Design of Cartridge-Based Ceramic Heat-Exchanger Microchannel

Feb 28, 2013 - This system employs a ceramic microchannel cartridge with catalyst ... hydrogen yields of ∼80% are achievable using the cartridge-bas...
0 downloads 0 Views 726KB Size
Article pubs.acs.org/EF

Design of Cartridge-Based Ceramic Heat-Exchanger Microchannel Reformers for Process Intensification: Experiments and Simulations Shalini Damodharan, Bhanu Vardhan Reddy Kuncharam, and Benjamin A. Wilhite* Artie McFerrin Department of Chemical Engineering, Texas A&M University, College Station, Texas 77843-3122, United States ABSTRACT: This paper details a combination of experimental and theoretical design analyses of a cartridge-based microchannel reformer system capable of integrating two or three separate reforming processes (reactant preheating, methanol steam reforming, and methanol combustion for autothermal operation) within a single monolithic device in a two-dimensional or radially layered distribution pattern. This system employs a ceramic microchannel cartridge with catalyst configurations tailored to enable stable autothermal operation over a broad range of reforming and combustion flow rates. Operation of the 25-channel prototype system coupling methanol combustion in air (13 mol % CH3OH and 17.3 mol % O2) with steam reforming of a dilute (2.6 mol %) methanol−water mixture at combustion and reforming overall flow rates of 300 standard cubic centimeters per minute (sccm) [gas hourly space velocity (GHSV) of 19 200 h−1] and 1800 sccm (GHSV of 14 400 h−1) achieved steam reforming hydrogen yields of ∼85%, corresponding to an overall hydrogen yield of 53%. When the outer layer of microchannels is employed for preheating of the reforming stream, the overall hydrogen yield was improved to 57%. A three-dimensional simulation of the microchannel reformer was constructed and validated through comparison to experimental data and then employed to predict the reformer performance using a concentrated (25 mol % CH3OH and 75 mol % H2O) steam reforming feed. Design simulations predict that hydrogen yields of ∼80% are achievable using the cartridge-based ceramic microchannel reformer.

1. INTRODUCTION The proliferation of portable electronic devices for both civilian1−3 and military4 use requires advances in energy densities and operational times of portable power systems. Hydrogen-driven fuel cells offer high energy densities, low cost, and reduced environmental impact, offering a promising alternative to batteries for meeting portable power consumer [e.g., laptop computers, personal digital assistants (PDAs), and cellular phones] and military (e.g., communication devices, microsensors, and actuators) needs.5 Compressed hydrogen fuel provides high gravimetric (39 700 W h kg−1) energy densities but poor volumetric energy density (2100 W h L−1 at 10 000 psi) alongside additional challenges associated with storage and transportation compared to liquid fuels (e.g., methanol). In practice, even the high gravimetric energy density associated with compressed hydrogen is significantly reduced by the weight of the containment vessel. For these reasons, onboard hydrogen production via catalytic reforming of energy-dense liquid fuels remains an attractive option for portable power systems. Efficient stand-alone reforming of hydrogen fuels requires effective thermal integration of multiple physical and/or chemical processes (e.g., preheating, reforming, and combustion), each with their own unique heat duty and optimal operating temperatures.6,7 Microchannel reactors have been identified as a leading technology for realizing all-in-one portable reformers capable of autothermal hydrogen generation via complex integration of several process volumes within a single compact device.7−9 Individual microchannels facilitate this process intensification by reducing the characteristic length scales associated with heat and mass transport, achieving order-ofmagnitude improvements in transfer rates between process volumes.10,11 Precision machining (e.g., lithography-based techniques and electric discharge machining) enables high pattern density and resolution for constructing microfluidic systems © XXXX American Chemical Society

