Article pubs.acs.org/EF
Ethanol Dehydration in a Pressure Swing Adsorption Process Using Canola Meal Mehdi Tajallipour, Catherine Niu,* and Ajay Dalai Department of Chemical Engineering, University of Saskatchewan 57 Campus Drive, Saskatoon, Saskatchewan S7N 5A9, Canada ABSTRACT: Canola meal was used as an adsorbent in a pressure swing adsorption (PSA) process for ethanol dehydration at different temperatures, vapor feed concentrations, and adsorbent particle sizes. Adsorption experiments were performed at breakthrough point and equilibrium. The results demonstrate that canola meal was able to break the ethanol−water azeotropic point 95.6 wt %, selectively adsorb water, and produce over 99 wt % pure ethanol. At elevated temperature and feedwater concentration, water mass transfer rate increased. In addition, the mass transfer rate decreased when the size of the adsorbent particles was increased. The water breakthrough curves were simulated by incorporating the Linear Driving Force model and the mass transfer resistances were evaluated. The internal mass transfer resistance was identified as the mass transfer limit. The watersaturated canola meal was regenerated at temperatures no higher than 110 °C under vacuum and successfully reused.
1. INTRODUCTION The present energy and environmental problems caused by fossil fuels have brought the public’s attention to renewable alternatives such as biofuels. Ethanol is a biofuel that has high energy values and can be produced from renewable resources.1 Ethanol can be blended with gasoline and replace methyl tertbutyl ether (MTBE) in the fuel, which requires ethanol of high purity over 99 wt % in order to avoid phase separation.2 Industrial production of anhydrous bioethanol (>99 wt %) starts with the fermentation process in which carbohydrates such as starchy or cellulosic materials are converted to ethanol using microorganisms. The end fermentation broth contains 6− 12 wt % of ethanol mixed with water, inorganics, and organics. Then, separation and purification of ethanol from the fermentation broth is usually by distillation of the fermentation-strength ethanol to obtain 70−95 wt % ethanol−water vapor below the azeotrope (95.6 wt %). This is followed by adsorption with molecular sieves in a pressure swing adsorption (PSA) process to remove the remaining water (dehydration) so as to achieve over 99 wt % fuel-grade ethanol.3,4 The PSA dehydration process draws more attention compared to other methods such as extractive and azeotropic distillation due to its lower energy separation requirement and has been regularly used in the bioethanol industry.5−7 In the PSA process, low grade ethanol−water vapor coming from the previous distillation step is fed into the adsorption column, then water is selectively adsorbed from the vapor by molecular sieves packed in the bed at moderate pressure, and finally, ethanol vapor of high purity exits the column and is then condensed and collected as the product. 3A molecular sieves have micropores that are too small for ethanol molecules to penetrate; therefore, water molecules are adsorbed and further enrichment of ethanol above the azeoptrope point can be easily achieved. The water saturated column is then regenerated at vacuum at a certain temperature.8 However, the high temperature requirement for regeneration of zeolite (200− 250 °C) has led to growing interest in searching for more energy efficient alternatives such as biomass derived adsorbents.9,10 Moreover, the use of bioadsorbents offers additional © 2013 American Chemical Society
advantages compared to synthetic adsorbents, for example, easy deposition in the environment and reusability as the fermentation feedstock to produce either ethanol or biogas when the regeneration does not seem feasible.11 Bioadsorbents such as cassava,12 cornmeal,3,13−15 wood chips,16 natural corncobs, natural and activated palm stone and oak 17 have been extensively studied for ethanol dehydration purposes. However, canola meal, a byproduct abundantly generated from oil extraction or the biodiesel industry has not been systematically investigated in the PSA process for ethanol dehydration though preliminary work on ethanol dehydration by canola meal was reported by Baylak and his colleagues.18 Canola meal is rich in cellulosic components and protein.19 It has high water adsorption capacity, being about 303−390% of its initial dry weight.19 Therefore, canola meal has a good potential to be used in the adsorption process for ethanol dehydration, which is the subject of this study.
