Applied Thermal Engineering 143 (2018) 988–1002

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Research Paper

A computational fluid dynamics model of a rotary regenerative heat T

exchanger in a flue gas desulfurization system
⁎
Koray Özdemira, , Mustafa Fazıl Serincanb
a
Tubitak Marmara Research Center Energy Institute, Gebze, Kocaeli, Turkey
b
Gebze Technical University, Mechanical Engineering Department, Gebze, Kocaeli, Turkey

H I GH L IG H T S

• ASophisticated
CFD model is presented for thermal-fluids analysis of a rotary heat exchanger.
• Operating conditions
geometric details of matrix are reduced to porous media parameters.
• and their effects are addressed by means of virtual tests.

A R T I C LE I N FO A B S T R A C T

Keywords: Rotary regenerative heat exchangers (RHEX) are commonly used in flue gas desulfurization (FGD) systems to
Rotary regenerative heat exchanger improve the dispersion of pollutants, reduce the visible plume, avoid liquid droplet rainout from the stack and
Flue gas desulfurization avoid corrosion problems on the system materials. In this study, transient behaviors of heating and cooling cycles
FGD of a rotary regenerative heat exchanger in an FGD system were investigated via a 3-D computational fluid
Computational fluid dynamics
dynamics (CFD) model. For this purpose, the rotary regenerative heat exchanger was modelled using porous
CFD
media approach for the heat transfer surfaces (i.e. matrix) inside the heat exchanger. A standalone channel
model as used to obtain porous media parameters and heat transfer coefficient. Numerical results confirmed by
published literature. Effects of different operating conditions on the performance of the heat exchanger have
been investigated. In conclusion treated gas outlet temperature increases with increasing angular velocity and
treated gas inlet temperature while it decreases with decreasing load. Consequently, untreated gas outlet tem-
perature decreases with increasing angular velocity and decreases with the decreasing treated gas inlet tem-
perature and load. By means of overall system performance it is observed that overall system performance
increases with decreasing angular velocity and treated gas inlet temperature while it increases with decreasing
load.

1. Introduction [1–13], semi-analytically [14] and numerically [15–38]. However heat

Rotary regenerative heat exchangers (RHEX) or simply regenerators
are widely used in many applications such as gas turbines, thermal
power plants, flue gas desulfurization systems and air conditioning
systems. RHEXs are commonly used in flue gas desulfurization (FGD)
systems to improve the dispersion of pollutants, reduce the visible
plume, avoid liquid droplet rainout from the stack and avoid corrosion
of system materials. Energy of the flue gas is used to raise the tem-
perature of the treated flue gas leaving the absorber before it is dis-
charged from the stack in the heating cycle. Temperature of the un-
treated flue gas entering the absorber is decreased in the cooling cycles.
Thermal design of the regenerators have been studied analytically
transfer and flow modeling of regenerators via computational fluid
dynamics has not been studied widely. Pascoli [39] has developed a
mathematical model describing the heat and fluid flow in a rotary re-
generator using porous media approach. He assumes the regenerator
operating at constant mass flow rates, rotation speed and inlet tem-
peratures, no thermal energy sources or sinks within the regenerator
walls, and single phase operation. The heat transfer coefficient at the
wall/fluid interface is assumed to be independent of the temperature
and time. His results agree well with Kays and London’s analytical
approach and experimental data. This model is preferred since the de-
tails of single channel are not required thanks to the porous medium
definition of the entire wheel.

⁎
Corresponding author.
E-mail address: koray.ozdemir@tubitak.gov.tr (K. Özdemir).

https://doi.org/10.1016/j.applthermaleng.2018.08.011
Received 9 January 2018; Received in revised form 2 June 2018; Accepted 4 August 2018
Available online 07 August 2018
1359-4311/ © 2018 Elsevier Ltd. All rights reserved.
K. Özdemir, M.F. Serincan
Nomenclature
c specific heat capacity (J/kg K)
E energy source
hgs heat transfer coefficient (W/m2 K)
k thermal conductivity
m mass flow rate
T temperature
w angular velocity
µ dynamic viscosity
γ porosity
ρ density
Kaydan and Hajidavalloo [40] investigated the thermal behaviour
of a full-scale rotary air-preheater using three-dimensional approach as
treating the preheater matrix as a porous media. Mass, momentum and
energy equations inside the matrix have been solved using moving re-
ference frame to represent the effect of the rotational speed. They have
obtained the porous media inputs experimentally. CFD results agreed
with the experimental data with a 7.5% error margin. According to the
analysis isotherms within the matrix are shown to be almost linear
except those close to the center of the matrix. They also show that
angular velocity of the matrix has significant effect on the efficiency up
to a certain limit. Analyzing the different types of matrix materials, it
has been concluded that material with low thermal diffusivity has
better thermal efficiency.
Corsini et al. [41] have described a methodology to account for the
effects of the Ljungström, the first efficient air preheater invented in
Sweden by Frederick Ljungström, they have implemented a CFD model
based on the work of Molinari and Cantiano. In order to reduce the
computational cost for modelling of the rotating matrix, which is filled
with hundreds of narrow channels, a synthetic description of the
Ljungström is applied. In this method a series of source terms have been
implemented in momentum and energy equations to account for the
detailed channel geometry. Also a porous medium approach is used to
simulate the flow inside the matrix. Standard high Reynolds k-ε Rey-
nolds-averaged Navier–Stokes equations (RANS) method was used to
solve the incompressible momentum equations. CFD results are com-
pared with available measurements in a real power plant where pres-
sure drops and temperature at the outlets are underestimated around
3%. Owing to fast calculation times authors suggest the approach to
study different ducting configurations Ljungstrom inlet to maximize the
effectiveness of the heat exchanger.
Sözen & Çiftçi [42] use a simplified model of a regenerator with
sinusoidal shaped ducts. They have only simulated the first shell
modeled from 0° to 360° using a commercial CFD software. Their model
consists of both fluid regions and the matrix geometry. Defining inter-
faces between the fluid regions and the matrix walls, they use sliding
mesh method to represent the rotation of the matrix. Alhusseny & Turan
[43] have carried numerical analysis of the fluid flow and heat
Untreated gas inlet Treated gas outlet
R R/10
Untreated gas outlet Treated gas inlet
Fig. 1. Corrugated sinusoidal ducts.
989
Applied Thermal Engineering 143 (2018) 988–1002
β packing density (m2/m3)
Subscripts
c treated gas
fg flue gas
g gas
h untreated gas
i Inlet
o outlet
s matrix solid material
transport phenomena through a rotary thermal regenerator using a
porous media approach. They have obtained the pressure drop and heat
transfer coefficient data to define the porous media parameters based
on the empirical equations for circular, square and triangular ducts.
They have presented the results by means of overall regenerator ef-
fectiveness, pressure drop, and the overall system performance (OSP)
and investigated the impact of different design parameters such as the
core geometrical features, and operating conditions. They have re-
vealed that OSP increases with increasing wall thickness, rotor length,
rotor frontal area and characteristic temperature difference; and de-
creases with increasing rotor hydraulic diameter.
In none of the above studies a complete 3D geometry of a rotary
heat exchanger in a FGD system has been modeled. Also transient
temperature variations inside the regenerator for different operating
conditions have not been studied rigorously.
2. Methodology
In this study, a rotary regenerative heat exchanger with corrugated
sinusoidal ducts have been pre-designed and modeled. In the design
corrugated sinusoidal ducts are selected as they the most common in
rotary regenerative wheels due to the simplicity of construction and
having large heat transfer surface area (Fig. 1).
RHEX operates with the flue gas composition of 13.91% CO2, 3.02%
O2, 14.46 H2O % and 68.61% N2. SO2 is not included in the flue gas
composition due to its negligible small content in this particular system.
Flue gas thermo-physical properties are obtained from a thermo-
dynamics software specially developed for flue gas calculations. The
thermo-physical property values obtained between 40 °C and 130 °C are
then expressed as a function of temperature as shown in Table 1. 0.5%
Carbon steel is defined as the material for the matrix with thermo-
physical properties given in Table 2. Geometric properties and oper-
ating conditions of the modeled RHEX are shown in Table 3. Treated
gas mass flow rate is lower than the untreated gas mass flow rate since
desulfurization occurs before treated gas enters the RHEX.
Channel geometry has been simplified by using the porous media
approach as described by Pascoli [39], and Kayden and Hajidavalloo
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Table 1
Flue gas properties.
Thermo-physical property Unit @40 °C @130 °C y = aT[K] + b

