Applied Thermal Engineering 22 (2002) 1347–1367

[Link]/locate/apthermeng

Performance comparisons of desiccant wheels for

air dehumidification and enthalpy recovery
L.Z. Zhang *, J.L. Niu
Department of Building Services Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong, China
Received 3 October 2001; received in revised form 13 March 2002; accepted 14 March 2002

Abstract
Desiccant wheels have two major applications: air dehumidification and enthalpy recovery. Since the
operating conditions are different, heat and mass transfer behaviors in the wheels are quite different. In this
paper, the performances of desiccant wheels used in air dehumidification and enthalpy recovery are
compared with each other. To accomplish this task, a two-dimensional, dual-diffusion transient heat and
mass transfer model which takes into account the heat conduction, the surface and gaseous diffusion in
both the axial and the thickness directions is presented. Effects of the rotary speed, the number of transfer
units, and the specific area on the performance of the wheel are investigated and compared in the two
situations. The cycles that the desiccant and air undergo in the wheel are plotted in psychrometric charts to
demonstrate the different heat and moisture transfer mechanisms during the dehumidification and enthalpy
recovery processes. Ó 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Desiccant wheel; Heat and mass transfer; Model; Dehumidification; Enthalpy recovery

1. Introduction

Normally the water vapor content of atmospheric air is small, some tens of grams per kilo of
air. Nevertheless, due to the very high heat of vaporization, the latent heat content in air-con-
ditioning is of the same order of the sensible one. The relative importance of latent load increases
with the ventilation rates to buildings. Higher ventilation rates are dictated both by better comfort
requirement and by the ASHRAE standard 62/1999.

*
Corresponding author. Tel.: +852-27665853; fax: +852-27746146.
E-mail address: belzhang@[Link] (L.Z. Zhang).

1359-4311/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 9 - 4 3 1 1 ( 0 2 ) 0 0 0 5 0 - 9
1348 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

Nomenclature

a pore radius (m)

As transfer area (m2 )
Bi Biot number
C constant in sorption curve
cp specific heat (kJ kg
1 K
1 )
de hydrodynamic diameter of a channel (m)
DA combined ordinary and Knudson diffusivity (m2 s
1 )
DS surface diffusivity (m2 s
1 )
f fraction of desiccant in the wheel material
h convective heat transfer coefficient (kW m
2 K
1 )
hm convective mass transfer coefficient (kg m
2 s
1 )
H specific enthalpy (kJ kg
1 )
k thermal conductivity (kW m
1 K
1 )
L length of a channel (m)
Le Lewis number
M1 molecular weight (kg kmol
1 )
md mass of the wheel (kg)
m_ g mass flow rate of air stream (kg s
1 )
N rotary speed (rpm)
NTU number of transfer units
P pressure (Pa)
qst isosteric adsorption heat (kJ kg
1 )
SDP specific dehumidification power (g kg
1 s
1 )
T temperature (K)
t time (s)
ug velocity of air stream (m s
1 )
V volume (m3 )
w water uptake in desiccant (kg water/kg dry desiccant)
wmax maximum water uptake of desiccant (kg kg
1 )
x, y coordinates (m)
Greek letters
a angle (rad)
b; k coefficients in Eq. (35)
h dimensionless temperature
e effectiveness
et total porosity
/ relative humidity
d half thickness of channel (m)
x moisture content (kg moisture/kg dry air)
q density (kg m
3 )
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1349

f tortuosity factor
s dimensionless time
n resistance coefficient
Superscripts
* dimensionless form of the variable
Subscripts
a air
c cooling, process stream for dehumidification and exhaust for enthalpy recovery
d desiccant, dehumidification
g gas
h heating, regeneration for dehumidification and process for enthalpy recovery
i inlet
L latent
m moisture
o outlet
opt optimum
s surface, sensible
w water
x in x-direction
y in y-direction

