Application of response surface methodology (RSM) for optimization of the sintering process of preparation calcia partially stabilized zirconia (CaO-PSZ) using natural baddeleyite Jing Li, Jinhui Peng ⇑ , Shenghui Guo, Libo Zhang Faculty of Metallurgy and Energy Engineering, Kunming University of Science and Technology, Kunming, Yunnan 650093, PR China Key Laboratory of Unconventional Metallurgy, Ministry of Education, Kunming University of Science and Technology, Kunming, Yunnan 650093, PR China Engineering Laboratory of Microwave Energy Application and Equipment Technology, Kunming, Yunnan 650093, PR China State Key Laboratory Breeding Base of Complex Nonferrous Metal Resources Cleaning Utilization, Yunnan Province 650093, PR China article info Article history: Received 11 December 2012 Received in revised form 16 May 2013 Accepted 16 May 2013 Available online 23 May 2013 Keywords: Stabilizer Response surface methodology Sintering Natural baddeleyite abstract Response surface methodology (RSM) was successfully applied to process of preparation calcia partially stabilized zirconia (CaO-PSZ). Besides that, natural baddeleyite was used as starting materials instead of chemical pure zirconia. The pressureless sintering process was optimized by the application of RSM. The independent variables, which had been found as the most effective variables on the relative density and bending strength by screening experiments, were determined as holding time, sintering temperature and heating rate. Two quadratic models were developed through RSM in terms of related independent vari- ables to describe the relative density and bending strength as the responses. Based on contour plots and variance analysis, optimum operational conditions for maximizing relative density and bending strength, at cooling rate of 3 °C/min, were 1540 °C of sintering temperature, 5 h of holding time and heating rate of 3 °C/min to obtain 98.57% for relative density and 165.72 MPa for bending strength. Crown Copyright Ó 2013 Published by Elsevier B.V. All rights reserved. 1. Introduction High-purity zirconia (ZrO 2 ) is a white crystalline powder. It exhibits three polymorphs depending on temperature. At room temperature pure zirconia can only be monoclinic [1,2]. As the temperature increase, monoclinic turns into a tetragonal phase at a certain temperature. Further increase in temperature leads to conversion of the tetragonal phase into cubic phase again, and it is a reversible process. The shift between tetragonal and monoclinic phase is of high interest in the transition of zirconia, and it is called martensitic transformation [3]. In view of pure ZrO 2 phase transformation characteristics, especially the volume expansion associated with transformation of tetragonal phase to monoclinic phase cannot be accommodated by zirconia grains, resulting in cracking of ZrO 2 material, unless some specific processes are used [4–6]. Otherwise, it would be difficult to prepare pure phase ZrO 2 mate- rial. In order to stabilize the zirconia phase, and ensure the exis- tence of high-temperature phase at room temperature, it is common to use stabilizer. Common zirconia stabilizers are rare earth or alkaline-earth oxides. It is possible to stabilize the ZrO 2 in the tetragonal and/or cubic forms at room temperature [7], by adding different stabiliz- ers, such as, MgO [8–11], CaO [12],Y 2 O 3 [13–15], CeO 2 [16,17], Al 2 O 3 [18] and even a combination of them [19,20]. Different quan- tities of stabilizer can cause zirconia to stabilize in different phase composition. If only part of the t-ZrO 2 metastable to room temper- ature, partially stabilized zirconia (PSZ) is formed. At present, the raw materials which are used to prepare PSZ ceramic are usually chemical pure or industrial pure. In this work, instead of chemical pure zirconia, natural baddeleyite was used as starting materials to prepare partially stabilized zirconia (PSZ) with calcia as stabilizer. Therefore, it not only decreased the cost, also reduced the social energy consumption and environmental pollu- tion [21]. The production process of partially stabilized zirconia using natural baddeleyite, stabilized by the addition of calcia, has been optimized. The natural baddeleyite was obtained after the flo- tation of baddeleyite ore. The traditional one-factor-at-a-time approach to optimization is time-consuming and incapable of reaching a true optimum be- cause of taking no account of interaction among factors. On the contrary, statistical methods can take into account the interaction of variables in generating the process response. Therefore, a statis- tically designed experiment with minimum experimental runs is greatly desired. Response surface methods (RSM) consist of a group 0925-8388/$ - see front matter Crown Copyright Ó 2013 Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jallcom.2013.05.102 ⇑ Corresponding author at: Faculty of Metallurgy and Energy Engineering, Kunming University of Science and Technology, Kunming, Yunnan 650093, PR China. Tel.: +86 871 5191046; fax: +86 871 5138997. E-mail address: [email protected](J. Peng). Journal of Alloys and Compounds 574 (2013) 504–511 Contents lists available at SciVerse ScienceDirect Journal of Alloys and Compounds journal homepage: www.elsevier.com/locate/jalcom
Transcript
Journal of Alloys and Compounds 574 (2013) 504–511
Contents lists available at SciVerse ScienceDirect
Journal of Alloys and Compounds
journal homepage: www.elsevier .com/locate / ja lcom
Application of response surface methodology (RSM) for optimizationof the sintering process of preparation calcia partially stabilized zirconia(CaO-PSZ) using natural baddeleyite
0925-8388/$ - see front matter Crown Copyright � 2013 Published by Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jallcom.2013.05.102
⇑ Corresponding author at: Faculty of Metallurgy and Energy Engineering,Kunming University of Science and Technology, Kunming, Yunnan 650093, PRChina. Tel.: +86 871 5191046; fax: +86 871 5138997.