Density, ρ g kg/m 3
1.1940 0.9273 ρ g = −0.0030 * T + 2.1220
Specific heat capacity, cpg J/kg·K 1003 1013 cpg = 0.1111 * T + 998.2056
Thermal conductivity, kg W/m·K 0.0259 0.0319 kg = 6.7111 * 10−5 * T + 0.0049
Dynamic viscosity, µg kg/m·s 1.79 * 10−5 2.17 * 10−5 µg = 4.2222*10−8 * T + 4.6881 * 10−6

Table 2 μg 1
∇p = −⎛ v + C ρg |v| v⎞
Matrix material properties. ⎝α 2 ⎠ (1)
Thermo-physical Property Unit Value
where viscous resistance coefficient is denoted as 1/α and inertial re-
Density, ρ s kg/m 3
7833 sistance coefficient denoted as C are to be determined [44]. This form of
Specific heat capacity, cps J/kg·K 465 momentum equation in porous media is preferred rather than Darcy’s
Thermal conductivity, ks W/m·K 54 law which omits the inertial terms. Inertial resistances have been in-
cluded due to high velocities through the channel. To estimate these
parameters, CFD simulations are carried out for a single channel geo-
Table 3
metry as seen in the left side of Fig. 2. Half of the material thickness is
Designed RHEX details.
included in this model to characterize the periodicity of whole matrix
Parameter Unit Full scale 1/10 scale geometry and to account for suction and discharge effects through the
channel.
Channel height/width ratio, a/b mm/mm 1.5 1.5
Channel hydraulic diameter, Dh mm 2.3648 2.3648 In this model only momentum equation is solved with normal inlet
Shaft diameter, Ds mm 200 200 velocity as the boundary condition. No slip condition is applied at the
Rotor diameter, D mm 6000 600 channel wall. SIMPLE scheme as the pressure-velocity coupling, least
Rotor height, h mm 300 300 square cell based gradient and second order spatial discretization is
Radial seal coverage % 10 10
used for pressure, momentum and energy to have a more accurate and
Porosity, γ 0.72 0.72
Packing density, β m2/m3 1120 1120 stable solution. A residual value below 10−5 for continuity equation as
Untreated gas inlet temperature, Th,i °C 130 130 the convergence criteria as well as the stationary state value of the
Untreated gas mass flow rate, mh kg/s 50 0.44494 pressure drop are applied. Corresponding pressure drop values have
Treated gas inlet temperature, Tc,i °C 40 40
been recorded for different inlet velocity values. Results of the virtual
Treated gas mass flow rate, mc kg/s 49.875 0.44383
Matrix angular velocity rpm 2.5 2.5
experiments have been shown in Fig. 3 where a second order poly-
nomial fit is obtained.
Porous media approach used in this study also utilizes a non-equi-
[40]. The flow thorough the RHEX matrix is only in the axial direction librium thermal model in which temperature between matrix and flue
as it is constrained by the walls of individual channels. This is ac- gas differs. This discontinuity is accounted by the heat transfer coeffi-
counted by introducing anisotropic porosity for the matrix material. cient across the surface area in the axial direction. Heat transfer coef-
Porosity in axial direction is assumed to be constant whereas porosity in ficients are estimated through another separate CFD model for a single
circumferential and radial directions are taken as very large numbers to channel geometry as shown in the right side of Fig. 2. The CFD model is
account for impermeable walls. repeated for both heating and cooling cycles. Boundary conditions de-
In addition to the porosity, packing density, viscous and inertial scribed below are changed for individual cycles as tabulated in Table 4.
resistance coefficients are the other parameters averaged over the flow Flow in the channel is laminar since Reynolds numbers are smaller than
domain in the porous media approach. Porosity and packing density are 2300 due to small velocities.
calculated utilizing equations of curves that characterize the matrix For the inlet:
design. These calculations are given in the Appendix.
u=uin, v=0, w=0, T=Tin (2)
On the other hand, the viscous and inertial resistance coefficients
are estimated from the single channel analysis. Pressure drop across the For the outlet pressure boundary condition is prescribed for the
matrix is estimated as momentum whereas convective outlet condition is given for the energy
equation.

Fig. 2. Model geometries of a standalone channel to estimate heat transfer coefficients (left), viscous and inertial coefficients (right).

990
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

10th portion. Also it is assumed that the variations in the transport

variables are more significant in the axial and circumferential direction
while the radial variations are comparably very small. Temperature
variations in radial direction is therefore negligible. As a result, the
same thermal boundary conditions can be applied to both full scale and
small scale models. Same Reynolds number, porosity and packing
density values are applied for the full and 1/10th scaled models.