capable of realizing complex two- and three-dimensional reactor architectures for greater portability, thermal management, and process intensification.7−11 However, construction method and material selection remain critical challenges to the realization of robust, cost-effective, and thermally efficient microchannel reformers. Existing construction methods based on the lamination of individually machined or patterned plates limits the complexity of radial distribution patterns to one-dimensional architectures by relying on the stacking of additional layers to increase overall capacity or to integrate multiple processes within the microreactor system.7,12 This “numbering up” approach also results in an economically unfavorable near-linear cost of scaleup. Incorporation of both the fluidic (or distribution) elements and catalytic reaction volumes within a single monolithic device adversely affects system maintenance, because failure of the latter via plugging or catalyst deactivation would require replacement of the entire system at significant cost. Material selection also plays an important role in maximizing thermal efficiency through managing both radial and axial heat conduction. High thermal conductivity materials (e.g., silicon or stainless steel) require the introduction of internal insulation layers to maintain significant thermal gradients within the microchannel network13−15 while also requiring complex architectures and vacuum packaging to minimize heat losses to surroundings.16,17 For these reasons, low thermal conductivity ceramics (e.g., alumina and cordierite) have emerged as promising material of construction Special Issue: Accelerating Fossil Energy Technology Development through Integrated Computation and Experiment Received: December 20, 2012 Revised: February 26, 2013

A

dx.doi.org/10.1021/ef3021359 | Energy Fuels XXXX, XXX, XXX−XXX

Energy & Fuels

Article

Figure 1. Schematic of the ceramic microchannel network employed for autothermal hydrogen production from methanol: (a) cross-section schematic of the 25-channel network illustrating two-dimensional radial distribution for integrating separate combustion, steam reforming, and heat recovery volumes and (b) axial cross-section schematic of the 25-channel network illustrating packed-bed catalyst configurations for each separate volume. SR, steam reforming; C, combustion; and HX, heat exchange, insulation volumes.

Figure 2. Heat-integrated microchannel reformer system: (a) brass distributors, in the unassembled view, (b) 25-channel ceramic cartridge, (c) assembled reformer system, with insulation removed, and (d) thermal image of the system during operation, highlighting surface temperatures of reactor, distributors, and insulation.

microchannel reformer achieved stable autothermal hydrogen production with an overall hydrogen yield of ∼13% while maintaining safe (15 h) at temperatures below the recommended maximum operating temperature (350 °C). The increased thermal capacitance of the combustion catalyst packed bed, coupled with the introduction of low thermal conductivity (open) entrance and exit regions to provide a thermal buffer region to prevent excess heat transport to the packaging/ distributors, resulted in hotspot stabilization at the axial midpoint over a broad range of process flow rates. Nickel mesh (100 μm, Alfa Aesar) of 0.4 cm length was used as a catalyst retainer on either side of each packed bed to ensure catalyst immobilization. 2.4. Microchannel Reformer Assembly. The microchannel reformer was assembled by aligning the brass distributors to either face of the extruded ceramic microchannel cartridge and reversibly sealing using a four-point compression chuck (Figure 2c), with silicon gaskets (1/16 in. McMaster-Carr) placed within the recessed face of either distributor block. The ceramic microchannel network was then encased in a two-piece insulating jacket cut from a firebrick block of 2.5 cm thickness × 13.2 cm length and 4.2 cm width with molded internal recess to allow for insertion of a multi-node thermocouple, for providing external insulation to the system. The two-piece insulating jacket was manufactured with a ∼3 mm gap that enabled monitoring of the reactor surface temperature via a thermal-imaging camera during operation (Figure 2d). 2.5. Microchannel Network Operation. The experimental apparatus used to investigate the performance of the assembled microchannel reformer over a range of process flow rates and flow

microchannel reformer design for autothermal production of hydrogen from methanol. Experiments are performed using a dilute methanol steam reforming mixture to facilitate gas metering and analysis. Experimental results are employed to provide validation of a robust three-dimensional reactor model constructed using the COMSOL version 3.5a programming environment. The resulting design model is employed to predict system performance over a range of process flow rates in terms of internal and external temperature gradients, individual and overall process methanol conversions, and hydrogen yields.

