chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

journal homepage: www.elsevier.com/locate/cherd

An imaging technique for characterization

of fluid flow pattern on industrial-scale column
sieve trays夽

Markus Schubert a,∗ , Michael Piechotta a , Matthias Beyer a ,

Eckhard Schleicher a , Uwe Hampel a,b , Jürgen Paschold c
a Helmholtz Zentrum Dresden-Rossendorf, Institute of Fluid Dynamics, Bautzner Landstraße 400,
01328 Dresden, Germany
b Technische Universität Dresden, AREVA Endowed Chair of Imaging Techniques in Energy and Process Engineering,

01062 Dresden, Germany

c BASF SE, GCP/TA – L540, 67056 Ludwigshafen, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Non-uniform flow patterns on distillation trays can result in significant losses in separation
Received 28 January 2016 efficiency compared to conditions with straight chordal flow paths. To improve flow pattern
Received in revised form 2 May 2016 via tray design revamps as well as for more accurate prediction of achievable separation effi-
Accepted 4 May 2016 ciency, better understanding of the flow formation is required. This is, however, not straight
Available online 11 May 2016 forward since the evolving patterns depend on column and tray design as well as on operat-
ing conditions and may cover configurations with continuous liquid being pushed through
Keywords: a curtain of rising bubbles (bubble regime) or a continuous gas phase carrying the liquid as
Distillation column droplets (spray regime).
Sieve tray In this study, the wire-mesh sensor technique was used to visualize flow patterns on larger
Wire-mesh sensor trays. The technique tracks conductivity tracer pulses during their passage across the tray. A
Tray flow pattern proper calibration routine was developed and a methodology is proposed to determine point
Residence time distribution liquid residence time distributions and weir-to-weir liquid velocities. Experiments were car-
Liquid velocity ried out in a tray column of 800 mm diameter. The feasibility study included experiments at
different liquid capacities and various outlet weir obstructions. Tray benchmark data at high
spatio-temporal resolution were obtained, which are suitable for validation of CFD models
for flow simulation of industrial-scale sieve trays.
© 2016 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction well-known that liquid channeling, bypassing and flow recir-

culation result in efficiency losses.
Sieve trays are widely used as contacting equipment in dis- To mathematically evaluate the efficiency of trays, the
tillation columns. The efficiency of the gas–liquid contacting point efficiencies across the tray have to be coupled with
and thus, the mass transfer and overall process perfor- liquid and gas motion models. As long as liquid plug flow
mance is strongly governed by the evolving flow pattern. It is is predominating the tray efficiency can be easily calculated

夽
Presented at the ProcessNet Annual Meeting of the Groups of Fluid Process Engineering and Membrane Technology, 26–27 March, 2015,
Bremen, Germany.
∗
Corresponding author. Tel.: +49 351 260 2627; fax: +49 351 260 2383.
E-mail address: m.schubert@hzdr.de (M. Schubert).
http://dx.doi.org/10.1016/j.cherd.2016.05.004
0263-8762/© 2016 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146 139

extracted isochrones of residence time as well as velocities

Nomenclature between the probes arranged in the same longitudinal row.
Although these data are very coarse, recent CFD approaches
Symbols used them as validation data (Gesit et al., 2003; Wang et al.,
a parameter in Eq. (2), l2 /mol2 2004; Hirschberg et al., 2005).
AC column cross-sectional area, m2 Several other techniques were developed to study the flow
AD downcomer area, m2 phenomena across the column trays. The simplest concept is
AAT active bubbling tray area, m2 to use floats such as table-tennis balls to track the flow field
b parameter in Eq. (2), l/mol (e.g. Porter et al., 1972). However, the method is subjected to
c concentration, mol/l severe errors caused by slip velocities between liquid surface
DC column diameter, m and balls as well as velocity gradients along the froth height.
DH hole diameter, m Furthermore, instead of a full 2D velocity field only individ-
FB vapor load factor referred to AAT , Pa0.5 ual trajectories can be extracted. Hydrogen bubble tracking
G conductance, 1/ was proposed by Müller et al. (1988), however, reliable data
HDC downcomer clearance, m were only taken without gas flow. The injection of dye is
HIW inlet weir height, m another common approach to visualize liquid dispersion and
HOW outlet weir heights, m major flow directions as well as stagnant zones (Porter et al.,
HTS tray spacing, m 1972; Bechtel et al., 2011). However, quantitative parameters
I current, A are hardly extractable. Temperature gradients across the tray
LFP flow path length, m resulting from gas–liquid contact with cold decompressed
LIW inlet weir length, m air allow drawing conclusions on possible concentration pro-
LOW outlet weir length, m files on the tray (Stichlmair and Ulbrich, 1985), however, this
m parameter shown in Fig. 5b, mol/l method does not explicitly provide data of the fluid flow pat-
n exponent shown in Fig. 5b, 1/s terns. Instead, flow patterns were deducted from isotherms
NH number of holes as frontier curve measured via thermal infrared camera after
t time, s a step input of liquid tracer of higher temperature (Li et al.,
U excitation voltage, V 2007). Also a strain probe technique was proposed (Biddulph
wL liquid load referred to AC , m3 /(m2 h) and Bultitude, 1990), however, adjusting defined velocities for
 conductivity, S/m various froth compositions for sensor calibration seems very
 point liquid residence time, s challenging.
To overcome the drawbacks of previous techniques, an
Indices
exploratory study on a measurement method, which pro-
i,j sensor matrix indicator
vides valuable information on the flow pattern with very high
k frame indicator
spatio-temporal resolution, has been performed.

