International Journal of Heat and Mass Transfer 126 (2018) 730–745

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Heat transfer in counterflow fluidized bed of oxide particles for thermal

energy storage
Daniel C. Miller, Christopher J. Pfutzner, Gregory S. Jackson ⇑
Department of Mechanical Engineering, Colorado School of Mines, Golden, CO, United States

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

Article history: The potential for inert oxide particles as a heat transfer and thermal energy storage (TES) media in con-
Received 22 February 2018 centrating solar power (CSP) depends in part on particle receiver designs that provide high wall-to-
Received in revised form 16 April 2018 particle heat transfer rates. This paper presents a novel continuous-flow approach to achieve high heat
Accepted 31 May 2018
transfer coefficients hw for particle receivers by fluidizing net-downward-flowing particles in a narrow
Available online 14 June 2018
vertical channel bounded by an external irradiated/heated wall and a parallel interior wall with a metal
mesh opening that allows the upward-flowing fluidizing gas to exit at the top of the channel. To demon-
Keywords:
strate the high hw of this flow configuration, a fluidized bed in a 10 cm
10 cm
0.64 cm deep channel
Fluidized bed
Bubbling fluidization
was heated through an external aluminosilicate wall with mid-IR quartz lamps that provided external
Particle heat transfer wall heat fluxes up to 20 W cm2. Extensive heat transfer measurements with fluidized Carbo Accucast
Thermal energy storage ID50 particles (diameters between 150 and 350 lm) in steady-state continuous downward flow and in
Concentrating solar power transient batch mode assessed total hw as functions of particle bed temperatures T b , bed solids volume
fractions ab , and superficial gas velocities U g . Results showed that the narrow-channel fluidized bed
can achieve overall hw as high as 1000 W m2 K1. The highest hw were measured at upward U g between
2 and 4 times the minimum bed fluidization velocities, U mf , which decreased to 0.12 m s1 for the mean
particle diameter at T b ¼ 600  C. Increasing U g further above U mf decreased hw due to an associated
decrease in ab . hw increased strongly with T b in part, because gas-phase conductivity and the radiative
heat transfer contribution increased with T b . The extensive measurements were fit to a modified version
of the Nusselt number correlation by Molerus (1992). For ab P 0:1, the Molerus correlation with
adjusted dependence on excess fluidization velocity (U g  U mf ) provided an excellent fit to the measured
convective fraction of hw (with < 10% error). Adding the radiation component with the Molerus correla-
tion provides an effective tool for calculating hw for this counterflow fluidized bed configuration. A simple
analysis explored the impact of such high hw for an indirect receiver design with angled external walls to
spread solar aperture fluxes. Results from the analysis indicated that total hw ¼ 1000 W m2 K1 can
enable solar collection efficiencies approaching 90% with external wall temperatures T w;ext  1020  C.
This potential performance motivates further exploration of this fluidized bed configuration for particle
receivers for CSP applications.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction and background
Concentrating solar power (CSP), as a non-CO2-emitting renew-
able energy resource, has unique potential in combination with
low-cost, large-scale thermal energy storage (TES) to provide dis-
patchable renewable electricity and thereby to enable higher grid
penetration of other intermittent renewable energy resources such
as wind and solar photovoltaics [3,4]. Much of recent CSP technol-
ogy development has focused on central receivers with solar
⇑ Corresponding author.
E-mail address: gsjackso@mines.edu (G.S. Jackson).
concentrations of 1000 suns (1 sun = 1000 W m2) or more to
achieve high temperatures in heat transfer fluids (HTFs) for cou-
pling to high-temperature TES and power cycles with better elec-
trical conversion efficiencies than typical steam Rankine power
cycles [5–7]. Central-tower CSP plants have a solar collector field
with heliostat mirrors that concentrate the solar flux onto a central
receiver where the solar energy is absorbed and transferred to the
HTF. A recovery heat exchanger extracts stored solar thermal
energy from the HTF into a power cycle that generates electricity.
New central-receiver CSP installations have incorporated TES sub-
systems based on molten nitrate salts, which also serve as the HTF,
to enable temporal shifting of electricity generation to make the
solar energy resource more dispatchable [8,9]. The nitrate salts

https://doi.org/10.1016/j.ijheatmasstransfer.2018.05.165
0017-9310/Ó 2018 Elsevier Ltd. All rights reserved.
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
have temperature stability limits below 600  C, and thus, can only
be integrated with lower-temperature power cycles such as Rank-
ine cycles with electric conversion efficiencies well under 40%. To
overcome these limitations of current molten salts, many research-
ers and developers are exploring HTFs that provide the potential
for higher-temperature TES to allow for CSP to couple with higher
efficiency power cycles, such as supercritical-CO2 cycles [10–12] or
recuperative air-Brayton cycles.
Alternative HTFs with higher temperature limits that are being
explored include molten carbonate salts [13,14], molten chloride
salts [15,16], liquid metals [7], inert solid oxide particles [17–20],
and redox-active oxide particles for thermochemical energy stor-
age [21–24]. Molten carbonate and chloride salt mixtures have
the promise of operating temperatures as high as 850 and 900  C
respectively, but the carbonates have stability issues at high tem-
peratures [14]. Furthermore, identifying cost-effective, corrosion-
resistant structural materials for containment of these molten salts
and liquid metal remains a significant challenge for operation
above 700  C [7,15]. Thermochemical energy storage (TCES) sys-
tems based on reducible oxides redox cycles have the potential
to provide very high storage temperatures and specific TES capac-
ity, but at the expense of system complexity that remains an issue
for substantially more research and development [23,25]. Inert
oxide particles also offer the potential for extremely high temper-
atures above 1000  C, but they only provide sensible energy stor-
age and as such lower specific TES (kJ kg1) for a given
temperature difference between the hot (T H ) and cold (T C ) storage
[17]. Nonetheless, low-cost, inert oxide materials such as Al2O3
[26], bauxite [27], or quartz sand [28] have the potential to provide
cost-effective energy storage at very high T H and any range of T C
without concerns related to high-temperature corrosion or lower
temperature freezing. The material stability and operational flexi-
bility of inert oxide materials for high-temperature TES motivates
the significant research on various particulate materials for heat
transfer media as summarized in recent reviews [20,29].
Particulate oxide materials as the HTF and TES storage media in
CSP plants require design and implementation of effective particle
transport systems and robust flow control for both the solar recei-
ver and the recovery heat exchanger to transfer the thermal energy
to the power cycle. Unlike high-temperature liquid HTFs such as
molten salts, particulate solids present unique challenges in terms
of flow distribution and effective heat transfer. Research on partic-
ulate HTFs has led to development of two different categories of
solar particle receivers, direct and indirect absorption receivers
[20,30]. Direct particle receivers allow solar radiation from a helio-
stat field to irradiate the particles directly either in an open envi-
ronment or through a transparent window, often proposed as
quartz. Sandia National Laboratories in the U.S. has been leading
significant efforts on developing direct absorption particle recei-
vers utilizing a falling-particle curtain [27,31,32]. Open direct-
receiver concepts rely on gravity-driven flows with or without
obstructions to slow the particle fall and increase the residence
time within the receiver. Other gravity-driven direct particle recei-
vers have been demonstrated [33,34] including designs with
quartz tubes to confine the directly irradiated particles [35]. With
open direct receivers, high efficiency solar energy capture is possi-
ble, but convective losses and particle containment remain chal-
lenges, and it can be difficult to balance the need for high solar
absorption efficiency and uniform particle heating in curtain-
based designs [20]. The concerns motivate designs with transpar-
ent windows like quartz, but the structural integrity at large scales
of such transparent window materials under the cyclic non-
isothermal conditions presents an unresolved challenge.
Concerns with direct receiver designs motivate research and
development on indirect particle receivers, where the particles
are confined within opaque wall materials that provide better
731
structural integrity under the cyclic high-temperature operation
of a central solar receiver [20]. One indirect particle receiver con-
cept uses a unique fluidized upward flowing suspension of parti-
cles to provide high heat transfer rates and particle flow control
[18,36,37]. Other gravity-driven indirect receivers have employed
granular flow over solar radiation-trapping cavities to increase par-
ticle residence times and surface area to spread solar fluxes and
lower requirements for hw [38,39]. Such designs with more com-
plex geometries increase risk for mechanical failure due to thermal
stresses and/or cyclic fatigue and as such designs that rely on sim-
pler configurations and rely on flows that promote higher hw
between confinement walls and the particles. Higher hw for exter-
nal receiver walls that absorb concentrating solar fluxes reduces
the necessary wall temperatures T w to capture a specific wall heat
flux q00 _ w decreases re-radiation losses
_ w . Lower T w for a given q00
from the receiver to the environment (that scale with T 4w ) and thus
increase receiver efficiency. To meet solar receiver efficiency goals
of 90% as set by the U.S. Department of Energy [11], particle recei-
vers with external walls will need to sustain very high hw as dis-
cussed below in the current study.
Bubbling fluidized beds offer an approach to high hw in indirect
particle receivers and other particle heat exchanger applications.
Many studies on heat transfer with fluidized particle flows have
focused on horizontal tubes or tube bundles inside the bed as
might be relevant in a recovery heat exchanger for CSP [40–45].
Other studies have addressed heat transfer between circulating flu-
idized beds and vertical surfaces [46–48], but the void volume frac-
tions and gas velocities for those studies were much higher than
typical for bubbling fluidized beds. Research groups in Europe
[18,36,30,49] have experimentally assessed heat transfer between
fluidized beds and vertical walls for central receiver applications.
Table 1 highlights those and a few other investigations and shows
significant variation in measured hw . The large variation in hw mea-
surements may be due in part, to variations in particle properties,
particle diameter dp , superficial gas velocity U g , and average bed
temperature T b . Nonetheless, these studies do indicate that some
conditions can provide very high hw P 1000 W m2 K1, and if
such hw can be sustained in a receiver design, fluidized bed config-
urations may achieve receiver efficiency targets > 90%.
Efforts in multi-phase flow CFD have explored the wall-to-
particle heat transfer in fluidized beds [35,53,54], but conventional
two-fluid CFD methods of modeling particle-laden flows have
shown limited success in predicting wall-to-bed heat transfer
[55,56]. Recent advances combining CFD with discrete element
methods (DEM) using high-performance computing have shown
promise in improving predictions of computational multi-phase
modeling of particulate flows [57,58]. Computational demands
for these approaches do not facilitate extensive parametric design
studies, and this has encouraged research into more efficient sim-
ulation approaches such as recurrence-CFD [59]. While advances in
fluidized bed modeling present opportunities for numerical design
of some aspects of fluidized-bed heat exchangers and solar particle
receivers, the complexities and computational costs of particle-
flow modeling at necessary length and time scales support further
careful experimental studies to measure hw for a wide range of
conditions and develop reliable correlations that support model
calibration as well as design of particle receivers and fluidized
bed heat exchangers for CSP and other thermal energy storage
applications.
The heat transfer studies for fluidized beds interacting with ver-
tical walls have generally explored particle flow regimes where the
net particle flow is in co-flow upward with the gas in a large flow
path. An alternative novel approach presented in this study
involves net-downward, continuous particle flow in counterflow
to the fluidizing gas in a narrow vertical channel. In this approach,
732 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745

