Professional Documents
Culture Documents
Experimental Evaluation of Distributor Technologies For Trickle Bed Reactors
Experimental Evaluation of Distributor Technologies For Trickle Bed Reactors
Experimental Evaluation of Distributor Technologies For Trickle Bed Reactors
pubs.acs.org/IECR
ABSTRACT: In this paper, different technologies of distributors for trickle bed reactors were compared, in terms of intrinsic
performance of distribution and resistance to tray unlevelness. γ-ray tomography has first been used, in a large column unit, to
visualize gas−liquid distribution just below several tray technologies. Results obtained show a better distribution performance
under multiaperture chimney trays, in comparison with gas-lift and bubble cap systems. Then, a specific cold mock-up was
designed in order to mimic tray tilt between two tray elements. Sensitivities to tray unlevelness of both gas-lift and multiaperture
systems are comparable, far better than the one for the bubble cap system. Multiaperture chimneys have indeed excellent
behavior at nominal conditions, while gas-lift technology is interesting for its ability to behave similarly over a large flow range.
Considering those results, some guidelines are finally given for the choice of the distribution technology.
1. INTRODUCTION have been the subject of recent reviews.8,9 Both reviews, mainly
Fixed-bed reactors operated under trickle-flow condition based on industrial advertising and communications such as
(TBRs) are one of the important classes of three-phase gas− Patel et al.,10 have highlighted promising performances of the
liquid−solid reaction systems encountered in industrial recent gas-lift technology. The objective of this work is to
practice. They are currently employed in hydrotreatments complete these reviews, by experimentally comparing different
and in chemical and biochemical processes for the petroleum, distributor tray types (chimney trays, bubble cap trays, and gas
petrochemical, and chemical industries. TBRs consist of a fixed assisted lift tube trays), in terms of both distribution
bed of solid catalyst particles contacted by a cocurrent gas− performance and resistance to unlevelness.
liquid downflow. The reaction occurs between the dissolved gas
As a first step, distribution performances of different
and the liquid phase at the catalyst surface or inside the catalyst
pellet. For its part, the trickling regime is characterized by small distributors have to be qualified and compared in a large cold
liquid flow rates and low to moderate gas flow rates. Liquid flow experimental unit. To this end, several techniques can
trickles over the catalyst bed, with the gas being a continuous potentially be used to map spatial distribution of liquid and gas
phase. at the top of the catalyst bed: wire mesh tomography11−13 or
Recently, increasing environmental standards of fuels have imaging techniques such as nuclear magnetic resonance
led to the development of high performance units with (NMR) imaging,3,14 electrical tomography,13 X-ray tomogra-
improved catalyst activities. However, TBR efficiency relies phy,15 and γ-ray computed tomography.12,16 γ-ray tomography
also on a good distribution of liquid feed over the catalyst bed was chosen for this benchmark because of its excellent spatial
cross section, which means both an effective liquid inlet resolution, its noninvasiveness, and its potential use on large
distributor and a perfect bed loading. Otherwise, channeling
reactors contrary to NMR imaging.
may appear, causing a significant loss in catalyst activity1 and an
As a second step, the impact of tray unlevelness on the global
earlier catalyst deactivation. When coupled with highly
exothermal reactions, gas and liquid maldistribution may be distribution has to be quantified. Tray tilt is industrially due to
amplified by the fast drying of the catalyst and may possibly its own bending and to a potential angle deviation from the
lead to local hot spots.2,3 horizontal plane during installation. For large reactor diameters,
Poor flow distribution generated by distributor technologies the height difference between two points of the tray may reach
can be partially corrected by optimized layers of grading 10 mm. A small scale mock-up was specifically implemented so
materials, with specific packing particles being able to enhance as to compare gas and liquid flows coming out of two tray
phase radial distribution.4−6 Unfortunately, this corrective elements placed with a definite height difference. In order to
approach is often not sufficient, with the catalyst bed having posttreat in the best way the experiments results achieved, the
in any case few transverse mixing properties.7
impact of the maldistribution measured with this experimental
All these facts have motivated recent developments of new
distributors in order to enhance their general performance. setup has also been quantified in terms of changes in reactor
Some specifications are indeed required: very good perform- performance.