Desiccant wheels have been widely used for air humidity treatment: dehumidification [1–4] and
enthalpy recovery [5–7]. In the first case, process air is dried after it flows through the wheel, which
rotates continuously between the process air and a hot regenerative air stream. The dried air can
either be used directly or be employed to make cooling following further psychrometric processes,
i.e. the so-called desiccant cooling. In the latter case, the desiccant wheel rotates between the
outside fresh air (process air) and the exhaust air from room. Heat and humidity would be re-
covered from the exhaust in winter and excess heat and moisture would be transferred to the
exhaust to cool and dehumidify the process air in the summer. However, due to different operating
conditions, heat and moisture transfer behaves quite differently in the wheels. Regretfully, the
performances for air dehumidification and enthalpy recovery have never been systematically
compared.
There have been many works in modeling heat and moisture transfer in desiccant wheels. Only
considering resistance at the gas–solid interface while neglecting heat conduction and moisture
diffusion in the solid, Jurinak and Mitchell [8] and Zheng and Worek [9] used one-dimensional
transient models for heat and moisture couplings between the solid desiccant and the air streams
in the wheel. The effect of matrix sorption properties was studied in the former case and a full-
implicit finite difference technique was employed to obtain the solution in the latter one. As a step
forward, San and Hsiau [10] and Simonson and Besant [11] adopted the one-dimensional solid
heat conduction equation in their respective models to account for the longitudinal thermal
resistance inside the solid. These two studies gave some insight into the heat mass transfer
1350 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

mechanisms in the desiccant. However, the model of Simonson and Besant neglected the internal
moisture resistance in the solid, and that of San and Hsiau considered only the axial surface
diffusion resistance by assuming axial surface diffusion as the dominant mass transfer mechanism
in the adsorbent.
The model presented here is a dual-diffusion model that takes into account both the heat and
the moisture resistance in two dimensions: axial and in-thickness directions of the solid. Moisture
transfer is expressed in two forms: surface diffusion and gaseous diffusion (Knudsen and ordinary
combined).

2. Mathematical model

2.1. Basic equations

A rotary regenerative desiccant wheel is shown in Fig. 1(a). It is a rotating cylindrical wheel of
length L and diameter dw and it is divided into two sections: adsorption section (angle fraction a0 )
and regeneration section (fraction 1
a0 ), where the adsorption and regeneration air streams are
in a counter-flow arrangement. The wheel generally consists of a matrix of numerous flow
channels which have, depending on the manufacturing process, a rectangular, triangular or si-
nusoidal shape. The channel walls are parallel to the axis of the wheel, and are made of composite
materials which have a desiccant content of f ¼ 0:7–0:8.

Fig. 1. Schematic of the desiccant wheel showing (a) the entire wheel, (b) a side view of one of the ducts.
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1351

For simplicity, some assumptions are made:

(1) One-dimensional air stream flow is considered.

(2) The axial heat conduction and mass diffusion in the fluid are neglected.
(3) No carryover of the fluid in the desiccant wheel is assumed.
(4) The thermodynamic properties in the solid are constants, and uniform.

Consequently, the model used in this study is transient and two-dimensional. Because of
symmetry, the mid-plane of a channel can be considered to be adiabatic, and a half-size channel
surrounded by dashed line is used as the physical model, as shown in Fig. 1(b).
Heat and mass conservation for the air stream
1 oTg oTg 4h

þ ¼ Ts
Tg ð1Þ
ug ot ox de ug qg cpg

1 oxg oxg 4hm


þ ¼ xs
xg ð2Þ
ug ot ox d e ug qg
where ug is the velocity (m s
1 ), Tg and xg are temperature (°C) and humidity ratio (kg kg
1 ) re-
spectively, t is time (s), x is axial coordinate (m), de is the hydrodynamic diameter of the channel (m),
qg is the density (kg m
3 ) and cpg is the specific heat (kJ kg
1 K
1 ), h and hm are the convective heat
transfer (kW m
2 K
1 ) and mass transfer (kg m
2 s
1 ) coefficients between the air stream and the
solid surface, respectively. In the equations, subscripts ‘‘s’’ and ‘‘g’’ refer to ‘‘surface’’ and ‘‘gas’’
respectively. By using heat mass transfer analogy, the relations between h and hm can be expressed as
h
hm ¼ ð3Þ
cpg Le
where Le is the Lewis number of air stream.
Boundary conditions
For adsorption section, i.e., 0 6 a < a0 :
( 
Tg x¼0 ¼ Tci
ð4Þ
xg x¼0 ¼ xci

For desorption section, i.e., a0 6 a < 1:

