Statistical Design Optimization of Hydrogen Production through Ethanol Steam Reforming using a Ni/Al 2 O 3 Catalyst

—Factorial experimental design and response surface methodology, together with central composite design, were employed to investigate the effect of the process variables in hydrogen production via ethanol steam reforming. The influence of temperature (T), water– ethanol molar ratio (MR), and liquid hourly space velocity (SV) on hydrogen yield ( 𝐘 𝐇 𝟐.𝐘 ), ethanol gasification ( 𝐘 𝐄𝐓𝐎𝐇 𝐂𝐨𝐧. ), and CO yield ( 𝐘 𝐂𝐎. 𝐘 ) were determined. In coded units, X 1 , X 2 , and X 3 represent T, MR, and SV, respectively. The multiple regression analysis results showed that temperature and water–ethanol molar ratio significantly affects the responses. By contrast, all the responses were not significantly affected by the liquid hourly space velocity. ANOVA indicated that the linear terms X 1 and X 2 , the quadratic term X 2 , and the interaction term X 1 X 2 exerted the highest influence on the 𝐘 𝐇 𝟐.𝐘 , and 𝐘 𝐄𝐓𝐎𝐇 𝐂𝐨𝐧. . 𝐘 𝐂𝐎. 𝐘 was most affected by the linear terms X 1 and X 2 and the interaction term X 1 X 2 . The optimal conditions for maximal 𝐘 𝐇 𝟐.𝐘 and 𝐘 𝐄𝐓𝐎𝐇 𝐂𝐨𝐧. , and minimal 𝐘 𝐂𝐎. 𝐘 were obtained at a temperature of 500 °C, a water– ethanol molar ratio of 20, and a liquid hourly space velocity of 0.72 h -1 . These optimal conditions resulted in a hydrogen yield of 4.7 mol/mol of ethanol, an ethanol gasification rate of 85%, and a CO yield of 0.25 mol/mol of


I. INTRODUCTION
Hydrogen production through ethanol steam reforming has been performed using different catalytic systems. Developing an active, selective, and stable catalytic system for producing hydrogen is a major challenge as far as ethanol steam reforming is concerned. Bshish et al [1] extensively reviewed various catalytic systems for hydrogen production through ethanol reforming. Nickelbased catalysts have been broadly investigated in ethanol steam reforming reaction on different supports because they are inexpensive and are extensively used in hydrocarbon hydrogenation and steam reforming [2][3][4][5][6][7][8][9][10] . Aside from catalytic properties, operating parameters such as temperature, water-ethanol molar ratio, and space velocity are also control catalyst activity. Therefore, determining the most important operating parameters and their values are important in enhancing ethanol utilization to maximize hydrogen production and minimize the percentage of carbon monoxide (CO) in the effluent gas. Several studies have investigated the effect of operating parameters such as steam to carbon (S/C) molar ratio, reaction temperature, feed flow rate, and space time on hydrogen production [11,12]. However, these studies assumed the operating parameters do not interact [13], which is inefficient for optimizing the reaction conditions [14].  113 Therefore, studying the interactions between the operating parameters is essential to optimizing hydrogen production.
Statistically, the experimental design is established and used to control the experiments [15]. Experimental design is used to analyze the influence of separate and interacting factors, and to develop models that simulate the responses with a minimum number of experiments. Response surface methodology (RSM) is used to examine the relationship between one or more responses and a set of quantitative experimental variables. Its simplicity and high efficiency make RSM an optimization technique that can be commonly applied to optimize process variables in different applications [16][17][18][19][20]. Moreover, RSM reduces the experimental trails needed to estimate the process variables [21].
In this paper, we used factorial design to optimize the variables that control hydrogen production via ethanol steam reforming, namely, temperature (X1), water-ethanol molar ratio (X2), and liquid hourly space velocity (X3). Hydrogen production via ethanol steam reforming was performed using a Ni/Al2O3 catalyst, where the alumina support, denoted as AlS.G, was prepared via a sol-gel process. The optimization process was performed via RSM using central composite design (CCD) to evaluate the maximum hydrogen yield ( 2. ) and ethanol gasification ( . ), and the minimum CO yield ( . ). Ni loading was maintained at 6 wt% for all experiments.