2. EXPERIMENTAL SECTION 2.1. Distributor Construction. Brass plates were used for the distributor construction to enable the use of in-house facilities. Four 1 /8 × 2 × 2 in. brass plates were machined with two-dimensional patterns (shown in Figure 2a) of size 1/16 in. on either side of the plate with an intermediate pattern to achieve fluidic connections across each plate. Among the four brass plates used for distributor construction, the three topmost plates were of 1/8 in. thickness to reduce the weight and size of the distributor; whereas the bottommost plate was of 1 in. thickness to accommodate standard 1/8 in. nominal pipe thread (NPT) fittings to provide fluidic connections to the microreactor assembly. The topmost plate was fabricated with a 1/16 in. deep recess to accommodate a silicon gasket for reversible compression sealing of the distributor to the ceramic microchannel network. Final assembly of the distributor was achieved by laminating the four-plate stack using compression sealing with a 0.065 in. thick silicon gasket (McMasterCarr) between each plate. Fluidic connections in the distributors were designed to compartmentalize reaction streams to their respective channels according to the radial distribution pattern detailed in Figure 1a. 2.2. Ceramic Microchannel Network Construction. Individual cartridges consisting of a 5 × 5 network of 25 parallel microchannels (Figure 2b) were cut from a dense-fired cordierite honeycomb monolith substrate (Rauschert Technical Ceramics, Denmark) with a cell density of 72 cells per square inch (CPSI). Each channel is of 2.5 mm width and 150 mm length, with the overall dimension of the ceramic microchannel reactor as 15 × 15 × 150 mm. Each face of the reactor was polished using 8 in. diameter 60/P60 grit abrasive paper to extrude the ceramic reactor to the brass distributor. The cordierite microchannels were washcoated with nanoalumina slurry to close the fine pores of the monolith, so that there would not be any crossover between channels. Nanoalumina solution was prepared by mixing aluminum oxide nanopowder (from Sigma Aldrich), 20% colloidal aluminum oxide in water (from Alfa-Aesar), distilled H2O, and methanol in a mass ratio of 2:1:21:21. The microchannel reactor was dipped into the nanoalumina slurry for 5 min and sonicated for 1 h at room temperature. The above procedure was repeated 3 times with C

dx.doi.org/10.1021/ef3021359 | Energy Fuels XXXX, XXX, XXX−XXX

Energy & Fuels

Article

Product stream elements were identified by retention times, and individual species peak areas were analyzed from GC samples to obtain product gas compositions. Mild degradation in reforming catalyst activity was observed over time spans of >15 h because of the reaction of CuO to Cu under reducing environments when reforming temperatures exceeded 350 °C. For this reason, a new monolith packed with fresh combustion and reforming catalyst was used for each set of experiments to ensure consistent catalytic activity during data collection. It is worth noting that, while the catalyst employed in the present study displayed mild degradation as a result of prolonged exposure to temperatures in excess of design temperatures (350 °C), the reactor performance was found to be highly reproducible between catalyst loadings, confirming that the catalyst activity under all experiments was consistent with an activated, stable performance. Methanol and methanol−water solutions employed by gas washers were replaced at every step change in the reforming flow rate, to ensure consistent reforming gas composition at each set of flow conditions. The reactor performance was measured over a range of combustion and reforming flow rates, such that the excess potential combustion heat relative to the steam reforming heat duty is varied. This mismatch in the heat balance between combustion and reforming processes is intentionally selected to compensate for heat losses to packaging and ambient for the present non-adiabatic self-insulating microreactor design. Methanol conversion in both combustion and reforming volume was calculated on the basis of carbon balance via eq 1. Individual combustion and reforming process product yields (eq 2), combined hydrogen yield (eq 3), and reforming process selectivity to carbon monoxide over methane and carbon dioxide (eq 4) were calculated from effluent molar flow rates. Thermal efficiency of the heatexchanger microreactor system was defined as the fraction of combustion heat used as sensible heat by the endothermic reaction (eq 5).