Jing Li, Jinhui Peng ⇑, Shenghui Guo, Libo ZhangFaculty of Metallurgy and Energy Engineering, Kunming University of Science and Technology, Kunming, Yunnan 650093, PR ChinaKey Laboratory of Unconventional Metallurgy, Ministry of Education, Kunming University of Science and Technology, Kunming, Yunnan 650093, PR ChinaEngineering Laboratory of Microwave Energy Application and Equipment Technology, Kunming, Yunnan 650093, PR ChinaState Key Laboratory Breeding Base of Complex Nonferrous Metal Resources Cleaning Utilization, Yunnan Province 650093, PR China
a r t i c l e i n f o a b s t r a c t
Article history:Received 11 December 2012Received in revised form 16 May 2013Accepted 16 May 2013Available online 23 May 2013
Response surface methodology (RSM) was successfully applied to process of preparation calcia partiallystabilized zirconia (CaO-PSZ). Besides that, natural baddeleyite was used as starting materials instead ofchemical pure zirconia. The pressureless sintering process was optimized by the application of RSM. Theindependent variables, which had been found as the most effective variables on the relative density andbending strength by screening experiments, were determined as holding time, sintering temperature andheating rate. Two quadratic models were developed through RSM in terms of related independent vari-ables to describe the relative density and bending strength as the responses. Based on contour plots andvariance analysis, optimum operational conditions for maximizing relative density and bending strength,at cooling rate of 3 �C/min, were 1540 �C of sintering temperature, 5 h of holding time and heating rate of3 �C/min to obtain 98.57% for relative density and 165.72 MPa for bending strength.
Crown Copyright � 2013 Published by Elsevier B.V. All rights reserved.
1. Introduction
High-purity zirconia (ZrO2) is a white crystalline powder. Itexhibits three polymorphs depending on temperature. At roomtemperature pure zirconia can only be monoclinic [1,2]. As thetemperature increase, monoclinic turns into a tetragonal phase ata certain temperature. Further increase in temperature leads toconversion of the tetragonal phase into cubic phase again, and itis a reversible process.
The shift between tetragonal and monoclinic phase is of highinterest in the transition of zirconia, and it is called martensitictransformation [3]. In view of pure ZrO2 phase transformationcharacteristics, especially the volume expansion associated withtransformation of tetragonal phase to monoclinic phase cannotbe accommodated by zirconia grains, resulting in cracking ofZrO2 material, unless some specific processes are used [4–6].Otherwise, it would be difficult to prepare pure phase ZrO2 mate-rial. In order to stabilize the zirconia phase, and ensure the exis-tence of high-temperature phase at room temperature, it iscommon to use stabilizer.
Common zirconia stabilizers are rare earth or alkaline-earthoxides. It is possible to stabilize the ZrO2 in the tetragonal and/orcubic forms at room temperature [7], by adding different stabiliz-ers, such as, MgO [8–11], CaO [12], Y2O3 [13–15], CeO2 [16,17],Al2O3 [18] and even a combination of them [19,20]. Different quan-tities of stabilizer can cause zirconia to stabilize in different phasecomposition. If only part of the t-ZrO2 metastable to room temper-ature, partially stabilized zirconia (PSZ) is formed.