3. Numerical method

Non-equilibrium thermal-fluids model is solved with ANSYS-

Fig. 3. Virtual experiment results for velocity and pressure drop.
p=pout =0 q''=dz ρc p u ·∇T
For the channel surface a logarithmic temperature variation is as-
sumed to account for the real operating conditions as below:
Ts=Tw,in ∗ (Tw,out /Tw,in ) y/L
FLUENT which employs Finite Volume Method. Orthogonal mesh has
been used in the entire geometry. Mesh dependency is analyzed for the
1/10 scale model with the given mesh numbers and mesh qualities as
shown in Table 5. According to the average mesh quality indicators,
even the coarsest mesh is in the acceptable range.
Since there is almost no difference in the results between these mesh
cases, the first mesh with the least number of elements is used (Fig. 5)
to minimize computational cost.
Mesh motion is considered for the rotational movement of the ma-
trix. SIMPLE scheme as the pressure-velocity coupling, least square cell
(3) based gradient and second order spatial discretization is used for
pressure, momentum and energy.
Non-equilibrium thermal model is used for the heat transfer in the
porous zone according to the governing conservation equations given in
(4) (8) for the fluid zone and (9) for the solid zone [44].

where y is the axial location in the direction of flow, L is the channel ∂ ⎛ ⎛ ⎞ ⎞

(γρg Eg ) + ∇·(→
v (ρg Eg+p)) = ∇·⎜γkg ∇Tg−⎜∑ hi Ji⎟ + (τ·→
v )⎟
length, Tw,in and Tw,out are the minimum and maximum temperature ∂t
⎝ ⎝ i ⎠ ⎠
values expected on the channel, respectively. Obtaining outlet tem-
perature value, average heat transfer coefficient is calculated by (4). + βhgs (Ts−Tg) (8)
q channel ∂
have,channel= ((1−γ)ρs Es) = ∇ ·((1−γ)k s∇Ts + βhgs (Tg−Ts))
A s,channel ∗ΔTLMTD (5) ∂t (9)
where Boundary conditions are described below:
q=ṁ channel c p(Tout-Tin) For the gas inlets velocity, temperature boundary conditions are
(6)
defined:
and
ṁ h
u=uh,in= , v=0, w=0, T=Th,in
(Tin-Tw,in)-(Tout-Tw,out) ρ@T h,in (10)
ΔTLMTD=
ln((Tin-Tw,in)/(Tout-Tw,out )) (7)
ṁ h
SIMPLE scheme as the pressure-velocity coupling, least square cell u=u c,in= , v=0, w=0, T=Tc,in
ρ@T h,in (11)
based gradient and second order spatial discretization is used. A re-
sidual value below 10−5 for continuity equation as the convergence For the gas outlets pressure boundary condition is aplied:
criteria as well as the stationary state value of the average heat transfer
p=p h,out =0 (12)
coefficient are applied.
Fig. 4 shows the comparison of the estimated Nusselt (Nu = hDh/kg) p=pc,out =0 (13)
and Poiseuille Numbers (f.Re) with Zhang & Niu’s study [45]. Note that
latter is only for the fully developed region while our study includes the For the matrix inner and outer walls insulation boundary condition
entrance region as well. The results for the fully developed regime is prescribed:
completely agree, i.e. Nu = 2.614 and fReDh = 13.934, respectively.
q"=q"wall =0 (14)
Table 4 shows the average heat transfer coefficient values for the
corresponding conditions obtained from the standalone analysis. As For the matrix zone, angular velocity, porosity, packing density,
observed, untreated gas cooling cycle and treated gas heating cycle viscous resistance coefficent, inertial resistance coefficent and heat
have negligibly small differences in terms of average heat transfer transfer coefficient are defined assuming heat transfer coefficient has a
coefficients. In fact these small deviations stem from thermo-physical constant average value in the whole cell zone.
properties calculated at corresponding stream temperature. Therefore In order to validate the CFD analyses analytical solutions carried
an overall mean value for the heat transfer coefficient is utilized by the according to the procedure described at Shah and Sekulic [10] which
arithmetic average of values for heating and cooling cycles as the input uses ε-NTU (effectiveness and number of transfer units) method. In this
for the single porous media zone. method ε and NTU are calculated by (15) and (18), respectively:
Model geometry for the full scale RHEX is shown in Fig. 1. Flue gas
inlet and outlet zones are distinguished in the figure to represent the ⎛ 1 ⎞
ε = εcf ⎜1−
actual operation. The model is further simplified by focusing on the 1/ ∗1.93 ⎟
⎝ 9Cr ⎠ (15)
10th inner portion of the geometry as shown in figure. Since the geo-
metry is composed of a repeating matrix structure same no-slip where εcf is the counterflow heat exchanger effectiveness denoted by
boundary conditions can be applied to the smaller model with the 1/ (15) and Cr∗ is total matrix heat capacity rate ratio as shown at (16).

991
 K. Özdemir, M.F. Serincan

Table 4
Boundary conditions and results of the single channel analysis.
Low angular Medium angular High angular Low treated gas Medium treated gas High treated gas Low load Medium load High load
velocity velocity velocity inlet temperature inlet temperature inlet temperature (mfg = 0.9 * mfg) (mfg = mfg) (mfg = 1.1 * mfg)
(ω = 0.5 rpm) (ω = 2.5 rpm) (ω = 12.5 rpm) (Tc,i = 20 °C) (Tc,i = 20 °C) (Tc,i = 20 °C)

Untreated gas Flue gas inlet 130.00 130.00 130.00 130.00 130.00 130.00 130.00 130.00 130.00
cools temperature, °C
down Flue gas inlet 5.92 5.92 5.92 5.32 5.92 6.51 5.92 5.92 5.92
velocity, m/s
Reynolds Number 591.21 591.21 591.21 591.21 591.21 591.21 532.09 591.21 650.33
Wall temperature at 110.00 110.00 110.00 107.00 110.00 117.00 110.00 110.00 115.00
the inlet, °C
Wall temperature at 60.00 60.00 60.00 40.00 60.00 72.00 65.00 60.00 65.00
the outlet, °C
Average heat 51.51 51.51 51.51 53.74 51.51 57.69 46.72 51.51 60.38

992
transfer coefficient,
W/m2K

Treated gas Flue gas inlet 40.00 40.00 40.00 20.00 40.00 60.00 40.00 40.00 40.00
heats up temperature, °C
Flue gas inlet 4.60 4.60 4.60 4.38 4.60 4.83 4.14 4.60 5.05
velocity, m/s
Reynolds Number 714.24 714.24 714.24 749.40 714.24 682.23 642.81 714.24 785.66
Wall temperature at 60.00 60.00 60.00 40.00 60.00 72.00 65.00 60.00 65.00
the inlet, °C
Wall temperature at 110.00 110.00 110.00 107.00 110.00 117.00 115.00 110.00 115.00
the outlet, °C
Average heat 49.67 49.67 49.67 50.45 49.67 56.44 44.93 49.67 48.76
transfer coefficient,
W/m2K

Average heat transfer coefficient for 50.59 50.59 50.59 52.10 50.59 57.06 45.82 50.59 54.57
porous media, W/m2K
Applied Thermal Engineering 143 (2018) 988–1002
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Fig. 4. Comparison of Nusselt Number (left) and Poiseuille Number (right).

