Water management studies in PEM fuel cells, Part II: Ex situ

Water management studies in PEM fuel cells, Part II: Ex situ investigation of flow maldistribution, pressure drop and two-phase flow pattern in gas channelsi n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 6
Avai lab le a t www.sc iencedi rec t .com
j ourna l homepage : www.e lsev ier . com/ loca te /he
Water management studies in PEM fuel cells, Part II: Ex situ investigation of flow maldistribution, pressure drop and two-phase flow pattern in gas channels
Z. Lua, S.G. Kandlikara,*, C. Ratha, M. Grimma, W. Domigana, A.D. Whitea, M. Hardbargera, J.P. Owejanb, T.A. Traboldb
aMechanical Engineering Department, Rochester Institute of Technology, 76 Lomb Memorial Dr., Rochester, NY 14623, USA bFuel Cell Research Laboratory, General Motors Corporation, 11 Carriage Street, Honeoye Falls, NY 14472, USA
a r t i c l e i n f o
Article history:
Keywords:
Channel-manifold interface
* Corresponding author. Fax: þ1 585 475 7710 E-mail address: [email protected] (S.G. Kan
0360-3199/$ – see front matter ª 2008 Intern doi:10.1016/j.ijhydene.2008.12.025
a b s t r a c t
Two-phase flow of water and reactant gases in the gas distribution channels of proton
exchange membrane fuel cells (PEMFCs) plays a critical role in proper water management.
In this work, the two-phase flow in PEMFC cathode parallel channels is studied over a wide
range of superficial air velocity (air stoichiometry) and superficial water velocity in
a specially designed ex situ experimental setup, which enables the measurement of
instantaneous flow rates in individual gas channels and simultaneous visualization of the
water flow structure. It is found that the two-phase flow at low superficial air velocities (air
stoichiometry below 5) is dominated by slugs or semi-slugs, leading to severe flow mal-
distribution and large fluctuations in the pressure drop. Slug residence time, measured
from the video observation and the instantaneous flow rate data, is found to be a new
parameter to describe the slug flow. At higher air velocities, a water film is formed on the
channel walls if they are hydrophilic. The pressure drop for the film flow is characterized
by smaller but frequent fluctuations, which are found to result from the water buildup at
the channel-exit manifold interface. As the superficial air velocity increases further, mist
flow is obtained where little water buildup is observed. The water buildup in the gas
channels at the two-phase flow is well described by the two-phase friction multiplier,
defined as the ratio of the two-phase pressure drop to the single gas phase pressure drop. It
is found that the two-phase friction multiplier increases with increasing water flow rate. A
flow pattern map is developed using superficial water and air velocities with clearly defined
transition regions.
ª 2008 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
reserved.
1. Introduction humidified air and hydrogen gas streams, must be present
Water management has been identified as one of the most
critical issues in the performance and longevity of a PEMFC
[1,2]. Sufficient water, often controlled by externally
. dlikar). ational Association for H
within the fuel cell to maintain the proton conductivity of the
polymer electrolyte membrane; however, excess water must
be removed from the cell to avoid flooding. Flooding is
a phenomenon in which liquid water accumulation inside
ydrogen Energy. Published by Elsevier Ltd. All rights reserved.
Re Real numbers
Deff Effective diffusion coefficient of oxygen, m2/s
L Thickness, m
a Half-angle of the channel corner
DP Pressure drop, kPa
DP2q Two-phase flow pressure drop, kPa
DPg Single gas phase pressure drop, kPa
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 63446
a fuel cell blocks gas transport pathways in the catalyst layers,
gas diffusion layers (GDLs), and the gas channels, inducing
large mass-transport losses. Due to water condensation from
the humidified gas streams and water product at the cathode
side, two-phase flow commonly exists in various PEMFC
components. In automotive applications, liquid water may
appear in the gas channels during the startup from the normal
(25 C) as well as freezing conditions due to the low water
saturation pressure in the gas streams. It is therefore impor-
tant to investigate the two-phase flow in a PEMFC and its
influence on water accumulation and flooding.
Two-phase flow in a fuel cell could be considered in three
sub-categories: catalyst layer flooding, GDL flooding, and two-
phase flow in gas channels. A number of studies are reported
on the investigation of water transport in the GDL, e.g. [3–7],
and the catalyst layers [8–10]. Two-phase flow in gas channels
is also very important and has recently received more atten-
tion with the application of several visualization techniques,
such as optical and neutron imaging. Tuber et al. [11] were the
first group to use optical visualization to study water buildup
in a cathode gas channel at low temperatures. This technique
was later widely used to study water transport in gas channels
under various fuel cell operating conditions [12–18]. As a result
of visual studies, different liquid water flow structures (i.e.
slug, film and mist flow) have been identified and their rela-
tions with fuel cell performance have been investigated.
Several important conclusions have been obtained from these
studies: (1) liquid water emerges in the form of droplets on the
hydrophobic GDL surface; (2) under normal fuel cell operating
conditions the film/slug flow is the primary flow pattern,
especially near the channel exit; (3) the liquid water columns
(slugs) accumulating in the gas channels result in cell
performance loss, likely due to reduction in the effective
electrochemical reaction area caused by water blockage.
Neutron radiography provided another in situ visualization
technique and was utilized by a number of groups to visualize
and quantify water retention in the GDL, under the lands, and
in the gas channels [19–22]. Higher water retention was found
at the U-bends and under the lands of the serpentine chan-
nels. In a recent publication, Owejan et al. [23] used this
technique to study the effects of flow field and diffusion layer
properties on water accumulation in a PEMFC and found that
flow field channels with hydrophobic coating retain more
water, but the distribution of a greater number of smaller
slugs in the channel improves fuel cell performance. Water
distribution in the anode side was also visualized in a few
studies [16,18,24]. It was found that flooding in anode
channels is at least as severe as in the channels on the
cathode side.