configurations is shown schematically in Figure 3. In all cases, the combustion gas was comprised of a dry gas blend of 20% O2/5% He/ 75% Ar, which was humidified to achieve 13 mol % methanol using a gas washer containing 500 mL of 99.9% purity methanol (J.T. Baker) maintained at 20 °C. Thus, the combustion gas entering the reformer is 13 mol % CH3OH, 17.3% O2, balance Ar. The reforming gas consisted of a 5% N2/95% Ar dry gas blend humidified to a stoichiometric composition of 2.6 mol % methanol and 2.6 mol % water via a second gas washer containing 118 mL of methanol and 282 mL of water maintained at 25 °C. Methanol composition exiting each bubbler was validated via separate packed-bed combustion experiments as follows. A mixture of 5% He/20% O2/75% Ar was supplied to each bubbler and then to a packed-bed reactor containing approximately 5 g of 1 wt % Pt−Al2O3 catalyst. Gas chromatographic analysis of the dry effluent CO2 and O2 composition enabled calculation of methanol composition provided by the bubbler, assuming complete combustion, as validated by the absence of CO. Combustion and reforming dry gases were blended from separate streams of ultrahighpurity Ar, 5% N2/95% Ar, and 5% He/20% O2/75% Ar using individual Alicat mass flow controllers. All flow rates in this paper are subsequently reported on a dry gas basis, i.e., prior to the addition of condensable reactants (steam and methanol). All gas mixtures were supplied by Airgas. The surface temperature was measured at seven equally spaced points along the axial length of the reactor via a 7-node thermocouple probe (Omega) attached to the outer surface of the ceramic microchannel cartridge. Thermal images of the microchannel reformer were obtained using a FLIR ThermoCAM infrared camera to capture the hotspot location and magnitude during steady-state operation (Figure 2d). Helium and nitrogen were employed as internal standards for determining combustion and reforming effluent flow rates on a dry gas basis, respectively. Both combustion and reforming product streams were analyzed on dry gas basis via a dual-column (MolSieve and Plot-Q) Agilent MS3000 gas chromatograph equipped with thermal conductivity detection and employing argon as the carrier gas. A sequence of ice (0 °C) and dry ice (−80 °C) condensers were used to remove any moisture content from each gas stream prior to sampling by a gas chromatograph. Digital collection and peak integration of resulting chromatographic data were achieved using the Cerity software package. In all experiments, reactor startup was achieved as follows. Prior to introducing reforming flow, the combustion process was self-ignited at a combustion flow rate of 500 sccm in the absence of any preheating of combustion fluid or external heating of the reactor. It has been noted by several authors that this self-ignition behavior of methanol−air mixtures in the presence of noble-metal catalysts makes methanol a uniquely facile fuel for portable power applications, by alleviating the need for preheating or external heating of the reformer during startup from ambient (∼20 °C) temperatures. Upon achieving a stable steady-state temperature profile along the axial length of the reactor, the flow rate of the combustion gas mixture was adjusted to the desired value and flow of the reforming gas mixture to the reactor introduced at 200 sccm and subsequently increased to 2000 sccm at intervals of 200 sccm while recording reactor performance at each set of flow conditions. In all experiments, combustion and reforming flows were introduced in a countercurrent flow configuration. For the first portion of this study, the heatexchange process volume (Figure 1) was sealed such that the outer 16 channels provide radial self-insulation to the inner 9 channels. The latter portion of this study explored the use of the heat-exchange channels for multiple recuperative heat-transfer schemes. Steady-state operation was verified by a combination of temperature profile monitoring and gas analysis of both combustion and reforming effluents, with typical stabilization times of 15 min employed between changes in process flow rates and initiation of data collection. Upon reaching steady-state operation, 10 samples each of combustion and reforming product streams were collected and analyzed via gas chromatography (GC). Gas chromatographs were calibrated using an external standard calibration gas with known concentrations of each species of interest (H2, CO, CO2, CH4, N2, O2, and He), to a measurement error of 5% and a lower detection limit of 500 ppm.