Table 5 where Cr is the heat capacity rate of a regenerator, Cmin is the minimum
Mesh dependency. heat capacity rate and C ∗ is the heat capacity ratio.
Mesh statics Mesh 1 Mesh 2 Mesh 3 UA
NTU =
Cmin (18)
Mesh numbers Nodes 610,698 859,404 1,253,178
Elements 581,535 821,730 1,202,870
where U is the overall heat transfer coefficient and A is the total heat
Orthogonal quality Minimum 0.38968 0.64800 0.56352 transfer area of the matrix.
Average 0.98517 0.98303 0.98543 Results are evaluated in terms of effectiveness, pumping power and
Maximum 0.99999 0.99999 1.00000
Standard deviation 0.03882 0.04054 0.03638
overall system performance defined by (19), (22) and (23), respectively
[43]. Overall regenerator effectiveness of a rotating heat exchanger can
Skewness Minimum 0.00131 0.00091 0.00148
be computed as:
Average 0.08267 0.09046 0.08369
Maximum 0.73090 0.55536 0.65375 ε c+εh
Standard deviation 0.09971 0.10313 0.09783 εt=
2 (19)
Aspect ratio Minimum 1.00550 1.00370 1.00350
Average 2.59140 2.63740 2.85580 where;
Maximum 12.44400 14.77400 18.83100
Qc ṁ c(c pc.o Tc.o-c pc.i Tc.i)
Standard deviation 2.18440 2.29040 2.68330 ε c= =
Q max,c ṁ c(c ph.i Th.i-c pc.i Tc.i) (20)
Treated gas outlet temperature (°C) 100.89 100.93 100.90
Untreated gas outlet temperature (°C) 69.51 69.48 69.49
Pressure drop through the matrix (Pa) 189.97 190.12 190.11 Qh ṁ h(c ph,i Th,i−c ph,o Th,o)
ε h= =
Q max,h ṁ h(c ph,i Th,i−c pc,i Tc,i) (21)

1−exp[−NTU (1−C ∗)] And pumping power is calculated as

εcf =
1−C ∗exp[−NTU (1−C ∗)] (16) ṁ c ṁ
PPt=PPc+PPh= Δp + h Δp
ρ c c ρ h h (22)
Cr Both heat transfer effectiveness and pumping power can be com-
Cr∗ =
Cmin (17) bined to give overall system performance as suggested in (23).

Fig. 5. Mesh structure.

993
K. Özdemir, M.F. Serincan
OSP=0.5(Qc+Q h )/PPt (23)
4. Results and discussion
Numerical model is solved for the design conditions given in
Table 3. Table 6 shows simulation results and comparison with the
analytical solution. It is found that, untreated gas outlet temperature is
overestimated by 2.23%, while treated gas outlet temperature is un-
derestimated by 2.98% and RHEX effectiveness is overestimated by
3.71%. These values can be considered as in the acceptable ranges for
further analysis.
Temperature distributions of untreated and treated flue gas through
the matrix are shown in Fig. 6 for the design point at the steady-state
conditions. Linear temperature variation is to be observed visually in
the flow direction for both treated and untreated gas sides. In the radial
direction (i.e. from the inner wall to outer wall of the matrix) there
seems to be the same temperature values since the flue gas enters
homogeneously. Temperature distribution within the matrix material
itself is also presented in Fig. 6. Because of the non-equilibrium thermal
phenomenon between the matrix and the gas, the matrix temperature
differs from the flue gas temperature. As expected, matrix temperature
decreases linearly from the untreated flue gas inlet to treated flue gas
inlet. Heat transfer between flue gas and the matrix occurs thanks to the
non-equilibrium thermal behavior. The same temperature values are
also observed through the radial direction results from the fact that the
matrix material has a high thermal conductivity and thermal diffusivity.
Further, pressure distribution of the flue gas at the middle plane of the
matrix is illustrated in Fig. 6. Similarly, a linear distribution is obtained.
Pressure drop for the untreated gas is higher than the treated gas due to
the differences in the thermo-physical properties and mass flow rate
values. Flue gas desulfurization captures the SO2 which leads to a de-
crease in the flue gas mass flow rate. As treated gas has a lower mass
flow rate, a lower pressure drop occurs in this side of the RHEX.
Angular velocity, treated gas inlet temperature and load (i.e. flue
gas mass flow rates) are important parameters in the operation of a
RHEX. Effects of these parameters on the thermal-fluids behavior of the
RHEX have been investigated in this study. Treated and untreated gas
outlet temperatures in transient, flue gas and matrix temperature dis-
tributions and pressure drop through the matrix have been studied for
each case. Effectiveness, pumping power and overall system perfor-
mances have been compared for each condition.
4.1. Effect of angular velocity
One of the important parameter on the RHEX effectiveness is the
angular velocity of the matrix. Higher angular velocity requires a higher
power to rotate the matrix. Therefore, it is important to know how the
heat recovery from the hot (i.e. untreated) flue gas changes as the an-
gular velocity of the matrix changes. Three different values for angular
velocity are given as the input for the rotational movement of the
matrix: low, medium (i.e. design) and high as 0.5 rpm, 2.5 rpm and
12.5 rpm, respectively.
Transient flue gas temperature variations from startup of the system
until the steady state conditions for different angular velocities are
shown in Fig. 7. Sinusoidal variations are observed in the treated and
untreated gas outlet temperatures as the heat exchange is periodic be-
tween the cold and hot cycles. The upper and lower temperature values
represent the maximum and minimum temperatures in the hot and cold
cycles, respectively. In addition, maximum temperature differences
between cold and hot cycles have the smallest value in the high angular
velocity case while they are higher in the low angular velocity case due
to the fact that higher RHEX rotational speeds require more number of
cycles. As seen in the figure, treated gas outlet temperature increases
with the increasing angular velocity while untreated gas outlet tem-
perature decreases with the increasing angular velocity. By increasing
994
Applied Thermal Engineering 143 (2018) 988–1002
the angular velocity, energy is transferred between the flue gases more
effectively which leads to an enhancement in the RHEX performance.
Rotating at a higher frequency implies that matrix will have a smaller
exposure time to either one of the alternating hot and cold cycles. This
results in temperature gradients across the matrix walls. Therefore, the
time-average of the temperature difference between the matrix and the
flue gas remains greater than those for the cases with lower angular
velocity. Sustaining non-equilibrium during alternating hot and cold
cycles results in better heat transfer performance. Consequently, steady
state conditions are reached faster.
Flue gas temperature and pressure distributions for low, medium
and high angular velocity cases are shown in Fig. 8. Treated gas outlet
temperature in the low angular velocity case is lower than those for the
medium and high angular velocity cases. Accordingly, untreated gas
outlet temperature in the low angular velocity case is higher than the
medium and high angular velocity cases due to lower heat recovery.
There is almost no difference in outlet temperature between medium
and high angular velocity. Temperature variations for both treated and
untreated gas from inlet to outlet have linear profiles. These trends are
also the same for the pressure distributions. There is almost no differ-
ence by means of pressure drop since the inlet velocities are the same.
Summary of the CFD analyses to examine the effect of angular ve-
locity are given in Table 7.
Effectiveness, in other words heat recovery capability, increases
with increasing angular velocity as shown in Table 7. Apart from in-
creasing temperature difference between the matrix and the flue gases,
higher rate of shuffling between hot and cold streams increases heat
transfer over a fixed time interval. As a result an increase in the RHEX
effectiveness is imminent. When the matrix rotates slowly, regenerator
effectiveness is reduced noticeably. On the other hand, if the matrix
rotates too fast, the carryover loss occurring between the two flue gases
caused by the amount of fluid trapped in the matrix walls, becomes
more pronounced. Pressure drop increases with increasing angular ve-
locity in the treated gas side while it decreases with increasing angular
velocity in untreated gas. Consequently, the pumping power increases
with increasing angular velocity in the treated gas side while it de-
creases with increasing angular velocity in untreated gas side. These
trends only stem from the differences in the thermo-physical properties
especially the densities. Because, as a general fact, pressure drop of flue
gas is directly proportional to the density which decreases as the tem-
perature increases. Therefore, as untreated gas has a higher mean
temperature thus a lower density than the treated gas, higher pressure
drop is observed since the mean temperature is lower. As a result,
overall system performance increases with increasing angular velocity.
However, there is no significant difference between medium and high
angular velocity cases. Thus it is not necessary to increase the angular
velocity after a certain value to achieve more heat recovery.
Table 6
CFD results for the design condition and validation.
Result (AS: Analytic Solution) Unit Design angular velocity
ω = 2.5 rpm
Treated gas outlet temperature-AS °C 98.69
Treated gas outlet temperature-CFD °C 100.89
CFD estimation wrt. AS % 2.23
Untreated gas outlet temperature-AS °C 71.64
Untreated gas outlet temperature-CFD °C 69.51
CFD estimation wrt. AS % −2.98
RHEX effectiveness-AS 0.6503
RHEX effectiveness-CFD 0.6744
CFD estimation wrt. AS % 3.71
Treated gas pressure drop-CFD Pa 170.24
Untreated gas pressure drop-CFD Pa 189.97
Pumping power required for treated gas W 68.86
Pumping power required for untreated gas W 83.53
Overall system performance 180.00
K. Özdemir, M.F. Serincan
Untreated Treated Untreated
gas inlet gas outlet gas inlet
Untreated Treated Untreated
gas outlet gas inlet gas outlet
Fig. 6. Temperature distributions of flue gas (left) and matrix (middle) and Pressure distributions of flue gas (right).
4.2. Effect of treated gas inlet temperature
In FGD systems, it is important to know how the heat recovery from
the untreated gas changes as the treated gas inlet temperature varies
during the start up of the absorber tower and thorugh wash of the flue
gas. In these systems, a higher value for the outlet temperature of the
treated gas is always favorable. Because in case of the lower treated gas
outlet temperature, condensation of the steam within the flue gas may
occur. As a result formation of corrosive acids is possible if this value is
lower than the dew point of that site in that particular operation. This is
very harmful for the surface of the matrix and the related surfaces
throghout the entire system. To prevent these circumstances and take
precautions it is crucial to know both the transient and steady-state
behavior of the treated gas outlet temperature. To examine this effect,
three different treated gas inlet temperature are given as the input as
low, medium (i.e. design) and high as 20 °C, 40 °C and 60 °C, respec-
tively.
Effect of flue gas temperature variations in treated gas inlet tem-
perature are shown in Fig. 9. There are sinusoidal variations observed
in treated and untreated gas outlet temperatures due to shuffling be-
tween the hot and cold cycles. Maximum temperature differences be-
tween hot and cold cycles nearly have the same value in whole cases. As
seen in the figure, treated gas outlet temperature increases with the
increasing treated gas inlet temperature while untreated gas outlet
temperature decreases with the decreasing treated gas inlet tempera-
ture. This is already expected due to the nature of the heat transfer
phenomenon in a heat exchanger. By decreasing the treated gas inlet
temperature, higher energy is transferred between two flue gases due to
Fig. 7. Flue gas outlet temperature variations due to angular velocity effect, treated gas (left), untreated gas (right).
995
Applied Thermal Engineering 143 (2018) 988–1002
Untreated Treated
Treated gas inlet gas outlet
gas outlet
Untreated Treated
Treated
gas outlet gas inlet
gas inlet
higher temperature gradients. In fact, by increasing treated gas inlet
temperature, matrix walls are exposed to a higher flue gas temperature
in the cold cycle, thus heat recovery rate between two streams de-
creases. Steady state conditions are reached almost with the same time
scale for all cases. Outlet temperature of the treated flue gas passes the
dew point temperature in about 60 s (i.e. 90 °C for this specific case).
Thus there is no need to preheat the treated gas to increase its inlet
temperature.
Flue gas temperature and pressure distributions for low, medium
and high treated gas inlet temperature are shown in Fig. 10. In the low
treated gas inlet temperature case, outlet temperatures of both treated
and untreated gases are lower than medium and high treated gas inlet
temperature cases. Temperature variation from inlet to outlet has a
linear profile. This trend is also the same for the pressure distribution.
Summary of the CFD results to examine the effect of treated gas inlet
temperature is given in Table 8. Heat exchanger effectiveness, or simply
the heat recovery of the RHEX, is the highest in the high treated gas
inlet temperature case, followed by low case and then the design case.
As the treated gas inlet temperature has a lower value higher heat
transfer rates are obtained because of higher temperature differences
between cold and high cycles. Similarly pressure drop through the
matrix, as well as the required pumping power also increases with the
increasing treated gas inlet temperature side since flue gas enters with
higher velocity due to reduced density for high treated gas inlet tem-
perature. Consequently, overall system performance increases with
decreasing treated gas inlet temperature. Therefore it is advantageous
to operate the absorber tower by decreasing the temperature of the
limestone slurry as long as dew point temperature is prevented.
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Fig. 8. Flue gas maen temperature distribution (left) and flue gas pressure distribution (right) in low, design and high angular velocity cases.