Two-phase flow in gas channels has been investigated
through modeling and numerical simulation as well. Quan
et al. [25] simulated the water flow behavior in an U-shaped
channel using volume of fraction model and varying initial
liquid water distributions. Jiao et al. [26] did similar work for
fuel cell stacks with varying preset water distributions.
Recently, Quan and Lai [27] studied the effects of channel
surface hydrophilicity together with channel geometry and air
inlet velocity on the water behavior in a serpentine channel,
and they concluded that the hydrophilicity of flow channel
surface plays an important role in water management of
a reactant flow channel. However, a common problem faced
in such models is that the liquid water must be placed a priori
at certain locations in the channels, thus making the simu-
lation less physically realistic. Wang et al. [28] developed
a two-phase flow model, with the analogy of the PEMFC
channels to the porous media, to estimate the liquid water
saturation profiles along the axial flow direction.
Most of these investigations focused on the behavior of
liquid water in gas channels but the flow dynamics of the gas
streams were largely neglected. Presence of liquid water is
found to influence the airflow and the two-phase flow pres-
sure drop. In a few studies, the pressure drop across the
channels was shown to be closely related to the PEMFC
performance [17,29], and was therefore proposed to be
a diagnostic tool for the detection of flooding [30–33]. Despite
these studies, the relationship between pressure drop and
water accumulation, in particular the effect of random droplet
emergence and liquid water clogging, is not clearly under-
stood. Part of the reason is that the pressure drop provides an
overall measure across the entire flow field, while the water
accumulation and water clogging may be highly localized to
certain regions in specific channels. On the other hand, vari-
ation of individual channel airflow rates provides a sensitive
measure of the liquid water in that channel. A real-time
monitoring of the instantaneous flow rate in individual
channels would therefore provide important information with
regards to water accumulation and two-phase behavior in
PEMFC gas channels.
parallel microchannels has been a long-standing technical
challenge. Recently, a novel technique to measure these
instantaneous flow rates was proposed by Kandlikar et al. [34].
This technique is based on pressure drop measurements in
the entrance region of each channel. With this method,
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 6 3447
measurement of individual channel flow rate variations
caused by accumulation of liquid water in the channel is now
possible.
The main focus of this work is to investigate the two-phase
flow in parallel PEMFC channels. For this purpose, the
instantaneous channel flow rate (flow maldistribution), pres-
sure drop, and the flow structure will be studied by using the
entrance region pressure drop method, differential pressure
transducers, and high-speed visualization, respectively. The
ex situ test setup is carefully calibrated in order to accurately
measure the instantaneous flow rates in individual channels.
A quantitative description of the channel flooding is estab-
lished and the pressure drop signatures for each flow pattern
are identified. The results obtained from these measurements
help us to better understand the relationships among flood-
ing, pressure drop, and flow structure.
2. Experimental
work. The water production is simulated by injecting water at
a desired flow rate through the GDL. The test section, shown
schematically in Fig. 1(a), comprises a gas channel plate, GDL
sample, a water channel plate, and gaskets to seal the
assembly. The gas channels consist of eight parallel channels,
Fig. 1 – (a) Exploded side-view of conceptual gas and
waterside manifold assembly. Not to scale. (b) Schematic of
ex situ experimental test section header design including
channel pressure tap locations, inlet pressure tap, air inlet,
and dowel pinholes. Not to scale.
each 183 mm long, 0.7 mm wide, and 0.4 mm deep with a land
width of 0.5 mm between adjacent channels. The channel
shape is weaving with a 5 weaving angle to avoid mechanical
shear on the GDL associated with straight channels. The
channel dimensions and geometry are taken from an actual
fuel cell flow design which ensures the best fuel cell perfor-
mance while meeting the Department of Energy targets for
automotive fuel cells [35]. The channels on the waterside
manifold have the same geometry and dimensions as the gas
channels except that the waterside channels are sectioned
into four segments corresponding to the four water chambers.
The use of four chambers allows for uniform water flow
through the entire GDL, and is superior to the use of a single
chamber along the entire channel length, due to the fact that
a single chamber would promote water flow primarily toward
the outlet end of the channels (due to highest available pres-
sure difference between the air and water sides). The water
flow rate in each chamber was controlled independently by
four individual syringe pumps, allowing each chamber to
support a different water injection rate if desired. Three holes
were drilled from each water channel to each water chamber,
resulting in a total of 24 holes per water chamber across the 8
channels, and 12 holes per channel across its entire length, as
shown in Fig. 1(a). Each hole had a diameter of 0.7 mm which
is equal to the gas channel width. Both the gas channels and
water manifold are machined out of Lexan plates, which are
then vapor polished to provide excellent optical clarity. This
test section simulates the water production at the PEMFC
cathode electrode as well as the transport through the
cathode GDL to the gas channels.
The header of the gas channels is specially designed to
allow the measurement of pressure drop in the entrance
region and across the channels. Three rows of holes are made
in the straight section of the channels to hold the pressure
taps, which allow measurement of the entrance region pres-
sure drop for individual channels. Downstream of the pres-
sure tap holes, another set of holes is drilled to create
a provision for dowel pins, which will be used to block all the
other channels while calibrating each channel individually.
Fig. 1(b) is a schematic of the header design. The instanta-
neous gas flow rate in individual channel can be obtained
from the measured individual channel pressure drop in the
entrance region provided that each channel is well calibrated.
The theory of the entrance region pressure drop method and
calibration of the ex situ setup can be found in Ref. [34].