At present, the raw materials which are used to prepare PSZceramic are usually chemical pure or industrial pure. In this work,instead of chemical pure zirconia, natural baddeleyite was used asstarting materials to prepare partially stabilized zirconia (PSZ) withcalcia as stabilizer. Therefore, it not only decreased the cost, alsoreduced the social energy consumption and environmental pollu-tion [21]. The production process of partially stabilized zirconiausing natural baddeleyite, stabilized by the addition of calcia, hasbeen optimized. The natural baddeleyite was obtained after the flo-tation of baddeleyite ore.
The traditional one-factor-at-a-time approach to optimizationis time-consuming and incapable of reaching a true optimum be-cause of taking no account of interaction among factors. On thecontrary, statistical methods can take into account the interactionof variables in generating the process response. Therefore, a statis-tically designed experiment with minimum experimental runs isgreatly desired. Response surface methods (RSM) consist of a group
Table 1Design matrix and responses (variables: A – sintering temperature (�C); B –holding time (h); C – heating rate (�C/min); responses: Y1 – relative density (%); Y2 – bending strength(MPa)).
J. Li et al. / Journal of Alloys and Compounds 574 (2013) 504–511 505
of empirical techniques devoted to the evolution of relations exist-ing between a cluster of controlled experimental factors and themeasured responses, useful for developing, improving and opti-mising processes by carrying out a limited number of experiments.In this study, ‘central composite design’ (CCD), was applied.
2. Experimental procedure and design
2.1. Experimental procedure
Instead of chemical pure zirconia, the natural baddeleyite (monoclinic ZrO2)which was obtained by flotation of baddeleyite ore (the ZrO2 content is not lessthan 99.30%) with the particle size of 152 lm and high purity CaO were selectedas the starting powders. Previously, the natural baddeleyite powder was milled toaverage particle size of 5.67 lm using crusher. The experimented composition con-sisted of 96.2 wt% ZrO2 (natural baddeleyite, 5.67 lm) with additions of 3.8 wt% ofCaO. The powders were rigorously wet mixed by means of planetary milling using
agate ball (ball-feed weight radio of 4–6:1) in ethanol for 12 h (average particle sizeof 1.92 lm), and oven-dried at 80 �C for 10 h. The dried mixture blended with adetermined quantity of binder, and was uniaxially pressed at 150 MPa by single ac-tion at a constant strain for 8 min. The size of samples was 3 mm � 4 mm � 40 mmand U15 mm � 3 mm, respectively.
Subsequently, the samples were heat treated at 850 �C in air at a heating rate of5 �C/min during 4 h in order to burn out the binder. After the discharge treatment,the furnace was allowed to cool at the cooling rate of 5 �C/min. Finally, the sampleswere placed into a corundum board with ZrO2 bed powder and sintered at the de-sign experiment condition. The bulk densities of the sintered ceramics were deter-mined geometrically and by Archimedes principle with distilled water as themedium. The bending strength of bars was measured by three point bending. Inthe tests crosshead speed was 0.5 mm/min. Five measurements were averaged tominimize the error.
2.2. Experimental design
Response surface method (RSM) can describe the relationship between factorsand the response accurately through using reasonable experiment design and fit-ting the relationship between multi-factor experiment factors and the level withthe polynomial. RSM is a solution to the problem of multi-variable statistical meth-ods. RSM usually contains three steps: (1) design and experiments; (2) responsesurface modelling through regression; and (3) optimization.
Central composite design (CCD) is a kind of RSM. It is a test design methoddeveloped on the basis of a two level full factorial and partial experimental design.It is well suited for fitting a quadratic surface, which usually works well for processoptimization. It is composed of a core factorial that forms a cube with sides that aretwo coded units in length (from �1 to +1). Therefore, it can evaluate the non-linearrelationship between the assessment of indicators and factors.
The CCD was applied using the Design Expert software. The total number ofexperiments with three variables was 20 (=2k + 2 k + 6), where k is the number ofindependent variables. Fourteen experiments were augmented with six replicationsat the center values (zero level) to evaluate the pure error.