4.3. Effect of load velocity.

The other important parameter affecting the performance of a RHEX
is the mass flow rate, in other words the inlet velocity, of flue gases. In
FGD systems, it is important to know how the heat recovery is affected
due to the boiler operating at different loads and generating flue gas at
different rates. In order to point out the effect of load in RHEX, three
different types of loading is investigated as low, medium and high flue
gas mass flowrates i.e. 90%, 100% and 110%. respectively.
Flue gas temperature variations as a result of load change are shown
in Fig. 11. Similar to the previous results sinusoidal variations are ob-
served here as well. As seen in the figure, treated gas outlet temperature
increases with the decreasing load while untreated gas outlet tem-
perature decreases with the decreasing load.
Flue gas temperature and pressure distributions for low, medium
and high load cases are shown in Fig. 12. Treated gas outlet tempera-
ture in the low load case is slightly lower than the medium and high
load cases while untreated gas outlet temperature is higher in the low
load case, as expected. There is almost no difference between medium
and high load cases in terms of both treated and untreated gas outlet
temperatures. Temperature variation from inlet to outlet has a linear
profile. This trend is also the same for the pressure distribution. When
flue gas flows through the channel, there is a collision between the
molecules due to the random molecular movements which results in the
decrease of the kinetic energy. This loss of kinetic energy is converted
into some other form of energy, i.e. internal energy. As it is known from
continuity equation that since the cross section of flow is not changing,
velocity of fluid has to be same. Therefore, pressure energy is converted
into kinetic energy to keep the velocity same. This is the reason why
pressure drop increases with increasing load, in other words increasing
Summary of the CFD analyses to examine the effect of load are given
in Table 9. RHEX effectiveness, has its highest value for the low load,
and decreases with increasing load. This reduction is due to lack of
matrix capacity for energy transfer in higher loads. The effectiveness is
enhanced at smaller heat capacity ratios especially when the solid heat
capacity is large. Moreover, it is clear that RHEX effectiveness is better
for lower mass flow rate as the flue gas heat capacity becomes smaller.
This is due to the fact that for lower mass flow rates, hence smaller inlet
velocities, better heat transfer occurs between flue gas and matrix wall.
Because in lower velocities, there is enough time for the matrix wall to
capture the heat from untreated gas in the cooling cycle and release this
heat to untreated gas in the heating cycle Also, high inlet velocities are
not practical since turbulence flow is not attending in the RHEX. That
is, in the laminar flow conditions it is practical to operate with lower
velocities. Increasing load lead to higher value of pressure drop and
required pumping power for both treated and untreated gas sides be-
cause of higher velocities results in higher carryover. Eventually overall
system performance increases with decreasing load.
5. Conclusion
In this study a computational fluid dynamics model is presented for
thermal-fluids analysis of a rotary regenerative heat exchanger in flue
gas desulfurization unit. A carbon steel matrix comprised of sinusoidal
channels is modeled as a porous medium. Heat transfer between fluid
and solid are modeled with a local thermal non-equilibrium assump-
tion. The sophisticated geometric details of the matrix are reduced
conveniently with porous media parameters such as permeability and
inertial flow coefficients. These parameters are estimated based on CFD