ance at design conditions, the flexibility to operate over a broad
range of gas/liquid flow rates (numerous industrial operating Received: February 15, 2013
conditions possible for a single reactor, difference in degree of Revised: July 4, 2013
feed vaporization from start of run to end of run, etc.), and a Accepted: July 18, 2013
low sensitivity to tray unlevelness. These tray improvements Published: July 18, 2013
© 2013 American Chemical Society 11189 dx.doi.org/10.1021/ie400504p | Ind. Eng. Chem. Res. 2013, 52, 11189−11197
Industrial & Engineering Chemistry Research Article
Flow − Fhigh
IQ =
Faverage (1)
where Flow and Fhigh are respectively phase flows through the
low and high distributors and Faverage represents the average
phase flow. The IQ criterion can be calculated for both gas and
liquid phases, with the liquid (L) being preferentially used to
characterize the maldistribution. A low value of IQL indicates a
low sensitivity of distributor to tray tilt. By contrast, a high
value of IQL points to a strong imbalance of distribution.
Figure 3 shows a typical evolution of IQL for various
distributors. Concerning results with multiaperture chimneys,
Figure 1. Different types of distributors: (a) perforated trays, (b) the evolution of IQL exhibits alternatively higher and lower
multiport chimney, (c) bubble cap, and (d) gas-lift tube.8 values corresponding respectively to liquid levels close to and
far from apertures. As a consequence, careful design must be
done to minimize these high value zones for IQL.
Bubble cap trays are significantly different, with chimneys
Multiaperture chimneys are generally composed of lateral consisting of a cap centered concentrically over a standpipe
openings (holes or slots/notches spaced vertically up the axis of (Figure 1). Vapor passes through slotted lateral openings,
the chimney) for liquid and superior apertures for gas, allowing lifting liquid into the upflow channel and then downflow into
vapor−liquid contacting of both phases in the conduit. As a an internal conduit. As a consequence, this technology operates
consequence, gravity is the main driving force for the liquid on a gas assist principle compared to the liquid weight principle
flow. employed by multiaperture chimneys. Operation with a wider
Good distribution under the tray is generally achieved by liquid flow range is believed possible. Yet sensitivity to tray
minimizing the distance between two chimneys, with a unlevelness seems not reduced, in spite of this gas driving force
triangular rather than square pitch preferred. Cost and (Figure 3). Moreover, because of their large diameters, center-
fabrication constraints practically limit this center-to-center to-center spacing for bubble cap trays is limited (fewer drip
spacing to about 100 mm. Some enhancements are proposed points).
by adding a dispersive system under the chimneys. The Shell Gamborg and Jensen19 proposed a modified bubble cap
HD tray may be designed with nozzles under the conduits,5 system, named “vapor lift”, with only one upflow tube leg
generating an overlapping of the created sprays, thus elevated above the tray. The operational principle is the same as
eliminating any grading material under the tray. IFPEN for bubble caps: gas flows through the nozzles and then by
EquiFlow technology17 is characterized by several dispersive gravity falls onto the catalyst bed. Following the authors,
devices with rims, such as grids (Figure 2). Both solutions because liquid flow is not sensitive to the local liquid level at the
allow, theoretically, having an excellent liquid and gas nozzle, this gas-lift tray a priori exhibits an excellent sensitivity
distribution, avoiding any disk-type discharge pattern under to unlevelness (Figure 3). Due to its rectangular configuration
the chimneys. and smaller/fewer slots, center-to-center spacing is close to the
As detailed by Maiti and Nigam,9 the surface area of lateral one of the chimney tray, with a square pitch, allowing much
openings is determined to ensure a certain liquid level on the better catalyst utilization than bubble cap trays. Lastly,
tray. The pressure balance establishes that the liquid velocity in according to the inventors, this design allows wide liquid flow
the holes is directly correlated to the liquid height above the ranges.
hole, with a 0.5 power dependence. As a result, when the liquid The main dimensions of trays evaluated during this study,
level is close to holes, this class of tray is also sensitive to i.e., two chimney trays with/without dispersion devices, a
unlevelness. Nevertheless, by keeping a minimum liquid head classical bubble cap tray, and a gas-lift tray, are given in Table 1.
over the holes, this sensitivity is drastically reduced. The distribution performance of the bubble cap tray was not
To describe this sensitivity, an indicator IQ is defined18 studied. Several types of gas-lift configurations, based on
between two chimneys submitted to an elevation difference (10 different geometric characteristics, were compared during
mm being generally considered): maldistribution measurements.