( 
Tg x¼L ¼ Thi
ð5Þ
xg x¼L ¼ xhi

where subscripts ‘‘c’’, ‘‘h’’, ‘‘i’’ and ‘‘o’’ refer to process air, regenerative air, inlet and outlet,
respectively. The dimensionless angle a ¼ a=2p.
Heat conservation in the desiccantz
 2 
oTd o Td o2 Td ow
qd ctot ¼ kd þ þ qst qd ð6Þ
ot ox2 oy 2 ot
1352 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

where qd is the density of dry desiccant (kg m
3 ), kd is the thermal conductivity of the solid
(kW m
1 K
1 ), w is the water content in the desiccant (kg kg
1 ), qst is the adsorption heat
(kJ kg
1 ), ctot is the total heat capacity of wet desiccant, which includes two parts: dry desiccant
and adsorbed water and is calculated by
ctot ¼ cpd þ wcpw ð7Þ

1
1
where cpd and cpw are the specific heats (kJ kg K ) of dry desiccant and liquid water respectively.
Moisture conservation
Two phases of water, namely, gas and adsorbed state, co-exist and diffuse in the pores of the
solid. There are three dominant diffusion mechanisms [12]: surface diffusion, ordinary diffusion,
and Knudsen diffusion. The first diffusion is in the form of adsorbed state and the latter two are in
gas state. If Fick’s law is used to express the diffusion dynamics, then the moisture conservation in
the solid can be expressed as
         
ox ow o ox o ox o ow o ow
et qa þ qd ¼ qa DA þ DA þ qd DS þ DS
ot ot ox ox oy oy ox ox oy oy
ð8Þ

where et is the total porosity of the desiccant. On the right hand side of Eq. (8), the first term is the
moisture transfer in gas (combined ordinary and Knudsen diffusion), and the second term is in the
adsorbed phase, namely, surface diffusion. DA and DS are the effective diffusivities (m2 s
1 ) of
the combined ordinary and Knudsen diffusion and surface diffusion, respectively. They are cal-
culated by the following equations [12,13].
 
1
et 1 1
DA ¼ þ ð9Þ
f DAO DAK

1

DS ¼ D0 exp
0:974  10
3 qst =Td ð10Þ
f

Td1:685
DAO ¼ 1:758  10
4 ð11Þ
Pa
 1=2
Td
DAK ¼ 97a ð12Þ
M1
where f is tortuosity factor that accounts for the increase in diffusional length due to the tortuous
path of the real pores, DAO is the ordinary diffusivity, DAK is the Knudsen diffusivity, D0 is a
constant for surface diffusion calculation, a is the pore radius of the adsorbent, Pa is the pressure
in atmospheres (Pa), Td is in K, M1 is the molecule weight of water.
Boundary conditions

oTd 


kd  ¼ h Td
Tg ð13Þ
oy y¼d
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1353
 
ox ow 


qa DA
qd DS  ¼ hm x
xg ð14Þ
oy oy y¼d
  
oTd  oTd  oTd 
¼ ¼ ¼0 ð15Þ
oy y¼0 ox x¼0 ox x¼L
  
ox  ox  ox 
¼ ¼ ¼0 ð16Þ
oy y¼0 ox x¼0 ox x¼L

where d is the half thickness of the channel wall (m), and L is the length of the channel (m).

2.2. Normalization of equations

Water content in the desiccant is governed by a general sorption isotherm as

fwmax
w¼ ð17Þ
1
C þ C=/
where wmax is the maximum water content (kg kg
1 ), C is a constant that determines the isotherm
shape, / is the relative humidity. Selecting T and x as two independent variables, a differential
form of adsorption content can be written in terms of humidity and temperature as
dw ¼ w dx þ u dT ð18Þ
where
 
ow
w¼ ð19Þ
ox T
 
ow
u¼ ð20Þ
oT x

Using Clapeyron equation to represent the saturation vapor pressure and assuming a standard
atmospheric pressure of 101 325 Pa gives the relation between humidity and relative humidity [13]
as
/
¼ 10
6 e5294=T
1:61/ ð21Þ
x
where T is in K. The second term on the right side of the equation will generally have less than a
5% effect, thus it can be neglected. Substituting Eqs. (17) and (21) into Eqs. (19) and (20), fol-
lowing equations can be obtained
fwmax C
w ¼ 10
6 e5294=T ð22Þ
ð1
C þ C=/Þ2 /2