2. MATERIALS AND METHODS The raw canola meal with irregular shape used in this work was purchased from Co-Op Feeds (Saskatoon, Canada). Canadian Standard Sieves Series (Combustion Engineering Canada Inc.) were used to sieve canola meal particles; the collected samples had particle sizes in the range 0.43−1.18 mm. The raw canola meal was also used to prepare cylindrical pellets. A California Pellet Mill (CPMLaboratory Model CL-5, California Pellet Mill Co., Crawfordsville, IN) was used to make cylindrical pellets with the uniform size of about 5 mm in diameter and 10 mm in length. Canola meal samples were dried in an oven at 110 °C for 24 h and then packed in the column. Information on the surface functional groups of canola meal samples was obtained using FTIR analysis (Jasco FT/IR-4100) at 1000−3500 1/cm. The canola meal sample (1−2 mg) was mixed with 100 mg of a solid infrared transparent substance (potassium bromide) and then pressed into a 7 mm disc. The elemental components (carbon, hydrogen, nitrogen, and sulfur) of canola meal samples were determined using a PerkinElmer Elemental CHNS analyzer. The Received: May 14, 2013 Revised: October 10, 2013 Published: October 16, 2013 6655
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
Figure 1. Process flow diagram of the PSA setup. Mastersizer particle size analyzer (Malvern Mastersizer-s long bench size distributor) was used to determine the size distribution of canola meal particles sieved in the range 0.43−1.18 mm. The ethanol−water solution was prepared by mixing 200 proof ethanol (reagent grade, Commercial Alcohol Inc., Canada) with distilled water. The water content for each collected sample was analyzed by an automated Karl Fischer Coulometer (Metler Toledo DL32). The ethanol mass fraction for each sample was determined as the difference between the total mass of the sample and the mass fraction of water. The ethanol mass fraction evaluated using this approach was consistent with the one analyzed by HPLC (Agilent, 1100 Series, Refractive Index Detection). 2.1. Adsorption and Desorption Processes. A pressure swing adsorption process was used in a laboratory scale to separate water and ethanol by canola meal. The schematic diagram of the experimental setup is shown in Figure 1. It includes the following: a pump (ColeParmer, RK-74930-05) to transfer the prepared mixture of water and ethanol to the nebulizer; a gas tank (N2); a gas flow meter (ColeParmer, PMR1-010360) to adjust the flow rate of the carrier gas (N2); a preheater (heated piping line); an evaporator, which is an approximately 15 m long copper tube coiled and immersed in a hot oil bath; a stainless steel adsorption column being 501 mm long and 46 mm ID equipped with a jacket; a heated pipe line connecting the outlet of the adsorption column to a back pressure regulator valve to ensure no condensation occurs in the stream before entering the back pressure regulator; a back pressure regulator (Parker Hannifin Corp, U.S.A.) that was used to pressurize the system according to the operating conditions; three glassware condensers to cool down the effluent and separate ethanol product from the carrier gas (N2). The temperature of the preheater in the setup was controlled by temperature controllers (Cole-Parmer, 89000-00, Canada) connected with heating tapes. The temperatures of the vapor stream and the bed were monitored at different points by four thermocouples (Omega K type, U.S.A.) labeled J-111, J-114, J-115, and J-126. The thermocouples J-115 and J-126 were placed inside the tube, where J-115 read the temperature of the vapor stream at the inlet of the column and J-126 read the temperature of the vapor stream entering the back pressure regulator. J-111 and J-114 were inserted at the middle and bottom of the column to monitor the bed temperature. Two pressure transducers (Honeywell, U.S.A.) were used to monitor pressure at the top and bottom of the adsorption column (J-112 and J113). The pressure transducers were attached to Omega DPiS32
outputs. The thermocouples J-115 and J-126 were connected to Omega DPi32 outputs, while the thermocouples J-111 and J-114 were attached to Omega UTC-USB Connectors and temperature data were recorded using a TRH Central Measurement and Data Logging Program (Omega). Prior to the adsorption step, the column (D-110) was kept under vacuum condition (25 kPa) at 110 °C and purged with nitrogen gas at a flow rate of 786 cm3/min from the bottom for 15 h to ensure that the bed was free of any moisture. In the adsorption step, the prepared mixture of ethanol and water was pumped into the nebulizer (J-135), where the mixture was broken into small aerosol droplets with the aid of N2 gas (Figure 1). After that, the mixture entered into the preheater where the tube was wrapped with heat tapes to warm up the mixture before entering into the evaporator (E-130). The mixture in the evaporator turned into vapor and reached the desired temperature corresponding to the operating conditions monitored by thermocouple (J-115). After the bed temperature had reached the desired set point, the adsorption step began. The temperature of the bed was controlled corresponding to the operating conditions using an oil bath (E-116) that circulated the heating oil through the jacket of the adsorption column. The adsorption process began once the vapor stream entered into the column (D-110) from the top. As the vapor stream passed through the bed, water was selectively adsorbed on the adsorbent and the ethanol product vapor along with N2 gas left the column from the bottom. The back pressure regulator valve (J-125) pressurized the system according to the operating conditions. The pressure drop along the column during the adsorption process was 2.1−3.4 kPa with an inlet vapor pressure of 243 kPa, which is negligible. The tube between the column and back pressure regulator was wrapped with heating tapes to keep the vapor temperature at 103 °C and avoid any condensation of the vapor stream before going into the pressure regulator. The effluent of the adsorption column was distributed between 3 condensers (E-120, E-121, and E-122), which were placed in a parallel pattern in order to separate the ethanol and water content of the vapor from the N2 gas. The collected samples were weighed and then analyzed to determine their water and ethanol contents. The adsorption process was terminated when the bed was saturated and the temperatures at the middle (J-111) and bottom (J-114) of the column (D-110) had reached the inlet temperature of the vapor. After the adsorption step, the saturated bed was regenerated. The first step in the desorption stage was depressurization, at which the 6656
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