A. Catalyst preparation
The sol-gel alumina support (AlS.G.) was prepared according to (Seo et al 2007 [22], where 96 g of precursor (aluminum sec-butoxide, Sigma-Aldrich) was dissolved in 828 ml of ethyl alcohol under constant stirring at 80 °C (solution 1). Up to 1.37 ml of HNO3 and 4.12 ml of distilled water diluted with 549 ml of ethyl alcohol (solution 2) were then mixed with solution 1. This mixture was fixed at 80 °C to form the sol. After cooling the sol, a transparent gel was obtained by adding drop wise 8.23 ml of distilled water and 68.66 ml of ethyl alcohol into the sol. Subsequently, the alumina gel was covered and kept for one day before it was dried overnight. The solid formed was calcined at 800 °C to obtain the alumina sol-gel. The support was denoted as AlS.G.

B. Instrumentation
To evaluate the optimum operating parameters for hydrogen production via ethanol steam reforming, all the runs were conducted in a Pyrex glass tube reactor (internal diameter: 8 mm; length: 50 cm) at 400 °C and atmospheric pressure. The schematic of the experimental set-up for the reforming reaction is shown in Fig. 1. Before the reaction, the catalyst was heated to 150 °C for 1 h, and subsequently reduced in situ under flowing H2 for another 1 h. A mixture of liquid water and ethanol was introduced into the reactor containing 0.5 g of catalyst together with the carrier nitrogen (40 ml/min) gas. The output gas stream was analyzed via gas chromatography (GC) (Model SRI 8610 C) using a molecular sieve and a silica gel column with a thermal conductivity detector with helium as the carrier gas. The condensed liquids were collected and analyzed using a gas chromatograph (SUPELCO) equipped with a flame ionization detector and an Equity-1 capillary column (30 m × 0.32 mm × 0.1 μm film thickness). The criteria used to determine catalyst performance included H2 yield, ethanol gasification, and CO yield. Equations (1), (2), and (3) were used to calculate Y H 2.Y , Y ETOH Con. , and Y CO. Y , respectively.
where F ETOH in and F ETOH out represent the molar flow rate of the ethanol inlet and outlet of the reactor, respectively, and F H 2 ,out and F CO ,out represent the flow rate of the hydrogen and carbon monoxide outlets of the reactor, respectively.

C. Experimental design
The three independent variables temperature (X1), water-ethanol molar ratio (X2), and liquid hourly space velocity (X3) were verified based on the experimental design as follows. The CCD method was implemented to investigate the influences of these operating factors on the 2. , . , and . of hydrogen production via ethanol steam reforming. Table 1 shows the three independent factors with their experimental ranges and the levels employed in this study. The center of the design is coded as (0), and the upper and lower levels of the design are coded as (1) and (-1), respectively. The factorial CCD of the independent factors is presented in Table 2. According to the CCD, 20 sets of runs were selected and each experiment was replicated twice. The experimental design and statistical analysis were conducted using the Minitab software package (version 14.12.0). The experimental data were modeled via an ANOVA, and a secondorder polynomial model was determined (Equation (4)) for each output response (Y H 2,Y , Y ETOH Con. , and Y CO,Y ). The observed experimental data were then compared with the data from the obtained models.
where Y i represents the response variables (Y H 2,Y , Y ETOH Con. , and Y CO,Y ); β 0 is a constant; β j , β jj , and β ij are the linear, quadratic, and interaction effect coefficients, respectively; and x i and x j are the input variables. For the three independent variables, the model represented is shown in Equation (5) The significance of the regression model was determined using an ANOVA with p-value less than 0.05. In addition, the coefficient of determination (R 2 ) was used to evaluate the performance of the model. In general, the higher the R 2 , the better the model fits the data. A reduced model was obtained by eliminating the insignificant coefficients from the completed model. The influence of the three experimental variables on each response was explored using RSM plots.

