152 ZHANG Jun-ping et al. thermal degradation and solvent contamination(Brunner The regression analysis of the response function with 1994).There are reports on SC-CO,extraction as an statistical analysis are given in Table 2.Statistical test- excellent alternative to the use of chemical solvents in ing of the model was performed in the form of ANOVA the extraction of oils from different plants such as corn Here,the value for Fand P(probability)(P<0.05 when (List et al.1984),soybean (Friedrich et al.1982)and significant)were 80.5708 and 0.0001,respectively,and cotton seeds(Kuk and Hron 1994;Bhattacharjee et al. the lack-of-fit of 0.7117 was not significant (P>0.05). 2007),cocoa beans(Saldafia et al.2002),green tea indicating that the generated model adequately explained (Chang et al.2000),and ginseng (Wang et al.2001). the data variation and significantly represented the ac- However,no studies have been reported on the oil ex- tual relationship between the reaction parameters.The traction from N.glandulifera seed by SC-CO.extrac- determination coefficient R2=0.9904 indicating that tion to our knowledge.Response surface methodol- 99.04%of the variability in the response could be ex- ogy(RSM)which combines mathematics with statis- plained by the model.From the P-values of each model tics is often used to design experiments,build models, term,we concluded that the linear coefficients of pres- and evaluate the effects of factors (Yue et al.2008). sure and CO.flow rate and the quadratic terms of pres- The main advantage of RSM is the small number of sure and temperature,had highly significant effects on experimental trials needed to evaluate multiple param- the oil yield at the 1%level(P<0.01).The interactions eters and their interactions(Chow et al.1998),and it between pressure and temperature,as well as tempera- has been used successfully in food processing opera- ture and CO,flow rate,had significant effects on the tions(Hierro and Santa-Maria 1992;Reverchon 1997; oil yield at the 5%level (P<0.05). Lee et al.2000;Huang et al.2008;Liu et al.2009). SC-CO,was used to extract seed oil from N.glanelulifera Response surface analysis in this work.The effects of independent factors (pressure,temperature,and CO,flow rate)on the oil From Eq.(1)we can see that the oil yield of N.glandulifera yield of N.glandulifera seed were investigated.RSM seed has a complex relationship with independent was employed to build a model between the oil yield variables.The best way of expressing the effects of and these independent factors as well as to develop a independent variables on the oil yield within the experi- model equation that will predict and determine the opti- mental space under investigation is to generate response mum conditions for the oil yield surface plots of the equation.The three-dimensional response surfaces curves and corresponding contour RESULTS AND DISCUSSION plots were obtained using the Design Expert and are shown in Figs.1,2,and 3 to illustrate the relationship between independent variables and the oil yield. Model fitting Fig.I shows response surface curve and its con- tour plot for the effects of pressure and temperature on Oil yields obtained from all the experiments are listed in the oil yield and their interaction at a fixed flow rate of Table 1.The experimental data were used to calculate 20 L h.The extraction pressure and temperature the coefficients of the second-order polynomial equation. showed a quadratic effect on the response.At low The application of RSM offered,based on parameter pressure,the oil yield was increased with the increase estimates,an empirical relationship between the response of pressure.This is most likely due to the improve- variable,and the test variables under consideration.By ment of oil solubility resulted from the increased CO, applying multiple regression analysis on the experimen- density with the rise of pressure(Lee et al.2000).When tal data,the response variable and the test variables were the pressure was increased to levels greater than ap- related by the following second-order polynomial proximately 30 MPa,the negative quadratic effect be- equation: gan to have an impact.Such effect of pressure is not Y=36.81+0.78x-0.014x,+1.27x,+0.41xx2-0.099xx3 unexpected,when the pressure becomes too high,a +0.32x-2.16x2-1.64x2+0.031x2 (1) reduction in the solvent diffusivity and mass transfer 2012,CAAS.All rights reserved.Published by Elsevier Ltd.152 ZHANG Jun-ping et al. © 2012, CAAS. All rights