The GDL and gasket materials used in this work are
provided by General Motors. In order to improve the water
management, the GDL is PTFE treated and coated with
a micro-porous layer (MPL). The GDL has a thickness of
approximately 230 mm. The hard-stop PTFE gasket ensures the
appropriate compression under the compressive load. The
test section is assembled under a compression of 2068 kPa
(300 psi). A compression ratio of about 20% for GDL is achieved
with this clamp pressure. During experiments the GDL is
placed in a vertical down flow orientation. The test section is
kept at ambient temperature (about 24 C) throughout the
experiments.
clean air generated from a Zero Air Generator (Parker HPZA-
30000, Haverhill, MA) flows through a bank of rotameters,
Fig. 2 – Experimental setup to investigate the two-phase
flow in the simulated PEMFC gas channels.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 63448
which control the total airflow rate, and then into the gas inlet
manifold. The input airflow rate is measured by a digital
flow meter (Omega FMA-1620 A) for the lower flow range of
0–1200 sccm (corresponding to a superficial air velocity range
of 0–8.9 m/s) and by rotameters for the higher flow rates. The
airflow rate in each individual channel is obtained through the
entrance region pressure drop in each channel measured with
eight differential pressure sensors (Honeywell FP2000, with
accuracy 0.25% in the range of 0–15 kPa). The total pressure
drop across the flow field is also measured with a differential
pressure sensor (Honeywell Sensotec FDW2AT) which has an
accuracy of 0.25% or better in a range of 0–35 kPa. All pressure
data are collected with a LabView program through a DAQ
system (National Instruments). Deionized water (18.2 MU,
Millipore) is independently delivered to each of the four water
chambers through four syringe pumps (Harvard). Before
acquiring data, the water injection and airflow at desired flow
rates are run for enough time until the steady state is reached
(normally within a few minutes). The data are taken over
a period of more than 10 min. The water flow patterns inside
the gas channels are captured by a Photron high-speed
camera with a long-distance microscopic lens. Videos were
recorded with a resolution of 1024 1024 and a frame rate
Table 1 – Air and water flow rates used in testing and correspo superficial velocities, calculated based on a total cross-sectiona the air velocity given in the table.
Water injection rate (mL/min)
Current densitya
0.00 0 0.0
0.02 66 0.2
0.04 132 0.4
0.10 330 1.0
0.20 660 2.0
a Current density values found by back calculating what the associated
channel number and dimensions (with an active area of 18.4 cm2).
range of 60–2000 fps. A halogen source dual fiber optic light
guide is used for illumination. All optical equipments and the
test setups are mounted on a vibration isolation table.
The water injection rates and airflow rates are selected to
match the water production rate and the gas flow rates at the
cathode side of a PEMFC. Table 1 shows the airflow and water
injection rates as well as the corresponding current densities.
The current density is not actually produced during ex situ
experimentation, rather it provides an idea of what a real fuel
cell with the same flow field would produce with the given
water generation rates. The superficial air velocity, Reynolds
number, and water superficial velocity are calculated based on
the designed channel cross-section area and are also listed in
the table. Table 1 displays airflow rates associated with
a stoichiometric ratio of 1 for each water flow rate; however,
the air stoichiometries ranging 1–45 are studied (larger ratios
applied only to lower water injection rates). The superficial air
velocities are varied accordingly.
3. Results and discussion
Flow maldistribution caused by the water accumulation in gas
channels is one of the most important water management
concerns in a PEMFC. In this work, the instantaneous gas flow
rates in individual channels are measured during two-phase
flow by means of the entrance region pressure drop method
[34]. It is worthy to point out that the entrance region is before
the introduction of water so that only single-phase gas flows
in this region.
dynamic) holdup is observed due to the lower air velocities
and severe flow maldistribution is observed. Fig. 3 shows the
flow rate in eight channels and their summation as a function
of time for a superficial water velocity of 3 104 m/s (water
injection rate of 0.04 mL/min) and a superficial air velocity of
1.7 m/s (corresponding to airflow rate of 198 sccm or air stoi-
chiometry of 2). The most significant result from this figure is
seen at 85 s, when the flow rate in channel 4 drops sharply
from 35 sccm to about 10 sccm, and the flow rates in other
channels (e.g. channels 2, 5 and 6) increase to maintain the
constant total flow rate. The simultaneous high-speed video
images of the flow structure are shown in Fig. 4. A large slug is
nding operating current density, Reynolds number and l area of 2.24 mm2. The Reynolds number of air is shown for
Superficial water velocity, UL (m/s)
Superficial air velocity, UG (m/s)
Reynolds number (Re)a
0.0 0.0 0.0
1.5 104 0.5 16.2
3.0 104 1.0 32.3
7.4 104 2.5 80.8
1.5 103 4.9 161.7
water and airflow rates would produce in a real fuel cell with same
Fig. 4 – The water flow structure, captured as a still picture
from a video, for the same experiment as Fig. 3 with
superficial water velocity of 3 3 10L4 m/s (water injection
rate 0.04 mL/min) and superficial air velocity of 1.7 m/s
(198 sccm) for (a) at the beginning of video recording and (b)
after 100 s. The numbers in the figures represent the
channel number.
Fig. 3 – The instantaneous flow rate in eight channels and
their summation taken over a period of 250 s for the two-
phase flow at superficial water velocity of 3 3 10L4 m/s
(water injection rate 0.04 mL/min) and superficial air
velocity of 1.74 m/s (198 sccm, corresponding to
stoichiometric ratio of 2). The numbers in the figures
represent the channel number.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 6 3449
clearly observed to form in channel 4 and it moves very
slowly. Figs. 3 and 4 clearly display the blockage of the gas flow
by slug formation in the channels and the resultant severe
flow maldistribution. In Fig. 4 stationary water holdup (large
droplets) is also observed in channels 5 and 6. They may affect
the flow distribution in these channels as well, but to a much
lesser degree. It is important to note that the flow rate in
channel 4 is not reduced to zero upon the slug formation. This
may indicate a semi-slug [36] in which the gas phase remains,
to some extent, continuous. Through careful investigation of
a large number of high-speed videos, water accumulation is
seldom observed to contact the hydrophobic GDL surface,
even with the combination of the highest water injection rate
and the lowest airflow rate.