Table 7
Results: Effect of angular velocity.
Result Unit Low angular velocity Design angular velocity High angular velocity
ω = 0.5 rpm ω = 2.5 rpm ω = 12.5 rpm

Treated gas outlet temperature °C 98.15 100.89 101.01

Untreated gas outlet temperature °C 72.95 69.51 69.25
Heat transfer rate in treated gas side kW 26.16 27.40 27.51
Heat transfer rate in untreated gas side kW 25.90 27.46 27.57
RHEX effectiveness 0.6400 0.6744 0.6771
Treated gas pressure drop Pa 168.89 170.24 170.66
Untreated gas pressure drop Pa 191.57 189.97 188.36
Pumping power required for treated gas W 68.06 68.86 69.05
Pumping power required for untreated gas W 84.66 83.53 82.79
Overall system performance 170.45 180.00 181.38

996
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Fig. 9. Flue gas outlet temperature variations due to treated gas inlet temperature effect, treated gas (left), untreated gas (right).

simulations of a single channel with the actual complexity. Heat
transfer coefficients are also estimated by means of CFD simulations of
a standalone channel investigating non-equilibrium phenomena. It is
shown that CFD model overestimates the treated gas temperatures by at
most 2.23% while untreated gas temperatures are underestimated by at
most 2.98%. The error bounds suggest the validity of the model to be
used as a virtual test tool. Effects of different operating conditions are
investigated such as angular velocity, treated gas inlet temperature and
load conditions. It is observed that treated gas outlet temperature in-
creases with increasing angular velocity while it decreases with de-
creasing load. On the other hand, untreated gas outlet temperature
decreases with increasing angular velocity and reduced load. By means
of overall system performance it is estimated that higher overall system
performance is achieved with lower angular velocity, lower treated gas
inlet temperature and reduced load.

Appendix

Scaling calculations

Same Reynolds number, porosity and packing density values are applied for the full and 1/10th scaled models. Reynolds number is expressed as:
ρUDh
Re =
μ (A1)
Since the geometry in the matrix structure is repeating, hydraulic diameters are same for a single channel, full scale and 1/10 scaled models as:
(Dh )single-channel=(Dh ) 1 scaled model=(Dh ) full-scale model
10 (A2)

⎛ 4n1A c ⎞
⎜ ⎟
4n A
=⎛ 2 c ⎞
⎜ ⎟
4n A
=⎛ 3 c ⎞
⎜ ⎟

⎝ 1 w ⎠single-channel ⎝ n2 Pw ⎠ 10
n P 1 scaled model ⎝ n3 Pw ⎠full-scale model (A3)

Fig. 10. Flue gas maen temperature distribution (left) and flue gas pressure distribution (right) in low, design and high treated gas inlet temperature cases.

997
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Table 8
Results: Effect of treated gas inlet temperature.
Result Unit Low treated gas inlet temperature Design treated gas inlet temperature High treated gas inlet temperature
Tc.i = 20 °C Tc.i = 40 °C Tc.i = 60 °C

Treated gas outlet temperature °C 95.16 100.89 109.03

Untreated gas outlet temperature °C 55.37 69.51 81.24
Heat transfer rate in treated gas side kW 33.73 27.40 22.13
Heat transfer rate in untreated gas side kW 33.82 27.46 22.16
RHEX effectiveness 0.6809 0.6744 0.6985
Treated gas pressure drop Pa 162.17 170.24 181.22
Untreated gas pressure drop Pa 183.22 189.97 195.74
Pumping power required for treated gas W 63.42 68.86 76.14
Pumping power required for untreated gas W 78.95 83.53 87.55
Overall system performance 237.23 180.00 135.27

where

n is the number of channel

n1 = 1 for single channel
n2 = number of channels for 1/10 scaled model
n3 = number of channels for full scale model

Axial velocity in each channel is also same for all the models due to the repeating channel geometry.
(U)single-channel=(U) 1 scaled model=(U) full-scale model
10 (A4)
Mass flow rate is expressed as:
ṁ =ρUA c (A5)
Where the cross sectional area, A c , is calculated as following:
A c=γ*A disk (A6)

γ : porosity
A disk : Area of disk which is expressed as:
π
A disk= (Do2-Ds 2) ∗ (1-SC)
4 (A7)
where:

Do : Outer diameter of the disk (Dofull-scale =6 m and Do1/10-scale = 0.6 m )

Di : Inner diameter of the disk (Difull-scale = 0.2 m and Di 1 = 0.2 m )
-scale
10
SC : seal coverage, % (SCfull-scale=SC1/10-scale =10\% )

Therefore below expression can be obtained:

Fig. 11. Flue gas outlet temperature variations due to load effect, treated gas (left), untreated gas (right).

998
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Fig. 12. Flue gas maen temperature distribution (left) and flue gas pressure distribution (right) in low, design and high load cases.

Table 9
Results: Effect of load.
Result Unit Low load mfg = 0.9mfg Design load mfg = mfg High load mfg = 1.1mfg

Treated gas outlet temperature °C 100.97 100.89 100.61

Untreated gas outlet temperature °C 69.42 69.51 69.79
Heat transfer rate in treated gas side kW 24.75 27.46 30.06
Heat transfer rate in untreated gas side kW 24.69 27.40 30.00
RHEX effectiveness 0.6753 0.6744 0.6712
Treated gas pressure drop Pa 151.86 170.24 188.81
Untreated gas pressure drop Pa 169.72 189.97 210.41
Pumping power required for treated gas W 55.28 68.86 83.97
Pumping power required for untreated gas W 67.16 83.53 101.81
Overall system performance 201.88 180.00 161.64

̇ full scale (A c ) full scale

(m)
=
(m)̇ 1 scale (A c ) 1 scale
10 10 (A8)

(A c ) 1 scale
̇ 1
(m) ̇
scale=(m) full scale
10
10 (A c ) full scale (A9)

̇ 1 ̇
(γ π
4
(Do 2-Di 2) ∗ (1-SC) ) 1 scale
10
(m) scale=(m) full scale
10
(γ π
4
(Do 2-Di 2) ∗ (1-SC) )
full scale (A10)

(A11)

((Do 2-Di 2)) 1 scale

̇ 1
(m) ̇
scale=(m) full scale
10
10 ((Do 2-Di 2)) full scale (A12)
For the hot side:
(0.62-0.22)
(m)̇ h 1 scale=50 (kg/s) = 0.44494 kg/s
10 (62-0.22) (m2) (A13)
For the cold side:
(0.62-0.22) (m2)
(m)̇c 1 scale = 49.875 (kg/s) = 0.44383 kg/s
10 (62-0.22) (m2) (A14)

Porosity calculations

Porosity can be calculated as the ratio of flow cross-section to the total cross-section as following:
Matrix material was designed by using circle and sine wave around a circle equations.