5294/ fwmax C
u¼
ð23Þ
T 2
ð1
C þ C=/Þ2 /2
1354 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

where w reflects the slope of w to x, and it is a dimensionless variable. Partial differential u reflects
the slope of w to T, and its unit is K
1 . For enthalpy recovery, the bigger the w, the better the
performance; for air dehumidification, the greater the u, the better the performance.
Consequently,
ow ox oT
¼w þu ð24Þ
ot ot ot
ow ox oT
¼w þu ð25Þ
ox ox ox
ow ox oT
¼w þu ð26Þ
oy oy oy
Introducing dimensionless temperature
T
Tci
h¼ ð27Þ
Thi
Tci
and time
tN
s¼ ð28Þ
60
as well as coordinates
x
x ¼ ð29Þ
L
y
y ¼ ð30Þ
d
energy Eq. (6) can be normalized to
oh o2 h o2 h ox
¼ kx 2 þ ky 2 þ qst ð31Þ
os ox oy os
where
60kd
kx ¼ ð32Þ
NL2 q d ðctot
qst uÞ

60kd
ky ¼ 2
ð33Þ
N d qd ðctot
qst uÞ

qst qd w
qst ¼ ð34Þ
qd ðctot
qst uÞðTh
Tc Þ
where N is the rotary speed of the wheel in rpm.
The mass Eq. (8) can be normalized to
   
ox  oh o ox oh o ox oh
þ b0 ¼ bx  kx  þ kT  þ by  kx  þ kT  ð35Þ
os os ox ox ox oy oy oy
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1355

where

qd uðTh
Tc Þ
b0 ¼ ð36Þ
et qa þ wqd

kx ¼ qa DA þ qd DS w ð37Þ

kT ¼ qd DS uðTh
Tc Þ ð38Þ

60
bx ¼ ð39Þ
NL2 ðe t qa þ wqd Þ

60
by ¼ 2
ð40Þ
Nd ðet qa þ wqd Þ

Accordingly, the boundary conditions for the solid become

  
oh  oh  oh 
¼  ¼  ¼0 ð41Þ
oy  y ¼0 ox x ¼0 ox x ¼1
  
ox  ox  ox 
¼  ¼  ¼0 ð42Þ
oy  y ¼0 ox x ¼0 ox x ¼1

oh 


 ¼ Bi h
hg ð43Þ
oy y  ¼1
 
ox  kT oh 



  ¼ Bim x
xg ð44Þ
oy y  ¼1 kx oy y  ¼1

where Bi is the Biot number for heat transfer

hd
Bi ¼ ð45Þ
kd
and Bim is the Biot number for mass transfer
hm d
Bim ¼ ð46Þ
kx
The heat and mass transfer equations for the air streams can be normalized as
ohg ohg

c1 þ  ¼ NTU hs
hg ð47Þ
os ox
oxg oxg

c1 þ  ¼ NTUm xs
xg ð48Þ
os ox
1356 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

where
NL
c1 ¼ ð49Þ
60ug

hAs
NTU ¼ ð50Þ
m_ g cpg

hm As
NTUm ¼ ð51Þ
m_ g
where NTU is the commonly recognized number of transfer units, As is the total contact area
between the air and the solid, m_ g is the mass flow rate of air stream (kg s
1 ).

2.3. Performance index

If the wheel is used for enthalpy recovery, two effectiveness are defined.
Sensible effectiveness, es
m_ c ðTci
Tco Þ
es ¼ ð52Þ
m_ min ðTci
Thi Þ
Latent effectiveness, eL
m_ c ðxci
xco Þ
eL ¼ ð53Þ
m_ min ðxci
xhi Þ
where m_ min is the least value of adsorption and regenerative mass flows.
Dehumidification effectiveness, ed
xci
xco
ed ¼ ð54Þ
xci
Specific dehumidification power on the basis of unit mass of desiccant (kg kg
1 s
1 )
m_ c ðxci
xco Þ
SDP ¼ ð55Þ
md
Specific area of the wheel
As
Av ¼ ð56Þ
Vw
where Vw is the volume of the wheel. The higher the specific area, the more compact the wheel is.
The pressure drop for the wheel can be written as
 
L qa u2g
DP ¼ 4nc þ nL ð57Þ
de 2
where nc is the friction factor of flow channels, and nL is a local friction coefficient that reflects the
entrance and exit losses.
For fully developed laminar flow in sinusoidal channels [14],
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1357

11:173
nc ¼ ð58Þ
Re
Eqs. (31), (35), (47) and (48) are the governing equations for heat and mass transfer in the
wheel, which are strongly coupled with each other.