v/v% of the distribution is above 0.57 ± 0.002 mm and 50 v/v % is below. 3.2. Temperature Effect. The effect of the bed temperature on water/ethanol adsorption was investigated at 100, 105, and 110 °C, while the water partial pressure in the vapor feed stream was kept at 24 kPa corresponding to 5 wt % water content in the ethanol−water mixture and the total pressure at the inlet was 243 kPa. Pressure drop across the column was negligible, being 2.1−3.4 kPa. The water breakthrough curves and ethanol production profiles are shown in Figure 3. As can be seen from Figure 3a, canola meal broke the azeotropic point (95.6 wt % EtOH) at the tested temperatures and produced over 99 wt % ethanol. As presented in Figure 3b, water breakthrough time decreased as the temperature was increased. The breakthrough times of 60, 54.8, and 41.9 min were observed at 100, 105, and 110 °C, respectively. The slopes of the curves are associated with the water mass transfer rate.4 In Figure 3b, the slope of water breakthrough curves increases as the temperature increases, which shows greater mass transfer rates at elevated temperatures. Table 1 summarizes the uptakes and recovery of 99 wt % ethanol and water separation factor at different bed temperatures. At the breakthrough point, water uptake decreased as the temperature was increased. Water uptakes for runs at 100, 105, and 110 °C were 0.85, 0.79, 0.58 (mol/kg adsorbent), respectively. In the case of ethanol, the uptake decreased from 1.74 to 1.16 (mol/kg adsorbent) when the temperature was increased from 100 to 110 °C. Furthermore, Table 1 shows that an increase in temperature caused a slight decrease in the ethanol recovery; the values were 75, 74, and 73% (amount of collected ethanol over 99% divided by amount ethanol entered into column until breakthrough point) for runs at 100, 105, and 110 °C, respectively. The results indicate that water/ethanol adsorption is an exothermic process. Comparison of water/ethanol uptake at the breakthrough and equilibrium points indicates that even though water uptake was higher at equilibrium than at the breakthrough point, water separation factors for all runs at equilibrium were lower. Due to the exothermic nature of adsorption processes, greater water adsorption by the adsorbent results in higher heat generation. This fact can be further observed from the temperature profile of the bed presented in Figure 4. The temperature of the packed bed in this work was controlled by an oil jacket. For each set, temperature values such as 100, 105, and 110 °C, the temperature of the vapor feed and oil jacket and the initial temperature of the packed bed were all controlled at the same specific value. However, as ethanol dehydration proceeded, the bed temperature had a pulse rise and gradually reduced to the set values, as shown in Figure 4. This is because the heat generated from adsorption could not be removed instantly by the oil jacket; however, the bed temperature was eventually controlled with time. The maximum temperature rise ΔT = Tmax − Tinlet (or hot spot) increased as the set temperature was decreased,4 where Tmax is the peak temperature profile and Tinlet is the temperature of the vapor stream at the inlet. This again confirms that the adsorption is exothermic. 3.3. Feed Concentration Effect. The effect of water/ ethanol feed concentration on adsorption performance was investigated by varying the water/ethanol partial pressure in the feed stream at temperature 110 °C, total pressure 243 kPa, and superficial velocity 0.9 cm/s. The water partial pressure in the feed vapor was adjusted to 24, 45, and 85 kPa corresponding to
pressure of the column was decreased to the atmospheric one by opening the valve (J-131) located on the top of the adsorption column (D-110). Next, the pressure of the column was reduced to 25 kPa and kept constant using a vacuum pump (G-134). Then, the temperature of the bed was kept constant (at 110 °C), and the column was purged with nitrogen gas at a flow rate of 786 cm3/min for 5 h. Nitrogen gas was directed by a bypass to enter the column from the bottom and leave from the top. Finally, the outlet stream went through a condenser (E-132) to separate water/ethanol from the nitrogen gas and prevent them from entering the vacuum pump (G-134). To study the effects of temperature, feed concentration, and pellet size on the dynamic adsorption of water/ethanol on canola meal, water breakthrough curves and ethanol production profiles were generated. Water breakthrough curves were generated by plotting the dimensionless water content (C/C0) versus time, where C is the water content (wt %) in the effluent at given time intervals and C0 represents the water content (wt %) in the feed. Ethanol production profiles represent the ethanol concentration in the output stream at the given time intervals. The breakthrough point in this work refers to the point where the concentration of water in the effluent reaches 1 wt %, which corresponds to 99 wt % ethanol. The water/ethanol uptake on canola meal at equilibrium conditions was also determined. The equilibrium conditions in each run were identified by considering both the water breakthrough curve and the temperature profile. At equilibrium, the bed reached its saturation point, at which the water content in the effluent equaled its feed value, and the bed temperature had restored its initial value. At breakthrough and equilibrium, water/ethanol uptake was calculated by the difference between the total mass of water/ethanol input into the column and the accumulated mass of water/ethanol in the effluent, being divided by the dry net weight of the canola meal in the column. The selectivity of water adsorption by the adsorbent was determined by calculating the separation factor α as follows:
α=
X w / Xe Yw /Ye
(1)
where Xw and Yw are the mass fractions of water in the adsorbed and vapor phases, respectively, while Xe and Ye are the corresponding ethanol mass fractions.15
3. RESULTS AND DISCUSSION 3.1. Particle Size Distribution. To determine the size distribution of sieved canola meal particles, the Mastersizer particle size analyzer was used. Figure 2 shows the size distribution of sieved canola meal particles in the size range 0.43−1.18 mm. The particle size analysis was duplicated and the average value of the volume median diameter D(v, 0.5) was determined. The volume median diameter D(v, 0.5) of fresh canola meal particles is approximately 0.57 ± 0.00 mm. This means that 50
Figure 2. Particle size distributions for fresh canola meal. 6657
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
Figure 3. (a) Ethanol production profiles. (b) Water breakthrough curves. Operating conditions: Pw = 24 kPa, u0 ≈ 0.9 cm/s, and dp = 0.43−1.18 mm.
Table 1. Water and Ethanol Adsorption on Canola Meal Particles (0.43−1.18 mm) at Various Temperatures (Total Pressure 243 kPa, Superficial Velocity 0.9 cm/s) operating condition
equilibrium adsorption
T (°C)
water content (wt %)
Pwater (kPa)
PEtOH (kPa)
H2O uptake
100 105 110
5 5 5
24 24 24
178 178 178
1.43 1.19 1.10
a
adsorption until breakthrough point (99 wt % ethanol)
EtOH uptake
α
H2O uptakea
EtOH uptakea
αb
recovery (%)c
3.17 2.67 2.52
3.35 3.30 3.34
0.85 0.79 0.58
1.74 1.58 1.16
3.60 3.69 3.65
75 74 73
a
a
mol/kg adsorbent. bSeparation factor. cAmount of collected ethanol (over 99 wt %) divided by amount of ethanol entered into column until breakthrough point.