C,R XCH = 3OH

Y HC,R = 2

SCO =

η=

C,R FCH 3OH,in

FHC,R 2,out C,R 3FCH 3OH,in

C,R = Y CO 2

YĤ 2 =

C,R C,R C,R FCO,out + FCO + FCH 2 ,out 4,out

;

C,R Y CO =

(1) C,R FCO,out C,R 1FCH 3OH,in

;

C,R FCO 2 ,out C,R 1FCH 3OH,in

(2)

FHC2,out + FHR2,out C R 3(FCH + FCH ) 3OH 3OH

(3)

FCO,out FCO,out + FCO2,out + FCH4,out

(4)

298 K R (ΔHSR )XCH FR 3OH CH3OH,in C (ΔHC298 K )XCH FC 3OH CH3OH,in

(5)

3. THEORETICAL SECTION 3.1. Model Assembly. A steady-state three-dimensional model of the 5 × 5 thermally integrated microchannel reactor was constructed using the COMSOL version 3.5a programming environment equipped with the chemical reaction engineering module. The model employs channel, substrate, and packed-bed dimensions identical to the above experimental system. All simulations assume countercurrent flow of combustion and reforming flows and a non-reacting inert gas sealed within each of the outer 16 channels. These conditions allow for model comparison to the majority of experimental data obtained in the present work. 3.2. Fluid-Phase Expressions. Individual species mass transport and reaction in each reforming and combustion volume is described using a steady-state, non-isothermal D

dx.doi.org/10.1021/ef3021359 | Energy Fuels XXXX, XXX, XXX−XXX

Energy & Fuels

Article

pseudo-homogeneous reacting fluid model assembled as follows. Darcy’s law governing momentum transport through porous media is employed in conjunction with the continuity equation to calculate velocities of reacting fluids in each microchannel ⎛ ⎛ −κ ⎞⎞ ∇⎜⎜ρf ⎜⎜ ∇P ⎟⎟⎟⎟ = 0, ⎠⎠ ⎝ ⎝ μf

where the rate of heat appearance (Q) is the product of methanol combustion, decomposition, and/or steam reforming rates and corresponding reaction enthalpies. The specific heat of the fluid phase is calculated using a mass-weighted average of individual species specific heats, while thermal conductivity is calculated from individual species thermal conductivities as follows:

∇ (ρ u ) = 0

n

(6)

kf =

where the permeability of packed-bed catalytic regions is determined using the method by Mason and co-workers κ=

εd p

i=1

2

(7)



Θij =

assuming a packed-bed porosity of 0.45, tortuosity of 1.5, and mean particle diameter of 0.6 mm. The dynamic viscosity for each reacting fluid is evaluated using the semi-empirical formula developed by Wilke23 with Lennard−Jones parameters taken from24 n

μf =

∑ i=1

xiμi n

xi + ∑ j = 1 xjϕij

where ϕij =

⎛ 8 ⎜1 + ⎝

⎤ Mj 1/4

( ) Mi

⎥ ⎥⎦

1/2 Mi ⎞ ⎟ Mj ⎠

2

and

Mi T σ 2 Ωμ

(8)

Individual species mass transport via convection and Maxwell− Stefan multi-component diffusion is described by casting mass flux in terms of mole and weight fractions following the procedure outlined by Curtiss and Bird25 and subsequently equating to the rate of individual species mass appearance by catalytic reaction assuming unity catalyst effectiveness ⎛ ⎞ n ∇P ⎞⎟⎟ eff ⎛ ⎜ ⎜ ∇⎜ρwi ∑ Dij ∇xj + (xj − wj) = R i − ρf u∇wi ⎝ P ⎠⎟⎠ ⎝ j=1