3. Results and discussion

The two-dimensional heat and mass transfer equations of the desiccant are numerically solved
by means of alternating direction implicit (ADI) method [15]. Because the equations are strongly
coupled and non-linear, iterations are necessary to get converged values for each time step. Before
numerical analysis can be performed, the physical domain of the problem as well as the equations
must be discretized. The whole calculating domain is divided into a number of equal-step discrete
elements. Each element is identified as a control volume by a nodal point. The numbers of nodes
are: 40 in axial, 5 in thickness, and 120 in time (angle). Two possible applications are studied. For
enthalpy recovery, the inlet temperature and humidity are: process air, 35 °C, 0.025 kg kg
1 ; re-
generative air (exhaust), 24 °C, 0.013 kg kg
1 . For air dehumidification, inlet properties are:
process air, 30 °C, 0.021 kg kg
1 ; regenerative air, 90 °C, 0.021 kg kg
1 . The base properties for the
simulated desiccant wheel are listed in Table 1. The desiccant material is silica gel. The geometry
of the channels in the wheel is sinusoidal with a width of 4.35 mm and a height of 1.74 mm.

3.1. Performance with rotary speeds

There exists an optimum rotary speed at which the efficiency reaches the climax. When a
desiccant wheel rotates much faster than the optimum speed, the adsorption and regeneration

Table 1
Configuration of the desiccant wheel
Symbol Unit Value
a m 11  10
10
C 1.0
D0 m2 s
1 1:6  10
6
f 0.75
k W m
1 K
1 0.20
L m 0.1
md kg 15.0
m_ g kg s
1 0.4
qst kJ kg
1 2650
wmax kg kg
1 0.25
a0 0.50
qd kg m
3 1129
n 2.8
d mm 0.1
et 0.70
1358 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

processes are too short, which results in poor performance. On the other hand, when the rotary
speed is lower than the optimum, the adsorption and regenerative processes are too long and
wasting more energy in sensible heating/cooling rather than in sorption processes, and therefore
are less effective. The variations of performance with rotary speed are plotted in Figs. 2 and 3 for
dehumidification and enthalpy recovery purposes, respectively. The wall thickness is 0.2 mm.
These figures indicate that the optimum rotary speed for air dehumidification is much lower than
those for sensible and latent heat recovery. Air dehumidification is more sensitive to rotary speed
than enthalpy recovery is. This is because temperature variations of the wheel during dehumid-
ification processes are larger than during enthalpy recovery processes. Wheels need more time to
fluctuate between the necessary temperature amplitude in the first case than in the second.
In the following analysis, the wheels are operated at the optimum speeds.

Fig. 2. Performance variations under various rotary speeds with a wall thickness 0.2 mm, for air dehumidification.

Fig. 3. Performance variations under various rotary speeds with a wall thickness 0.2 mm, for enthalpy recovery.
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1359

3.2. Effects of number of transfer units

3.2.1. (a) Various wall thickness

The number of transfer units has a great influence on the system performance. There are many
ways to modify NTU. When the wheel mass and channel size are fixed, changing the wall
thickness will change the channel numbers, and consequently the contact area and NTU. When
the wheel volume and wall thickness are fixed, changing channel sizes will change the number of
channels and the packing density, consequently the NTU. Fig. 4 shows the variations of dehu-
midification effectiveness and specific dehumidification power with increasing NTU through
changing wall thickness. When the NTU is in 0–2.5, both the ed and the SDP increase rapidly with
an increasing NTU, however, when the NTU is bigger than 2.5, a further increase of NTU, i.e., a
larger transfer area or a lower mass flow rate, will have little merit in further increasing the
performance. These trends are similar to those for a counter-flow sensible heat exchanger. Fig. 5
shows the influence of NTU on enthalpy recovery efficiencies. The NTU of 2.5 is needed for

Fig. 4. Effects of NTU on dehumidification performance.

Fig. 5. Effects of NTU on performance for enthalpy recovery.

1360 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

sensible and latent efficiencies higher than 0.6. It is noted that the latent effectiveness is usually
smaller than the sensible effectiveness, because the moisture transfer resistance is usually larger
than the heat transfer resistance. In other words, mass diffusion in the solid is far less than the
thermal diffusion.