achieved for experimental runs with water partial pressures of 24, 45, and 84 kPa, respectively. The corresponding ethanol recovery (with concentration over 99 wt % EtOH) were 73, 69, and 68% indicating the production of ethanol with concentration above 99 wt % decreased as the ethanol partial pressure in the feed stream was decreased (see Table 2). It can be seen from the water breakthrough curves in Figure 5b that an increase in the water partial pressure of the feed stream (corresponding to higher water content in the feed) resulted in an increase in the slope of the profiles. This is indicative of a higher mass transfer rate. The same information can be obtained from the temperature profiles in Figure 6, where higher and steeper temperature curves correspond to runs with more water content in the feed stream and higher water uptake. This finding was expected due to the exothermic nature of adsorption processes. Table 2 shows that the water uptake at breakthrough increased as the water content was increased in the feed stream, while the ethanol uptake decreased. At breakthrough, water uptakes of 0.58, 0.89, and 1.32 (mol/kg adsorbent) were obtained at runs with water partial pressures of 24, 45, and 85 kPa. The corresponding values for ethanol uptake at breakthrough were 1.16, 0.99, and 0.64 (mol/kg adsorbent), respectively. At equilibrium, higher values for water and ethanol uptake were achieved compared to those at breakthrough point. Separation factor does not show a consistent trend when water partial pressure was increased. 3.4. Particle Size Effect. The performance of adsorption can be affected by the particle size of the adsorbents by changing the packing density, bed porosity and mass transfer resistance.4 Experimental runs using particles of sizes 0.43−1.18 mm and pellets of 5 mm were conducted to investigate the effect of adsorbent size on water/ethanol adsorption. The remaining operating conditions were kept constant at T = 100 °C, Ptotal = 243 kPa, u0 = 0.9 m/s, and Pw = 45 kPa. The results are shown in Table 3 and Figure 7. From Table 3, it can be seen that the separation factors at breakthrough decreased as the 5 mm pellet was applied. This
Figure 4. Variation of temperature with time in the middle of column read from (J-111).
5, 10, and 15 wt % water content in the ethanol−water mixture. Ethanol production profiles and water breakthrough curves at different feed concentrations are shown in Figure 5. The calculated values for water and ethanol uptake at breakthrough and equilibrium are listed in Table 2. From Figure 5a, it can be seen that the breakthrough time decreased as the water concentration was increased in the feed stream. Breakthrough times of 42, 30, and 21 min were
Figure 5. (a) Ethanol production profiles. (b) Water breakthrough curves. Operating conditions: T = 110 °C, u0 = 0.9 cm/s, dp = 0.43− 1.18 mm. 6658
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
Table 2. Water and Ethanol Adsorption on Canola Meal Particles (0.43−1.18 mm) at Various Water/Ethanol Partial Pressures (Total Pressure 243 kPa, Superficial Velocity 0.9 cm/s) operating conditions
equilibrium adsorption
adsorption until breakthrough point (99 wt % ethanol)
T (°C)
water content (wt %)
Pwater (kPa)
PEtOH (kPa)
H2O uptakea
EtOH uptakea
α
H2O uptakea
EtOH uptakea
α
recovery (%)b
110 110 110
5 10 15
24 45 85
178 160 125
1.10 2.03 4.31
2.52 1.64 1.59
3.34 4.38 4.04
0.58 0.89 1.32
1.16 0.99 0.64
3.65 3.17 3.10
73 69 68
a
mol/kg adsorbent. bAmount of collected ethanol (over 99 wt %) divided by amount of ethanol entered into column until breakthrough point.
Figure 7. (a) Ethanol production profiles. (b) Water breakthrough curves. Operating conditions: T = 100 °C, u0 = 0.9 cm/s, and Pw = 45 kPa. Figure 6. Temperature profiles at middle of column read from (J-111) for runs with different water partial pressure.
3.5. Simulation of Water Breakthrough Curves. Simulation of water breakthrough curves may assist to identify mass transfer resistance and provide modeling results for engineering design. In general, the mathematical models used to simulate water breakthrough curves in an adiabatic packed bed adsorption process involve a set of partial differential and algebraic equations, including mass, energy, and momentum balance and an adsorption rate equation,20,21 which usually require complex computation. For a packed bed with temperature controlled, a simplified model has been developed by applying mass balance of the fluid phase on a differential element of the bed (dz), in which the fluid stream contains a sorbate with concentration varying with axial position z and time t, c(z,t). In this model, negligible axial dispersion, constant fluid velocity, and the linear driving force (LDF) model were assumed.21,23
22
conforms to the results obtained by Kim et al., which show that adsorbents with smaller particle sizes have better separation capacity. Consequently, the ethanol recovery slightly increased from 65% for the 5 mm pellet to 68% for the adsorbent with particle sizes in the range 0.43−1.18 mm. Further, the separation factors at breakthrough decreased as the 5 mm pellet was applied. This conforms to the results obtained by Kim et al22 that adsorbents with smaller particle sizes have better separation capacity. Water uptake at breakthrough was 1.37 (mol/kg adsorbent) for a run using particles of size 0.43−1.18 mm and 1.49 (mol/ kg adsorbent) for 5 mm pellet. In the case of ethanol, the corresponding values were 1.53 and 1.91 (mol/kg adsorbent). Similarly, a trend of slight increase in water/ethanol uptake at equilibrium was observed when the particle size was increased. This increase in water/ethanol uptake could be explained by the fact that packing density rose from 499.78 to 543.17 kg/m3 as the particle size was increased, since canola meal particles are compressed together through the pellet making process to prepare pellets with 5 mm diameter from raw canola meal particles. Increase of the packing density probably improved the contact between the sorbate and the adsorbent. The fact that water uptake on 5 mm canola meal pellets was not reduced compared to that of 0.43−1.18 mm particles is a positive result toward applying canola meal adsorbent in the industrial process for ethanol dehydration.
