were done. All the CO2 sor-

ption tests were performed following the same proce-

dure.

The CO2 capture capacity of the sorbent was evaluated

by the breakthrough curve for CO2 adsorption. In this

study, the CO2 capture capacity of sorbent was calculated

from its breakthrough curve during CO2 adsorption in the

presence of 9.0 vol.% H2O and 1 vol.% CO2. The CO2

capture capacity describes the amount of CO2 absorbed

until the output concentration of CO2 reached 1 vol.%,

which was is the same value as that of the inlet.

2.3. Design of Experiments

The RSM has several classes of designs, with its own

properties and characteristics. Central composite design

(CCD), Box-Behnken design (BBD) and three-level fac-

torial design are the most popular designs applied by the

researchers. In this work, the Box-Behnken design was

used to study the effects of the variables towards their

responses and subsequently in the optimization studies.

This method is suitable for fitting a quadratic surface and

it helps to optimize the effective parameters with a mi-

nimum number of experiments, as well as to analyze the

interaction between the parameters. In order to determine

if there exist a relationship between the factors and the

response variables investigated, the data collected must

be analyzed in a statistically manner using regression. A

regression design is normally employed to model a re-

sponse as a mathematical function (either known or em-

pirical) of a few continuous factors and good model para-

meter estimates are desired [18].

Replicates of the test at the center are very important

as they provide an independent estimation of the experi-

mental error. Each variable is investigated at two levels.

Meanwhile, as the number of factors, n, increases, the

number of runs for a complete replicate of the design in-

creases rapidly. In this case, main effects and interactions

may be estimated by fractional factorial designs running

only a minimum number of experiments. Individual sec-

ond-order effects cannot be estimated separately by 2n

factorial designs. The responses and the corresponding

parameters are modeled and optimized using ANOVA to

estimate the statistical parameters by means of response

surface methods.

Basically this optimization process involves three ma-

jor steps, which are, performing the statistically designed

experiments, estimating the coefficients in a mathemati-

cal model and predicting the response and checking the

adequacy of the model.

It is assumed that the independent variables are con-

tinuous and controllable by experiments with negligible

errors. It is required to find a suitable approximation for

the true functional relationship between independent va-

riables and the response surface [19].

The experimental sequence was randomized in order

to minimize the effects of the uncontrolled factors. Re-

gression analysis was performed to determine the surface

response as function of second order polynomial equa-

tion:

1=1

=+ ++

kk k

i iii iiij ij

ii iij

 





 

x (1)

where Q is the predicted response (sorbent capture ca-

pacity), βi, βii, βij represent linear, quadratic and interac-

tion effects. β0 is the intercept term and xi, xj, …, xk are

the input variables which affect the Q value [20].

In this work the influence of four sorbent preparation

parameters involve initial solution concentration, impreg-

nation time and calcination step temperature and time on

sorbent CO2 capture capacity has been studied. The range

of these variables selected base on previous works of

other researchers [6,7,13,14] (Table 1).

3. Results and Discussion

3.1. Development of Regression Model Equation

The design of experiment is given in Table 2, together

with the experimental results. Eight replicate runs were

observed at the centre of the design to allow the estima-

tion pure error.

Using multiple regression analysis, the second order po-

lynomial model (Equation (1)), which characterizes the

relationship between sorbent capture capacity and studi-

ed variables, was obtained. The coefficients of polyno-

mial are shown in Table 3.

Response surface plots were described by the regres-

sion model for BBD which was developed using MINI-

TAB 14 software. The student’s t-test was performed to

determine the significance of the regression coefficients.

The results of statistical analysis including the regression

coefficient, t and p values for linear, quadratic and com-

bined effects of the variables are given in Table 3. The

statistical significance of the model was also determined

by F-test for analysis of variance (ANOVA) and re-

Table 1. Experimental range and le vels of indepe ndent var i-

ables for preparation of sorbents.