These data acquired from the experimental runs are then used to optimize sin-tering process. In this study, the response variables measured were relative densityand bending strength.
3. Results and discussion
Experimental design along with the observed responses isshown in Table 1.
Fitting the data to various models (linear, two factorial, qua-dratic and cubic) and their subsequent ANOVA showed that rela-tive density and bending strength were most suitably describedwith quadratic polynomial model (Eqs. (1) and (2)):
Y1 ¼ 96:57þ 2:19Aþ 0:58B� 0:54C � 0:54ABþ 0:020AC
� 0:14BC � 1:23A2 þ 0:069B2 þ 0:20C2 ð1Þ
Table 3ANOVA analysis for responses Y1 [relative density (%)], and Y2 [bending strength (MPa)].
Source Sum of squares DF Mean square F value Prob > F
For Y1
Model 100.02 9 11.11 25.92 <0.0001 SignificantResidual 4.29 10 0.43Lack of Fit 4.26 5 0.85 159.37 <0.0001 SignificantPure error 0.027 5 5.35E�03R2 = 0.96
Adeq precision = 19.598
For Y2
Model 1766.36 9 196.26 12.97 0.0002 SignificantResidual 151.36 10 15.14Lack of fit 143.94 5 28.79 19.42 0.0027 SignificantPure error 7.41 5 1.48R2 = 0.92
Adeq precision = 13.438
506 J. Li et al. / Journal of Alloys and Compounds 574 (2013) 504–511
Y2 ¼ 156:47þ 8:82Aþ 3:76B� 2:84C � 2:70AB� 1:01AC
þ 0:57BC � 4:48A2 � 0:47B2 þ 1:24C2 ð2Þ
The estimated model coefficients and their significance in rela-tion to the experimental scatter are shown in Table 2. At a 90% (Ta-ble 2) confidence level, the model equation in terms of actualfactors is (Eqs. (3) and (4)):
With the quadratic factors such as B2 and C2 and then the inter-action factors AC and BC are not significant.
Y2 ¼ 156:47þ 8:82Aþ 3:76B� 2:84C � 4:48A2 ð4Þ
With the quadratic factors such as B2 and C2 and then the inter-action factors AB, AC and BC are not significant.
The statistical significance of the model equation was evaluatedby the F-test for analysis of variance (ANOVA). ANOVA evaluationsof this model, shown in Table 3, imply that this model can describethe experiments. As can be seen in Table 3 the prob > F-values forrelative density and bending strength are lower than 0.05 indicat-ing that quadratic models were significant. The coefficient of deter-mination (R2) that was found to be close to 1 (0.96 for both Y1, and0.92 for Y2) also advocated a high correlation between observedand predicted values. The ‘‘lack of fit tests’’ compares the residualerror to the ‘‘Pure Error’’ from replicated experimental designpoints. The p-values, greater than 0.05, for both the responses indi-
Fig. 1. Predicted response
cate that lack of fit for the model was insignificant. Adequate pre-cision measures the signal to noise ratio and a ratio greater than 4is desirable. The adequate precision for Y1 and Y2 were 19.598, and13.438, respectively. These high values of adequate precision dem-onstrated that models are significant for the process.
On the model analysis of variance, the correlation coefficients ofquadratic regression equations of relative density and bendingstrength are R2 = 0.96 and R2 = 0.92. And that indicate model fitvery well the actual situation. F-values which were 25.92 and12.97 imply that the models are significant.
The actual and predicted relative density and example illus-trates the isoresponse curves are plotted in Figs. 1–3, respectively.Figs. 2 and 3 depicted the change of relative density and bendingstrength with sintering temperature, holding time and heatingrate, plotted for the case where the cooling rate is 3 �C/min. Andthe effect caused by factors on response variation was showed inFig. 4.
The sintering temperature and heating rate have different effecton the responses (relative density and bending strength). Sinteringtemperature is positive, heating rate is negative. High temperatureand low heating rate were profitably for reducing the rate of poreso as to increase the relative density of sintered body. The relativedensity and bending strength increase gradually with lengthenedholding time and then become flat. Extending the holding timehas good effect on the responses, but later nearly no effect whenthe transformation from monoclinic to tetragonal reaching equilib-rium. Extension of holding time could help grain growth and den-
vs. actual response.