Table 1
Measured heat transfer coefficients between particle fluidized beds and vertical and inclined surfaces.

Refs. Materials dp (lm) Surface hw (W m2 K1)

[30] SiC 63.9 Vertical wall 430–1120

[18,36] SiC 63.9 Vertical tube 150–1116
[50] Corundum 730–1410 Vertical wall 157–400
[51] Al2O3, SiC 280–1000 Vertical wall 250–725
[49] Glass, ZrO2 491–750 Vertical wall 100–300
[52] Sand 150–350 Inclined surface 60–450

the particulate HTF flows from the top to the bottom of the bed
where the particles exit and are naturally separated from the
upward flowing gas. The inherent gas-particle separation with this
approach would enable more effective particle collection with
reduced heat loss for TES subsystems. The viability for such a flu-
idized bed concept as a central particle receiver depends in large
part on achievable hw at receiver operating conditions in this flow
configuration. Experiments demonstrating this particle-gas coun-
terflow regime as presented here help to assess the operating win-
dow of this flow configuration and to determine if the high hw
observed in fluidized beds with upward particle flow configura-
tions can be matched in this net-downward flow configuration.
This study presents experimental heat transfer measurements
in a lab-scale 10 cm tall
10 cm wide
0.64 cm deep fluidized
bed to assess effective flow regimes and associated hw for net
downward particle flow. The experiments used high-temperature
Fe-doped aluminosilicate particles (Carbo Accucast ID50), with dp
between 150 and 350 lm, as a viable particulate HTF and TES
media in high-temperature CSP applications. The experiments Fig. 1. Major components of particle receiver heat transfer test stand. Detailed view
reveal valuable relationships between hw and important operating of near-wall fluidized bed receiver with flow path diagram showing particle and gas
parameters including bed temperatures T b , solids volume fraction flow paths.
ab , and superficial gas velocity U g . These trends provide a basis
for identifying operating conditions where such a flow configura-
tion may achieve adequately high hw for high efficiency solar recei-
p ¼ 0:754 based on packed bed (and not individual particle) mea-
surements [20,32].
vers and for effective recovery heat exchangers as part of a TES
The particles were stored in a supply hopper above the fluidized
sub-system in a CSP plant. The results are fitted quite well to a cor-
bed channel and preheated to desired particle inlet temperatures
relation modified from the work of Molerus [1,2] with an addition
T b;in as high as 600  C. During continuous particle-flow experi-
of the contribution from the radiative wall-to-particle exchange.
The resulting correlation provides a basis for future work to assess ments, the particles flowed out the bottom of the hopper through
potential receivers and heat exchanger designs based on counter- an orifice with a sliding gate that provided particle flow rate con-
flow fluidized beds. trol. As particles entered the narrow-channel test region, gas
injected up from the bottom of the bed through orifices in a distrib-
utor across the width of the channel provided particle fluidization.
This fluidizing gas flow exited the narrow-channel via a mesh on
2. Experimental methods
the upper region of back wall.
A lab-scale fluidized bed was designed to simulate flow condi-
tions in a single panel of a central receiver with a reduced incident 2.1. Particle flow path
flux that is spread from the aperture onto walls angled to the nor-
mal solar irradiation entering a receiver cavity. The reduced flux on All materials in contact with the preheated hot particles are
the angled wall can be simulated in an experimental setup by irra- stainless steel 304 or 316 unless otherwise noted. A 4-liter V-
diation of a vertical panel with near-IR quartz lamps. The particle shape, particle-feed hopper provided gravity-driven, particle flow
and gas flows for the fluidized bed on the opposing side of the into the vertical test channel. Six 300 W cartridge heaters inserted
panel or wall extract the heat flux through the irradiated wall. into tubes that are welded across the width of the hopper provided
Fig. 1 highlights the primary components of the experimental rig preheating of the particles. A unique split-sheath cartridge (Dalton
and the particle and gas flow paths, all of which are described Electric) allowed reliable operation to heater temperatures close to
below and in more detail in a supporting thesis [60]. The vertical 1000°C, but the particles where never heated above 600°C. The
flow path geometry presents many degrees of freedom, but the hopper was heavily insulated with external aluminosilicate fiber
focus in this study on the wall-to-particle heat transfer led to use to minimize heat loss and loads on the heaters. Fins on the heater
of a single plate geometry. Carbo Accucast ID50 particles (75% insertion tubes distributed the heat into the stored particles. Tem-
Al2O3, 11% SiO2, 9% Fe2O3) with high sphericity,  0:9 [32] and a perature controllers (Watlow) delivered power to the cartridge
mean Sauter dp ¼ 260 lm were used for all experiments, and this heaters, with several K-type thermocouples (TC’s) inserted into
mean Sauter value was used for dp in subsequent calculations. the hopper bed to monitor T b;in .
Properties of the oxide particle composition include: density At the bottom of the hopper, a plate attached to a lead screw
qp ¼ 3230 kg m3, specific heat capacity C p ¼ 1:218 kJ kg1 K1, served as a gate for the particle flow exiting the hopper. This gate
thermal conductivity kp ¼ 0:7 W m1 K1 and radiative emissivity was fully opened or closed during testing and not used to modulate
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
particle flow rates. Rather the inlet flow was modulated for differ-
ent runs by plates inserted into a rectangular pocket cut into the
top of the fluidized bed channel just below the gate. Various orifice
plates could be inserted to vary the set particle inlet flow rate. Pre-
vious studies on particle flow from an orifice under gas counter-
flow showed that the particle flow can stagnate due to forces
from the gas flow countering the gravitational forces, which results
in particle bridging [61]. To avoid this, a single rectangular hole (as
wide as 3.2 mm) was used as the orifice, rather than a narrow slit
or distributed holes across the width of the receiver. Because the
fluidized particles behave like a fluid, particles entering from the
single orifice were distributed quickly across the width of the
6.35 mm wide vertical channel as illustrated in the schematic in
Fig. 2(a). A simple Beer’s law analysis of the 6.35 mm width was
shown to be optically thick for the bed down to a solids fraction
ab ¼ 0:10 based on the equations for transmissivity for particle
beds [62]. Particle inlet temperatures are measured with a K-type
TC just above this orifice where the particles are moving as a
packed flow.
Particles exited the bed through another orifice plate on the
bottom of the bed. The size of this central exit orifice was adjusted
to provide a desired volume of particle holdup in the bed as the
exit mass flow was proportional to the pressure build-up, which
increased with fluidized bed height. A variable gate controlled
the orifice size to allow the particle bed height to increase, and
then the gate opening was tuned to a final size such that the aver-
age bed height remained relatively constant amidst periodic spout-
ing during steady-state operation. A K-type TC located in the dense
particle bed flow just above the outlet orifice provided particle exit
temperatures T b;out . Particles discharged into a collection hopper
that rests on a load cell (Loadstar Sensors), calibrated to 10 kg with
a resolution of 10 g. The increase in the collection hopper mass
with time provided a measure of the particle mass flow rate ðm _ p Þ.
2.2. Gas flow path
Fluidizing gas delivered through a Brooks 5850E (up to 38 slpm)