Code levels

Independent

variables SymbolsUnit −1 0 1

Initial solution

concentration ISC wt% 10 25 40

Impregnation time ITI hr 1 12 23

Calcination time CTI hr 3 4 5

Calcination

temperature CTE ˚C 300 400 500

J. ESMAILI, M. R. EHSANI

Table 2. Box-Behnken design matrix, the experimental and

predicted values of sorption capture capacity.

Sorbent preparation variables

(

Code levels

)

Sorbent capture capacity (mg/g)

Run

ISC ITI CTI CTE Experimental Predicted

1 1 0 −1 0 71.61 75.49

2 0 −1 0 1 69.63 64.98

3 0 0 0 0 74.25 74.19

4 −1 0 0 −1 52.89 55.07

5 0 1 0 1 72.19 68.32

6 1 0 1 0 79.09 76.66

7 1 0 0 −1 74.14 77.37

8 1 −1 0 0 73.19 76.84

9 −1 0 1 0 55.31 54.72

10 0 0 1 1 64.89 65.81

11 −1 1 0 0 54.97 56.99

12 1 0 0 1 71.34 71.67

13 0 1 −1 0 69.87 71.49

14 −1 0 −1 0 49.67 52.82

15 0 0 0 0 71.26 74.19

16 1 1 0 0 75.35 78.83

17 −1 0 0 1 52.69 49.37

18 −1 −1 0 0 53.19 54.07

19 0 1 1 0 75.51 73.06

20 0 −1 0 −1 69.46 71.56

21 0 0 −1 1 66.45 63.96

22 0 1 0 −1 70.04 73.14

23 0 0 1 −1 70.45 71.19

24 0 −1 1 0 72.51 70.57

25 0 0 0 0 76.07 74.19

26 0 0 −1 −1 66.85 69.98

27 0 −1 −1 0 68.71 69.07

Table 3. Estimated regression coefficients of second order

polynomial model for optimization of sorbent preparation

(R2 = 0.971).

Coefficient Estimated

coefficient t-Value p-Value

β0 −57.431 −2.553 0.025

β1 (ISC) 2.3232 5.599 0.000

β2 (ITI) 0.1163 −0.290 0.777

β3 (CTI) 20.014 2.717 0.019

β4 (CTE) 0.2803 3.757 0.003

β11 (ISC2) −0.0303 −8.900 0.000

β22 (ITI2) −0.0057 −0.566 0.582

β33 (CTI2) −2.4512 −2.432 0.032

β44 (CTE2) −0.0004 −3.781 0.003

β12 (ISC*ITI) −0.0014 0.090 0.930

β13 (ISC*CTI) −0.0120 0.434 0.672

β14 (ISC*CTE) 0.0000 −0.613 0.551

β23 (ITI*CTI) 0.0018 0.434 0.672

β24 (ITI*CTE) 0.0004 0.467 0.649

β34 (CTI*CTE) 0.0016 −1.217 0.247

siduals analysis was performed to validate the model at

97% of confidence level. The model fitted well with

amylase activity and the optimal values from the model

was justified (p = 0.000). The ANOVA given in Table 4

indicates that the linear, quadratic terms in second order

polynomial Model were highly significant (p < 0.001)

and adequate to represent the relationship between sor-

bent capture capacity (mg/g) and initial solution concen-

tration, impregnation time and calcination step temperature

and time.

Figures 2 and 3 show the three-dimensional response

surface which were constructed to show the interaction

effect of the sorbent preparation variables on the sorbent

capture capacity.

Table 4. Analysis of variance (ANOVA) of second order po-

lynomial model for optimization of sorbent capture capaci-

ty.

Source Degree of

freedom

Sum of

square

Mean of

square F-value P-value

Regression 14 1794.5 128.182 28.54<0.001

Linear 4 1387.6 43.678 9.73 <0.001

Square 4 395.87 98.967 22.04<0.001

Interaction 6 11.06 1.843 0.41 0.858

Residual error12 53.89 4.491

Lack-of-Fit 10 42.09 4.209 0.71 0.709

Pure error 2 11.80 5.898

Total 26 1848.4

Figure 2. The combined effect of initial solution concentra-

tion and impregnation time on sorbent capture capacity

(CTE = 400˚C, CTI = 4 hr).