(1 − εb) ∂q ̅ ∂c ∂c +u + =0 ∂t ∂z εb ∂t
(2)
The first term represents accumulation rate of the sorbate, the second one represents the convection term, and the third one is adsorption rate based on q,̅ the volume average adsorbent loading per unit mass (uptake), which accounts for the variation of q throughout the adsorbent particle, by averaging the rate of adsorption over the adsorbent particle. εb is bed porosity, z is the bed depth (m), and u is the interstitial velocity of vapor, calculated by
Table 3. Water/Ethanol Adsorption on Canola Meal Particles at Various Sizes (Total Pressure 243 kPa, Superficial Velocity 0.9 cm/s, and 100 °C) operating condition
equilibrium adsorption
adsorption until breakthrough point (99 wt % ethanol)
water content (wt %)
Pwater (kPa)
PEtOH (kPa)
dp (mm)
H2O uptakea
EtOH uptakea
α
H2O uptakea
EtOH uptakea
α
recovery (%)b
10 10
45 45
160 160
0.43−1.18 5.00
2.92 3.38
3.38 4.15
3.06 2.95
1.37 1.49
1.53 1.91
3.18 2.78
68 65
a
mol/kg adsorbent. bAmount of collected ethanol (over 99 wt %) divided by amount of ethanol entered into column until breakthrough point. 6659
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels u=
Article
u0 εb
erf( −x) = −erf(x) (3)
∂q ̅ = kLDF(q* − q ̅ ) ∂t
To correlate the sorbate uptake in the solid phase with its concentration in the fluid phase, a linear adsorption equilibrium model was used.24,27 (5)
q ̅ = Kc
(6)
RT =
t > 0,
c(t , 0) = c0 , c(t , L) = ce
R ex =
ξ +
(8)
Re = Sc =
(9)
1 8 τ
⎤ 1 ⎞⎥ ⎟⎟ 8 ξ ⎠⎥⎦
kLDFKz ⎛ 1 − εb ⎞ ⎜ ⎟ u ⎝ εb ⎠
⎛ z⎞ τ = k ⎜t − ⎟ ⎝ u⎠
R pK 3kc
(18)
Dpρu0 μ(1 − εb)
(19)
μ pDm
(20)
Sh = 2 + 1.1Sc1/3Re 0.6
(21)
The external mass transfer coefficient kc of particles in the fixedbed in eq 18 was determined from the following correlation: Sh =
(11)
where ξ is the dimensionless distance coordinate and τ is the dimensionless time coordinate corrected for displacement. The equations for ξ and τ are
ξ=
(17)
where μ is the viscosity of vapor in kg/(m s), ρ is the vapor density in kg/m3, DP is the equivalent diameter of a spherical particle in m, u0 represents the superficial velocity of vapor in m/s, and Dm is the molecular diffusivity for a gas mixture in m2/ s. Then, using the values of Re and Sc, the Sherwood number Sh was calculated through eq 21:21
(10)
+
(16)
where kc is the external mass transfer coefficient (m/s), Rp is the adsorbent particle radius (m), and K is the equilibrium constant at temperature T. Thus, the internal mass transfer resistance is evaluated as the difference of the external mass transfer resistance and the overall mass transfer resistance. To evaluate the external mass transfer coefficient kc, it was necessary to calculate Reynolds number Re and Schmidt number Sc:
where z (bed depth) varies from 0 to L, and c0 and ce are the concentrations of the sorbate in the feed stream and effluent, respectively. The value of the bed depth L was 0.5 m for all experimental runs. The simulation of a breakthrough curve requires solving eq 8 subject to initial and boundary conditions (eqs 9 and 10). The following approximate solution was obtained by Klinkenberg:23,24 ⎡ ⎛ ce 1 ≈ ⎢1 + erf⎜⎜ τ − c0 2 ⎢⎣ ⎝
KLDF
The external mass transfer resistance was evaluated by
The initial and boundary conditions for an initially dry bed that was exposed to a step change in sorbate concentration at the inlet at time zero are as follows:21 q ̅ (0, z) = c(0, z) = 0
1
30
where kLDF is the overall mass-transfer coefficient (1/s), which includes both external and internal transport resistances. Combining eqs 2, 3, and 7 gives
t = 0,
(15)
RT = R ex + R in
(7)
(1 − εb)(kLDFK ) ∂c ∂c +u + (c * − c ) = 0 ∂t ∂z εb
2
e −η dη
The overall mass transfer resistance is also correlated to the external (Rex) and internal (Rin) mass transfer resistances by30
where q* is the saturated sorbate loading in equilibrium with the sorbate concentration, c* in the bulk fluid, c is the sorbate concentration in equilibrium with average loading q,̅ and K is the adsorption equilibrium constant. Combining eqs 4, 5, and 6 gives ∂q ̅ = kLDFK (c* − c) ∂t
∫0
(14)
Chang et al. used this model to simulate the water breakthrough curves in a corn meal packed bed with the temperature controlled.28,29 Through determination of kLDF by fitting eq 11 to the experimental data, the overall mass transfer resistance RT, measured in seconds, was calculated:30
(4)
q* = Kc*
2 π
erf(x) =
where u0 is the superficial velocity in the column (m/s). Then, the LDF model introduced by Glueckauf,24−26 was applied to replace the adsorption rate in eq 3 by
x
Dpkc Dm
(22)
In the correlations mentioned earlier, DP was introduced as the equivalent diameter of a spherical particle. There are some correlations used to calculate DP from the geometric properties of the particles.24 For crushed particles of irregular surface, with no obvious longer or shorter dimension, Dp equals 4 times the hydraulic radius rH, where for a packed bed 4rH equqls 1.0D,24 where D is the average diameter of particles (m). Thus,
(12)
Dp = 1.0D
(13)
ξ and τ are defined as coordinate transformations for z and t in order to convert the equations to an equation with a much simpler form of the error function erf(x). The error function is defined by
(23)
In the case of 0.43−1.18 mm canola particles, D was taken to be D(v, 0.5) determined by the results in Section 3.1, being 0.57. Once kc was calculated, the external mass transfer resistance was determined by eq 18, and the internal resistance 6660
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
was calculated as the difference between the total resistance and the external resistance. As mentioned before, in this work, the experiments were carried out in a packed bed with temperature controlled through an oil jacket. In order to investigate the applicability of the model (eq 11) to simulate the water breakthrough curves achieved in this work, the water adsorption equilibrium constant K was determined through fitting the equilibrium water uptake data with a linear isotherm. The parameters are listed in Table 4. It is noted that simulation of the water adsorption isotherm using sophisticated models is necessary and could be an area of future work.