3.2.2. (b) Various Av

For desiccant wheels, when the total volume of the wheel is fixed, higher specific area can be
achieved by constructing with smaller size channels and thinner channel walls. With higher Av the
wheel will be more compact and both the transfer area and the heat mass transfer coefficients will
become larger. These factors all contribute to an increased performance, as shown in Fig. 6 for
dehumidification and Fig. 7 for enthalpy recovery. As can be seen, the effectiveness rises almost
linearly with increasing specific area. Therefore, in practice, the honeycomb-type desiccant wheels
which have large contact areas should be recommended. However, increased performance is
achieved at the price of increased pressure drop, as shown in Fig. 8. The pressure drop rises
rapidly with Av . In other words, fan power requirement will be increased for honeycomb wheels to

Fig. 6. Effects of specific area on dehumidification performance.

Fig. 7. Effects of specific area on performance for enthalpy recovery.

 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1361

Fig. 8. Pressure drop under various specific area, m_ g ¼ 0:4 kg s
1 , md ¼ 15 kg.

realize higher performance. Eventually, a compromise between increased Av and increased pres-
sure drop has to be maintained. It is obvious that NTU increases with Av .

3.3. Temperature and humidity profiles in the wheel

Temperature and humidity fields in the wheel can be seen from Figs. 9 and 10. Variations of the
temperatures (a) and humidities (b) of air inlet and the channel mean are plotted against the wheel
angle for dehumidification in Fig. 9 and for enthalpy recovery in Fig. 10 respectively. The values
at two local positions in the channel wall (one is near cool air inlet, the other is near warm air
inlet) are also plotted in the figures. The coordinates for the two points are: point 1, x ¼ 0:01,
y  ¼ 0:5; point 2, x ¼ 0:99, y  ¼ 0:5.
From Fig. 9, it can be seen that in dehumidification, desiccant near the warm air inlet has a
larger temperature wave than that near the cool air inlet, but the amplitudes of humidity waves
are almost the same at these two positions. The mean temperatures at positions near the warm air
inlet are higher than those neat the cool air inlet. However, the mean humidities near the warm air
inlet are lower than those near the cool air inlet. These facts reveal that the higher the temper-
ature, the lower the humidity ratio for desiccant. During the heating period, the mean humidity of
desiccant is higher than that of the air inlet, causing moisture to be desorbed from desiccant to air;
while during the cooling period, the mean humidity of desiccant is lower than that of air inlet,
causing moisture to be adsorbed from the air to the desiccant.
From Fig. 10, it is seen that during cooling period, humidities in desiccant (both local and
mean) are higher than those in air stream and moisture desorbs from the desiccant to the cooling
air, while during the heating period, humidities in desiccant are lower than those in air stream and
moisture adsorbs from the air to the desiccant. These phenomena in enthalpy recovery are just
opposite to those in air dehumidification. Furthermore, the amplitudes of temperature and hu-
midity waves at different locations are not in much disparity, which discloses a fact that tem-
perature and humidity fields in the wheel are more homogeneous in enthalpy recovery than those
in dehumidification.
1362 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

Fig. 9. Variations of the temperature (a) and humidity (b) of two local points, wheel mean and air inlet across the wheel
angle for dehumidification. Point 1: near inlet; point 2: near outlet.

3.4. Cycles in psychrometric charts

The cycle the desiccant undergoes during a revolving is certainly of interest. The evolutions of
the states of desiccant mean and air outlet in relation to wheel angle are plotted in psychrometric
charts, as shown in Fig. 11 for dehumidification and Fig. 12 for enthalpy recovery. In the two
figures, the dashed lines are constant enthalpy lines and the inlet states of air streams are rep-
resented by points Hi for warm air and Ci for cool air. The arrows indicate the directions of angle
increasing or wheel revolving. Points Ho and Co represent the two average outlet air states for
heating stream and cooling stream respectively. For the desiccant, three properties, namely, water
content, air humidity in the pores, and temperature, are plotted in the graph simultaneously.
Let us first see the cycles in air dehumidification. Fig. 11(a) shows the variations of desiccant
mean humidity and water uptake with temperature in a cycle. Here the mean values of desiccant
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1363

Fig. 10. Variations of the temperature (a) and humidity (b) of two local points, wheel mean and air inlet across the
wheel angle for enthalpy recovery. Point 1: near inlet; point 2: near outlet.