11190 dx.doi.org/10.1021/ie400504p | Ind. Eng. Chem. Res. 2013, 52, 11189−11197
Industrial & Engineering Chemistry Research Article
Figure 3. Sensitivity to tray tilt for different technologies, with a 10 mm elevation difference.18
Figure 6. Liquid retentions at the top of the catalyst bed for different distributors trays, for a superficial gas velocity of 2 cm·s−1, in comparison with
the model of Boyer et al.21
Figure 8. (a) Distribution efficiencies at the top of the catalyst bed for different distributor trays, for superficial gas velocities of (a) 2 and (b) 10
cm·s−1.
important than the one of the multiaperture system A, contrary A and B have good distribution capabilities and a correct
to what was exposed by several authors.8,9,18,19 resistance to nonhorizontality, except at some points of the
An explanation for these results below expectations for gas- liquid flow range (turndown regime, the case when the liquid
lift distributors can be found in Figure 9d (effect of liquid flow level approaches hole rows). Distribution obtained under gas-
rate on gas maldistribution IQG values). While there is no gas lift trays D is less interesting (more heterogeneities). Their
maldistribution for the multiaperture chimney A, it is resistance to tray tilt is comparable with those of trays A and B,
catastrophic for the bubble cap distributor C, and significant lower at nominal conditions but with the advantage of being
for all the gas-lift configurations D (IQG ∼ 15−30%). It is relatively constant over the liquid flow range.
believed that bubble caps and gas-lift distributors are less It is clear that few current distributor technologies respect
dependent on the liquid flow rate in comparison with classical this ideal IQL criterion of 15−20%, particularly when the design
chimneys. However, there exists a dependence on the gas flow is done for a large liquid flow range. Considering all the facts,
rate due to the gas assisted principle. It is then likely that the the authors propose the choice of multiaperture chimney trays
gas maldistribution observed in Figure 9d caused the liquid for the classical liquid flow range, with the gas assisted system
imbalance established in Figure 9b,c. Besides, this gas being interesting only for the high liquid flow range.
maldistribution can have a dramatic impact for processes with Motivations for this choice of multiaperture chimney trays
hydrodynamic operating conditions close to the bubble flow are a better distribution under this tray, an excellent resistance
regime (higher gas flow rates). to unlevelness at nominal conditions (i.e., presumable unit
3.3. Discussion. Results previously exposed, for both operating conditions), and a correct resistance for all liquid
distribution performance and resistance to tray unlevelness, velocities encountered. To support these guidelines, Table 3
have shown that no clear direction exists for the choice of the presents several examples of design of multiaperture chimney
distribution technology. The sole certain fact established is that trays, submitted to different liquid flow ranges. This gives some
the bubble cap distributor C must be avoided because of its indications for the selection between classical chimney and gas-
catastrophic resistance to tray tilt. Multiaperture chimney trays lift technologies (here for a classical chimney height of 300
11194 dx.doi.org/10.1021/ie400504p | Ind. Eng. Chem. Res. 2013, 52, 11189−11197
Industrial & Engineering Chemistry Research Article
Figure 9. (a) Effect of liquid load on the liquid distribution for A multiaperture chimneys, with a tray tilt of 10 mm, at low superficial gas velocity (vsg
= 2 cm·s−1). Experimental results (□). Theoretical prediction: 250 mm height chimney (); 500 mm height chimney (---). (b) Effect of liquid load
on the liquid distribution for different trays, with a tray tilt of 10 mm, at low superficial gas velocity (vsg = 2 cm·s−1). Experimental results: A,
multiaperture chimney (□); C, bubble cap chimney (×); D1, gas-lift chimney (▲); D2, gas-lift chimney (●); D3, gas-lift chimney (◆). Theoretical
predication: A, multiaperture chimney (). (c) Effect of liquid load on the liquid distribution for different trays, with a tray tilt of 10 mm, at high
superficial gas velocity (vsg = 15 cm·s−1). Experimental results: A, multiaperture chimney (□); C, bubble cap chimney (×); D1, gas-lift chimney (▲);
D2, gas-lift chimney (●); D3, gas-lift chimney (◆). Theoretical predication: A, multiaperture chimney (). (d) Effect of liquid load on the gas
distribution for different trays, with a tray tilt of 10 mm, at high superficial gas velocity (vsg = 15 cm·s−1). Experimental results: A, multiaperture
chimney (□); C, bubble cap chimney (×); D1, gas-lift chimney (▲); D2, gas-lift chimney (●); D3, gas-lift chimney (◆).