It is interesting to note that in Fig. 3 the channel flow rate is
not uniformly distributed even before the slug formation in
channel 4. Generally, the flow maldistribution in a PEMFC flow
field could be due to several reasons, including water holdup
in gas channels during PEMFC operation, improper design of
manifold and flow field, and non-uniform GDL intrusion. It is
thus of importance to delineate the flow maldistribution
caused by each mechanism. For this purpose, it is necessary to
compare the flow distribution pattern under two-phase flow
conditions to that under single-phase gas flow condition. Fig. 5
lists the flow distribution patterns during two-phase flow of
Fig. 3 at time of 16.5 s and 200 s, corresponding to before and
after the slug formation, respectively. The flow rate in channel
4 significantly decreases upon the formation of the slug, while
the flow rates in some of the other channels increase to
balance the total flow rate. For comparison, the flow distri-
bution patterns under the single gas phase conditions are
shown in Fig. 6 for three airflow rates: 310, 1000 and
3000 sccm. Fig. 6 shows a similar pattern for different airflow
rates, for example, with channel 6 having the highest airflow
rate and channel 8 the lowest flow rate at all three input flow
rates. A non-uniform flow distribution is clearly observed in
Fig. 6. Two potential reasons could contribute to this flow
maldistribution under single-phase gas flow conditions:
a non-uniform flow field unit and non-uniform GDL intrusion.
The results of multiple experiments (not shown here) indicate
that the non-uniform GDL intrusion is the most probable
cause for the maldistribution during single-phase gas flow. A
more detailed analysis on the GDL intrusion and its effects on
the flow maldistribution are available in Ref. [37]. Since the
non-uniform GDL intrusion induced flow maldistribution
pattern has been established, it is reasonable to compare the
flow distribution under two-phase flow conditions in Fig. 5 to
those under single-phase gas flow conditions (Fig. 6). It is
found that, under two-phase flow conditions, the flow
1 2 3 4 5 6 7 8 0
10
20
30
1 2 3 4 5 6 7 8 0
10
20
30
40
w r a t e ( s c c m
)
Fig. 5 – Flow distribution patterns under two-phase flow
conditions for Fig. 3. (a) Flow distribution at t [ 16.5 s which
is prior to the formation of slug in channel 4. (b) Flow
distribution at t [ 200 s when a large slugexists in channel 4.
Fig. 7 – Water buildup at the channel-outlet manifold
interface for the same experiment as in Figs. 3 and 4.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 63450
distribution patterns are significantly different from those at
single-phase gas flow. For example, in Fig. 6 channel 6 shows
the highest flow rate at a single-phase gas flow condition,
while the flow rate in channel 6 is among the lowest at two-
phase conditions. This implies that, in addition to the non-
uniform GDL intrusion, there is another factor to contribute to
the flow maldistribution. A possible such factor may be the
instantaneous water holdup in the channel in the form of
water films or stationary droplets.
Another interesting phenomenon in Fig. 3 is the occa-
sionally occurring sharp spikes in the airflow rate, for
example, at times of 100, 150 and 225 s. These spikes are
induced due to the water removal processes at the channel-
outlet manifold interface. Fig. 7 shows the water buildup at
the transition area from channel to the exit header. This water
buildup, in the form of a film, provides a means for water to
1 2 3 4 5 6 7 8 0
100
200
300
400
500
w r a t e ( s c c m
)
310 sccm, 1000 sccm and 3000 sccm.
drain out of the flow field through a common exit. Since this
water film bridges all the channels, the drainage process will
affect all channel flow rates in the same way. Each spike in the
flow rate reflects a water drainage process. This finding
implies that the parallel channel-manifold interface plays an
important role in the water transport in PEMFC, especially at
the stack level where much more water is transported
through this interface and should be taken into account in the
flow field design.
for all water injection rates studied at lower air velocities
(below 5.2 m/s). These test conditions correspond to current
densities in the range of 0.2–2 A/cm2 and cathode stoichio-
metric ratios from 1 to 5. In this work all the water injected is
transported through the gas channels. In an operating fuel
cell, part of the water produced at the cathode electrode is
absorbed by the electrolyte membrane to increase its water
content and some water may diffuse through the membrane
to the anode. The result is that the net water transport in the
cathode gas channels is less than the water production,
although condensation from the air stream may increase the
water flow. Nevertheless, the ex situ experiment provides
several important observations which are not readily avail-
able with an in situ fuel cell experiment. One such observation
is the slug residence time, which is defined as the period over
which the flow rate in one or several channels is significantly
decreased (see Fig. 3). The slug residence time is an important
parameter in describing the two-phase flow dynamics and
provides an important indication of channel flooding. A slug
(or semi-slug) can reside in a channel for a time duration from
several seconds to hundreds of seconds and its residence time
decreases non-linearly with the increasing airflow rate, as
shown in Fig. 8. The slug residence time does not seem to have
an apparent relation to the superficial water velocity. A
0 2 4 6 8 0
50
100
150
200
UL = 2.98 x 10-4 m/s
UL = 7.44 x 10-4 m/s
UL = 14.9 x 10-4 m/s
S l u
e n
e ( s )
U G
(m/s)
Fig. 8 – Slug residence times as a function of superficial air
velocity and water velocity.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 6 3451
significant consequence of slug flow is a reduction in reactant
gas concentration at reaction sites. The oxygen diffusion time
can be roughly estimated with the following equation:
so2 diff ¼
(1)
where s is the oxygen diffusion time, LGDL is the thickness of
the GDL, LCL is the thickness of the catalyst layer, Deff GDL and Deff
CL
are the effective diffusion coefficients for O2 in GDL and
catalyst layer, respectively. Using the characteristic length
scale for each component and the corresponding effective gas
diffusion coefficient (e.g. DGDL¼ 101 cm2/s, DCL¼ 106 cm2/s)
yields s in the order of 0.1 s [38], which is much less than the
slug residence time in the channels. Therefore, slug flow
Fig. 9 – Two-phase flow dynamics at intermediate airflow rate.
summation for the two-phase flow at water flow rate of 0.04 mL/
stoichiometric ratio of 8). (b) Enlarged water flow structure capt
readily leads to fuel cell flooding and poses an important
water management concern.