Abbreviations
ADC analog digital converter
2. Experimental setup and measurement
CFD computational fluids dynamics
technique
DP differential pressure
FM flow meter
2.1. Tray column mock-up
Na2 SO4 sodium sulfate
P(i,j) sensor matrix point
The tray column mock-up and corresponding periphery are
Q3 3rd quartile
schematically shown in Fig. 1 together with embedded figures
RTD point liquid residence time
for design and dimensions of sieve tray and outlet weirs as
well as for the tray arrangement used in this study.
The column consists basically of three acrylic glass flange-
(Lewis, 1936). In addition, Bell and Solari (1974) proposed mounted segments with heights of 800 mm, 400 mm and
solutions for symmetric non-uniform and retrograde flows. 635 mm from bottom to top, respectively. The inner column
Stichlmair and Ulbrich (1987) related point and tray efficiency diameter is 800 mm. Two trays are installed between the seg-
describing the pattern as multiple liquid plug flow channels. ments with a tray spacing of 400 mm. The trays are standard
However, it is known that liquid flow on trays may differ sieve trays with a fractional active area of approx. 0.77 referred
from such rather ideal cases. Thus, CFD approaches are more to the column cross-section. The trays, also made of acrylic
and more used to model sieve tray hydraulics (Krishna et al., glass, are equipped with holes of 5 mm size. The distance
1999; van Baten and Krishna, 2000; Krishna and van Baten, from inlet to outlet weir is 584 mm. Column and sieve tray
2003; Noriler et al., 2008) and even to couple them with mass dimensions are summarized in Table 1.
and energy transfer (Rahimi et al., 2006; Noriler et al., 2010). Four different outlet weir configurations (see upper right
However, these attempts exceed hardly column diameters of part of Fig. 1) were applied in order to manipulate the flow
30 cm, where pattern non-uniformity is still less pronounced pattern. Beyond the standard weir (0), asymmetric weirs with
and the computational effort is tractable. partial obstructions were applied. The dimensions of the inlet
However, experimental benchmark data at larger scale and outlet weirs are summarized in Table 2.
are very limited. At FRI (Fractionation Research, Inc.), Solari Air was used as the gas phase and adjusted up to a
and Bell (1986) studied fluid flow patterns on a sieve tray of flow rate of 900 m3 /h, which corresponds to a vapor load
1.22 m diameter. They applied a tracer technique with a 4 × 4 factor of FB = 0.72 Pa0.5 referred to the active (bubbling) tray
matrix of optical probes installed at the bubbling tray area and area. Water was supplied as liquid phase via an adjustable
140 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146

Fig. 1 – Schematics of the column tray mock-up.