are calculated across a channel. Fig. 11(b) shows evolutions of the humidity and temperature of
the outlet air stream. The whole cycle for the desiccant at a fixed position comprises two processes:
Process abc: This is the dehumidification process for the cool air whose inlet state is point Ci.
The outlet states of the air stream across wheel angle lie on the line abc whose average is denoted
by Co. Therefore the average process for cool air is the thick line Ci ! Co. The two figures show
that the desiccant temperature drops continuously and the water uptake rises steadily as the wheel
revolves. However, the humidity variations are more complicated. From a to b, the humidities in
both the desiccant and the outlet air drop quickly to their lowest values. This period is very short
(1/5 of abc) and the water content in desiccant changes very little, so it is called isosteric cooling
period. However, desiccant temperature decreases substantially in this period, which causes hu-
midities to drop sharply based on sorption isotherms. The heat transferred in this phase is mainly
1364 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

Fig. 11. Variations of the states of desiccant mean (a) and outlet air (b) during wheel revolving for dehumidification:
abc, during dehumidification; cda or c0 da0 , during heating, dashed lines: constant enthalpy.

sensible heat due to the fact that little moisture is adsorbed. From b to c, the humidities in the
desiccant and the outlet air go up slowly from the lowest values. In this period, a large amount of
moisture is adsorbed by desiccant and therefore it is called adsorption period. Heat transferred in
this phase is mainly sorption heat. Since the desiccant goes toward saturation with increasing
quantities of moisture being adsorbed, humidities in desiccant and air stream rise continuously.
Process cda (c0 da0 for air): This is the re-activation period for the desiccant with a warm air
stream whose inlet state is point Hi and whose mean outlet state is Ho. The outlet states of air
distributed along line c0 da0 , depending on the wheel angle. From c to d (c0 to d for air), this is the
isosteric heating phase during which desiccant temperature rises swiftly with water uptake slightly
changed. The humidities in the desiccant go up quickly for the preparation of moisture desorp-
tion. From d to a (d to a0 for air), this is the desorption phase during which a large quantities of
moisture desorbs from desiccant to air stream. Since the desiccant has less and less water, the
humidities decrease with wheel revolving. As in the isosteric cooling process, heat transferred in
the isosteric heating phase is mainly sensible heat, while in the desorption phase heat is mainly
sorption heat.
During adsorption process for air Ci ! Co, enthalpy increases and during re-activation process
Hi ! Ho, enthalpy decreases for air stream. However, from the point of view of wheel angle,
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1365

Fig. 12. Variations of the states of desiccant mean (a) and outlet air (b) during wheel revolving for enthalpy recovery:
ac, during cooling; ca or c0 a0 , during heating; dashed lines: constant enthalpy.

during process abc, enthalpy of outlet air decreases, and during process c0 da0 , enthalpy of outlet
air increases with increasing angle. Furthermore, most of the enthalpy decrease/increase occurs in
ab and c0 d sections. Processes bc and da0 are almost parallel to constant enthalpy lines. This proves
that ideal adsorption and desorption processes are constant enthalpy processes. It is the isosteric
heating and cooling processes that makes the real processes in desiccant wheels non-constant
enthalpy. In other words, sensible heat leads to the deviation of adsorption/desorption processes
from constant enthalpy lines. It is also observed that outlet air states move toward constant
enthalpy lines that go through air inlet states with wheel revolving (angle increasing), i.e., pro-
cesses will become more and more constant enthalpy with wheel revolving.
Fig. 12 shows the variations of the states of desiccant mean and outlet air with respect to wheel
revolving for enthalpy recovery. The desiccant mean states evolve along the same line (ac) during
the cooling process as during the heating process (ca), but with an opposite direction. The outlet
states of air distribute along line ac for cooling process whose inlet is Ci and mean outlet is Co,
and along line c0 a0 for the heating process, where the inlet is Hi and the mean outlet is Ho, re-
spectively. The two inlet states Hi and Ci are on the same line of ac or c0 a0 . During heating, both
the temperature and the humidity increase, while during cooling, both the temperature and the
1366 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367

humidity decrease with wheel revolving. As a result, enthalpy increases for Ci ! Co (cooling air
inlet to outlet) and decreases for Hi ! Ho (heating air inlet to outlet). The variation of water
uptake in desiccant during a wheel working cycle is relatively very small. This is the reason why
wheels for enthalpy recovery applications should be rotated much faster than those for dehu-
midification to realize optimized performances.