It has been found that slugs are more readily formed in dry
channel sections. This is understandable because the
dry channel wall has a larger contact angle than the wetted
channel wall where a thin water film is already formed, as
surface tension provides the main resistance to the slug
movement. When a slow moving slug comes into contact with
water film or other droplets in the channel, its velocity sharply
increases. Once the slug starts to move it cannot maintain the
slug form and usually transforms into a water film until it is
cleared out of the channel. The channel surface properties
(surface tension and contact angle) play an important role in
the two-phase flow dynamics and are planned for future
study.
the flow maldistribution decreases significantly and the slug
or semi-slug flow disappears. Fig. 9 shows an example of the
flow maldistribution and the corresponding flow structure at
intermediate airflow rates, where the superficial water
velocity is 3.0 104 m/s (water injection rate 0.04 mL/min)
and air velocity is 7.9 m/s (corresponding to a stoichiometric
ratio of 8). Under these conditions, a water film is formed and
moves along the channel sidewall (Fig. 9(b)). The water film
formation along the channel is dictated by the Concus–Finn
condition [39]:
2 (2)
where q is the contact angle of water on the channel wall and
a is the half-angle formed by the channel corner. The contact
angle of the Lexan channel is measured to be around 60. For
the rectangular channel (a¼ 45) used in this experiment, Eq.
(2) is not satisfied and therefore water cannot be wicked into
the channel corner to form a stable, continuous film, and
liquid water collects on the channel walls in the form of large
(a) Instantaneous flow rate in eight channels and their
min and air velocity of 7.9 m/s (1057 sccm, corresponding to
ured as a still picture from a video.
Fig. 10 – Instantaneous flow rate plot for a superficial water
velocity of 3.0 3 10L4 m/s (water flow rate 0.04 mL/min)
and a superficial air velocity of 24.63 m/s (3302 sccm,
corresponding to stoichiometric ratio of 25).
1 2 3 4 5 6 7 8 0
50
100
150
200
w r a t e ( s c c m
)
1 2 3 4 5 6 7 8 0
200
400
600
input air flow = 3302 sccm (mist flow) b
Fig. 11 – Flow distribution patterns under two-phase flow
conditions at water flow rate of 0.04 mL/min and at (a)
airflow rate of 1057 sccm (corresponding to water film flow)
and (b) airflow rate of 3302 sccm (corresponding to mist
flow).
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 63452
droplets. Due to the drag force of the downward airflow as well
as gravity, large droplets deform along the channel wall and
form a film-like structure. When the drag force and gravity
overcome surface tension forces, the film-like structure starts
to move, leading to a thick leading face and a much thinner
tail, as shown in Fig. 9(b).
As superficial air velocity further increases, corresponding
to air stoichiometry above 10, little or no water accumulation
in the channels is observed. Liquid water (in the form of
droplets) is removed from the GDL surface directly by the
shear force of the gas flow, followed by mist flow. Fig. 10
shows the individual channel flow rate plot at a superficial
water velocity of 3.0 104 m/s (water injection rate 0.0.4 mL/
min) and a superficial air velocity of 24.6 m/s (corresponding
to air stoichiometry of 25), as a typical example of mist flow.
No flow rate variation is observed, which agrees with flow
structure visualization. In mist flow tiny water droplets are
suspended in the gas stream. Evaporation is also expected to
contribute significantly to the water removal at such high
airflow rates.
At intermediate to high superficial air velocities (above
7.4 m/s), the thin water film present on the walls does not
significantly reduce the channel cross-sectional area. There-
fore, little variation, as low as 4.3% for the maximum deviation
from the mean channel flow rate in Fig. 9, is observed in the
airflow rate in each channel, unlike the case of the slug flow
where large changes, as high as 75% for the channel 4 in Fig. 3,
in airflow rate are seen. Fig. 11 shows the flow distribution
patterns for the film flow (UL¼ 3.0 104 m/s, UG¼ 7.9 m/s,
Fig. 9) and the mist flow (UL¼ 3.0 104 m/s, UG¼ 24.6 m/s,
Fig. 10). However, a different distribution pattern is found for
the film flow and the mist flow, for example, channels 2 and 5
show higher flow rates in the film flow, while the flow rates
are lower during the mist flow. This difference may be due to
the effects of the water film in the channels. By comparing
Figs. 11 and 6, it is found that the flow distribution pattern for
the mist flow is same to that of the single-phase condition.
This is expected because during mist flow no additional mal-
distribution is introduced except for the intrinsic maldis-
tribution caused by the non-uniform GDL intrusion.
3.2. Pressure drop
The flow maldistribution in the parallel gas flow channels is
undesirable as it leads to water holdup and performance
degradation (non-uniform current density, power reduction,
etc.). However, extensive test setup is required to accomplish
the instantaneous channel flow rate measurements [34]. On
the other hand, the pressure drop across all gas channels
provides a convenient and reliable tool to monitor the two-
phase flow.
Fig. 12 shows a comparison of the pressure drop behavior
for the two-phase flow at a superficial water velocity of
3.0 104 m/s (water injection rate 0.04 mL/min) and at three
different air velocities, 1.7 m/s (Fig. 3), 7.9 m/s (Fig. 9) and
24.6 m/s (Fig. 10) which correspond to a typical slug flow,
annular/film flow and mist flow, respectively. The most
significant observation from these figures is that a large fluc-
tuation in the pressure drop is noted for the slug flow, smaller
oscillations for film flow, and no variation for mist flow.
Pressure drop (DP) oscillations for slug flow (Fig. 12(a)), in the
range 0.1–0.3 kPa, are apparently related to the formation and
removal of slugs and semi-slugs within the channels. DP for
film flow pattern (Fig. 12(b)) shows periodic oscillations, which
are thought to result from water buildup at the channel-
manifold interface. On the other hand, mist flow shows little
DP variation during tested periods (Fig. 12(c)), which is
consistent with channel flow rate measurements.