Figure 3. The combined effect of calcination temperature

and calcination time on sorbent capture capacity (ISC =

25%, ITI = 12 hr).

J. ESMAILI, M. R. EHSANI 61

As can be seen in Figure 2, the sorbent capacity gen-

erally increases with increase in initial solution concen-

tration and impregnation time. This trend is the result of

increase of potassium carbonate loading on sorbent with

the increase of these parameters; but the rate of increas-

ing is slow at higher values since the K2CO3 loading is

limited. The initial solution concentration is more effec-

tive than impregnation time; it shows that the rate of im-

pregnation of K2CO3 on alumina support is high.

As can be seen in Figure 3, the sorbent capacity has a

maximum point with increase in calcination temperature

and time and after this point it decreases. This trend is

the result of improvement of sorbent structure in calcina-

tion step that increases the sorbent capture capacity. At

calcination step, the higher temperature and longer time

damage some formed structures and decreases the sor-

bent capacity.

3.2. Process Optimization

In the production of commercial sorbent, relatively high

sorbent capture capacity are expected. Therefore, in order

to optimize preparation condition, Minitab 14 is used.

The sorbent was prepared under the experimental condi-

tions given in Table 5, together with the predicted and

experimental values for sorbent capture capacity. The

optimum sorbent prepared was obtained by using initial

solution concentration of 32.3 wt%, impregnation time of

13.4 hr, calcination temperature of 367˚C and calcination

time of 4.1 hr. The optimum sorbent showed capture ca-

pacity of 77.21 mg CO2/g sorbent. It was observed that

the experimental values obtained were in good agreement

with the values predicted from the model, with relatively

small errors between the predicted and the actual values

(Table 5).

Figure 4 shows the variation of sorbent capture capac-

ity with the parameters at the optimum conditions. For

the two parameters involve calcination time and tempe-

rature, the optimum point locate at the curve maximum

point. The overall trend of four curves represents that the

initial solution concentration is the most effective pa-

rameters on the final sorbent capture capacity.

Figure 5, shows the breakthrough curves of the sor-

bent prepared at the optimum conditions during CO2 ad-

sorption in the presence of 9.0 vol.% H2O and 1 vol.%

CO2 at 60˚C. It was observed that the breakthrough time

was 16 min. Further study is required to enhance the CO2

adsorption rate.

Table 5. Model validation.

Sorbent capture

capacity (mg/g)

Initial

solution

concentration

(wt%)

Impregnation

time (hr)

Calcinations

temperature

(˚C)

Calcination

time (hr) Predicte

Experimental

32.3 13.4 367.0 4.1 78.66 77.21

(a)

(b)

(c)

(d)

Figure 4. The effect of four parameters on sorbent capture

capacity at optimum conditions (× symbol shows optimum

point for each curve).

J. ESMAILI, M. R. EHSANI

Figure 5. The breakthrough curves of the sorbent prepare

at optimum conditions during CO2 adsorption.

4. Conclusions

The sorbent capacity for carbon dioxide capture from a

gas stream is important for the industrial application of

solid sorbents. The present investigation was carried out

to study combined effects of the initial solution concen-

tration, impregnation time, calcination temperature, and

calcination time on the sorbent capture capacity using a

Box-Behnken design under the Response surface metho-

dology (RSM). The obtained results demonstrate that

sorbent capacity increased with increasing the initial so-

lution concentration and impregnation time. Sorbent ca-

pacity has a maximum point for variation of calcination

temperature and time and further increases of these vari-

ables lead to decrease of sorbent capacity. However,

ANOVA analysis as well as 3D surface plots revealed

that initial solution concentration has the greatest effect

on sorbent capacity. On the basis of the results it can be

concluded that RSM presents an excellent tool which

enables the evaluation of interactions and competitive

effects in multivariable systems and reduces the number

of needed experiments in contrast to the classical method

of changing one variable at a time.