mass flow controller (MFC) was preheated with a large cartridge
heater heat exchanger. The hot gas from the preheater flowed into
Fig. 2. (a) A vertical slice through the depth of the fluidized bed showing the locations of TC’s and pressure transducers and the basic geometry of the gas and particle flow
paths in continuous flow operation. The schematic shows the locations of TCs to measure temperature drop across the wall ðDT w Þ for calculating local heat fluxes into particle
bed and the locations for pressure taps to get the average solids fraction in the bed ðab Þ. (b) Geometry looking across the width of the channel to the back plate showing the
location of the mesh and the geometries of the gas and particle flow paths.
733
an Inconel 6.35 mm diameter tube with 50 holes (0.5 mm in diam-
eter), which distributed the gas injection evenly across the bottom
width of the bed. At one end of the injection tube, a K-type TC
inserted through a cap measured the gas inlet temperature (T g;in )
and provided a control signal for the preheater. The 50 holes dis-
tributed the gas flow uniformly across the width of the bed based
on visual inspection and the minimal temperature difference
(DT g;in  4  C). The 38 slpm mass flow controller provided superfi-
cial velocities (U g ) up to 1.15 m s1 at ambient conditions and
much higher U g at elevated temperatures.
The vertical, parallel walls of the test section constrained the
upward gas and net-downward fluidized particle flows. The stain-
less steel backplate has a rectangular section cut out and covered
with a 316 stainless mesh (woven cloth – 120
120 mesh with
0.9 mm diameter wire). The mesh allowed upward fluidizing gas
flow to exit the back of the channel. Various heights of mesh were
tested and it was found that a robust fluidized bed could be sus-
tained up to the bottom of the mesh as illustrated in the flow visu-
alization image of the rig in Fig. 3. The final test configuration of
the mesh shown in Fig. 2(b) occupied approximately the top 20%
of the channel height. If the expanding bed reached the bottom
of the mesh region before becoming fully fluidized, the gas
bypassed the rest of the bed through the mesh, creating unstable
fluidizing behavior. This resulted in defluidization of most, if not
all of the bed. Thus, the bed height was controlled by the particle
exit orifice gate to insure that the average bed height did not rise
above the bottom of the mesh. Behind the mesh, the gas exhausts
through three 6.35 mm tubes out the back of the receiver. Gas out-
let temperature was measured from the middle exhaust tube, very
close to the backside of the mesh.
2.3. Heat transfer measurements
Front wall materials included a TC-instrumented aluminosili-
cate plate (0.32 cm thick) for heat flux measurements or a
fluorine-doped tin oxide (FTO) glass (0.30 cm thick) for flow visu-
alization. Electrical conductivity of the FTO glass mitigated con-
cerns of particle charging affecting the flow and permitted clear
sight during flow visualization tests as shown in Fig. 3. The front
734 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
Fig. 3. A frontal view of the fluidized bed operating in continuous particle-flow
mode with the FTO glass for flow visualization. Image shows location for particle
inlet flow through a central orifice at the top of the channel and for particle outlet
flow out an orifice through the bottom plate of the channel. Upward pointing
arrows indicate the distributed fluidizing air injection from the gas distributor. The
dashed lines indicate the location of the woven mesh for gas exit at the top of the
back wall of the channel.
walls were mounted with ceramic paper seals under compression
around the edges so that the plates could be readily interchanged.
The aluminosilicate plate was machined with four grooves (0.5
mm wide by 0.5 mm deep) on both the internal and external sur-
faces. Eight fast-response K-type TCs (0.5 mm diameter) were
placed in the grooves and potted in place with an alumina paste
to insure good thermal communication with the plate. The TC junc-
tions were placed at the end of each groove 3.2 cm from the side
walls. Two pairs of opposing TCs were positioned 3.2 cm from
the bottom of the bed cavity and the other two pairs were located
3.2 cm from the top of the cavity as illustrated in Fig. 2.
For heat transfer measurements, four near-IR quartz lamps
(Fannon FP-10081G) controlled by a phase-angle controller with
a solid-state relay heated/irradiated the external face of the alumi-
nosilicate plate. The lamps had just over 10 cm of heated length,
and each lamp provided up to 500 W of radiation for 2 kW of radia-
tive heating on the external face. The lamps were located in a plane
2.5 cm from the front plate per manufacturer recommendations for
achieving the most constant and uniform heating profile, minimiz-
ing the variation in flux concentrations across the external wall
surface. The variation in surface heat fluxes into the wall q00w were
from top to bottom with variations up to 3.0 W m2. The top to
bottom variations did not influence the experimental analysis
since the hw at the top and bottom regions of the bed were ana-
lyzed independently. Aremco 840-CMX coating was applied to
the external face of the aluminosilicate to enhance emissivity. A
quartz window between the quartz lamps and front plate elimi-
nated forced convective losses from the lamps’ cooling fan. A top
and bottom shroud minimized natural convection losses from the
external face of the front plate. The opposing pairs of thermocou-
ples measured the local temperature drop across the wall
ðT w;ext  T w;int Þ and provided a means to calculate the local conduc-
tive q00w through the wall into the particle bed by Eq. (1).


T w;ext  T w;int
q00w ¼ kw ð1Þ
DxTC
DxTC is the distance between the two TC’s in the wall and kw is the
thermal conductivity of the aluminosilicate which is relatively con-
stant (kw ¼ 2:0 W m1 K1) over the range of temperatures tested.
With this low kw , temperature drops across the wall thickness –
ðT w;ext  T w;int Þ P 10  C – provided a relatively accurate measure of
local q00w according to Eq. (1). High-temperature metal alloys and
stainless steels have conductivities closer to 20–25 W m1 K1 at
600 °C, and as such, would not provide adequately large tempera-
ture differences to mitigate noise in the TC measurements for accu-
rate q00w .
Two TC’s inserted into the bed at heights consistent with the
wall TC heights provided local particle bed temperatures (T b ).
Gas temperature (T g ) and T b equilibrate within a very short dis-
tance from the gas distributor such that the bed-inserted TC’s are
assumed to give a relatively accurate assessment of the local T b
[50]. With the local T b , local wall-to-particle heat transfer coeffi-
cients hw are calculated from Eq. (2).
q00w
hw ¼
 ð2Þ
T w;int  T b
hw measurements at both the lower height (subscript bot) and the
upper height (subscript top) in the bed channel were taken at both
the left (subscript 1) and right sides (subscript 2) of the bed for a
wide range of T b and T w . Under some flow conditions, the bed
height was not high enough to have the top measurements within
the bed and under such conditions hw;top values were quite small
due to the low particle concentrations. As such, those hw;top values
were not included in the assessment of fluidized bed correlations
for hw .
To assess the confidence in hw measurements, a Kline-
McClintock error analysis was conducted considering each of
the variables used in the calculation of Eq. (2). These include ther-
mal conductivity of the front plate, kw , distance between the front
plate TC’s, DxTC , and the relevant temperatures, T w;ext ; T w;int and
T b . The TC measurements have an accuracy of 1:1 C, while kw
has an accuracy of  0.1 W m1 K1. The accuracy of the place-
ment of the DxTC was 0.05 mm. Calculated hw from Eq. (2) were
most sensitive to the temperature measurements and particularly