999
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Circle equation is:

x c = Rcosθ (A15)

yc =Rsinθ (A16)
Sine wave around a circle equation is:
x s=[R+asin(nθ)]cosθ (A17)

ys=[R+asin(nθ)]sinθ (A18)
where;

R: circle radius
a: sinusoidal amplitude
n: number of sinusoids on circle

Design was conducted for a channel which’s channel height/width ratio, a/b with 1.5 and hydraulic diameter with 2.3648 mm. This values are
only our choice that does not mean the optimum values.
To obtain the geometries below equations of the curves were used (i: inner, o: outer):
First circle inner:
x c,i,first = Rcosθ (A19)

yc,i,first =Rsinθ (A20)

First circle outer:
x c,o,first = (R+t)cosθ (A21)

yc,o,first=(R+t)sinθ (A22)
sinusoidal inner:
x s,i = (R+t+a+asin(nθ))cosθ (A23)

ys,i = (R+t+a+asin(nθ))sinθ (A24)

sinusoidal outer
x s,o = (R+2t+a+asin(nθ))cosθ (A25)

ys,o = (R+2t+a+asin(nθ))sinθ (A26)

Second circle inner
x c,i,second = [R+2(t+a)]cosθ (A27)

yc,i,second = [R+2(t+a)]sinθ (A28)

Second circle outer
x c,o,second = [R+t+2(t+a)]cosθ (A29)

yc,o,second = [R+t+2(t+a)]sinθ (A30)

where;

R: inner radius (100 mm)

A: sinusoidal amplitude (2.5 mm)
n: number of sinusoids on circle (190.395)
t: tcikness of the plate (1 mm)
θ: angle from 0 to 2π

Area between the first circle outer and first circle inner is calculated as:
2π 1
Area first-circle = ∫0 2
(([(R+t)cos(θ)]2 +[(R+t)sin(θ)]2 )-([Rcos(θ)]2 +[Rsin(θ)]2 ))dθ
(A31)
Area between the sine wave outer and sine wave inner is calculated as:
2π 1
Area sine wave = ∫0 2
(([[R+2t+a+asin(nθ)]cos(θ)]2 +[[R+2t+a+asin(nθ)]sin(θ)]2 )-([[R+t+a+asin(nθ)]cos(θ)]2 +[[R+t+a+asin(nθ)]sin(θ)]2 ))dθ

(A32)
Total transverse area of the solids thorough the flow direction is calculated as:
Areatotal-solid=Area first-circle+Area sine wave=642.46 mm2 (A33)

1000
K. Özdemir, M.F. Serincan Applied Thermal Engineering 143 (2018) 988–1002

Area between the second circle inner and first circle outer is calculated as:
2π 1
Area whole = ∫0 2
(([[R+2(t+a)]cos(θ)]2 +[[R+2(t+a)]sin(θ)]2 )-([Rcos(θ)]2 +[Rsin(θ)]2 ))dθ=2276.08 mm2
(A34)
Thus, porosity of the matrix in the flow direction is calculated as:
Areatotal-solid 642.46
γ=1- =1- =0.72
Area whole 2276.08 (A35)

Packing density calculations

Packing density, total heat transfer area in the flow flow, is calculated with below procedure:
Length of a curve is found as below:
2 2
⎛ dx ⎞ +⎛ dy ⎞ dt
b
Lengthcurve = ∫a ⎝ dt ⎠ ⎝ dt ⎠ (A36)
Length of the first circle outer:
2 2
⎛ d[(R+t)cos(θ)] ⎞ +⎛ d[(R+t)sin(θ)] ⎞ dθ
2π
L first circle outer =∫0 ⎝ dθ ⎠ ⎝ dθ ⎠ (A37)
Length of the sinusoidal inner:
2 2
⎛ d[[R+t+a+asin(nθ)]cos(θ)] ⎞ +⎛ d[[R+t+a+asin(nθ)]sin(θ)] ⎞ dθ
2π
L sinusoidal inner = ∫0 ⎝ dθ ⎠ ⎝ dθ ⎠ (A38)
Length of the sinusoidal outer:
2 2
⎛ d[[R+2t+a+asin(nθ)]cos(θ)] ⎞ +⎛ d[[R+2t+a+asin(nθ)]sin(θ)] ⎞ dθ
2π
L sinusoidal outer = ∫0 ⎝ dθ ⎠ ⎝ dθ ⎠ (A39)
Length of the second circle inner:
2 2
⎛ d[[R+2(t+a)]cos(θ)] ⎞ +⎛ d[[R+2(t+a)]sin(θ)] ⎞ dθ
2π
L second circle inner = ∫0 ⎝ dθ ⎠ ⎝ dθ ⎠ (A40)
Total wetted surface area is calculated as:
Area wetted surface=(L first circle outer+L sinusoidal inner+L sinusoidal outer+L second circle inner )*h (A41)
Whole volume is calculated as:
Volume whole=Area whole∗h (A42)
Thus, packing density of the matrix is calculated as:
Area wetted surface m2
β= = 1120 3
Volume whole m (A43)