The pressure drop is mainly a result of friction during the
passage of reactant gases through the channels. As the pres-
ence of liquid water hampers gas flow in fuel cell channels,
0 50 100 150 200 250 0.6
0.8
1.0
1.2
1.4
1.6
5.0
5.2
5.4
5.6
5.8
4.96
4.97
4.98
4.99
19.2
19.4
19.6
19.8
20.0
at superficial water velocity of 3.0 3 10L4 m/s (water
injection rate 0.04 mL/min) and three different air
velocities: (a) 1.7 m/s (198 sccm, corresponding to air
stoichiometry of 2); (b) 7.9 m/s (1057 sccm, corresponding
to air stoichiometry of 8); and (c) 25.6 m/s (3302 sccm,
corresponding to air stoichiometry of 25). These three air
velocities correspond to typical slug flow, film flow and
mist flow, respectively. The inserted plot in (b) is the
enlarged figure at smaller scale.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 6 3453
the gas pressure drop in two-phase flow is greater than that in
single-phase flow. This increased pressure drop due to liquid
water buildup is a key cause of flow maldistribution,
dramatically reducing the PEMFC performance and durability.
Due to the complexity of two-phase flow, it is not an easy task
to form an analytical model for the two-phase pressure drop.
However, Wang and Wang [28] have demonstrated that the
ratio of two-phase flow pressure drop and the single gas phase
pressure drop can be used as a simple indicator of the liquid
buildup in PEMFC channels, with a higher ratio corresponding
to a higher water buildup in the channels. This ratio, known as
the two-phase friction multiplier [40], is defined by Eq. (3):
F2 g ¼
Dpg (3)
where DP2ø and DPg are the pressure drop with two-phase flow
and with only single-phase gas flow in the channel, respec-
tively. Fig. 13 shows the two-phase friction multiplier as
a function of water injection rate and airflow rate. DP2ø and
DPg are taken as the average pressure drop over the tested
period (about 10 min). It can be seen that at low air velocities
(or air stoichiometric ratios) the two-phase friction multiplier
is greater than unity due to liquid water buildup. Slug flow
shows the highest Fg 2, indicating the greatest water buildup.
As the air velocity increases, Fg 2, decreases and approaches
unity. The air velocities for Fg 2¼ 1 correspond to the mist flow
range, indicating that there is a little pressure drop difference
between the mist flow and the single airflow. The superficial
water velocity has a twofold effect on the pressure drop.
Firstly, the two-phase friction multiplier increases with
increasing superficial water velocity, implying higher water
buildup in channels at higher water injection rates. Secondly,
higher water velocity shifts the transition point to mist flow
toward a higher air velocity.
Pressure drop fluctuations can also provide important
information about the flow system. They arise from many
factors such as flow rate fluctuations, water holdup fluctua-
tions, density fluctuations, etc. In the present work, the
1 10
UL = 3.0×10-4 m/s
UL = 7.4×10-4 m/s
UL = 14.9×10-4 m/s
velocities.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 63454
percentage deviation is used to measure pressure fluctua-
tions. The percent deviation of the pressure drop fluctuation
decreases dramatically with increasing air velocity, from
a deviation above 3% for the slug flow to a deviation below
0.5% for the mist flow. In addition to the magnitude of the
pressure drop fluctuation, the pattern (or frequency) of the
pressure drop fluctuation (see Fig. 12 for example) is also
characteristic to the flow distribution pattern. Traditionally,
this information can be analyzed by using the Discrete Fourier
Transform (DFT) [41]. Following Eq. (4), the normalized pres-
sure drop (i.e. the percent deviation), xi(t), is transformed into
the Fourier domain, F{xi(t)}, where n is the number of samples
and xi is the percent fluctuation. At each point x, F(k) has both
real and imaginary parts, Re and Im, shown in Eq. (5). The
magnitude of F(k) is then found using Eq. (6).
FfxiðtÞg ¼ Xn1
FðkÞ ¼ Reþ jIm (5)
M ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Re2 þ Im2
p (6)
It is thus possible to calculate the power spectrum, that is,
the power of each frequency by squaring its magnitude. It
is found that most of the spectral signal is below 0.5 Hz. This is
understandable because the slug and film flow are in fact slow
processes, as shown in Fig. 12(a) and (b). The summation of
the power spectrum overall frequencies indicates the amount
of slug and film flow events. Fig. 14 shows the calculated total
powers at different airflow rates. A decreasing logarithmic
trend emerges for increasing airflow rate. Clear DFT power
boundaries have been identified to match the transition
between the three flow patterns identified from the high-
speed imaging. A DFT power of 5.0 105 corresponds to the
transition from the slug flow to the film flow and a power of
1.0 104 to the transition from the film flow to the mist flow.
DFT power analysis of the pressure drop thus provides
0 6 12 18 24 30 1x102
1x103
1x104
1x105
1x106
1x107
1x108
1x104
divisions mark transition between flow patterns. The
symbol (x) represents slug flow, (,) for film flow and (O)
for mist flow.
a useful tool to quickly identify dominant flow patterns where
video and individual channel airflow rate measurements are
not available. It should be noted that these DFT power
boundaries are specific to the system studied in this work.
3.3. Flow pattern map
Channel airflow rate behavior coupled with high-speed visu-
alization has been used to identify three major forms of water
transport, namely slug flow, annular/film flow and mist flow.
Fig. 15 shows the two-phase flow pattern map, in terms of
superficial water velocity and logarithmic superficial air
velocity, which illustrates the operating conditions at which
these patterns emerge. It can be seen that slug flow is the
primary form of water transport at lower air velocities (flow
rates). As the velocity increases, the flow pattern shifts from
slug flow to film flow in the intermediate air velocity range.
Finally, as the air velocity continues to rise, the flow pattern
shifts to mist flow. The transitions are a function of the water
velocity as well as the air velocity. As water injection rate
(velocity) increases, the transitions take place at higher air
velocities which indicate an increase in the level of water
buildup. It should be noted that the flow pattern boundaries
are an approximation of a transition region over which both
flow patterns may exist at different times. Using this map, the
primary method of water transport can be determined at any
combination of water and airflow rates within the test region.