2.2. Measurement technique and data acquisition

Table 1 – Column and sieve tray dimensions.
Symbol (unit) Dimension
The measurement method is based on the use of a con-
Column diameter DC (mm) 800 ductivity wire-mesh sensor to detect and visualize the flow
Column cross-sectional area AC (m2 ) 0.503 pattern via concentration of a tracer on tray (1), which is
Downcomer area AD (m2 ) 0.025 added upstream the inlet weir. The tracer volume was added
Tray spacing HTS (mm) 400
upstream the inlet weir from an elongated chute to ensure a
Active (bubbling) tray area AAT (m2 ) 0.387
uniform distribution along the inlet weir. In addition, care was
Downcomer clearance HDC (mm) 20
Flow path length LFP (mm) 584 taken to not splash the tracer volume into the downcomer area
Hole diameter and spaces DH (mm) 5 × ∼21 ×  in order to avoid a wavy tracer front distorting the flow on the
Number of holes NH 1024 tray.
Open area fraction referred to AAT FOA 0.051 The wire-mesh sensor is a low-invasive high spatio-
temporal imaging instrument and consists basically of a
matrix-like arrangement of the measuring points. Therefore,
centrifugal pump at flow rates up to 3 m3 /h, which corre- electrode wires are stretched across the cross-section of a ves-
sponds to a liquid load of wL = 6.0 m3 /(m2 h) referred to the sel or pipe in two planes, the transmitter and the receiver
column cross-sectional area. At these conditions, no weep- plane, with a small axial gap. While the wires of one plane run
ing occurred during operation. While the gas was released to parallel to each other, the transmitter wires and receiver wires
atmosphere, the liquid was recycled via the water reservoirs. normally cross each other in an angle of 90◦ forming a matrix-
During conductivity tracer experiments (see below), liquid was like set of virtual crossing points. By applying an excitation
fed from water reservoir (1) to the column and discharged signal to the receiver wires one by one, the electrical current
to water reservoir (2) and vice versa to prevent looping of flowing towards the receiver wires is recorded fully in paral-
the conductivity tracer in the feed. The water reservoirs were lel by means of transimpedance amplifiers followed by analog
equipped with heaters to counterbalance cooling effect of the to digital converters (ADC). This is carried out by the corre-
decompressed air while contacting the liquid. To ensure a sponding electronic circuitry and can be speed up to 10,000
homogeneous gas distribution in the cross-section, a baffle frames per second (Prasser et al., 1998). Two types of elec-
plate was installed at the column bottom. The liquid phase tronics are available, a conductivity-based (Prasser et al., 1998)
was fed to the column top on a plate installed above tray 1 to and a capacitance-based (da Silva et al., 2007) type. While the
homogenize the flow in the downcomer and to tray 1, respec- conductivity-based device is suitable for experiments on mix-
tively. Pressure drop of the trays are measured via u-bend tures containing conducting liquids, the capacitance based
manometer. electronics has been mainly developed to allow also measure-
ments in non-conductive mixtures as oil and gas. Such sensors
were already applied to visualize gas–liquid flow structures
Table 2 – Inlet and outlet weir dimensions. and to study bubble size distributions in pipes (da Silva et al.,
Symbol (unit) Dimension 2010, 2011; Lucas et al., 2005; Vieira et al., 2014a, 2014b;
Schleicher et al., 2015) as well as to identify flow regimes
Inlet weir length LIW (mm) 533 and to quantify phase fraction distributions in bubble col-
Inlet weir height HIW (mm) 35
umn reactors (Rollbusch et al., 2015; Nedeltchev et al., 2014;
Outlet weir length LOW (mm) 474
Outlet weir heights HOW,0/1/2/3 (mm) 20/35/35/70
Nedeltchev and Schubert, 2015). Wire-mesh sensors were
also embedded in catalyst packings of trickle bed reactors
 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146 141

the local instantaneous conductance Gi,j,k in the single cross-

ing points

Ii,j,k
Gi,j,k = = ADCi,j,k . (1)
U

Applying a geometry factor for each crossing point one

could – in principle – calculate the local instantaneous con-
ductivities  i,j,k . Unfortunately, the geometry of the crossing
points is not exactly known and the electrical field distribu-
tion is quite complicated. Moreover, the local conductance is
not only a function of the electrical conductivity of the liquid
but depends strongly on the local gas content, too. Therefore,
a direct estimation of the local instantaneous conductivi-
ties, and thereby liquid tracer concentrations, is impossible.
Fig. 2 – Photography of the tray-embedded wire-mesh
However, the data are corresponding to each other and thus,
sensor with close-up of wire-grid and holes.
the local instantaneous liquid conductivity values are direct
proportional to the measured ADC counts. In the following
to study the gas–liquid distribution (Mohammed et al., 2015; chapter the procedure to generate a calibration function for
Schubert et al., 2010) and liquid residence time and disper- the relation of ADC values to local liquid conductivities is
sion via tracer injection (Mohammed et al., 2013; Härting explained. More details on wire-mesh sensors and their mea-
et al., 2015). Wire-mesh sensors were also used in nuclear surement principles can be found by Prasser et al. (1998) and
engineering to study the coolant liquid mixing in pressure da Silva et al. (2007).
vessel downcomer and primary circuit loops of a pressur-
ized water reactors mockup via a salt tracer (Prasser et al., 3. Data calibration, visualization and
2003). Meanwhile, the wire-mesh sensors are commercialized parameter extraction
(HZDR Innovation GmbH) and can be built in many customized
geometry. 3.1. Data calibration
The wire-mesh sensor in this study is composed of two
planes of perpendicularly arranged wires with 64 receivers Na2 SO4 solution was used as tracer. To quantify the tracer
perpendicular to the weirs and 52 transmitters parallel to concentration of the liquid phase at every crossing point dur-
the weirs (thereof 50 within the active bubbling tray area ing operation, proper calibration is needed. Therefore, prior to
and one wire in the downcomer area and in the receiving the tracer dispersion experiments, the Na2 SO4 concentration
pan area, respectively). Each wire has a diameter of 0.5 mm in the liquid reservoir was adjusted at different values up to
and the lateral distance between the wires is 12.0 mm. The cNa2 SO4 = 4.8E − 0.4 mol/l (corresponds to the maximum con-
distance between the two planes is 10 mm. The sensor is centration of the tracer applied in this study) in increments
installed at tray 1 (Fig. 2). Here, the lower wire plane is of approx. cNa2 SO4 = 4.0E − 0.5 mol/l. Calibration data were
embedded in the tray in routed notches aligned with the recorded during column operation at identical conditions as
main flow path (weir-to-weir) to reduce the intrusiveness in the subsequent tracer study, i.e. gas and liquid flow rates
of the sensor and thus, the resistance and impact on the as well as respective outlet weir configuration. Fig. 3a shows
flow. examples of the ADC time-series signals at sensor crossing
The sensor is connected to a WMS200 electronics point P(05,25) for three different Na2 SO4 concentrations of
(Teletronic Rossendorf GmbH) applying the above mentioned the liquid phase. As the mixture (liquid and gas) conductivity
measurement principle. The measurement frequency (frame in the crossing points determine the measured signals (ADC
rate) was set to 2,500 Hz to allow an oversampling of the data counts), crossing gas bubbles overlap the signal in a way that
for later statistical evaluations. Thus, a three-dimensional no clear ‘liquid-only’ points can be detected in the plot. Thus,
data set of ADC values ADCi,j,k corresponding to the measured to unambiguously relate the characteristic distribution and
current flow Ii,j,k in the single crossing points Pi,j for each frame dynamics of the ADC signal belonging to a certain liquid tracer
k is recorded across the tray area. Due to the constant excita- concentration at defined operating conditions, detailed sta-
tion voltage U the measured current is direct proportional to tistical analysis is required. Different ADC signal parameters