Table 5. Mass Transfer Coefficients at Various Temperaturesa
dp (mm)
K
R2
100 105 110
0.43−1.18 0.43−1.18 0.43−1.18
269 198 157
0.96 0.96 0.99
kc (× 10 m/s)
RT (s)
Rex (s)
Rin (s)
100 105 110
0.29 0.39 0.40
0.64 0.65 0.67
370.37 256.41 250.00
0.36 0.26 0.20
370.01 256.15 249.80
Canola meal particle size 0.43−1.18 mm, water partial pressure 24 kPa, total vapor pressure 243 kPa, superficial velocity 0.9 cm/s.
The results show that the model reasonably well simulated the experimental breakthrough curves. As was mentioned before, there was a temperature rise during the dehydration process because the generated heat could not be removed promptly from the bed by the oil jacket though the temperature was eventually controlled at the set values. Fortunately, the maximum temperature rise in the bed in comparison with the set temperature for the experiments shown in Figure 4, which are associated with the experimental water breakthrough results shown in Figure 8 was lower than 6.4%. In addition, feedwater content in the ethanol−water mixture for these runs was 5 wt %. Water adsorption from the vapor stream did not cause significant variation of the fluid velocity. As a result, the model (eq 11) is able to simulate the water breakthrough curves. The obtained values of the overall mass transfer coefficient kLDF, presented in Table 5, increased from 2.9 to 4 ( × 103 1/s) as the temperature was elevated from 100 to 110 °C. This indicates that the mass transfer rate increased as the temperature was increased. This is also reflected by the slight increase in the slopes of water breakthrough curves as can be seen in Figure 8. The larger slope suggests a higher mass transfer rate.27 The overall mass transfer coefficient obtained from the modeling results were further used to calculate the overall mass transfer resistance using eq 17. The calculated values for different resistances are also presented in Table 5. The results show that the overall external and internal mass transfer resistances decreased as the temperature was increased. Simo and his co-workers4 reported a similar decreasing trend of the resistances with an increase in the temperature for water adsorption on 3A zeolite. From Table 5, it can be also seen that more than 98% of the overall mass transfer resistance was due to the internal region, which indicates that the internal mass transfer resistances governed the adsorption process. Simulations for water breakthrough curves obtained at 5, 10, and 15 wt % feedwater content corresponding to water partial pressures of 25, 45, and 85 kPa, respectively, and 100 °C are shown in Figure 9. Though satisfactory simulation was achieved at a water content of 5 wt % in the ethanol−water feed mixture with a coefficient of determination of 0.99, the model was not able to accurately simulate the data obtained at feedwater contents of 10 and 15 wt %. This is because water adsorption is exothermic. More water input led to more water adsorbed and more heat generated. The limitation of prompt heat removal by the oil jacket made higher temperature rise in the bed unavoidable. The maximum bed temperature rise in the run with 10 wt % water in the feed increased to 9% of the set bed temperature (100 °C here), which further increased to 15% in the run with 15 wt % water in the feed. In addition, fluid velocity changed more significantly during water adsorption in the cases of 10
Then, the water breakthrough model (eqs 11, 12, and 13) was used to fit the experimentally obtained water breakthrough curves. The overall mass transfer coefficient kLDF was estimated by minimizing the following objective function: ⎡⎛ ⎞ ⎤2 ⎛ ⎞ c c − ⎜ e⎟ ⎥ min f = ∑ ⎢⎜ e ⎟ ⎢⎝ c 0 ⎠ ⎝ c0 ⎠exp⎥⎦ ⎣ model
kLDF (× 102 1/s)
a
Table 4. Equilibrium Constants for Water Adsorption on Canola Meal T (°C)
T (°C)
(24)
In the simulation of the water breakthrough curves using eq 11, the coefficient of determination R2 was also calculated in order to evaluate how well the model fits the experimental data. The simulated breakthrough curves for the data achieved at 5 wt % water content in the feed ethanol−water mixture and temperatures 100, 105, and 110 °C are presented in Figure 8 and parameters are shown in Table 5. The obtained values of R2 at 100, 105, and 110 °C were 0.99, 0.98, and 0.98, respectively.