Table 3. Effect of Liquid Flow Range on the Design of a contrary to what is exposed in recent reviews mainly based
Multiaperture Chimney and Impact on Maldistribution on industrial communications, other tests show that gas-lift
Criterion IQL for the Liquid Phasea systems do not outclass multiaperture trays in terms of
resistance to nonhorizontality.
liquid flow range 2 3 4 5 6
(vsl,max/vsl,min) Thus, some guidelines were finally proposed, based on
chimney height (mm) 300 experimental conclusions and modeling, favoring the choice of
chimney diameter (mm) 50 multiaperture chimneys for current operating conditions of
hole rows 1 2 3 3 3 trickle-bed reactors (i.e., classical liquid flow ranges), with gas-
hole number 3 3+2 3+3+3 3+3+3 3+3+3 lift systems being competitive only in the case of very large
hole diameter (mm) 10 9+4 7+4+4 7+5+5 6+6+6 liquid flow ranges (because of their constant acceptable
IQmean (%) 6.5 5.6 7.0 8.2 9.0 performances vs liquid velocity).
IQmax (%) 14.0 15.2 17.1 24.3 30.9 By the way, developments are still required in terms of
IQnominal (%) 3.5 2.3 2.9 3.2 3.4 distributor internals so as to propose an ideal technology:
a
ρL = 600 kg·m−3, ρG = 30 kg·m−3, and vsl,nominal = 0.8 cm·s−1.
excellent distribution under the tray and perfect resistance to a
potential unlevelness, for all operating conditions.
Incontestably, more work should be done with cold flow
mm): up to a liquid flow range of 5, the performance of testing. Priority will be given to two experimental axes: (1)
multiaperture chimney trays is good enough at turndown experiments at high pressure and/or with high density gases
conditions (IQL,max < 20%) and excellent at design conditions. and (2) evolution of initial liquid maldistribution over the
4. CONCLUSION catalytic bed. Developments of computational fluid dynamics
modeling are also required, particularly for gas assisted systems.
■
Current state-of-art trickle bed reactor distributors were tested
and compared in terms of distribution performance and
resistance of this performance when the tray is submitted to APPENDIX: “TWO PARALLEL PLUG REACTORS”
a nonhorizontality. MODEL OF MALDISTRIBUTION
Experiments quantifying the intrinsic performance of The impact of large scale maldistributions on the performances
distribution have revealed an advantage for the technology of of fixed bed reactors is estimated by using a simple numerical
multiaperture chimneys with grids/deflectors. Moreover, model. Liquid and gas flow rates are considered as constant
11195 dx.doi.org/10.1021/ie400504p | Ind. Eng. Chem. Res. 2013, 52, 11189−11197
Industrial & Engineering Chemistry Research Article
along the axis of the reactor. Only the case of liquid considered, assuming no mass transfer limitation and a zero
maldistribution is discussed; the gas phase is supposed to order kinetics on hydrogen. Following Varga et al.,22 the energy
flow homogeneously in the reactor at constant superficial gas of activation Ea is taken as 80−120 kJ·mol−1. The enthalpy of
velocity. reaction ΔHr is taken as 75 kJ·mol−1, a value in the range of
Heterogeneities of flow rates at the outlets of chimneys or magnitude reported by Murali et al.23 The pre-exponential
distributing devices are modeled by dividing the section of the factor is fitted in order to process the reaction at T = 650 K, a
reactor into two equal areas, fed by two different liquid flow realistic temperature in comparison with industrial operating
rates. The spatial scale of maldistribution between high and low conditions. The liquid superficial velocity U is 0.01 m·s−1 in the
flow rates is supposed to be sufficiently high to avoid reactor.
attenuation of maldistribution along the vertical axis, and the The assumptions listed above simplify the model which can
reactor is assumed to behave like two parallel plug reactors, as be easily solved by using a simple first order numerical method.