The optimized values obtained for initial solution con-

centration, impregnation time and calcination step tempe-

rature and time were 32.3 wt%, 13.4 hr, 367˚C and 4.1 hr,

respectively, with the predicted maximized response of

sorbent capture capacity (78.66 mg CO2/g sorbent). Stu-

dy on other sorbent characterization such as mechanical

strength require for application of this procedure for pre-

paration of industrial sorbents.

REFERENCES

[1] P. Hagewiesche, S. S. Ashour, H. A. Al-Ghawas and O. C.

Sandall, “Absorption of Carbon Dioxide into Aqueous

Blends of Monoethanolamine and N-Methyldiethanola-

mine,” Chemical Engineering Science, Vol. 50, No. 7,

1995, pp. 1071-1079.

doi:10.1016/0009-2509(94)00489-E

[2] M. Mavroudi, S. P. Kaldis and G. P. Sakellaropoulos,

“Reduction of CO2 Emissions by a Membrane Contacting

Process,” Fuel, Vol. 82, No. 15-17, 2003, pp. 2153-2159.

doi:10.1016/S0016-2361(03)00154-6

[3] R. V. Siriwardane, M. S. Shen, E. P. Fisher and J. A. Po-

ston, “Adsorption of CO2 on Molecular Sieves and Acti-

vated Carbon,” Energy & Fuels, Vol. 15, No. 2, 2001, pp.

279-284. doi:10.1021/ef000241s

[4] H. Hayashi, J. Taniuchi, N. Furuyashiki, S. Sugiyama, S.

Hirano, N. Shigemoto and T. Nonaka, “Efficient Recov-

ery of Carbon Dioxide from Flue Gases of Coal-Fired Po-

wer Plants by Cyclic Fixed-Bed Operations over K2CO3-

on-Carbon,” Industrial and Engineering Chemistry Re-

search, Vol. 37, No. 1, 1998, pp. 185-191.

doi:10.1021/ie9704455

[5] N. Shigemoto, T. Yanagihara, S. Sugiyama and H. Haya-

shi, “Material Balance and Energy Consumption for CO2

Recovery from Moist Flue Gas Employing K2CO3-on-

Activated Carbon and Its Evaluation for Practical Adapta-

tion,” Energy & Fuels, Vol. 20, No. 2, 2006, pp. 721-726.

doi:10.1021/ef058027x

[6] S. C. Lee, B. Y. Choi, S. J. Lee, S. Y. Jung, C. K. Ryu

and J. C. Kim, “CO2 Absorption and Regeneration Using

Na and K Based Sorbents,” Studies in Surface Science

and Catalysis, Vol. 153, No. 1, 2004, pp. 527-530.

doi:10.1016/S0167-2991(04)80307-0

[7] S. C. Lee, B. Y. Choi, C. K. Ryu, Y. S. Ahn and J. C. Kim,

“Absorption and Regeneration of Alkali Metal-Based So-

lid Sorbents,” Catalysis Today, Vol. 111, No. 3-4, 2006,

pp. 385-390. doi:10.1016/j.cattod.2005.10.051

[8] J. S. Hoffman and H. W. Pennline, “Study of Regenerable

Sorbents for CO2 Capture, First National Conference on

Carbon Sequestration, Washington,” 2001.

[9] Y. Liang, D. P. Harrison, R. P. Gupta, D. A. Green and W.