More work on the two-phase flow patterns and the effect of
GDL structure, thickness and hydrophobicity is reported
elsewhere [42].
Mist flow is an efficient mode of liquid water removal from
the gas channels; however, it requires higher airflow rates and
consequently higher parasitic pumping power. At the other
end of the spectrum, slug flow, which is the most common
flow pattern at normal fuel cell operating condition, readily
leads to severe performance degradation despite the fact that
it requires the lowest pressure drop. Slug flow is therefore
1 10 0
Fig. 15 – Flow regime map of two-phase flow patterns as
a function of superficial air velocity (UG) and superficial
water velocity (UL). Division lines are guidance for your
eyes.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 6 3455
undesirable for the fuel cell operation and should be avoided.
Film flow becomes the most preferred method of water
transport in a PEMFC because of the relatively high water
removal capacity and the fact that it does not require a very
high pressure drop. Film formation and removal are strongly
dependent on the surface properties of the channel and GDL.
These properties should be optimized to reduce the instability
of the film flow (especially the instability of the film thickness)
and this work is currently underway.
4. Conclusion
An ex situ experimental setup is designed to investigate the
two-phase water transport in PEMFC parallel cathode gas
channels. The two-phase flow dynamics are studied in terms
of the instantaneous airflow rate in each channel, the total
pressure drop across the entire flow field, and high-speed
video observations of the water flow structure, all of which are
measured simultaneously.
At low airflow rates, corresponding to air stoichiometries
employed in the normal fuel cell operation, i.e. in the range of
1–5, the two-phase flow is dominated by the slug flow. The
individual channel flow rate measurement indicates that
a water slug seldom completely blocks a channel. The slugs or
semi-slugs are randomly formed in the channels, leading to
severe flow maldistribution. The slugs or semi-slugs reside in
the channels from several seconds to several hundreds of
seconds, decreasing with increasing airflow rate. This slug
residence time is much greater than the time scale for the
reactant gas to diffuse to the catalyst layer. Therefore, the slug
flow will inevitably result in PEMFC performance degradation
and possibly durability problems. Large fluctuations, in the
order of 5% of the average pressure drop, are observed in the
pressure drop for the slug flow. The DFT power analysis shows
an extremely high power, above 5.0 105.
At higher airflow rates, the two-phase flow is dominated by
annular/film flow for the hydrophilic channel walls studied in
this work. A stable, continuous film along the channel corners
is not observed because the contact angle of the hydrophilic
channel does not satisfy the Concus–Finn condition. Instead,
liquid water collects on the channel walls in the form of large
droplets, which are then removed in a film-like structure
along the sidewalls by the drag forces of the core airflow and
gravity. Since the channel is not significantly blocked by the
water film, severe flow maldistribution is not found under the
film flow pattern.
At extremely high airflow rates, little water buildup is
observed and mist flow is obtained. In mist flow, no variation
in channel flow rate and pressure drop can be detected, which
indicates that tiny water droplets move at the similar velocity
with the airflow. Water evaporation may also contribute
significantly to the water removal since the air velocity is very
high and dry air is used in this study.
It is found that water buildup in the gas channels and at the
channel-exit manifold interface can be quantitatively
described by a two-phase pressure drop factor, which is
defined as the ratio of the two-phase pressure drop to the
single-phase pressure drop. This factor decreases with
increasing airflow rates and decreasing water flow rates. The
effect of the water and airflow rates on the two-phase flow has
been presented in a flow pattern map.
Acknowledgments
and Fuel Cell Laboratory in the Mechanical Engineering
Department at Rochester Institute of Technology and is sup-
ported by the US Department of Energy under contract No. DE-
FG36-07G017018.
r e f e r e n c e s
[1] Wang C-Y. Fundamental models for fuel cell engineering. Chem Rev 2004;104:4727–66.
[2] Kandlikar SG, Lu Z. Fundamental research needs in combined water and thermal management within a proton exchange membrane fuel cell stack under normal and cold- start conditions. J Fuel Cell Sci Technol 2009, in press.
[3] Litster S, Sinton D, Djilali N. Ex-situ visualization of liquid water transport in PEM fuel cell gas diffusion layers. J Power Sources 2006;154:95–105.
[4] Chen F, Chang MH, Hsieh. Two-phase transport in the cathode gas diffusion layer of PEM fuel cell with a gradient in porosity. Int J Hydrogen Energy 2008;33:2525–9.
[5] Sinha PK, Wang C-Y. Liquid water transport in a mixed-wet gas diffusion layer of a polymer electrolyte fuel cell. Chem Eng Sci 2008;63:1081–91.
[6] Sinha PK, Wang C-Y. Pore-network modeling of liquid water transport in gas diffusion layer of a polymer electrolyte fuel cell. Electrochim Acta 2007;52:7936–45.
[7] Nam JH, Kaviany M. Effective diffusivity and water- saturation distribution in single- and two-layer PEMFC diffusion medium. Int J Heat Mass Transfer 2003;46:4595–611.
[8] Ge S, Wang C-Y. Characteristics of subzero startup and water/ice formation on the catalyst layer in a polymer electrolyte fuel cell. Electrochim Acta 2007;52:4825–35.
[9] Song D, Wang Q, Liu Z, Eikerling M, Xie Z, Navessin T, et al. A method for optimizing distributions of Nafion and Pt in cathode catalyst layers of PEM fuel cells. Electrochim Acta 2005;50:3347–58.
[10] Eikerling M. Water management in cathode catalyst layers of PEM fuel cells – a structure-based model. J Electrochem Soc 2006;153:E58–70.
[11] Tuber K, Pocza D, Hebling C. Visualization of water buildup in the cathode of a transparent PEM fuel cell. J Power Sources 2003;124:403–14.
[12] Yang XG, Zhang FY, Lubawy AL, Wang C-Y. Visualization of liquid water transport in a PEFC. Electrochem Solid State Lett 2004;7:A408–11.