presented in Figure 10. Solved equations are
∂C
Ui = −K exp( −Ea /RT )
∂z (3)
∂T
UiρCp = ΔHrK exp( − Ea /RT )
∂z (4)
where Ui is the liquid superficial velocity in the reactor i (i = 1
or 2). Ui is calculated as follows:
⎡ ⎛ 3 ⎞⎤
Ui = ⎢1 ± ⎜i − ⎟⎥U̅ IQ L
⎣ ⎝ 2 ⎠⎦ (5)
As the liquid velocity is different in each side of the reactor,
the conversion is not equal between the two parallel plug
reactors, and the resulting mean conversion is calculated at the
Figure 10. Modeling maldistribution by two parallel plug reactors. The outlet of the reactor. When the maldistribution indicator IQL is
gray scale schematizes the sulfur content in the liquid phase. not null, a difference appears between temperature and
concentration profiles of the two parallel reactors. The global
Each side of the reactor is modeled as an adiabatic one- sulfur content at the outlet of the whole reactor is also
dimensional reactor. Heat and mass diffusion phenomena are increased, as reported in Figure 11a. In Figure 11 is also
also neglected as convection by the liquid flow rate is reported the maximal difference of temperature in the
predominant in the evolution of concentration and temperature transversal direction, between reactors, and obtained at
in the vertical direction. approximately 1.3 m from the inlet.
The case of diesel fuel hydrodesulfurization is considered, In spite of its simplicity, this example shows how
inside a single bed reactor of 6 m height. The liquid phase maldistribution can induce radial temperature gradients and
properties are those of hexadecane C16H34 (ρ = 800 kg·m−3, Cp impact performances of fixed bed reactors. As a content of 10
= 1800 J·kg−1·K−1). The only reactant is benzothiophene ppm atomic sulfur is required in the diesel fuel, the loss of
(C8H6S) present in the liquid phase only at 2% weight conversion due to maldistribution is thwarted by an increase of
regarding atomic sulfur. Sulfur specification at the reactor outlet the temperature at the inlet of the reactor, ΔT0. The
is 10 ppm. A first order apparent kinetics on benzothiophene is temperature adjustment depends on operating conditions of
Figure 11. (a, left) Loss of conversion induced by maldistribution and maximal temperature difference between both plug reactors. (b, right)
Increase of inlet temperature necessary to reach the target [S]out = 10 ppm.
the reactors and kinetics. The energy of activation has obviously (14) Sederman, A. J.; Gladden, L. F. Magnetic resonance imaging as a
an important impact on the temperature increase linked to the quantitative probe of gas-liquid distribution and wetting efficiency in
maldistribution, as shown in Figure 11b. trickle-bed reactors. Chem. Eng. Sci. 2001, 56, 2615.
An important result obtained with the simple model “two (15) Marchot, P.; Toye, D.; Crine, M.; Pelsser, A.-M.; L’Homme, G.
Investigation of liquid maldistribution in packed columns by X-ray
parallel plug reactors” is that, below an equilibrium index IQL of tomography. Chem. Eng. Res. Des. 1999, 77, 511.
15−20%, the effect of maldistribution remains under control (16) Boyer, C.; Fanget, B. Measurement of liquid flow distribution in
with inlet temperature adjustments below 1 °C. Higher inlet trickle bed reactor of large diameter with a new gamma-ray
temperature adjustments are not desired as they could tomographic system. Chem. Eng. Sci. 2002, 57, 1079.
accelerate catalyst deactivation and finally shorten run cycle (17) Harter, I.; Agoero, R. Device for distributing a polyphase
lengths. mixture on a granular solid bed comprising a porous antisplash nozzle
■
element with flanges. U.S. Patent 2005/0062178, 2005.
(18) Haldor Topsoe. Reactor internalsoptimizing catalyst perform-
AUTHOR INFORMATION ance; 2000.
Corresponding Author (19) Gamborg, M. M.; Jensen, B. N. Two-phase downflow liquid
distribution device. U.S. Patent 5,942,162, 1999.
*E-mail: frederic.bazer-bachi@ifpen.fr.
(20) Koudil, A.; Boyer, C.; Bentolila, Y. Measuring of liquid and gas
Notes flow distribution in a fixed bed reactor using a gamma-ray tomographic
The authors declare no competing financial interest. system. Proceedings of the Third World Congress on Industrial Process
■
Tomography; Virtual Centre for Industrial Process Tomography:
Glossop, U.K., 2003.