Fig. 3 – (a) Examples of the ADC time series signal (10 s shown) at sensor crossing point P(05,25) for different concentrations
at wL = 6.0 m3 /(m2 h), FB = 0.72 Pa0.5 and standard weir configuration and (b) corresponding boxplots obtained for a window
length of 1.2 s (red line indicates calibration curve using the third quartiles Q3).
142 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146

(e.g. mean, median, maximum, minimum, quartiles, etc.) were 3.3. Determination of point liquid residence time and
tested at various time span (to ensure statistical confidence) weir-to-weir liquid velocity
and evaluated with respect to the slope of the calibration
curve. Eventually, the third quartiles (Q3) obtained for a win- 3.3.1. Point liquid residence time
dow size of 3000 frames corresponding to 1.2 s time span were The concentration data of four replicated tracer experiments
identified as the most reliable statistical parameter for the cal- for the sensor crossing point P(27,27) as indicated by the red
ibration at same time ensuring sufficient temporal resolution. square in Fig. 4 are shown in a dimensionless manner in
Fig. 3b shows the box plot for the calibration data obtained Fig. 5a. The number of points per second given is reduced by
from the sensor crossing point P(05,25) in Fig. 3a. a factor of 5 for better illustration. The data are placed at the
The corresponding calibration curve is defined according time axis such as that the first appearances of the tracers coin-
to Eq. (2) as cide. In addition, the mean value is plotted in the same figure.
The data show a fair agreement between the experimental
2 replicates.
ADCQ3 = a · (cNa2 SO4 ) + b · cNa2 SO4 . (2)
Bell (1972) suggested determining the point liquid resi-
dence time (RTD) according to Eq. (3) as
It should be mentioned that the parameters a and b were
determined individually for every crossing point of the sen-
v
t · cNa2 SO4 (t)dt
sor to recognize the local sensor behavior due to wire lengths,  = 0 ∞ . (3)
cNa2 SO4 (t)dt
leakage current etc. as well as local gas–liquid structure and 0

dynamics. Further, it is worth mentioning that the calibration

In order to avoid errors resulting from the noisy fluctua-
cycle is individually required for every operating condition, i.e.
tions in the tail of the tracer pulse, an exponential fit was
gas and liquid throughputs and weir configuration, because
applied (see Fig. 5b). The parameter m and the exponent n
the dynamics of the ADC time series data and thus, the
were also determined individually to the tracer pulse of each
extracted third quartiles (Q3) depend on the local fluid dynam-
crossing point of the sensor.
ics.
The distributions of the point liquid residence time  on the
sieve tray for the four replicated runs are shown in Fig. 6a. The
3.2. Visualization of tracer pulse passage pattern show comparable residence time behavior with only
minor deviations in the column wall region, in particular close
The application of the calibration curves for all crossing to the outlet weir (see dashed square).
points of the wire-mesh sensor allows for the visualization In addition, the mean residence time pattern was
of the 1500 ml tracer pulses traveling from inlet to outlet calculated as average from the four runs (Fig. 6b). For quan-
weir indicated by the spatio-temporal Na2 SO4 concentration titative comparison, the corresponding RTD histograms were
distribution. Therefore, data of a ‘sliding window’ of 3000 extracted. Except for the fourth experiment (Exp 4), all RTD
frames was progressively shifted around 250 frames, i.e. incre- histograms collapsed into virtually one distribution trend. No
ments of 0.1 s in the time domain, and statistically analyzed. discernable reasons for the deviations were observed during
Fig. 4 shows exemplarily the tracer pulse passage at differ- the experiment for run Exp 4. It is speculated that deviations
ent time steps obtained at FB = 0.72 Pa0.5 and liquid load of might be traced back to a flow instability induced by the pump.
wL = 6.0 m3 /(m2 h) using the standard outlet weir. The main liq- Thus, subsequently mean data are used to characterize the
uid flow direction is from left (inlet) to right (outlet). From the fluid dynamic characteristics in detail.
instantaneous concentration pattern it can be concluded that
the liquid flow on the sieve tray is rather symmetric. Further- 3.3.2. Weir-to-weir liquid velocity
more, the weir-to-weir liquid velocity peaks at the centerline. The velocity distribution on the sieve trays determines the
In addition, the corresponding video of the whole tracer evolution of flow pattern and hence, the mass transfer and
pulse passage on the tray is given in the supporting informa- overall process performance. Ideally, the liquid flows rather
tion (video file: wL6.0m3m-2-h FB0.72Pa-0.5 weir0). The color straight from inlet to outlet weir without liquid channeling,
in the video corresponds with the color bar of Fig. 4. This video bypassing and recirculation. Thus, the weir-to-weir-velocity,
has a temporal resolution of 10 frames per second according i.e. the velocity component perpendicular to the weirs, gives
to the incremental shift of the sliding window as explained a clear indication about any asymmetric flow behavior. Lat-
above. Subsequently, the data matrix containing the time eral components of the velocity have not been determined. To
series data of the Na2 SO4 concentration can be further utilized determine weir-to-weir liquid velocities on the tray perpendic-
for the extraction of fluid dynamic characteristic. ular to the weirs, the time delay between tracer concentration