1400.00 1450.00 1500.00 1550.00 1600.002.00
3.00
4.00
5.00
6.00Relative density
9292.5
9393.5
9494.5
9595.5
9696.5
97
97
97.597.5
98
98
98.5
1400.00 1450.00 1500.00 1550.00 1600.002.00
3.00
4.00
5.00
6.00Relative density
9292.5
9393.5
9494.5
9595.5
96
96
96.5
96.5
9797
97.5
97.5
98
(a) Effect caused by sintering temperature and holding time on relative density variation with heating rate of
3 /min and 5 /min, respectively.
1.50
3.25
5.00
6.75
8.50Relative density
9292.5
9393.5
9494.5
9595.5
96 96.5 97
97
97.5
98
98.5
1.50
3.25
5.00
6.75
8.50Relative density
9292.5
9393.5
9494.5
9595.5
95.5
96
96
96.5
96.5
9797
97.5
98
98.5
99
(b) Effect caused by sintering temperature and heating rate on relative density variation with holding time of 3h
and 5h, respectively.
1.50
3.25
5.00
6.75
8.50Relative density
92
92.5 93 93.5 94 94.5
95
95.5
9696.5
97
97.5
6.005.004.003.002.001.50
3.25
5.00
6.75
8.50Relative density
97.5
98
98.5
99
(c) Effect caused by holding time and heating rate on relative density variation with sintering temperature of
1450 and 1550 , respectively.
A: Sintering temperature/
B: H
olding time/h
A: Sintering temperature/
B (Holding time/h) =3 B (Holding time/h) =5
A (Sintering Temperature/ ) =1550
B: Holding time/h
A (Sintering Temperature/ ) =1450
C (Heating rate/ /min) =3 C (Heating rate/ /min) =5
B: H
olding time/h
1400.00 1450.00 1500.00 1550.00 1600.00
A: Sintering temperature/
1400.00 1450.00 1500.00 1550.00 1600.00
A: Sintering temperature/
6.005.004.003.002.00
B: Holding time/h
C: H
eating rate//m
in
C: H
eating rate//m
inC
: Heating rate/
/min
C: H
eating rate//m
in
Fig. 2. Example illustrates the isoresponse curves (Y1).
J. Li et al. / Journal of Alloys and Compounds 574 (2013) 504–511 507
Fig. 4. The effects curves of three factors on responses (Y1 and Y2).
J. Li et al. / Journal of Alloys and Compounds 574 (2013) 504–511 509
Table 4Predicted values vs. validation experiment values.
Sintering temperature x1 (�C) Holding time x2 (h) Heating rate x3 (�C/min) Relative density (%) Bending strength (MPa)
Predicted value Experiment value Predicted value Experiment value
1540 5 3 98.61 98.57 166.124 165.72
Fig. 5. The XRD of samples sintered at the optimized process.
Fig. 6. SEM picture of samples sintered at the optimized process.
510 J. Li et al. / Journal of Alloys and Compounds 574 (2013) 504–511
sification of the sintered bodies. Because the sintering samplescontain dozens of stomata and there exists only point contact be-tween particles. On the influence of high temperature with lowheating rate and extension of holding time, particles aggregateand connected pores turned into isolate and gradually reduced,and even disappeared to reach the final densification. But if densi-fication was finished, increasing temperature could enter overburnstate which would produce more fluxing holes and lower the den-sity of sintered body. The results were obtained that the maximumrelative density of 98.48% and bending strength of 169.951 weregot at the 3 �C/min of cooling rate.
The mathematical model generated during RSM implementa-tion was validated by conducting experiment on given optimalmedium setting. Process parameters of experimental optimiza-tion are shown in Table 4. The optimized parameters are sinter-ing temperature of 1540 �C, holding time of 5 h, heating rate of3 �C/min.
The experimental values are anastomosis with predicted value.The XRD spectrum and SEM picture of samples sintered at the opti-mized process are shown in Figs. 5 and 6.