so at lower q00w , when T w;ext  T w;int across the plate was small
compared to the thermocouple accuracy. At higher q00w , the uncer-
tainty is reduced. Although the total hw uncertainty based on the
error analysis were large for smaller q00w measurements, the
repeatability of the hw values for any set of conditions, as shown
in the results below, gives confidence in the measurements even
for smaller q00w ,
2.4. Pressure and solid fractions measurements
Pressure taps in the side walls along the height of the fluidized
bed channel provided a means for calculating bed solids fractions
(ab ). In bubbling fluidized beds as tested here, pressure drop
through a bed can be related to U g which is the superficial velocity
of the fluidizing gas based on the entire channel cross-sectional
area independent of the gas void fraction, ð1  ab Þ. Increases in
U g from 0 expand the bed upward, and the pressure drop DP b
through the bed per unit bed height Dyb increases according to
the well-known Ergun Equation (Eq. (3)) up to the point of incipi-
ent fluidization.
DP b a2b lg U g ab qg U 2g
¼ 150 þ 1:75 ð3Þ
Dy b ð1  ab Þ3 d2p ð1  ab Þ3 dp
where qg and lg are the density and dynamic viscosity of the gas
phase.
U g at the point of incipient fluidization is called the minimum
fluidization velocity, U mf . Increases in U g above U mf do not further
raise DPb per unit Dyb , which remains constant in the bubbling flu-
idized bed regime. Rather, further increases in U g increase DP b and
Dyb proportionally. When the bed is fluidized ðU g P U mf Þ; DPb is
determined by the weight of the particles based on the following
relationship with Dyb
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
DP b  
¼ ab qp  ð1  ab Þqg g
Dyb
where qp is the solid density of the particles. Since qp is typically
more than 1000 times qg , the second term in the parentheses in
Eq. (4) associated with the gas weight can be neglected. With that
approximation, Eq. (4) can be rearranged to provide the ab as a
function of an experimentally measured DPb .
DP b
ab ¼
qp g Dyb
Equating the right hand sides of Eqs. (3) and (4) provides a quadra-
tic expression for estimating U mf as a function of bed and gas prop-
erties, which depend on T b . As such, U mf depends on T b and in fact,
drops significantly with T b , The equation for U mf requires knowl-
edge of ab;mf just before fluidization. Flow visualization experiments
determined when DP b per unit bed height Dyb stopped increasing
when U g ¼ U mf ¼ 0:25 m s1 at T b ¼ 25  C. That DPb measurement
provided a means to calculate ab;mf ¼ 0:46, which was assumed to
be independent of T b . This value was used to fit U mf as a function
of T b for all other measurements. The calculated U mf ¼ 0:13 m s1
at T b ¼ 600  C shows the importance of considering the variation
of U mf with T b in relevant calculations.
Increases in U g too much above U mf results in slug flow where
gas bubbles in the bed coalesce to form slugs that push particles
out of the way like a piston. Even further increases in U g results
in transitions to turbulent fluidization where the large bubbles or
slugs break apart with chaotic motion of the bed. With turbulent
fluidization, DP b fluctuates erratically but also begins to decrease
because of the significant drop in ab . In the current study, all tests
were performed with U g in the bubbling fluidized bed regime
where ab is higher. Higher ab improves hw and as such, provides
superior heat transfer at lower gas flow power requirements.
The three pressure taps on one of the side walls as shown in
Fig. 2 measured DP b and thus gave average ab according to Eq.
(5) as a function of Dyb between the taps. Differential transducers
(Setra) with a range of  2500 Pa provided adequate resolution for
the small DPb . The taps were located at heights above and below
the DT w measurements as illustrated in Fig. 2. The pressure tap
locations provided localized average ab in the regions close to
the q00w measurements and therefore provided a basis for correlat-
ing hw with ab as discussed in the Results section.
2.5. Data acquisition and testing procedure
A National Instruments cDAQ-9174 with analog input and out-
put modules processed TC and pressure-transducer signals and
interfaced with National Instruments LabView software. TC signals
were amplified using a 10 volt Dataforth signal conditioner (model
SCM7B37). A custom LabView virtual instrument (VI) captured the
data and provided controls for the lamp power, gas preheater, and
the gas electronic mass flow controllers. The VI read data at 240 Hz
and averaged every 40 samples, saving a reading every 1/6 of a sec-
ond. The MFC control was calculated real time in order to maintain
a desired U g , which was calculated from qg determined from the
ideal gas law and T g;in measurements.
Two modes of heat transfer experiments were conducted: a par-
ticle batch mode and a continuous particle-flow mode. Batch tests
started with a fixed quantity of particles in the fluidized bed chan-
nel. The bottom and top of the bed was sealed so that the particles
do not exit the channel. Air, as the fluidizing gas, was preheated
with an inline electric flow heater to the desired T g;in for testing.
Power to the quartz lamps was slowly ramped up to minimize
thermal stresses on the aluminosilicate plate and other compo-
nents. The particles began to heat up due to the incoming gas
735
and flux through the ceramic plate. Data was logged as the desired
ð4Þ
particle temperature was approached, ramping lamp power until a
steady state was attained. An approximate steady state was
defined when T g and T b were within 5  C of the desired tempera-
ture and maintained steady state within 2  C for more than 60 s.
Continuous tests started by preheating particles in the top hop-
per. As the particles approached the desired T b;in , the quartz lamps
were ramped up similar to the batch tests, preheating the test
region to the particle temperature in the hopper. Fluidizing gas
ð5Þ was also preheated during this time. Once T b;in reached the desired
set point, the gate opened to flow particles into the narrow-
channel test section. For continuous flow tests, the top orifice
gap was set to 3.2 mm wide by 4.0 mm long. The bottom gap
was varied in size depending on DPb . A bottom orifice 1.5 mm wide
by 10 mm long provided relatively steady bed operation for the
corresponding top orifice. Initially, a large fraction of the particles
passed straight through and collected in the lower hopper. The sec-
ondary gate at the bottom of the test section was closed momen-
tarily to build up particles inside. DP b measurements indicated
when ab reached a desired value, at which point the secondary gate
was re-opened and adjusted until m _ p , and (T b;out  T b;in ) became
steady. Particles passing through the bed transferred heat from
the irradiated aluminosilicate plate. Unlike the batch tests, the
quartz lamp power remained near maximum power for continuous
flow tests in order to maximize heat into the particles.
Achieving steady particle flow for the counterflow fluidized
beds with automated real-time controls requires further develop-
ment beyond the scope of this current study. However, because
these studies focused on assessing hw , which was similar for batch
tests and continuous flow tests at similar U g and ab , the majority of
testing to measure hw was conducted with batch tests. A smaller
sample of continuous tests were conducted to verify that similar
hw were obtained at the same U g and ab and further, to provide a
basis for understanding how to operate a continuous flow fluidized
bed with upward gas flow and net downward particle flow.
3. Results
Experiments were conducted with varying T b ; U g , and ab in
order to determine how these parameters affect the wall-to-bed
heat transfer coefficient hw . The experimental conditions were
designed to find a correlation to predict hw as a function of exper-
imentally derivable parameters. T b and T w measurements from all
experiments were processed at each of the four locations
(T b;top;1 ; T b;top;2 ; T b;bot;1 , and T b;bot;2 ) to provide local hw . The mean
hw for the two top and the two bottom locations of the bed gave
an average for the specific bed height, which was matched to local
bed conditions for correlation fitting. Reported values of hw include
all modes of heat transfer from the internal wall to the particle bed.
These modes include particle convection hw;pc , gas convection hw;gc ,
and radiation hw;r .
Many test conditions were run with particle build-up such that
very small and/or fluctuating ab < 0:05 in the top third of the chan-
nel where the top measurements were taken. Thus, the bulk of the
data analysis here focuses on the bottom of the bed where for most
conditions, ab P 0:2. Results from the bottom of the bed indicate
that batch mode tests have similar hw to those from the continuous
particle-flow experiments. The continuous flow tests are presented
here first because they reveal interesting characteristics in terms of
operation and hydrodynamics of the unique flow configuration of a
counterflow, narrow-channel fluidized bed. Flow visualization of
the bubbling bed using FTO glass on the front wall also showed
the similarity of the batch and continuous flow modes and pro-
vided insight into the observed trends for bed operation and local
hw measurements. Plots of hw versus each of the varied operating
736 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
parameters provided the basis for comparing the results to existing
correlations for bubbling fluidized bed heat transfer. The extensive
results provided adequate data to modify a correlation from
Molerus [2] to fit the measurements.
3.1. Continuous flow tests
For the continuous flow testing, U g was maintained at 0.5 m s1
for all continuous particle flow tests. U mf = 0.25 m s1 at room tem-
perature and decreases approximately with l1 g as T b increases.
Based on the ratios of l1
g ; U mf decreases to approximately half of
the room temperature value. As such, U g ¼ 0:5 m s1 ensures flu-
idization for these tests. Particle mass flow rates (m _ p ) fluctuated
by approximately 10% during steady-state flow as illustrated
by the transient plot of m _ p for a continuous particle flow test in
Fig. 4. The rapid fluctuations in the particle mass flow rate arise
from the differentiation of the exit hopper mass as a function of
time. T b;in varied over a range of about 40  C during each test due
to temperature gradients in the upper feed hopper. But for all con-
tinuous flow tests, steady conditions were achieved for 10 min or
more for each test as illustrate in Fig. 4.
Particles entered the receiver from the small central orifice
below the top hopper and exited from the orifice below the flu-
idized region as shown in Fig. 2. Even with the limited size of the
particle inlet and outlet, fluidization across the 10 cm width of
the channel was uniform and vigorous. Room-temperature visual-
ization of the continuous particle flow as represented by the sam-
ple bed snapshots on the top row in Fig. 5 showed consistent
formation of sizable bubbles of 1–2 cm in length rising through
the bed and disrupting the top bed surface as the gases escaped
out of the bed through the rear wall mesh. Nonetheless, the net-
downward flowing particles behaved much like a fluid and rapidly
distributed across the width of the channel, even with the rela-
tively narrow inlet and outlet orifices for the particles. The gas
injection through the many orifices along the length of the gas dis-
tributor at the bottom of the bed (as illustrated in Fig. 2(a)) dis-
tributed bubble formation well across the width of the bed. The
spouting at the top bed surface as illustrated in Fig. 5 was not sig-
nificantly impacted by the inlet or outlet conditions of the particle
flow.
Fig. 4 shows the m _ p and the particle mass accumulation in the
collection hopper for one of the test conditions with continuous
counterflow particles. Upon reaching near steady state conditions,
m_ p remained at 3.0 g s1 to within 10% for over 1500 s of the
entire experiment. Once the bed height, m _ p , and T b and T w values
stabilized, hw for the bottom two wall locations as calculated from
Eq. (2) stabilized as indicated in Fig. 6(a). In the bottom half of the
bed, ab stayed at 0:34  0:01 during steady state conditions, and
this sustained relatively constant and high hw around 600 W m2
Fig. 4. Mass accumulation and flow rate (m_ p ) for a continuous test example. m
_ p is
steady near 3 g s1. (U g ¼ 0:5 m s1 ; ab = 0.25–0.35, T b = 340–380 °C,
wo;t ¼ 3:175
4 mm)
K1 with very close agreement between the right and left halves
of the bed (as depicted in Fig. 6). On the other hand the top T w
measurements were very near the height of the top surface of
the bed, and thus they were subject to lower average ab and with
much larger variability due to spouting at the top of the fluidized
bed. The much lower ab in the top region (average  0:1) led to lar-
ger fluctuations in hw with lower mean values around 300 W m2
K1 (as shown in Fig. 6). Furthermore, when ab in the top of the bed
declined near the end of the tests, hw also dropped showing the
strong correlation between hw and ab . The smaller and more fluc-
tuating hw and ab in the top region of the bed were observed for all
continuous particle-flow tests. Thus, assessment of hw as a function
of ab and U g for the continuous flow tests focused on measure-
ments in the bottom half of the bed.
3.2. Batch tests
To corroborate simpler and more timely particle batch mode
experiments with the continuous particle-flow tests, a subset of
batch experimental test results with a similar range of ab ; U g ,
and T b were compared to the continuous particle flow measure-
ments. For the batch mode tests, the top and bottom orifices for
particle flow are blocked so that particles did not flow in or out
of the vertical channel. The gas flow path still exited through the
back mesh after bubbling up through the bed as in the continuous
particle flow tests. Fig. 7 shows that batch mode and continuous
mode tests follow very similar trends with good agreement of hw
for similar T b for each range of ab . In addition, Fig. 5 shows that
the nature of the bubbling fluidization, spouting, and bed height
were very similar for the continuous particle flow and for batch
mode at conditions with similar ab and U g . Because the continuous
particle flow fluidization is well represented by batch mode exper-
iments, a majority of results were taken in batch mode as discussed
in this section.
The error bars in Fig. 7 represent the uncertainty associated
with hw from the Kline-McClintock error analysis. As discussed ear-
lier, the largest uncertainty occurred at the smallest values of q00w