A. Empirical models
Two replicates of 20 sets of experiments were performed to investigate the influence of reaction temperature, ethanol to water molar ratio, and liquid hourly space velocity on hydrogen production during ethanol steam reforming. Prior to the reaction, all runs were conducted under the same preheating and reduction procedure. Table 3 illustrates the experimental (observed) and predicted data obtained from the model for the three responses (Y H 2,Y , Y ETOH Con. , and Y CO,Y ). Standard deviations and mean values of the three observed responses are also shown in Table 3.
The estimated regression coefficients of reduced and nonreduced models with their corresponding p-values are shown in Tables 4 and 5, respectively. These tables also provide the R 2 and adjusted R 2 values, which were used to evaluate the goodness of fit of the regression model. ANOVA were conducted to identify the terms that significantly influenced the responses at a p-value less than 0.05. Tables 4 and 5 indicate that the most important terms are X1, X2, X22, and X1X2, which represent the linear terms of temperature, molar ratio, quadratic term of molar ratio, and the interaction term of temperature and molar ratio, respectively. These terms significantly influenced all responses (Y H 2,Y , Y ETOH Con. , and Y CO,Y ). Notably, the liquid hourly space velocity (LHSV), representing X3, had negligible influence on the responses when all the p-values were greater than 0.05.   As shown in Tables 4 and 5, R 2 and adjusted R 2 for Y H 2,Y and Y ETOH Con. indicated that the proposed model adequately fitted the experimental data. For Y CO,Y , the R 2 and adjusted R 2 values of the regression model were quite low, which suggests the presence of other important parameters.

‫األسمرية‬ ‫الجامعة‬ ‫مجلة‬
The high R 2 (0.965) for Y H 2,Y in Table 5 indicates that the proposed model can account for most of the observed variations. The high R 2 value of 0.947 and the small p-value (p < 0.05) for the Y ETOH Con. indicated that the regression model had a good correlation between the independent variables and the response. However, Table 5 shows that the linear terms of X1 and X2, the quadratic term of X22, and the interaction term of X1X2 significantly influenced both Y H 2,Y and Y ETOH Con. with p-value less than 0.05. By contrast, Y CO,Y was mostly affected by the linear terms of X1 and X2 and the interaction term of X1X2.
The following second-order polynomial equations (Equations (6), (7), and (8)) describing the relationship between the process variables and their predicted responses for all the three dependent variables (Y H 2,Y , Y ETOH Con. , and Y CO,Y ,) were generated. 119 Equations 6 and 8 show that X2 had the greatest influence on Y H 2,Y and Y CO,Y . The terms X3 and X1 also affected Y H 2.Y , but to a lesser extent than X2. The interaction between X1X2 only slightly affected Y H 2.Y . As shown in Equation (7), Y ETOH Con. was mostly affected by the linear term of X2, followed by X3 and X1. However, the quadratic term X22 and the interaction term X1X2 had lower effect on Y ETOH Con. than the linear terms. Figure 2 presents the 3D response surface plots of the relationships between the independent variables and Y H 2,Y . The advantage of using response surface plots with two independent variables at a time is to enable the examination of the relation of the main and the interaction factors with the response. Figure 2a illustrates the effect of X1 and X2 on Y H 2.Y . The surface plot shows that the highest hydrogen yield was obtained when both independent variables (X1 and X2) were high (interaction effect). This finding was in agreement with other data [22], which indicate that high temperatures and high water-ethanol ratios enhance H2 production. On the other hand, Fig. 2b shows that Y H 2,Y varied only with X1, whereas X3 had no effect on Y H 2,Y . Figure 2c also shows that Y H 2,Y was significantly affected by X2 up to 18, but was not significantly changed beyond this ratio. Generally, Y H 2,Y increased as X1 and X2 increased, which was expected because the steam reforming reaction is endothermic and positively affected by high temperature.