[13] Zhang FY, Yang XG, Wang C-Y. Liquid water removal from a polymer electrolyte fuel cell. J Electrochem Soc 2006;153: A225–32.
[14] Weng FB, Su A, Hsu CY, Lee CY. Study of water-flooding behavior in cathode channel of a transparent proton-exchange membrane fuel cell. J Power Sources 2006;157:674–80.
[15] Liu X, Guo H, Ma C. Water flooding and two-phase flow in cathode channels of proton exchange membrane fuel cells. J Power Sources 2006;156:267–80.
[16] Spernjak D, Prasad AK, Advani SG. Experimental investigation of liquid water formation and transport in a transparent single-serpentine PEM fuel cell. J Power Sources 2007;170:334–44.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 3 4 4 5 – 3 4 5 63456
[17] Liu X, Guo H, Ye F, Ma CF. Water flooding and pressure drop characteristics in flow channels of proton exchange membrane fuel cells. Electrochim Acta 2007;52:3607–14.
[18] Ge S, Wang C-Y. Liquid water formation and transport in the PEFC anode. J Electrochem Soc 2007;154:B998–1005.
[19] Trabold TA, Owejan JP, Jacobson DL, Arif M. and Huffman PR. In situ investigation of water transport in an operating PEM fuel cell using neutron radiography: part I – experimental method and serpentine flow field results. Int J Heat Mass Transfer 2006;49:4712–20.
[20] Owejan JP, Trabold TA, Jacobson DL, Baker DR, Hussey DS, Arif M. In situ investigation of water transport in an operating PEM fuel cell using neutron radiography: part 2 – transient water accumulation in an interdigitated cathode flow field. Int J Heat Mass Transfer 2006;49:4721–31.
[21] Turhan A, Heller K, Brenizer JS, Mench MM. Quantification of liquid water accumulation and distribution in a polymer electrolyte fuel cell using neutron imaging. J Power Sources 2006;160:1195–203.
[22] Park J, Li X, Tran D, Abdel-Baset T, Hussey DS, Jacobson DL, et al. Neutron imaging investigation of liquid water distribution in and the performance of a PEM fuel cell. Int J Hydrogen Energy 2008;33:3373–84.
[23] Owejan JP, Trabold TA, Jacobson DL, Arif M, Kandlikar SG. Effects of flow field and diffusion layer properties on water accumulation in a PEM fuel cell. Int J Hydrogen Energy 2007; 32:4489–502.
[24] Chen YS, Peng H, Hussey DS, Jacobson DL, Tran DT, Abdel-Baset T, et al. Water distribution measurement for a PEMFC through neutron radiography. J Power Sources 2007;170:376–86.
[25] Quan P, Zhou B, Andrzej S, Liu Z. Water behavior in serpentine micro-channel for proton exchange membrane fuel cell cathode. J Power Sources 2005;152:131–45.
[26] Jiao K, Zhou B, Quan P. Liquid water transport in straight micro-parallel-channels with manifolds for PEM fuel cell cathode. J Power Sources 2006;157:226–43.
[27] Quan P, Lai M-C. Numerical study of water management in the air flow channel of a PEM fuel cell cathode. J Power Sources 2007;164:222–37.
[28] Wang Y, Basu S, Wang C-Y. Modeling two-phase flow in PEM fuel cell channels. J Power Sources 2008;179:603–17.
[29] Liu X, Guo H, Ye F, Ma CF. Flow dynamic characteristics in flow field of proton exchange membrane fuel cells. Int J Hydrogen Energy 2008;33:1040–51.
[30] He W, Lin G, Nguyen TV. Diagnostic tool to detect electrode flooding in proton-exchange-membrane fuel cells. AIChE J 2003;49:3221–8.
[31] Ma HP, Zhang HM, Hu J, Cai YH, Yi BL. Diagnostic tool to detect liquid water removal in the cathode channels of proton exchange membranefuel cells. J Power Sources 2006;162:469–73.
[32] Bosco AD, Fronk MH. Fuel cell flooding detection and correction. U.S. patent 6103409; 2000.
[33] Barbir F, Gorgun H, Wang X. Relationship between pressure drop and cell resistance as a diagnostic tool for PEM fuel cells. J Power Sources 2005;141:96–101.
[34] Kandlikar SG, Lu Z, Domigan WE, White AD, Benedict MW. Measurement of flow maldistribution in parallel channels and its application to ex-situ and in-situ experiments in PEMFC water management studies. Int J Heat Mass Transfer doi:10.1016/j.ijheatmasstransfer.2008.09.025.
[35] Owejan JP, Gagliardo JJ, Sergi JM, Kandlikar SG, Trabold TA. Water management studies in PEM fuel cells: part I – fuel cell design and in-situ water distributions. Int J Heat Mass Transfer, submitted for publication.
[36] Grolman E, Fortuin JMH. Liquid hold-up, pressure gradient, and flow patterns in inclined gas-liquid pipe flow. Exp Therm Fluid Sci 1997;15:174–82.
[37] Kandlikar SG, Cooke D, Lin TY, Lu Z. Evaluation of GDL intrusion in gas channel arrays of PEM fuel cells. J Power Sources, submitted for review.
[38] Wang Y, Wang CY. Transient analysis of polymer electrolyte fuel cells. Electrochim Acta 2005;50:1307–15.
[39] Concus P, Finn R. On the behavior of a capillary surface in a wedge. Appl Math Sci 1969;63:292–9.
[40] Lockhart RW, Martinelli RC. Proposed correlation of data for isothermal two-phase, two-component flow in pipes. Chem Eng Prog 1949;45:39–48.
[41] Walker JS. Fast fourier transforms. 2nd ed. New York: CRC Press; 1996.
[42] Lu Z, Grimm M, Rath C, Kandlikar SG. Water management studies in PEM fuel cells, part III – effects of GDL hydrophobicity, thickness, and structure on two-phase flow behavior in PEMFC gas channels, in preparation.
Introduction
Experimental
Pressure drop