reserved. Published by Elsevier Ltd. thermal degradation and solvent contamination (Brunner 1994). There are reports on SC-CO2 extraction as an excellent alternative to the use of chemical solvents in the extraction of oils from different plants such as corn (List et al. 1984), soybean (Friedrich et al. 1982) and cotton seeds (Kuk and Hron 1994; Bhattacharjee et al. 2007), cocoa beans (Saldaña et al. 2002), green tea (Chang et al. 2000), and ginseng (Wang et al. 2001). However, no studies have been reported on the oil ex￾traction from N. glandulifera seed by SC-CO2 extrac￾tion to our knowledge. Response surface methodol￾ogy (RSM) which combines mathematics with statis￾tics is often used to design experiments, build models, and evaluate the effects of factors (Yue et al. 2008). The main advantage of RSM is the small number of experimental trials needed to evaluate multiple param￾eters and their interactions (Chow et al. 1998), and it has been used successfully in food processing opera￾tions (Hierro and Santa-Maria 1992; Reverchon 1997; Lee et al. 2000; Huang et al. 2008; Liu et al. 2009). SC-CO2 was used to extract seed oil from N. glanelulifera in this work. The effects of independent factors (pressure, temperature, and CO2 flow rate) on the oil yield of N. glandulifera seed were investigated. RSM was employed to build a model between the oil yield and these independent factors as well as to develop a model equation that will predict and determine the opti￾mum conditions for the oil yield. RESULTS AND DISCUSSION Model fitting Oil yields obtained from all the experiments are listed in Table 1. The experimental data were used to calculate the coefficients of the second-order polynomial equation. The application of RSM offered, based on parameter estimates, an empirical relationship between the response variable, and the test variables under consideration. By applying multiple regression analysis on the experimen￾tal data, the response variable and the test variables were related by the following second-order polynomial equation: Y=36.81+0.78x1 -0.014x2 +1.27x3 +0.41x1 x2 -0.099x1 x3 . +0.32x2 x3 -2.16x1 2 -1.64x2 2 +0.031x3 2 (1) The regression analysis of the response function with statistical analysis are given in Table 2. Statistical test￾ing of the model was performed in the form of ANOVA. Here, the value for F and P (probability) (P<0.05 when significant) were 80.5708 and 0.0001, respectively, and the lack-of-fit of 0.7117 was not significant (P>0.05), indicating that the generated model adequately explained the data variation and significantly represented the ac￾tual relationship between the reaction parameters. The determination coefficient R2 =0.9904 indicating that 99.04% of the variability in the response could be ex￾plained by the model. From the P-values of each model term, we concluded that the linear coefficients of pres￾sure and CO2 flow rate and the quadratic terms of pres￾sure and temperature, had highly significant effects on the oil yield at the 1% level (P<0.01). The interactions between pressure and temperature, as well as tempera￾ture and CO2 flow rate, had significant effects on the oil yield at the 5% level (P<0.05). Response surface analysis From Eq. (1) we can see that the oil yield of N. glandulifera seed has a complex relationship with independent variables. The best way of expressing the effects of independent variables on the oil yield within the experi￾mental space under investigation is to generate response surface plots of the equation. The three-dimensional response surfaces curves and corresponding contour plots were obtained using the Design Expert and are shown in Figs. 1, 2, and 3 to illustrate the relationship between independent variables and the oil yield. Fig. 1 shows response surface curve and its con￾tour plot for the effects of pressure and temperature on the oil yield and their interaction at a fixed flow rate of 20 L h-1. The extraction pressure and temperature showed a quadratic effect on the response. At low pressure, the oil yield was increased with the increase of pressure. This is most likely due to the improve￾ment of oil solubility resulted from the increased CO2 density with the rise of pressure (Lee et al. 2000). When the pressure was increased to levels greater than ap￾proximately 30 MPa, the negative quadratic effect be￾gan to have an impact. Such effect of pressure is not unexpected, when the pressure becomes too high, a reduction in the solvent diffusivity and mass transfer