References
[1] W. Nusselt, Die Theorie des Winderhitzers, Z. Deutscher Ingenieure 71 (1927)
85–91.
[2] H. Hausen, Über die Theorie des Warmeaustausches in Regeneratoren, Z. Angew.
Math. Mech. 9 (1929) 173–200.
[3] T.E.W. Schumann, A liquid flowing through a porous prism, Heat Transfer (1929)
405–416.
[4] B.S. Baclic, A simplified formula for cross-flow heat exchanger effectiveness, J. Heat
Transfer 100 (1978) 746–747.
[5] A. Burns, A.J. Willmott, Transient performance of periodic flow regenerators, Int. J.
Heat Mass Transfer 21 (1978) 623–627.
[6] E.M. Smith, General integral solution of the regenerator transient test equations for
zero longitudinal conduction, Int. J. Heat Fluid Flow 1 (1979) 71–75.
[7] T. Yılmaz, E. Cihan, Enerji Geri Kazanımında Etkin Bir Araç: Döner Tip
Rejeneratörler, Tesisat Mühendisliği Dergisi 10 (1) (1993) 29–33.
[8] T. Yılmaz, O. Büyükalaca, L. Candar, Calculation of rotary heat exchangers with
various matrix geometries. First Trabzon International Energy and Environment
Symposium. Trabzon, (1996).
[9] E. Doğruyol, Döner Rejeneratif Isı Eşanjöründe Devir Sayısına Bağlı Olarak Verimin
Ölçülmesi, Çukurova Üniversitesi Fen Bilimleri Enstitüsü Makina Mühendisliği
Anabilim Dalı, Adana, 1997.
[10] R.K. Shah, D.P. Sekulic, Fundamentals of Heat Exchanger Design, John Wiley &
Sons. Inc., 2003.
[11] O. Büyükalaca, T. Yılmaz, Design of regenerative heat exchangers, Heat Transfer
Eng. 24 (4) (2003) 32–38.
[12] S. Sanaye, S. Jafari, H. Ghaebi, Optimum operational conditions of a rotary re-
generator using genetic algorithm, Energy Build. 40 (2008) 1637–1642.
[13] S. Sanaye, H. Hajabdollahi, Multi-objective optimization of rotary regenerator using
genetic algorithm, Int. J. Therm. Sci. 48 (2009) 1967–1977.
[14] H.Y. Wang, L.L. Zhao, Z.G. Xu, W.G. Chun, H.T. Kim, The study on heat transfer
model of tri-sectional rotary air preheater based on the semi-analytical method,
Appl. Therm. Eng. 28 (2008) 1882–1888.
[15] T.J. Lambertson, Performance Factors of a Periodic-flow Heat Exchanger, ASME
Transaction, Monterey, California, 1957.
[16] H.H. Sogin, K.-E. Hassan, A Design Manual for Regenerative Heat Exchangers of the
Rotary Type, Illinois Institute of Technology, Ohio, 1958.
[17] A.J. Willmott, Digital computer simulation of a thermal regenerator, Int. J. Heat
Mass Transfer 7 (1964) 1291–1302.
[18] F.W. Larsen, Rapid calculation of temperature in a regenerative heat exchanger
having arbitrary initial solid and entering fluid temperatures, Int. J. Heat Mass
Transfer 10 (1967) 149–168.
[19] A.J. Willmott, Simulation of a thermal regenerator under conditions of variable
mass flow, Int. J. Heat Mass Transfer 11 (1968) 1105–1116.
[20] A.J. Willmott, The regenerative heat exchanger computer representation, Int. J.
Heat Mass Transfer 12 (1969) 997–1014.
[21] A.J. Willmott, Digital computer simulation of a thermal regenerator, Int. J. Heat
Mass Transfer 12 (1969) 1291–1302.
[22] P. Razelos, M.K. Benjamin, Computer model of thermal regenerators with variable
mass flow rates, Int. J. Heat Mass Transfer 21 (1977) 735–743.
[23] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, McGraw Hill, 1980.
[24] Y.L. Bae, Performance of a Rotating Regenerative Heat Exchanger - A Numerical
Simulation, Oregon State University, 1986.

1001
K. Özdemir, M.F. Serincan
[25] D.R. Atthey, An approximate thermal analysis for a regenerative heat exchanger,
Int. J. Heat Mass Transfer 31 (1988) 1431–1441.
[26] A. Hill, A.J. Willmott, A robust method for regenerative heat exchanger calcula-
tions, Int. J. Heat Mass Transfer 30 (1987) 241–249.
[27] Accurate and rapid thermal regenerator calculations, Int. J. Heat Mass Transfer 32
(1989) 465–476.
[28] Y. Varol, Rejeneratif Isı Değiştireçleri Yardımıyla Enerji Geri Kazanımı, Fırat
Üniversitesi Fen Bilimleri Enstitüsü Makina Eğitimi Anabilim Dalı, Elazığ, 1991.
[29] Ş. Ünal, Döner Tip Rejeneratörlerin Etkinliğinin Nümerik Olarak Hesaplanması,
Çukurova Üniversitesi Fen Bilimleri Enstitüsü Makina Mühendisliği Anabilim Dalı,
Adana, 1996.
[30] B. Atalay, Waste Heat Recovery Using Regenerative Heat Exchanger, Çukurova
University Institute of Natural and Applied Sciences Department of Mechanical
Engineering, Adana, 1998.
[31] N. Ghodsipour, M. Sadrameli, Experimental and sensitivity analysis of a rotary air
preheater for the flue gas heat recovery, Appl. Therm. Eng. 23 (2003) 571–580.
[32] F. Demiralp, Döner Tip Rejeneratörlerin Etkinliğinin Sonlu Farklar Yöntemiyle
Hesaplanması, Fırat Üniversitesi Fen Bilimleri Enstitüsü Makine Eğitimi Anabilim
Dalı, 2003.
[33] T. Skiepko, R.K. Shah, A comparison of rotary regenerator theory and experimental
reults for an air preheater for a thermal power plant, Exp. Therm. Fluid Sci. 28
(2004) 257–264.
[34] S. Alagic, N. Stosic, A. Kovacevic, I. Buljubasic, Numerical analysis of heat transfer
and fluid flow in rotary regenerative air pre-heaters, J. Mech. Eng. 51 (2005)
1002
Applied Thermal Engineering 143 (2018) 988–1002
411–417.
[35] B. Drobnic, J. Oman, M. Tuma, A numerical model for the analyses of heat transfer
and leakages in a rotary air preheater, Int. J. Heat Mass Transfer 49 (2006)
5001–5009.
[36] Z. Wu, R.V.N. Melnik, F. Borup, Model-based analysis and simulation of re-
generative heat wheel, Energy Build. 38 (2006) 502–514.
[37] C. Falk, Predicting performance of regenerative heat exchanger, Lund (2009).
[38] J. Dallaire, L. Gosselin, A.K. da Silva, Conceptual optimization of a rotary heat
exchanger with a porous core, Int. J. Therm. Sci. 49 (2010) 454–462.
[39] O. Pascoli, Analysis and Characterization of Heat and Mass Transfer in Rotary
Exchangers, Universita Degli Studi di Udine, Udine, 2012.
[40] A.H. Kaydan, E. Hajidavalloo, Three-dimensional simulation of rotary air preheater
in steam power plant, Appl. Therm. Eng. 73 (2014) 399–407.
[41] A. Corsini, G. Delibra, G.D. Meo, M. Martini, F. Rispoli, A. Santoriello, A CFD-based
virtual test-rig for rotating heat exchangers, Energy Procedia 82 (2015) 245–251.
[42] A. Sözen, E. Çiftçi, Determination of performance of the heat wheel via CFD, J.
Polytech. 19 (4) (2016) 547–554.
[43] A. Alhusseny, A. Turan, An effective engineering computational procedure to
analyse and design rotary regenerators using a porous media approach, Int. J. Heat
Mass Transfer 95 (2016) 593–605.
[44] Ansys Fluent User Guide.
[45] J.L. Niu, L.Z. Zhang, Heat transfer and friction coefficients in corrugated ducts
confined by sinusoidal and arc curves, Int. J. Heat Mass Transfer 45 (2002)
571–578.