for the batch mode tests at lower T b when T w;ext  T w;int was
small. Nonetheless, the repeatability of hw derived from the mea-
surements at the lower T b as shown in Fig. 7 provided confidence
in the measurements even at those conditions.
Batch mode experiments were conducted with
100  C 6 T b 6 670  C and 0.25 m s1 6 U g 6 1:7 m s1. ab varied
by changing the batch volume of particles placed into the receiver
and by varying U g . Different static bed volumes provided tests over
a range of ab : 0:1 6 ab 6 0:45. U g was limited to 6 1:7 m s1 as
further increases in gas velocity resulted in defluidization and
gas channeling through the bed. The lack of a disengagement or
freeboard zone at the top of the bed is thought to be the cause of
this behavior. This was consistent with observations from the con-
tinuous particle flow tests, where high values of U g > 1:7 m s1
also resulted in the fluidizing gas channeling through the bed
before exhausting out the mesh, leaving defluidized regions of
solids held up at the top of the bed.
Fig. 8 shows the transient evolution of local hw for all four loca-
tions and the average T b for a batch mode test leading up to steady
state operation when the heat loss through the back and side walls
balances the heat input from the lamps. Temperature and pressure
measurements were averaged every 30 s in order to smooth out
fluctuations due to the bubbling fluidization. At the beginning of
the batch mode experiments, the IR lamps were turned on and
T w;ext was much higher than T w;int while the aluminosilicate wall
heated up transiently, and the apparent hw from Eq. (2) at early
times were artificially high as the equation does not include
the transient wall heating. As the external wall approached
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745 737

Fig. 5. Visualization of bubbling fluidized beds with FTO glass for flow visualization operating in continuous flow and batch flow tests shows. The three images for each mode
are separated by 1 s each and show the similarity in fluidization characteristics between the two modes. Downward facing arrows indicate particle flow. Upward facing
arrows indicate gas flow.

Fig. 7. Comparison of measured hw vs. T b for continuous particle flowing tests

(solid symbols) and batch mode tests (hollow symbols) over a range of
0:25 6 ab 6 0:35 and 0:35 6 ab 6 0:55. Results show good agreement for hw
between the two modes. U g varies for batch tests, but is fixed at 0.5 m s1 for
continuous. Error bars are derived from a Kline-McClintock analysis.

Fig. 6. Transient evolution of heat transfer coefficient (hw ) and bed solids fraction
(ab ) for a continuous particle flow test with constant U g ¼ 0:5 m s1 and
T b ¼ 340—380  C: (a) at the two wall locations in top third of the bed and (b) at
the two wall locations in the bottom third of the bed. The measurements in the
bottom region demonstrated more stable and higher hw due to the more consistent
and higher ab than in the top region, which was impacted by spouting at the top
bed surface.

steady-state conditions, hw dropped toward its asymptotic steady-

state values as illustrated in Fig. 8. Typically, steady-state local hw
were taken from values averaged over 60–120 s after 900 s of heat-
ing. The average U g ; T b , and calculated ab (from the pressure tap
measurements and Eq. (5)) during those 60–120 s provided the Fig. 8. Transient data for the local hw from a typical batch test approaching a
conditions for hw used in subsequent plots and correlation analysis steady-state T b ¼ 100  C and U g ¼ 0:45 m s1.

for hw .
Fig. 9 shows the steady-state hw plotted as a function of average Clusters of points close together for similar T b and U g values in
T b for a range of U g values up to the 1.7 m s1. hw increases in an Fig. 9 are due to variations in ab , which were controlled by chang-
approximately linear fashion with respect to T b for a fixed U g . ing the initial batch particle volume loading. Fig. 9 clearly shows
738 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
Fig. 9. Average heat transfer coefficient (hw ) as a function of T b from batch
experiments for bottom of bed. 0:1 6 ab 6 0:45.
that increases in U g result in decreases in hw . The hw data is replot-
ted with respect to U g in Fig. 10. The decrease in hw with increasing
U g lessens for U g above 1.0 m s1.
The dependence of hw on T b shown in Fig. 9 agrees with previ-
ous fluidized bed studies [50], but many fluidized bed heat transfer
(Nusselt number, Nu) correlations in the literature limit tempera-
ture dependence through gas-phase properties and principally
gas-phase thermal conductivity kg , which appears explicitly in
the definition of Nu [50,51,63–65]. For the small particle size in this
study (dp < 0:35 mm), it can be assumed that the gas temperature
was equilibrated with the local T b . The gas between the particles in
the bubbling fluidized bed is the strongest resistance to particle
convective heat transfer characterized by hw;pc . The heat transfer
resistance between the particles is reduced by the parallel pathway
of particle collisions, but with ab < 0:5 in the bubbling fluidized
bed, gas-phase mediated exchange between particles remains the
dominant resistance to convection away from the wall. The gas
phase resistance decreases with increasing kg as T b increases. This,
in part, explains the increasing hw with T b as shown in Fig. 10. Gas-
phase convection independent of the particles (characterized by
hw;gc ) does not usually account for a substantial fraction of total
convection, and thus hw;pc is the primary contribution to the con-
vective portion of the total hw . Thus, because the particle and
gas-only convective components could not be distinguished
through the measurements, hw;c is written henceforth to refer to
the entire convective portion of the total wall-to-particle hw .
The measurements of hw did not distinguish between convec-
tive and radiative heat exchange between the wall and the parti-
cles. The contribution from radiative exchange becomes
significant at higher T b . Radiative exchange is characterized by
an equivalent radiative wall-to-bed heat transfer coefficient hw;r
which can be approximated by Eq. (6).
 