Fig. 4 – Example of a tracer pulse passage (wL = 6.0 m3 /(m2 h), FB 0.72 Pa0.5 , standard weir configuration).
 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146 143

Fig. 5 – (a) Dimensionless tracer concentrations obtained at sensor crossing point (27,27) and (b) Na2 SO4 concentration pulse
(Exp 3) with exponentially fitted tail (wL = 6.0 m3 /(m2 h), FB = 0.72 Pa0.5 , standard weir).

Fig. 6 – (a) Point liquid residence time obtained from four replicated tracer experiments at wL = 6.0 m3 /(m2 h), FB = 0.72 Pa0.5
using the standard weir, (b) mean point liquid residence time after averaging and (c) corresponding RTD histogram plot.

pulses from two sensor crossing points aligned in the respec- points with a distance of 75 mm were analyzed. In this way,
tive straight flow path is utilized, which is extracted via the weir-to-weir velocity at every position on the tray can
cross-correlation. Subsequently, the liquid velocity in the be extracted. For the sake of clarity, weir-to-weir velocities
main flow direction can be calculated according to the dis- are shown only for selected positions and subsequently dis-
tance of the selected crossing points. In this study, crossing cussed. Fig. 7a shows the liquid velocity exemplarily at four

Fig. 7 – (a) Weir-to-weir liquid velocities obtained from four replicated tracer experiments at wL = 6.0 m3 /(m2 h), FB = 0.72 Pa0.5
using the standard weir, (b) velocity data and mean velocity profile at L2, and (c) velocity data and mean velocity profile at L4.
144 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146

Fig. 8 – Point liquid residence time distribution and RTD histograms at FB = 0.72 Pa0.5 using the standard weir for (a)
wL = 6.0 m3 /(m2 h), (b) wL = 4.0 m3 /(m2 h), (c) corresponding liquid point residence time at the centerline (solid symbols) and
close to the column wall (empty symbols).

lines parallel to the weirs equally distributed across the tray for Fig. 9 shows the corresponding profiles of the weir-to-weir
the four replicated experimental runs described above (Fig. 6). liquid velocity at position (L2) and (L4), respectively (for the
The lengths of the arrows indicate the velocity magnitudes exact positions refer to Fig. 7). At position (L2), which is the
similar to a vector plot. end of the widening part of the tray, rather flat velocity profiles
While the velocity data of the four runs still scatter slightly are observed, while in the column constriction zone at posi-
at flow path position (L2), excellent agreement was found at tion (L4), parabolic profiles evolve as mentioned above. Similar
flow path position (L4). The analysis of the liquid velocities at velocity distributions close to the outlet weir were reported by
the tray indicates further that lateral fluctuations close to the Solari and Bell (1986), while results of Biddulph and Bultitude
inlet weir get shaped towards a parabolic profile while travel- (1990) imply that introduced velocity profiles at the inlet weir
ing to the outlet weir. For the analysis of the effects of liquid persist across the tray almost to the outlet. The low velocities
load and partial outlet weir obstruction, averaged data from close to the wall at position (L4) is naturally a result of local liq-
the replicated experimental runs are subsequently analyzed. uid pooling. However, it should be mentioned that other liquid
velocity components than the one directed from inlet to out-
4. Characterization of fluid flow pattern let weir are not accessed by this measurement method, thus
any vortex or swirling flow with possible liquid recirculation
4.1. Effect of the liquid load cannot be visualized.