Fig. 5 shows that partially stabilized zirconia ceramic was ob-tained after sintering process. The SEM image (Fig. 6) shows thatthe product obtained at the optimized process with the homoge-neous structure and nearly no pore which result high relative den-sity and good mechanical properties. That is because the samplesbefore sintering typically contain dozens of stomatal, they are onlypoint contact between particles. Strength of samples is very low.However, it will happen that the contact area expansion, particlesgathering and volume contraction in high temperature. And grainboundaries forms with the shortening distance between particles.In the meantime, the stomatal become isolated from connected,gradually reduced, and escaped in the high temperature. Therefore,the densification of the PSZ was achieved eventually with finehomogeneous structure.
J. Li et al. / Journal of Alloys and Compounds 574 (2013) 504–511 511
4. Conclusions
(1) Optimization of the production of calcia partially stabilizedzirconia (CaO-PSZ) by using natural baddeleyite shows thatall the three reaction variables have their effects on theproperties of CaO-PSZ (e.g. relative density and bendingstrength). The predicted model fits well with the experimen-tal results. The temperature, 1540 �C; holding time, 5 h; andheating rate, 3 �C/min were found to be the optimumconditions to achieve the maximum relative density ofCaO-PSZ.
(2) In this paper the natural baddeleyite which was obtained byfloating of baddeleyite ore, was used as raw material to pre-pare partially stabilized zirconia instead of chemical purezirconia. Therefore, it can shorten the process and reduceenergy consumption.
Acknowledgment
Project supported by the International Science & TechnologyCooperation Program of China (No. 2012DFA70570), and the Yun-nan Provincial International Cooperative Program (No. 2011IA004).
References
[1] S. Popovic, B. Grzeta, I. Czakó-Nagy, S. Music, J. Alloys Comp 241 (1996) 10–15.[2] R. Pazhani, H. PadmaKumar, A. Varghese, A. MosesEzhilRaj, S. Solomon, J.K.
Thomas, J. Alloys Comp 509 (2011) 6819–6823.[3] G.M. Wolten, J. Am. Ceram. Soc. 46 (1963) 418–422.[4] K.A. Sense, J. Am. Ceram. Soc. 44 (1963) 465.[5] S. C Carniglia, S.D. Brown, T.F. Schroeder, J. Am. Ceram. Soc. 54 (1971) 13–17.[6] H.J. Garrett, R. Ruh, Amer. Ceram. Soc. Bull. 47 (1968) 578–579.[7] O.C. Standard, C.C. Sorrel, Key. Eng. Mater. 153–154 (1998) 251–300.[8] T. Czeppe, P. Zieba, A. Pawlowski, J. Eur. Ceram. Soc. 22 (2002) 35–40.[9] T. Sakuma, Key. Eng. Mater. 153–154 (1998) 75–96.
[10] R.R. Hughan, R.H.J. Hannnk, J. Am. Ceram. Soc. 69 (1986) 556–563.[11] H.H. Luo, Q.Z. Wei, B.K. Wei, B. Yu, J. He, D.J. Li, J. Alloys Comp. 474 (2009) 551–
556.[12] S. Kumar, P. Pramamik, Trans. J. Br. Ceram. Soc. 94 (1995) 123–127.[13] B. Basu, J. Vleugels, O. Van Der Biest, J. Eur. Ceram. Soc. 24 (2004) 2031–2040.[14] M.A. Borik, E.E. Lomonova, V.V. Osiko, V.A. Panov, O.E. Porodinkov, M.A.
[15] M. Yoshimura, S.T. Oh, M. Sando, K. Niihara, J. Alloys Comp. 290 (1999) 284–289.
[16] S. Deville, H.E. Attaoui, J. Chevalier, J. Eur. Ceram. Soc. 25 (2005) 3089–3096.[17] Z.L. Wei, H.M. Li, X.Y. Zhang, S.H. Yan, Z. Lv, Y.Q. Chen, M.C. Gong, J. Alloys
Comp. 455 (2008) 322–326.[18] A. Rittidech, T. Tunkasiri, J. Alloys Comp. 38 (2012) S125–S129.[19] Y.L. Zhang, W.J. Jin, Y.H. Rong, T.Y. Hsu, D.Y. Jiang, J.L. Shi, Acta Mater. 54 (2006)
1289–1295.[20] J. Moon, H. Choi, H. Kim, C. Lee, Surf. Coat. Technol. 155 (2002) 1–10.[21] J. Li, J.H. Peng, S.H. Guo, L.B. Zhang, Ceram. Int. 39 (2013) 197–202.