hw;r ¼ rw;p T 2b þ T 2w ðT b þ T w Þ ð6Þ
where w;p is the system emissivity involving the wall and bed
emissivities.
Fig. 10. Average heat transfer coefficient (hw ) as a function of U g from batch
experiments for bottom of bed. 0:1 6 ab 6 0:45.
Fig. 11. (a) The radiative heat transfer coefficient hw;r as a function of T b (red
circles) and T w (blue squares). (b) Plot of ratio of hw;r to total hw as a function of T b .
(For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
1 1
¼
 ð7Þ
w;p 1=p þ 1=w  1
Eq. (7) represents the maximum possible heat transfer between the
bed and the wall by modeling the bed as a surface [66], and this has
been used to model wall-to-bed radiative heat exchange in many
studies [67,30]. Studies that incorporate scattering and absorption
in the particle bed have shown that the effective bed emissivities
at bed densities similar to the counterflow fluidized bed here are
generally not less than  90% of the measured bed emissivity
[68]. As such, Eq. (7) represents a good approximation and was used
here to calculate hw;r particularly in light of the fact that hw;r was
always < 15% of the total hw .
Fig. 11 plots the calculated hw;r for the range of experimental
text conditions as a function of T b and T w with p ¼ 0:754 and
w ¼ 0:78. At the highest tested T b  650  C, hw;r was  14% of
the total hw . Previous studies that considered contributions of
hw;r have shown similar trends and values for hw;r [69,70].
For the dependence of total hw on U g , literature suggests that hw
peaks at U g values just above U mf when the bed is fluidized. For the
tests in this study, U mf is approximately 0.25 m s1 based upon
when measured DDyP reaches its peak value. For this study, all
b
reported values of U g are above U mf , and hw deceases with increas-
ing U g for the test conditions as shown in Fig. 9.
The decrease in hw with increasing U g above U mf results from
the reduction in ab due to the increased frequency of gas bubbles
and the higher effective bed height Dyb in the narrow-channel flu-
idized bed. In wider beds, bubbles tend to coalesce toward the
middle of the bed away from the walls [71], but with the narrow
vertical channel between the front plate and mesh backplate, lar-
ger gas bubbles occupy the entire thickness of the bed. The
increased bubble frequency and size with higher U g exposes the
wall to more time during which the low gas-convective heat trans-
fer coefficient hw;gc becomes the dominant contribution to hw . This
obviously reduces overall hw averaged over time and explains the
trends demonstrated in Fig. 9.
4. Discussion
4.1. Correlation fitting
The literature presents several studies on wall-to-particle heat
transfer for fluidized beds, but to the authors’ knowledge, no
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
studies have explored the heat transfer in the configuration here of
a narrow-channel, counterflow fluidized bed, which may be an
effective configuration for particle receivers in CSP applications.
Many previously reported fluidized bed studies discussed how
overall heat transfer coefficient hw in fluidized beds must incorpo-
rate the multiple modes of heat transfer between the wall and the
particles, including particle (cluster) convection hw;pc , gas-only
convection hw;gc and radiative exchange between the walls and
particles hw;r . The particle convection occurs during the fraction
of time f p when clusters or packets of particles contact the heat
transfer surface. During the other time, the lean gas phase compo-
nent contributes [72]. As such, overall hw can be approximated as a
time-weighted average of each phase, weighted by f p .
hw ¼ f p hw;pc þ ð1  f p Þhw;gc þ hw;r
Since hw;pc is typically much larger than hw;gc , convective heat trans-
fer tends to be enhanced for flow operating conditions that encour-
age higher f p . For example, small increases in U g above U mf for
fluidization increases hw;c , but further increases lower hw;c due to
the increased presence of bubbles and the reduced f p as well as
ab [40,44,43,52,73].
Many studies on heat transfer in fluidized beds for various con-
figurations and conditions have developed Nusselt number (Nu)
correlations for hw . The correlations were generally derived at
low enough T b such that hw;r remained less than 5% of the total
hw . As such, these correlations are designed to fit the variation in
hw;c , the convective portion of the heat transfer coefficient. Chen
discussed the lack of agreement of commonly cited Nu correlations
for hw;c in bubbling fluidized beds with experimental data [72]. As
such, it was important to explore if any of the wall-to-particle cor-
relations worked well for the current study.
For the fluidized bed heat transfer correlations, most Nusselt
number correlations (Nud ) use particle diameter dp as the charac-
teristic length scale and kg as the relevant thermal conductivity
because it is associated with the principal thermal resistance to
convection between particles. As such,
hw;c dp
Nud ¼
kg
The use of dp as the length scale comes from consideration that gas
flows around the particles limit the transfer of heat between parti-
cles away from the wall [71] and in particular, through the thermal
boundary layer near the wall [55,74]. More theoretical length scales
based on contact time with the so-called packet renewal model [75]
require experimental measurements, which reduces their applica-
bility. In general, Nud correlations predict that hw;c increases with
decreasing dp because of the principal dp dependency in the Nud
definition [51,50,40].
Many proposed Nud correlations have a further dependence on
dp through the Archimedes number, which is a non-dimensional
expression of the ratio of gravitational (or buoyant) forces to vis-
cous forces. Many correlations in the literature have correlated
heat transfer in fluidized beds with the Archimedes number,
Arint , which represents the ratio of the body force due to buoyancy
to the viscous force.
dp gðqp  qg Þqg
3 the surface to the bulk includes the resistance through the intersti-
Arint ¼
l2
Molerus showed that this common expression provided good corre-
lations only when the particles are large enough such that the Rey-
nolds number Red (based on dp and terminal particle velocity) is in
the intermediate flow regime between fully laminar and fully tur-
bulent flow [1]. Hence, the subscript was added to Arint to signify
the flow regime. Through analysis of flow conditions with smaller
739
particles where Red is in the laminar viscous flow regime, Molerus
showed that an alternative expression for the Archimedes number
Arlam as defined here in Eq. (11) correlated more effectively with
experimental results in that flow regime.
qffiffiffiffiffiffiffiffi
dp g ðqp  qg Þ
3
Arlam ¼ ð11Þ
l
Molerus established that fluidization regimes can generally be dis-
tinguished by the magnitude of Arlam and Arint . For conditions in this
study where average dp is relatively small, the distinction in the dif-
ferent Archimedes number is important.
A few commonly cited Nud correlations are provided in Table 2,
which presents only a small sample of fluidized bed correlations
ð8Þ from the literatures. Many studies on fluidized bed heat transfer
compare to these correlations, fitting new coefficients or non-
dimensional numbers as needed to provide improved fits to the
specific experimental conditions [52,73,40,45]. A broad review of
these and other relevant studies can be found elsewhere [60],
and a very brief review is given here.
Flamant et al. explored the impact of dp and T b in a stationary
fluidized bed with heat exchange to a flat wall enclosing the bed
[51,50]. This was one of the few fluidized bed heat transfer studies
done at very high temperatures (500  C 6 T b 6 900  C) that repre-
sent conditions for solar receivers that motivate the current study.
As shown in Table 2, Flamant et al. correlated the increase in hw
with T b associated with radiative heat transfer through the Planck
number, N ¼ 3kb;eff ab =8rT 3b , and the effective bed thermal conduc-
tivity can be estimated as a function of kp ; kg , and ab [76]. The cor-
relation by Flamant et al. provided an estimated Nud as a function
of N and Arint but the combination of radiation and convection into
a single product limits its accuracy for the broad range of proposed
applicable Arint 6 1:3
104 (Table 2).
Borodulya et al. [78] performed a comprehensive review of flu-
idized bed heat transfer studies before 1990 and developed a Nud
correlation that included a so-called conduction term through
the particle cluster and a convective term due to the movement
of fresh particles into the near-wall region. The correlation shown
ð9Þ
in Table 2 which has been designed to cover a broad range of Arint
values provides good approximations, but showed poorer relative
accuracy at conditions with Nud < 10, characteristic of the condi-
tions in this study.
Molerus extensively investigated the dependence of fluidized
bed heat transfer on non-dimensional groups for different fluidiza-
tion regimes [80,81,1,2,77] and provided correlations to fit his data
over a range of conditions based on Arlam and Arint as shown in
Table 2. Of particular interest to the current research, his studies
explored the specific regime, referred to as a viscous flow regime,
where fluidized particles fall through the fluidizing gas. This flow
regime identified by Arint < 103 characterized almost all of the con-
ditions tested in this study. From visualization experiments involv-
ing particle mobility, Molerus concluded that particles in this
regime remain near the surface for a relatively long duration and
that particle convective heat transfer is governed by the exchange
of heat from the hot near-wall particles to particles in the bulk of
the fluidized bed. As such, a heat transfer resistance model from
ð10Þ tial gas and the effects of gas-phase viscosity on particle mobility.
Molerus related his model to the Zabrodsky correlation [79], and
then developed comprehensive correlations with respect to Arlam
– the second and fourth Nud correlations listed in Table 2.
For higher Arint P 103 , referred to as the intermediate (between
viscous and turbulent) flow regime, Molerus fitted his data to a dif-
ferent Nud correlation (third in Table 2) that included the increased
effects of gas convection in this regime. Molerus showed that the
740 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745