ACKNOWLEDGMENTS (21) Boyer, C.; Volpi, C.; Ferschneider, G. Hydrodynamics of trickle
This work was financially supported by IFP Énergies Nouvelles, bed reactors at high pressure: two-phase flow model for pressure drop
to whom the authors wish to express their gratitude. and liquid holdup, formulation and experimental validation. Chem.
■
Eng. Sci. 2007, 62, 7026.
(22) Varga, Z.; Hancsok, J.; Nagy, G.; Polczmann, G.; Kallo, D.
REFERENCES Upgrading of gas oils: the HDS kinetics of dibenzothiophene and its
(1) McManus, R. L.; Funk, G. A.; Harold, M. P.; Ng, K. M. derivatives in real gas oil. Top. Catal. 2007, 45, 203.
Experimental study of reaction in trickle-bed reactors with liquid (23) Murali, C.; Voolapalli, R. K.; Ravichander, N.; Gokak, D. T.;
maldistribution. Ind. Eng. Chem. Res. 1993, 32, 570. Choudary, N. V. Trickle bed reactor model to simulate the
(2) Lopes, R. J. G.; Quinta-Ferreira, R. M. Evaluation of multiphase performance of commercial diesel hydrotreating unit. Fuel 2007, 86,
CFD models in gas−liquid packed-bed reactors for water pollution 1176.
abatement. Chem. Eng. Sci. 2010, 65, 291.
(3) Lysova, A. A.; von Garnier, A.; Hardy, E. H.; Remiert, R.;
Koptyug, I. V. The influence of an exothermic reaction on the spatial
distribution of the liquid phase in a trickle-bed reactor: direct evidence
provided by NMR imaging. Chem. Eng. J. 2011, 173, 552.
(4) Moller, L. B.; Halken, C.; Hansen, J. A.; Bartholdy, J. Liquid and
gas distribution in trickle-bed reactors. Ind. Eng. Chem. Res. 1996, 35,
926.
(5) Ouwerkerk, C. E. D.; Bratland, E. S.; Hagan, A. P.; Kikkert, B. L.
J. P.; Zonnevylle, M. C. Performance optimization of fixed bed
processes. Pet. Technol. Q. 1999, 2, 21.
(6) Tsochatzidis, N. A.; Karabelas, A. J.; Giakoumakis, D.; Huff, G. A.
An investigation of liquid maldistribution in trickle beds. Chem. Eng.
Sci. 2002, 57, 3543.
(7) Boyer, C.; Koudil, A.; Chen, P.; Dudukovic, M. P. Study of liquid
spreading from a point source in a trickle bed via gamma-ray
tomography and CFD simulation. Chem. Eng. Sci. 2005, 60, 6279.
(8) Alvarez, A.; Ramirez, S.; Ancheyta, J.; Rodriguez, L. M. Key role
of reactor internals in hydroprocessing of oil fractions. Energy Fuels
1987, 21, 1731.
(9) Maiti, R. N.; Nigam, K. D. P. Gas-liquid distributors for trickle
bed reactors: a review. Ind. Eng. Chem. Res. 2007, 46, 6164.
(10) Patel, R. H.; Bingham, E.; Christensen, P.; Müller, M.
Hydroprocessing reactor and process design to optimize catalyst
performance. Presented at The First Indian Refining Roundtable,
1998.
(11) Llamas, J. D.; Perat, C.; Lesage, F.; Weber, M.; D’Ortona, U.;
Wild, G. Wire mesh tomography applied to trickle beds. A new way to
study liquid maldistribution. Chem. Eng. Process. 2008, 47, 1765.
(12) Bieberle, A.; Schubert, M.; Da Silva, M. J.; Hampel, U.
Measurement of liquid distributions in particle packings using wire-
mesh sensor versus transmission tomographic imaging. Ind. Eng. Chem.
Res. 2010, 49, 9445.
(13) Matusiak, B.; Da Silva, M. J.; Hampel, U.; Romanowski, A.
Measurement of dynamic liquid distributions in a fixed bed using
electrical capacitance tomography and capacitance wire-mesh sensor.
Ind. Eng. Chem. Res. 2010, 49, 2070.