Figure 8. Prediction of water breakthrough curves at (a) 110 °C, (b) 105 °C, and (c) 100 °C. Operating conditions: Pw = 24 kPa, dp = 0.43−1.18 mm, u0 ≈ 0.9 cm/s. τ is the dimensionless time coordinate corrected for displacement (τ = k(t − z/u)). The data points represent experimental results. 6661
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
performed on fresh canola meal and on regenerated canola meal (after 32 cycles), and the results are displayed in Table 6. Table 6. CHNS Analysis of Fresh and Used Canola Meal Samples adsorbent
N%
C%
S%
H%
fresh canola meal used canola meal
6.51 6.62
48.03 47.02
0.63 0.63
6.34 6.28
The results demonstrate that the contents of the major elements in canola meal, after being used for 32 cycles, were almost the same as those of fresh canola meal, indicating that canola meal is a stable material for ethanol dehydration. Figure 11 shows that the functional groups of canola meal did not change through the adsorption cycles; FTIR spectra of
Figure 9. Prediction of breakthrough curves for runs with water partial pressures 45 and 85 kPa. Operating conditions: Ptotal = 273 kPa, dp = 0.43−1.18 mm, T = 110 °C, u0 = 0.9 cm/s. τ is the dimensionless time coordinate corrected for displacement (τ = k(t − z/u)).
and 15 wt % water content compared with that of 5 wt % water content. Furthermore, the water adsorption isotherm deviated from linear form at elevated water content in the vapor. The results demonstrate that the model, without considering variations of temperature and fluid velocity during water adsorption, and assuming the linear equilibrium isotherm, is inapplicable to the system with water feed content at the level of 10 wt % and above. Similarly, the model was not able to simulate the water breakthrough curves achieved at different particle sizes and 10 wt % water input (data not shown here). 3.6. Regeneration and Reusability. In order to investigate the regeneration and reusability of the canola meal adsorbent, the water-saturated bed was regenerated by purging nitrogen at 110 °C under vacuum for 4 h. Furthermore, the weight of the canola meal packed in the bed was measured after desorption and compared with the initial dry weight before adsorption; the difference in weight was about 0.44%, which confirms the bed was completely dried. Then, the regenerated bed was reused for ethanol dehydration. The water breakthrough curves obtained using the bed packed with fresh canola meal and after regeneration are presented in Figure 10a. The
Figure 11. Fourier transform infrared (FTIR) spectra of fresh and regenerated canola meal. T: transmittance.
canola meal were nearly identical for fresh and regenerated canola meal samples, which confirms the stability of canola meal as the adsorbent in ethanol dehydration.
4. CONCLUSIONS In summary, the lab-scale pressure swing adsorption process (PSA) apparatus was designed and built to investigate water and ethanol adsorption behaviors of canola meal. Canola meal showed a good potential for water adsorption and it can be used for the separation of water−ethanol mixture. The experimental results demonstrate that canola meal was able to break the azeotropic point (95.6 wt %) and produce ethanol with concentration over 99 wt % in the effluent. The highest ethanol recovery was 75%, achieved at 100 °C, total pressure of 243 kPa, feed concentration of 95 wt % EtOH (5 wt % water), and particle size of 0.43−1.18 mm. The effects of the temperature, feed concentration, and particle size on water breakthrough curves and the ethanol production profile were investigated. As the set operation temperature for ethanol dehydration was increased, water and ethanol uptake were decreased. The adsorption demonstrated an exothermic nature. The water separation factor was about 3.3−3.6. Higher water uptake and higher mass transfer rates were achieved at higher water content in the feed. Increased canola meal particle size slightly decreased the water mass transfer rate, however, it slightly increased the water uptake. This is probably because the bigger size 5 mm pellets were made by a mill through compression. As a result, the packing
Figure 10. (a) Water breakthrough curves. (b) Temperature profiles. Operating conditions: T = 110 °C, Pw = 24 kPa, u0 = 0.9 cm/s, dp = 0.43−1.18 mm. Test 1: fresh bed. Test 2: regenerated bed. Tm: temperature at the middle of the bed. Tb: temperature at the bottom of the bed.
temperature profile is shown in Figure 10b. The overlapped water breakthrough curves and temperature profiles demonstrate that the regenerated bed has similar water adsorption performance to the fresh one. The canola meal has been used for over 32 cycles without deteriorated quality, demonstrating that canola meal is stable and reusable in the application for ethanol dehydration. Furthermore, CHNS analysis was 6662
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
density was increased, which probably improved the contact between the sorbate and the adsorbent. The Klinkenberg model simulates reasonably well the water breakthrough curves achieved at 5 wt % water content in the feed stream, 100−110 °C, total vapor pressure 243 kPa, and superficial velocity 0.9 cm/s. The modeling results indicate that internal mass transfer resistance governed the water adsorption process. However, it failed to accurately simulate the experimental data with water feed content at the level of 10 wt % and above. The water-saturated packed bed was successfully regenerated by purging nitrogen at 100 °C under vacuum for 4 h. The canola meal has been used for over 32 cycles of dehydration and regeneration without deteriorated quality, demonstrating that canola meal is stable and reusable in the application for ethanol dehydration. However, though the water uptake of canola meal is high, the selectivity of water over ethanol is low. To enhance the water selectivity could be a focus of future work in this area.
■
■
AUTHOR INFORMATION
Corresponding Author
*Tel.: 1 306 9662174. Fax: 1 306 9664777. E-mail: catherine.
[email protected]. Notes
■
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Financial assistance from the Natural Science and Engineering Research Council of Canada, Saskatchewan Canola Development Commission, Saskatchewan Agriculture Ministry and Pound-Maker Investments is acknowledged.