Table 2
Notable Nusselt number correlations for wall to particle convective heat transfer in fluidized beds at various conditions.

Refs. Correlation
0:14
[50] Nud ¼ 1:14Ar0:12
int N 102 6 Arint 6 1:3
104
3ab kb;eff 0:2 6 N 6 10
N¼ 8dp rT 3b
 
[1,2] k b g Þ 0:54 Ar0:346
Nud 1 þ 2C p;pg l ¼ f ð U Arlam P 300; Arint 6 103
g
lam

Nud ¼ Nup þ Nug 103 6 Arint 6 105

   0:1
k b g Þ 0:69 Arlam qp qg
Nup 1 þ 2C p;pg l ¼ f ð U
g q g

Nug ¼ 0:4257Ar0:2323
int Pr 0:33
g
[77]  
0:125ab;mf Arlam Prg qg 0:33
2=3
Nud ¼ B þ 0:165 qp qg
½1þB2 ðkg =2C p;p lg Þ
1
B3
" 1=3 #1
ðU g U mf Þ4 qp C p;p
B1 ¼ 1 þ 33:3 U mf gkg

 q 1=2 q C 2=3
B2 ¼ 1 þ 0:28a2b;mf ðU g  U mf ÞU mf q gq p p;p
gkg
  p g

U mf
B3 ¼ 1 þ 0:05 U g U mf
q 0:14  0:24
[78]
Nud ¼ 0:74Ar0:1 p C p;p a0:67
ab0:67 þ 0:46Red Prg 1b ab 140 6 Arint 6 1:1
107
int q g C p;g
[79] Nud ¼ 35:7q 0:2 0:4 0:64
p kg dp

effects of excess gas velocity above the minimum fluidization
velocity (U g  U mf ) needed to be explicitly captured in the correla-
tions for both the viscous and intermediate flow regimes. As shown
in Table 2 for his specific correlations for each flow regime, a func-
b g Þ was used to empirically fit the dependency of hw;c on a
tion f ð U
b g as defined in Eq. (12) [82].
non-dimensional excess gas velocity U
 13
bg ¼ qp C p;p
 hw;c (measured hw  hw;r ) from all of the conditions that fell within
U U g  U mf
gkg
The results collected here were compared with several Nud cor-
relations from similar fluidized bed configurations listed in Table 2.
The correlations (with one exception [50]) were generally derived
at low enough T b to consider convective components of hw without
the effects of radiation. Therefore, to compare the correlations with
results in this study, hw;r according to Eq. (6) was added to the hw;c
derived from each correlation to compare with the experimentally
measured hw . In general, no correlation represented the experi-
mental data sets well (presented in Figs. 7, 9, and 10), and most
correlations regularly deviated from the experiments by more than
25% [60].
The data in this study was taken in what Molerus called the vis-
cous regime – Arlam P 300; Arint 6 103 – characteristic of bubbling
fluidization, where hw;pc dominates the heat transfer. Molerus’s
correlation specifically designed for this range of conditions [1,2]
in Table 2 fits the experimental data in this study for one or more
superficial velocity U g very well, but showed an offset from the
experimental data for other values of U g as demonstrated in
Fig. 12 which presents parity plots for hw for different ranges of
b g Þ depen-
U g and of T b . This suggested a modification to the f ð U
dence in Molerus correlation to improve the correlation fits over
b g.
the full range of U
b g was calculated for each test condition using a temperature
U
dependent U mf as derived by setting DP b from Eqs. (3) and (4) equal
to each other and solving for U g ¼ U mf . At room temperature, U mf
was measured to be 0.25 m s1, and this measured value provided
a basis to determine an approximate ab;mf ¼ 0:46 needed in Eqs. (3)
and (4). That ab;mf was assumed constant for all conditions. The cal-
culated U mf decreases non-linearly with T b such that at 600  C, U mf
has dropped to approximately half of its room temperature value,
due principally to the increased lg with T.
The modified f ð Ub g Þ needed to represent more effectively Nud
over the full range of U g in this study must capture the non-
linear monotonic decrease of hw;c with increasing U g  U mf shown
in Fig. 9. This trend agrees with the hw;c in upward flowing particle
fluidized bed modeled in a recent solar receiver study [35]. All con-
ditions in this study had adequately high U b g P 15, such that a
 
b b nU
power-law form for f U g ¼ C U U g fit well the non-linear depen-
b g . Values for C U and nU were found by fitting the
dence of Nud to U
ð12Þ the Arlam P 300 and Arint 6 103 . The fits were done using the non-
linear least squares solver (’lsqnonlin’) in MATLABÒ. For the viscous
flow regime, the fit provided C U ¼ 1:6263 and nU ¼ 0:1366. The
resulting Nud correlation is given here in Eqs. (13).
 
k b 0:1366 Ar0:346
Nud 1 þ 2C p;pg l ¼ 0:8782 U g lam
g
ð13Þ
for Arlam P 300; Arint 6 103 and b g P 20
U
 
b g derived from the experimental fits goes
The function for f U
b g ! 0, i.e. U g ! U mf . The fits works well for the
toward infinity as U
conditions with Ub g P 15, which motivates the lower limit placed
on Ub for the updated correlation. Molerus originally proposed
 g
b
f U g reached a maximum at U b g  40 but as shown in Fig. 9,
 
the data here showed that hw;c and hence f U b g continued to
increase with decreasing U b g down to 15–20. Further testing at U g
closer to U mf , will help to assess the behavior of Nud in that region.
Fig. 13 shows comparisons of the experimental data and this fit-
ted correlation where the radiation heat transfer coefficient hw;r is
added to the correlation hw;c . The comparisons in Fig. 13(a) and (b)
plot the same data with respect U g and T b respectively. The error
between fitted and measured values is < 10% for the vast majority
of data points with the exception of a few points taken at
T b  100  C where the experimental error is largest due to the
smallest T b  T w for the measured hw calculations. Fig. 13(b) indi-
cates that for T b P 300  C, the agreement between the modified
Nud correlation and the experiments improves. For T b P 500  C
where a concentrating solar particle receiver might operate, the
error in the correlation was consistently < 6%, which suggests
the modified correlation may serve as an effective vehicle for
design analysis and studies of particle receivers and heat exchang-
ers designed around this concept of a counterflow, narrow-channel
fluidized bed.
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745 741

Fig. 12. Comparison of Molerus (1992) correlation with experimental data for different ranges of (a) superficial velocity U g and (b) particle bed temperature T b . The 0.45 m/s
superficial velocity data matched very well while data from most other velocities deviated well above 20%. hw;r was added to the correlation to compare with experimental hw .

Fig. 13. Comparison of the modified Molerus (1992) correlation for different ranges of (a) superficial velocity U g and (b) particle bed temperature T b . Correlation was
b g Þ ¼ 0:8782 U
modified by fitting a power law function ðf ð U b 0:1366 Þ for non-dimensional velocity dependence of Nud . Almost all data deviated less than 10%. hw;r was added to
g
the correlation to compare with experimental hw .