B. Effect of process variables on the responses
The response surface in Fig. 3 (a, b, and c) illustrates the influence of the interaction between varying temperature and water-ethanol molar ratio, between temperature and liquid hourly space velocity, and between water-ethanol molar ratio and liquid hourly space velocity on ethanol gasification, respectively. Figures 3a and 3b indicate that Y ETOH Con. significantly increased as X1 increased. From Fig. 3a, the orientation of the principal axes of the surface plot indicate that the interaction between X1 and X2 significantly affected Y ETOH Con. , and the optimum values of both variables occurred in the experimentally explored area. The effect of the interaction between X1 and X2 on Y ETOH Con. shown in Fig. 3a was higher than that between X1 and X3 shown in Fig. 3b, wherein X3 had no effect on Y ETOH Con. . Figure 3c shows that Y ETOH Con. was significantly influenced by X2, especially at low and medium levels. Meanwhile, X3 had a lower effect on Y ETOH Con. . Therefore, at a constant X2, Y ETOH Con. slightly increased as X3 increased. Generally, Figs. 2 and 3 indicate that the ethanol gasification and hydrogen yield expectedly increased as the temperature and molar ratio increased because the steam reforming reaction is endothermic and favors at high temperature. The ethanol gasification and hydrogen yield varied non-significantly with the LHSV range in this study. Response surface plots describing the effects of the main and the interaction variables on Y CO,Y are shown in Fig. 4. The effect of the interaction between X1 and X2 on Y CO,Y is illustrated in Fig.  4a. Y CO,Y decreased as X2 increased at high X1 values, but it did not exceed 0.18 mol/mol for all values of X2 at low X1 values. The surface plot also shows that the lowest Y CO,Y was obtained at the lowest X1 value. Figure 4b shows that the Y CO,Y values significantly varied with high X1 values, whereas the variation in X3 values had no effect on Y CO,Y . At a constant X1, Y CO,Y significantly varied with X2. Figure 4c shows that the lowest Y CO,Y was obtained at high X2 levels.  However, the low Y CO,Y at high X2 can be attributed to the acceleration of the water gas shift reaction caused by the catalyst (Equation (9)). Figure 4. Response surface plot of the interaction effects of process variables on conversion into gaseous product: (a) temperature and molar ratio, (b) temperature and LHSV, (c) molar ratio and LHSV.

C. Response optimization
The optimum levels of the three independent variables, namely, temperature, water-ethanol molar ratio, and liquid hourly space velocity were determined using the response optimization method. The optimum conditions that maximized ethanol gasification and hydrogen yield and minimized carbon monoxide yield were a temperature of 500 °C, a water-ethanol molar ratio of 20, and a liquid hourly space velocity of 0.72 h -1 . Applying these optimum conditions resulted in hydrogen yield of 4.7 mol/mol of ethanol, an ethanol gasification rate of 85%, and a carbon monoxide yield of 0.25 mol/mol of ethanol.

IV. CONCLUSION
This study focused on hydrogen production via ethanol steam reforming using a Ni/Al2O3 catalyst. The hydrogen yield, ethanol gasification and CO yield were optimized and the effects of temperature, water-ethanol molar ratio, and liquid hourly space velocity variations as independent variables were analyzed through statistical analysis. A second-order polynomial model was ‫األسمرية‬ ‫الجامعة‬ ‫مجلة‬ : ‫ا‬ ‫والتطبيقية‬ ‫األساسية‬ ‫لعلوم‬ ‫الم‬ ‫جلد‬ ( 2 ) ( ‫العدد‬ 1 ،) ‫يونيو‬ 2017 122 developed using an ANOVA. According to the reduced models, temperature (X1) and waterethanol molar ratio (X2) significantly affect the responses. All of the responses were not significantly affected by the liquid hourly space velocity (X3), possibly because of the small LHSV range applied. The response optimization method was used to identify the combination of input variable settings that jointly optimized the three responses. The optimum values for the independent variables were a temperature of 500 °C, a water-ethanol molar ratio of 20, and a liquid hourly space velocity of 0.72 h -1 . These conditions result in an ethanol gasification rate of 85%, a hydrogen yield of 4.7 mol/mol of ethanol, and CO yield of 0.25 mol/mol of ethanol. In conclusion, special consideration must be given to temperature and the water-ethanol molar ratio for hydrogen production via ethanol steam reforming using Ni/Al2O3 catalysts.