The point liquid residence time distributions obtained at 4.2. Effect of partial outlet weir obstruction
different liquid capacities are mapped in Fig. 8a and b. In
addition, the RTD histogram data are embedded. The pat- In order to manipulate the flow pattern and to adjust
terns’ color coding as well as the histograms indicates lower unsymmetrical flow conditions on the tray, three out-
point residence times at the higher liquid load with a pretty let weirs with partial flow obstructions were applied (see
narrow distribution. The average point liquid residence time Fig. 1 and Table 2) at a vapor load factor FB = 0.72 Pa0.5
for wL = 6.0 m3 /(m2 h) is approximately 5 s. At lower load of and a liquid load wL = 5.2 m3 /(m2 h). For the outlet weir
wL = 4.0 m3 /(m2 h), it increases to approximately 7.5 s. How- configurations 1 and 2 with only a small obstruction
ever, a distinct feature of both patterns is the monotonous height (HOW = 35 mm), visual observation of the gas–liquid
increase of the residence time from inlet to outlet weir at the flow on the tray did not reveal any evidence of over-
centerline shown in Fig. 8c. Similar observations were found flow hindrance. Tracer studies were performed and the
by Stichlmair and Ulbrich (1987) and Solari and Bell (1986). Fur- corresponding videos of the tracer pulse passage on the
thermore, a conspicuous zone of extended residence time near tray for all weir obstructions are given in the supporting
the column wall at the onset of constriction towards the outlet information (video files: wL5.2m3m-2-h FB0.72Pa-0.5 weir1,
weir is observed. Accordingly, the slope of the point residence wL5.2m3m-2-h FB0.72Pa-0.5 weir2, wL5.2m3m-2-h FB0.72Pa-
time from inlet to outlet weir near the wall increases (Fig. 8c). 0.5 weir3). The color coding in the videos corresponds with
The high local residence time in this area, in particular at the the color bar of Fig. 4.
lower liquid load, indicates local pooling of the liquid as also Fig. 10 summarizes the extracted point liquid residence
observed by Solari and Bell (1986). time distribution and related RTD histograms. In contrast to

Fig. 9 – Effect of liquid load on weir-to-weir liquid velocity at FB = 0.72 Pa0.5 using the standard weir at different tray
positions (see Fig. 7), (a) L2, (b) L4.
 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146 145

Fig. 10 – Point liquid residence time distributions and RTD histograms at FB = 0.72 Pa0.5 and wL = 5.2 m3 /(m2 h) using outlet
weir with different partial obstruction (a) weir 1, (b) weir 2, (c) weir 3.

Fig. 11 – Effect of partial weir obstruction on weir-to-weir liquid velocity at FB = 0.72 Pa0.5 and wL = 5.2 m3 /(m2 h) at different
tray positions (see Fig. 7), (a) L2, (b) L4.

the visual flow observation, the residence time distribution measured. However, at position (L4), the weirs decrease the
patterns show clearly an effect of the obstructions for weir weir-to-weir liquid velocity considerably. It can be concluded
1 and 2. A zone of extended residence time in front of the that the lower weir obstructions with partial liquid overflow
outlet weir obstruction is observed. For the longer obstruction influence only the liquid flow near the outlet. Contrary, if over-
(weir 2) the zone expands further towards the inlet weir. Sur- flow in the obstructed part of the outlet weir is fully blocked,
prisingly, a second zone of extended residence time is also the whole tray pattern is dramatically altered.
observed in front of the ‘open’ part of the weir as known
from the unhindered flow (see Fig. 8). A further increase of 5. Conclusions
the obstruction weir height (Fig. 10c) boosts the effect on
the residence time distribution pattern. Here, the obstruction A wire-mesh sensor measurement technique has been
fully prevent liquid overflow. Accordingly, the point residence proposed as an excellent tool for high spatio-temporal visual-
time in front of the obstruction reaches up to 20 s. The RTD ization of the conductivity tracer passage on larger sieve trays.
histograms further highlights the effect of the obstruction. Further, a proper calibration and data treatment procedure has
While the zone near the inlet weir shows similar RTD behavior been developed, which allows extracting quantitative point
(represented by the left-hand part of the distribution), increas- liquid residence time patterns for the whole column tray area
ing flow hindrance further decelerates the flow at the outlet as well as weir-to-weir liquid velocities. High reproducibility of
region and obfuscates straight flow paths (represented by the the results was demonstrated with experimental replicates.
widened right-hand part of the distribution). Although not The feasibility of the methodology was proven for various
explicitly studied here, the obstruction may certainly increase operating conditions, i.e. liquid load and modified outlet weir
the gas–liquid contact height on the tray and thus, the liquid designs.
holdup, which will also slightly raise the RTD level. The results obtained from the experiments with standard
Stichlmair and Ulbrich (1987) studied also the effect of weir suggest that fluid flow conditions on sieve trays can be
modifications of the outlet weir on the flow pattern. How- far from being uniform, which requires such high-resolution
ever, their experiments with blocked sections of the outlet measurement techniques compared to conventional punc-
weir revealed only slight influence. Flow patterns comparable tual data acquisition. It was found that the point residence
to the one shown in Fig. 10 were determined by Stichlmair and time increases monotonously from inlet to outlet weir at
Weisshuhn (1973) with a sieve tray, which was slightly inclined the centerline while an extended residence time zone was
to the horizontal. Although the two-phase layer remained found near the column wall at the outlet region. The char-
visually the same across the whole tray, the tilting disturbed acteristic features of the flow pattern resulting from outlet
notably the flow pattern. weir obstructions, i.e. decelerated flow at the outlet region,
Fig. 11 shows the corresponding weir-to-weir liquid veloc- extended flow paths and distinct profiling of the weir-to-weir
ities at two tray positions, i.e. at the end of the widening part velocity, are also unveiled. The point residence time and weir-
(L2) and in the column constriction zone (L4). For the lower to-weir velocity data obtained in this study can be utilized as
obstructions (weir 1 and 2) no difference can be observed additional benchmark for CFD development.
at position (L2). Although the liquid load is little lower In the future, the technique will also be applied to local
(wL = 5.2 m3 /(m2 h)) compared to the experiment shown in tracer studies to visualize and quantify liquid recirculation
Fig. 7 (wL = 6.0 m3 /(m2 h)), a similar flat symmetric profile is as well as two-directional liquid tracer dispersion. Future
146 chemical engineering research and design 1 1 1 ( 2 0 1 6 ) 138–146