4. Conclusions

The two-dimensional, dual-diffusion transient heat and mass transfer model presented in this
study is superior to other one-dimensional ones. The advantages of such a model lie in the fact
that it considers the heat conduction, the surface and gaseous diffusion in both the axial and the
thickness directions simultaneously. The effects of the channel wall thickness can be investigated.
Many other structural and operating parameters of the wheel could also be studied.
The results indicate that the optimum rotary speed for air dehumidification is much lower than
that for enthalpy recovery. In both cases, honeycomb-type wheels are recommended since a thick
channel wall is harmful to the heat mass transfers in the solid. An NTU of 2.5 is needed for
desiccant wheels to have a good performance. The wheel for enthalpy recovery has more ho-
mogeneous temperature and humidity fields than a wheel for dehumidification. In dehumidifi-
cation, desiccant near the warm air inlet has a larger temperature wave than that near the cool air
inlet, but the amplitudes of humidity waves are almost the same at these two positions. Cycles in
psychrometric charts show that for air dehumidification, there are distinct four phases, i.e.,
isosteric heating, desorption, isosteric cooling and adsorption. However, for enthalpy recovery,
the heating and the cooling processes can be represented by a single straight line, but in opposite
directions.

Acknowledgements

This research is funded by the postdoctoral fellowship of the Hong Kong Polytechnic Uni-
versity.

References

[1] A. Kodama, M. Goto, T. Hirose, T. Kuma, Experimental study of optimal operation for a honeycomb adsorber
operated with thermal swing, Journal of Chemical Engineering of Japan 26 (1993) 530–535.
[2] R. Tauscher, U. Dinglreiter, B. Durst, F. Mayinger, Transport processes in narrow channels with application to
rotary exchangers, Heat and Mass Transfer 35 (1999) 123–131.
[3] W. Zheng, W.M. Worek, Numerical simulation of combined heat and mass transfer processes in a rotary
dehumidifier, Numerical Heat Transfer, part A: Applications 23 (1993) 211–232.
[4] W. Zheng, W.M. Worek, D. Novolsel, Effect of operating conditions on optimal performance of rotary
dehumidifiers, ASME Journal of Energy Resources Technology 117 (1995) 62–66.
[5] J.Y. San, S.C. Hsiau, Effect of axial solid heat conduction and mass diffusion in a rotary heat and mass regenerator,
International Journal of Heat and Mass Transfer 36 (1993) 2051–2059.
 L.Z. Zhang, J.L. Niu / Applied Thermal Engineering 22 (2002) 1347–1367 1367

[6] C.J. Simonson, R.W. Besant, Heat and moisture transfer in energy wheels during sorption, condensation, and
frosting conditions, ASME Journal of Heat and Transfer 120 (1998) 699–708.
[7] C.J. Simonson, R.W. Besant, Energy wheel effectiveness: part I––development of dimensionless groups,
International Journal of Heat and Mass Transfer 42 (1999) 2161–2170.
[8] J.J. Jurinak, J.W. Mitchell, Effect of matrix properties on the performance of a counterflow rotary dehumidifier,
ASME Journal of Heat Transfer 106 (1984) 638–645.
[9] W. Zheng, W.M. Worek, Numerical simulation of combined heat and mass transfer processes in a rotary
dehumidifier, Numerical Heat Transfer, part A: Applications 23 (1993) 211–232.
[10] J.Y. San, S.C. Hsiau, Effect of axial solid heat conduction and mass diffusion in a rotary heat and mass regenerator,
International Journal of Heat and Mass Transfer 36 (1993) 2051–2059.
[11] C.J. Simonson, R.W. Besant, Heat and moisture transfer in energy wheels during sorption, condensation, and
frosting conditions, ASME Journal of Heat Transfer 120 (1998) 699–708.
[12] A.A. Pesaran, A.F. Mills, Moisture transport in silica gel packed beds. I––theoretical study, International Journal
of Heat and Mass Transfer 30 (1987) 1037–1049.
[13] P. Majumdar, Heat and mass transfer in composite desiccant pore structures for dehumidification, Solar Energy 62
(1998) 1–10.
[14] L.Z. Zhang, J.L. Niu, A Numerical study of laminar forced convection in sinusoidal ducts with arc lower
boundaries under uniform wall temperature, Numerical Heat Transfer, part A: Applications 40 (2001) 55–72.
[15] A.A. Samarskii, P.N. Vabishchevich, Computational Heat Transfer, John Wiley & Sons Inc., New York, 1995.