J. McMichael, “Carbon Dioxide Capture Using Dry So-

dium-Based Sorbents,” Energy & Fuels, Vol. 18, No. 2,

2004, pp. 569-575. doi:10.1021/ef030158f

[10] Y. W. Seo, S. H. Jo, C. K. Ryu and C. K. Yi, “Effects of

Water Vapor Pretreatment Time and Reaction Tempera-

ture on CO2 Capture Characteristics of a Sodium-Based

Solid Sorbent in a Bubbling Fluidized-Bed Reactor,”

Chemosphere, Vol. 69, No. 5, 2007, pp. 712-718.

doi:10.1016/j.chemosphere.2007.05.036

[11] C. Zhao, X. Chen and C. Zhao, “CO2 Absorption Using

Dry Potassium-Based Sorbents with Different Supports,”

Energy & Fuels, Vol. 23, No. 9, 2009, pp. 4683-4687.

doi:10.1021/ef900182d

[12] S. Hirano, N. Shigemoto, S. Yamaha and H. Hayashi, Cy-

clic Fixed-Bed Operations over K2CO3-on-Carbon for the

Recovery of Carbon Dioxide under Moist Conditions,”

Bulletin of the Chemical Society of Japan, Vol. 68, No. 3,

1995, pp. 1030-1035. doi:10.1246/bcsj.68.1030

[13] A. G. Okunev, V. E. haronov, Y. I. Aistov and V. N. Parmon,

“Sorption of Carbon Dioxide from Wet Gases by K2CO3-

in-Porous Matrix: Influence of the Matrix Nature,” Reac-

tion Kinetics and Catalysis Letters, Vol. 71, No. 2, 2000,

pp. 355-362. doi:10.1023/A:1010395630719

[14] A. G. Okunev, V. E. Sharonov, A. V. Gubar, I. G. Danilo-

va, E. A. Paukshtis, E. M. Moroz, T. A. Kriger, V. V. Ma-

lakhov and Y. I. Aistov, “Sorption of Carbon Dioxide by

J. ESMAILI, M. R. EHSANI

the Composite Sorbent: Potassium Carbonate in Porous

Matrix,” Russian Chemical Bulletin, Vol. 52, No. 2, 2003,

pp. 359-363.

[15] I. A. W. Tan, A. L. Ahmad and B. H. Hameed, “Optimiza-

tion of Preparation Conditions for Activated Carbons

from Coconut Husk Using Response Surface Methodolo-

gy,” Chemical Engineering Journal, Vol. 137, No. 3,

2008, pp. 462-470. doi:10.1016/j.cej.2007.04.031

[16] M. Tir and M. Moulai-Mostefa, “Optimization of Oil Re-

moval from Oily Wastewater by Electrocoagulation Us-

ing Response Surface Method,” Journal of Hazardous

Materials, Vol. 158, No. 1, 2008, pp. 107-115.

doi:10.1016/j.jhazmat.2008.01.051

[17] J. N. Sahu, J. Acharya and B. C. Meikap, “Response Sur-

face Modeling and Optimization of Chromium(VI) Re-

moval from Aqueous Solution Using Tamarind Wood

Activated Carbon in Batch Process,” Journal of Hazard-

ous Materials, Vol. 172, No. 2-3, 2009, pp. 818-825.

doi:10.1016/j.jhazmat.2009.07.075

[18] A. A. Ahmad, B. H. Hameed and A. L. Ahmad, “Remo-

val of Disperse Dye from Aqueous Solution Using Waste-

Derived Activated Carbon: Optimization Study,” Journal

of Hazardous Materials, Vol. 170, No. 2-3, 2009, pp.

612-619. doi:10.1016/j.jhazmat.2009.05.021

[19] V. Gunaraj and N. Murugan, “Application of Response Sur-

face Methodologies for Predicting Weld Base Quality in

Submerged Arc Welding of Pipes,” Journal of Materials

Processing Technology, Vol. 88, No. 1-3, 1999, pp. 266-

275. doi:10.1016/S0924-0136(98)00405-1

[20] R. H. Myers and D. C. Montgomery, “Response Surface Me-

thodology: Process and Product Optimization Using De-

signed Experiments,” John Wiley and Sons, New York,

2002.