technical developments will focus on portable solutions for Nedeltchev, S., Donath, T., Rabha, S., Hampel, U., Schubert, M.,
larger column with diameters of more than 2 m to further 2014. New evidence for the mixing length concept in a narrow
support column tray design as well as trouble-shooting. bubble column operated in the transition regime. J. Chem.
Eng. Japan 47, 722–729.
Nedeltchev, S., Schubert, M., 2015. Statistical validation of the
Appendix A. Supplementary data mixing length concept in bubble columns operated in the
transition flow regime. J. Chem. Eng. Japan 48, 107–111.
Supplementary data associated with this article can be found, Noriler, D., Meier, H.F., Barros, A.A.C., Wolf Maciel, M.R., 2008.
in the online version, at doi:10.1016/j.cherd.2016.05.004. Thermal fluid dynamics analysis of gas–liquid flow on a
distillation sieve tray. Chem. Eng. J. 136, 133–143.
Noriler, D., Barros, A.A.C., Wolf Maciel, M.R., Meier, H.F., 2010.
References Simultaneous momentum, mass, and energy transfer
analysis of a distillation sieve tray using CFD techniques:
Bechtel, M., Paschold, J., Shilkin, A., Großmann, C., Kessel, J., prediction of efficiency. Ind. Eng. Chem. Res. 49,
2011. Untersuchungen gasgetriebener Fluiddynamik auf 6599–6611.
Böden. In: ProcessNet Annual Meeting of the Groups of Fluid Porter, K.E., Lockett, M.J., Lim, C.T., 1972. The effect of liquid
Process Engineering, Fulda, Germany. channeling on distillation plate efficiency. Trans. Inst. Chem.
Bell, R.L., 1972. Experimental determination of residence time Eng. 50, 91–101.
distribution on commercial scale distillation trays using a Prasser, H.-M., Grunwald, G., Höhne, T., Kliem, S., Rohde, U.,
fiber optic technique. AIChE J. 18 (3), 491–497. Weiss, F.-P., 2003. Coolant mixing in a PWR – deboration
Bell, R.L., Solari, R.B., 1974. Effect of nonuniform velocity fields transients, steam line breaks and emergency core cooling
and retrograde flow on distillation tray efficiency. AIChE J. 20 injection – experiments and analyses. Nucl. Technol. 143 (1),
(4), 688–695. 37–56.
Biddulph, M.W., Bultitude, D.P., 1990. Flow characteristics of a Prasser, H.-M., Böttger, A., Zschau, J., 1998. A new electrode-mesh
small-hole sieve tray. AIChE J. 36 (12), 1913–1916. tomograph for gas/liquid flows. Flow Meas. Instrum. 9,
da Silva, M.J., Schleicher, E., Hampel, U., 2007. Capacitance 111–119.
wire-mesh sensor for fast measurement of phase fraction Rahimi, R., Rahimi, M.R., Shahraki, F., Zivdar, M., 2006. Efficiency
distributions. Meas. Sci. Technol. 18, 2245–2251. of sieve tray distillation columns by CFD. Chem. Eng. Technol.
da Silva, M.J., Thiele, S., Abdulkareem, L., Azzopardi, B., Hampel, 29 (3), 326–335.
U., 2010. High-resolution gas-oil two-phase flow visualization Rollbusch, P., Becker, M., Ludwig, M., Bieberle, A., Grünewald, M.,
with a capacitance wire-mesh sensor. Flow Meas. Instrum. 21, Hampel, U., Franke, R., 2015. Experimental investigation of the
191–197. influence of column scale, gas density and liquid properties
da Silva, M.J., dos Santos, E.N., Hampel, U., Rodriguez, I.H., on gas holdup in bubble columns. Int. J. Multiphase Flow 75,
Rodriguez, O.M.H., 2011. Phase fraction distribution 88–106.
measurement of oil–water flow using a capacitance Schleicher, E., Aydin, T.B., Vieira, R.E., Torres, C.F., Pereyra, E.,
wire-mesh sensor. Meas. Sci. Technol. 22, 104020. Sarica, C., Hampel, U., 2015. Refined reconstruction of
Gesit, G., Nandakumar, K., Chuang, K.T., 2003. CFD modeling of liquid–gas interface structures for stratified two-phase flow