( ) Mi

2

⎥ ⎦

1/2 Mi ⎞ ⎟ Mj ⎠

1/2

(1/Mi + 1/Mj)

P(υi1/3 + υj1/3)2

ε τ

Nu fs =

hfsd p kf

= (0.24)Pr 1/3Re 0.8 ,

⎛ d pρ U ⎞ Cp,mixμf f ⎟⎟ and Pr = where Re = ⎜⎜ kf ⎝ μf ⎠

(10)

where empirical values for diffusion volume (υ) are employed for CH3OH (29.9), O2 (16.6), CO (18.9), CO2 (26.9), H2 (7.07), H2O (12.7), and Ar (17.9). Heat transport by fluid-phase convection and conduction is described by the following steady-state expression: ∇( −k f ∇Tf ) = Q − ρf Cp,f u∇Tf

(14)

where the fluid−solid interphase heat-transfer coefficient is estimated using Dixon’s theoretical predictions of effective heattransfer parameters in packed beds27 for Reynold’s numbers between 40 and 2000.

i

−7 1.75

(12)

(13)

qfs = hfs(Tf − Ts)

n

∑ xiMi

Effective Maxwell−Stefan binary diffusivities are estimated using the Fuller, Schettler, and Giddings method as detailed by Geankoplis,26 accounting for catalyst bed tortuosity and porosity as follows: Dijeff =

⎛ 8 ⎜1 + ⎝

Mj 1/4 ⎤

assuming a thermal conductivity of 3 W m−1 K−1 for the fired cordierite substrate. Heat flux between the fluid and substrate phases is equated over all internal fluid−substrate boundaries

(9)

1 × 10 T

⎡ ⎛ k ⎞1/2 ⎢1 + ⎜ i ⎟ ⎝ kj ⎠ ⎣

∇( −ks∇Ts) = 0

where

P RT

xiki n xi + ∑ j = 1 xj Θij

Inlet boundary conditions for steam reforming and combustion fluid volumes assume uniform inlet velocities determined from individual channel dry basis volumetric flow rates normalized by the channel cross-sectional area. Uniform inlet temperatures of ambient (298.15 K) are assumed for both reforming and combustion channels. Outlet boundary conditions for both combustion and steam reforming volumes assume a constant pressure of 1 bar and are open to convective heat and mass transport. In all simulations, combustion molar feed composition was fixed at 13 mol % CH3OH and 17.3% mol % O2, while reforming feed compositions of 2.6 mol % H2O and CH3OH are employed for predicting experimental data, and 25 mol % CH3OH and 75 mol % H2O are employed for simulating system performance under concentrated feed conditions. Inlet and outlet conditions for the stagnant gas-filled outer channels correspond to uniform temperatures of 298.15 K, zero velocity, and pure inert gas. 3.3. Solid-Phase (Ceramic Substrate) Expressions. The ceramic substrate comprising the 25-channel cartridge is described assuming steady-state heat conduction in the absence of convective or radiative heat transport

,

⎡ ⎛ μ ⎞1/2 ⎢1 + ⎜ i ⎟ ⎝ μj ⎠ ⎢⎣

μi = 2.67 × 10−5

ρf =



(15)

Heat flux from the solid substrate to ambient over the four exposed outer surfaces of the microchannel reformer was described using the following heat flux boundary condition:

qas = has(Ts − Ta)

(11) E

(16)

dx.doi.org/10.1021/ef3021359 | Energy Fuels XXXX, XXX, XXX−XXX

Energy & Fuels

Article

species molar flow rates through each fluid channel are calculated via integration of total molar fluxes across inlet and outlet boundary surfaces. Conversions and yields are calculated from resulting values using eqs 2−6. Individual atomic species (H, O, and C) mass balances were conserved in all simulations within