Rp = adsorbent particle radius (mm) RT = overall mass transfer resistance (s) rH = hydraulic radius (m) Re = Reynolds number (dimensionless) Sc = Schmidt number (dimensionless) Sh = Sherwood number (dimensionless) Xe = mass fraction of ethanol in the adsorbed (dimensionless) Xj = mole fraction of component j (dimensionless) Xw = mass fraction of water in the adsorbed (dimensionless) Ye = mass fraction of ethanol in the vapor (dimensionless) Y w = mass fraction of water in the vapor (dimensionless) ybar = the average of yi,exp u = interstitial velocity of vapor (m/s) u0 = superficial velocity of vapor (m/s) z = bed depth (m)
phase phase phase phase
GREEK LETTERS α = separation factor εb = bed porosity μ = viscosity of vapor (kg/m·s) ρ = vapor density (kg/m3) REFERENCES
(1) Kumar, S.; Singh, N.; Prasad, R. Renewable Sustainable Energy Rev. 2010, 7, 1830−1844. (2) Frolkova, A.; Raeva, V. Theor. Found. Chem. Eng. 2010, 4, 545− 556. (3) Vareli, G. D.; Demertzis, P. G.; Akrida-Demertzi, K. Zeitschrift für Lebensmitteluntersuchung und Forschung A 1998, 2, 122−127. (4) Simo, M.; Sivashanmugam, S.; Brown, C. J.; Hlavacek, V. Ind. Eng. Chem. Res. 2009, 20, 9247−9260. (5) Simo, M.; Brown, C. J.; Hlavacek, V. Comput. Chem. Eng. 2008, 7, 1635−1649. (6) Black, C. Chem. Eng. Prog. 1980, 78−85. (7) Tindall, B. M.; Natarajan, R. S. In In Production of Anhydrous Ethanol by Pressure Swing Adsorption; AICHE: New York, 1987. (8) Carmo, M. J.; Gubulin, J. C. Adsorption 2002, 3, 235−248. (9) Okewale, A.; Etuk, B.; Igbokwe, P. Int. J. Eng. Technol. 2011, 11, 81. (10) Beery, K. E.; Ladisch, M. R. Enzyme Microb. Technol. 2001, 7, 573−581. (11) Boonfung, C.; Rattanaphanee, P. World Academy of Science, Engineering, and Technology 2010. (12) Kim, Y.; Hendrickson, R.; Mosier, N.; Hilaly, A.; Ladisch, M. R. Ind. Chem. Eng. Res. 2011, 14, 8678−8685. (13) Ladisch, M. R.; Voloch, M.; Hong, J.; Bienkowski, P.; Tsao, G. T. Ind. Eng. Chem. Process Des. Dev. 1984, 3, 437−443. (14) Hu, X.; Xie, W. Sep. Sci. Technol. 2001, 1, 125−136. (15) Chang, H.; Yuan, X. G.; Tian, H.; Zeng, A. W. Ind. Eng. Chem. Res. 2006, 11, 3916−3921. (16) Benson, T. J.; George, C. E. Adsorpt J. Int. Adsorpt Soc. 2005, 697−701. (17) Al-Asheh, S.; Banat, F.; Al-Lagtah, N. Chem. Eng. Res. Des. 2004, 7, 855−864. (18) Baylak, T.; Kumar, P.; Niu, C. H.; Dalai, A. Energy Fuels 2012, 8, 5226−5231. (19) Aider, M.; Barbana, C. Trends Food Sci. Technol. 2011, 1, 21−39. (20) Chahbani, M.; Tondeur, D. Sep. Purif. Technol. 2000, 2, 185− 196. (21) Ruthven, D. M. Principles of Adsorption and Adsorption Processes; Wiley: New York, 1984. (22) Khan, A. A.; Singh, R. Colloids Surf. 1987, 1, 33−42.
■
NOMENCLATURE C = water content at time t (wt %) c = concentration of water at time t and position z along the bed (mol/L) c0 = water content in the feed stream (wt %) ce = water content in the effluent (wt %) c* = sorbate concentration in the bulk fluid in equilibrium with q* (mol/m3) Dm = diffusion coefficient for a multicomponent gas mixture (m2/s) D = average diameter of particles (m) Dp = equivalent diameter of a spherical particle (m) dp = particle size diameter (mm) DAB = diffusion coefficient in binary gas mixture (m2/s) erf(x) = error function K = equilibrium thermodynamic constant (dimensionless) kc = external mass transfer coefficient (m/s) kLDF = overall mass transfer coefficient (1/s) P = pressure (kPa) PEtOH = ethanol partial pressure (kPa) Pwater = water partial pressure (kPa) Ptotal = total pressure (kPa) q = water uptake (mol H2O/kg adsorbent) q* = saturated sorbate loading in equilibrium with c* (mol/ m3) q̅ = average sorbate loading in equilibrium with c (mol/m3) R2 = coefficient of determination Rex = external mass transfer resistance (s) Rin = internal mass transfer resistance (s) 6663
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
Energy & Fuels
Article
(23) Klinkenberg, A. Ind. Eng. Chem. 1954, 11, 2285−2289. (24) Seader, J. D.; Henley, E. J. Separation Process Principles; Wiley: New York, 1998. (25) Glueckauf, E. Trans. Faraday Soc. 1955, 1540−1551. (26) Glueckauf, E.; Coates, J. J. Chem. Soc. 1947, 1315−1321. (27) Sereno, C.; Rodrigues, A. Gas Sep. Purif. 1993, 3, 167−174. (28) Chang, H.; Yuan, X.; Tian, H.; Zeng, A. Chem. Eng. Technol. 2006, 4, 454−461. (29) Chang, H.; Yuan, X.; Tian, H.; Zeng, A. Chem. Eng. Process 2006, 9, 747−754. (30) Gorbach, A.; Stegmaier, M.; Eigenberger, G. Adsorption 2004, 1, 29−46.
6664
dx.doi.org/10.1021/ef400897e | Energy Fuels 2013, 27, 6655−6664