flow patterns and hydraulics of commercial-scale sieve trays. using wire-mesh sensor. Flow Meas. Instrum.,
AIChE J. 49 (4), 910–924. http://dx.doi.org/10.1016/j.flowmeasinst.2015.06.002.
Härting, H.U., Berger, R., Lange, R., Larachi, F., Schubert, M., 2015. Schubert, M., Kryk, H., Hampel, U., 2010. Slow-mode
Liquid backmixing in an inclined rotating tubular fixed bed gas/liquid-induced periodic hydrodynamics in trickling
reactor–Augmenting liquid residence time via flow regime packed beds derived from direct measurement of
adjustment. Chem. Eng. Process: Process Intensification, cross-sectional distributed local capacitances. Chem. Eng.
http://dx.doi.org/10.1016/j.cep.2015.02.003. Process: Process Intensification 49 (10), 1107–1121.
Hirschberg, S., Wijn, E.F., Wehrli, M., 2005. Simulating the Solari, R.B., Bell, R.L., 1986. Fluid flow patterns and velocity
two-phase flow on column trays. Trans. IChemE, Part A 83, distribution on commercial-scale sieve trays. AIChE J. 32 (4),
1410–1424. 640–649.
Krishna, R., van Baten, J.M., Ellenberger, J., Higler, A.P., Taylor, R., Stichlmair, J., Weisshuhn, E., 1973. Untersuchungen zum
1999. CFD simulations of sieve tray hydrodynamics. Trans. Bodenwirkungsgrad unter besonderer Berücksichtigung der
IChemE, Part A 77, 639–646. Flüssigkeitsvermischung. Chem. Ing. Tech. 45 (5), 242–247.
Krishna, R., van Baten, J.M., 2003. Modelling sieve tray hydraulics Stichlmair, J., Ulbrich, S., 1985. Einfluss der Maldistribution auf
using computational fluid dynamics. Trans. IChemE, Part A die Trennwirkung großer Kolonnenböden. Chem. Ing. Tech. 57
81, 27–38. (5), 468–470.
Lewis, W.K., 1936. Rectification of binary mixtures. Plate efficiency Stichlmair, J., Ulbrich, S., 1987. Liquid channeling on trays and its
of bubble cup columns. Ind. Eng. Chem. 28 (4), 399–402. effect on plate efficiency. Chem. Eng. Technol. 10, 33–37.
Li, Y., Wang, L., Yao, K., 2007. New technique for measuring fluid van Baten, J.M., Krishna, R., 2000. Modelling sieve tray hydraulics
flow patterns on a multiple downcomer tray. Ind. Eng. Chem. using computational fluid dynamics. Chem. Eng. J. 77, 143–151.
Res. 46, 2892–2897. Vieira, R.E., Kesana, N., Torres, C.F., McLaury, B.S., Shirazi, S.A.,
Lucas, D., Krepper, E., Prasser, H.-M., 2005. Development of Schleicher, E., Hampel, U., 2014a. Experimental investigation
co-current air-water flow in a vertical pipe. Int. J. Multiphase of the effect of 90◦ standard elbow on horizontal gas–liquid
Flow 31, 1304–1328. stratified and annular flow using dual wire-mesh sensor. Exp.
Mohammed, I., Bauer, T., Schubert, M., Lange, R., 2013. Therm. Fluid Sci. 59, 72–87.
Hydrodynamic multiplicity in a tubular reactor with solid Vieira, R.E., Kesana, N., Torres, C.F., McLaury, B.S., Shirazi, S.A.,
foam packings. Chem. Eng. J. 231, 334–344. Schleicher, E., Hampel, U., 2014b. Experimental investigation
Mohammed, I., Bauer, T., Schubert, M., Lange, R., 2015. Gas–liquid of horizontal gas–liquid stratified and annular flow using
distribution in tubular reactors with solid foam packings. wire-mesh sensor. ASME J. Fluids Eng. 136 (12), 121301.
Chem. Eng. Process: Process Intensification 88, 10–18. Wang, X.L., Liu, C.J., Yuan, X.G., Yu, K.T., 2004. Computational
Müller, E.A., Cavero, A., Estevez, L.A., 1988. Improving flow fluid dynamics simulation of three-dimensional liquid flow
pattern in a distillation tray by modifying downcomer apron and mass transfer on distillation columns trays. Ind. Eng.
shape. Chem. Eng. Commun. 74, 195–208. Chem. Res. 43, 2556–2567.