4.2. Solar receiver performance
The results from this study on narrow-channel, counterflow flu-
idized beds indicate promise that merits further studies for solar
receiver applications as well as other potential particle heat trans-
fer systems. This promise stems from three key aspects of the
results summarized here.
 The operation of the counterflow arrangement appears robust
over a broad range of conditions with very low gas-to-particle
mass ratios. This suggests good operability within this flow
regime with the potential for high fractions of heat recovery
to the particles.
 Average hw as high as 1000 W m2 K1 implies the possibility of
very high heat fluxes into the particles to reduce the size and
hopefully cost of solar particle receivers or other particle heat
exchanger designs.
 hw follows the modified Molerus correlation very well, and this
presents the opportunity for preliminary receiver or heat
exchanger design over the bubbling fluidized bed operating
conditions.
These results, in particular the very high hw values, prompted an
exploration into the impacts of higher hw on the potential perfor-
mance of a concentrating solar particle receiver using this flow
configuration with high incident radiative heat fluxes.
For a concentrating solar particle receiver in a central tower,
local T b will likely range between 400 and 800°C if the particle
TES subsystem is coupled to supercritical-CO2 power cycles
[10,12]. Since this study demonstrated that a particle receiver
could operate in the viscous bubbling fluidization regime, the
modified Molerus Nud correlation in Eq. (13) provides a good basis
for estimating the convective heat transfer coefficient hw;c for a
range of operating conditions characteristic of the receiver. Add-
ing the wall-particle radiation heat transfer coefficient hw;r from
Eq. (6) to the correlation hw;c provides a basis for estimating the
overall hw for various bubbling fluidization conditions. Fig. 14
plots the predicted total hw over a range of T b characteristic of
a central-tower solar particle receiver for three different dp ¼
160, 260, and 360 lm. The tests in this study were all taken for
a particle mix with a mean Sauter diameter of dp ¼ 260 lm, and
Fig. 14 shows that reducing dp to 160 lm can significantly
742 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
Fig. 14. Plots of total hw (hw;c þ hw;r ) versus T b for different dp . hw;c was determined
b g P 20.
by the modified Molerus correlation for Arlam P 300; Arint 6 1000, and U
The hw;c were calculated for U b g ¼ 40 for different dp which correspond to U g of
0.258, 0.351, and 0.485 m s1 for dp ¼ 160, 260, and 360 lm respectively.
increase hw;c P 200 W m2 K1 over the range of T b . For the cur-
rent oxide particle density, dp ¼ 160 lm is likely near the bottom
of the size range where particles can still be effectively trans-
ported controllably at large scales for a solar receiver and TES
subsystem. All the same, Fig. 14 shows that a bubbling fluidized
bed has the possibility of sustaining total hw well above 1000
W m2 K1.
To assess the impact of the high hw on solar receiver perfor-
mance, a simple analysis of heat transfer through an irradiated
external receiver wall to a counterflow fluidized bed was per-
formed. An important design consideration for a solar receiver
involves maintaining T w below wall material temperature limits,
which for high-Ni alloys such as HaynesÒ 230Ò is likely
< 1100  C [83]. To assess the efficiency of the solar-to-particle heat
transfer for a particle receiver, a proposed receiver design based on
angled panels as shown in Fig. 14(a) provides a basis for assessing
the benefits of the high hw of the narrow-channel bubbling flu-
idized bed. For the receiver design in Fig. 14(a), increased solar effi-
ciency will be achieved if solar fluxes at the front plane are
concentrated as high as possible. Concentrating factors of 1000
suns (i.e., P 106 W m2) are highly beneficial for solar receiver effi-
ciency. For a solar flux of 1000 suns ¼ 106 W m2, an hw ¼ 1000 W
m2 K1 would require an unacceptably high T w;int  T b ¼ 1000  C.
Hence, the design in Fig. 14(a) has angled walls to spread the aper-
ture solar flux by a factor of aflux ¼ 1= sinðhÞ. With aperture fluxes
of that magnitude, normal or near normal locations of the external
receiver wall near the leading edge (in Fig. 14(a)) can reach extre-
mely high values (T w;ext 1100  C), that are not sustainable with
high-nickel alloys like HaynesÒ 230Ò or equivalent metals. Coat-
ings or structures are required at the leading edge to avoid the
excessive T w;ext , and discussion of strategies for leading edge ther-
mal management will be reserved for another study. The current
study considered only the heat transfer and T w;ext over the majority
of the receiver area in the cavity with the normal flux spread by the
angled surface.
Heat transfer through the angled surface for a given aperture
_ solar and local bed temperature T b was calculated by
solar flux q00
simultaneously solving the energy balance on the external and
internal receiver wall surfaces. The two energy balances for the
external and internal angled surfaces of the receiver wall are
shown in Eqs. (14) and (15) respectively.
q_ 00solar kw
 rw  4 
0¼  T w;ext  T w;int  T w;ext  T 4amb
aflux Dxw aflux
 hw;c;amb ðT w;ext  T amb Þ ð14Þ
kw


0¼ T w;ext  T w;int þ hw T w;int  T b ð15Þ
Dxw
where kw ; Dxw ; T amb and hw;ext are the receiver wall thermal con-
ductivity, wall thickness, external ambient temperature, and exter-
nal ambient convective heat transfer coefficient respectively. In Eq.
(14), the angled external wall view factor to the ambient was
approximated as 1/aflux . In the analysis presented here, the wall
properties were held fixed with Dxw ¼ 0:005 m and kw ¼ 25:0 W
m1 K1, which is taken from the high-temperature property data
of HaynesÒ 230Ò [83]. The two above equations can be used as part
of a design study for preferred operating conditions of a counter-
flow fluidized bed heat exchanger, but the preliminary analysis pre-
sented here focused on one condition characteristic of a central
solar receiver with fixed values for T b ¼ 700  C, aflux ¼ 5:0
ðh ¼ 11:4 Þ; T amb ¼ 22  C, hwc;amb ¼ 10 W m2 K1, and w ¼ 0:9.
These allowed for an effective assessment of how the high total
internal hw could benefit the operation of a central particle receiver
designed for coupling to high-temperature TES subsystems and
power cycles like supercritical-CO2 Brayton cycles.
Eqs. (14) and (15) were simultaneously solved using the non-
linear equation solver ‘fsolve’ in MATLABÒ for T w;ext and T w;int with
q_ 00solar and hw specified by the user or for q_ 00solar and T w;int with T w;ext
and total hw specified. The plots in Fig. 15(b) show results from
the former case with T w;ext plotted over a range of total hw feasible
for the counterflow fluidized bed configuration. Fig. 15(b), which
shows T w;ext for different solar concentrations at the specified bed
conditions, indicates the significant value of the high total hw
observed in this study for enabling the receiver to maintain T w;ext
within structural material limits of high-temperature Ni-based
alloys at solar concentrations P 1000. For example, if T w;ext limits
were set at 1000 C, at the fixed aflux ¼ 5:0 and T b ¼ 700  C,
hw P 650 W m2 K1 would permit operation at 1000 suns and
hw P 1100 W m2 K1 would permit operation at 1500 suns.
Fig. 15(b) also plots the solar efficiency gsolar for the different hw
at the specific operating conditions. gsolar is defined by the fraction
of solar energy captured in the particles as given here by Eq. (16).


aflux hw T w;int  T b
gsolar ¼ ð16Þ
q_ 00solar
Losses from the fluidized particles to the gases and back wall of the
fluidized bed were not considered. This was justified by considering
the ratio of gas-phase mass to particle mass flow rates. With the low
fluidization velocities required and the counterflow arrangement,
the energy losses associated with heating the small gas flow has
been estimated to be <0.5% of the rate of energy captured by the
particles in the fluidized bed. This preliminary analysis did not
account for other losses at the receiver leading edge or back walls.
These losses may drop gsolar plotted in Fig. 15(b) a couple of percent-
age points, but they were not accounted for here to avoid losing the
general nature of the results. The plots of gsolar vs. hw do highlight
the significant benefits of higher hw permitting higher solar concen-
trations. Higher solar concentrations at the same T w;ext increases
heat input through the receiver wall with the same heat loss back
to the ambient, which leads to higher efficiencies. In particular,
for a fixed T w;ext ¼ 1100  C, gsolar rises from 85:0% to 90:3% when
an increased hw (from 650 to 1100 W m2 K1) permits solar con-
centrations to rise from 1000 to 1500 suns.
The plots of gsolar in Fig. 15(b) show that as hw > 1200 W m2
K , there is limited benefit to gsolar , but there are still benefits with
1
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
Fig. 15. (a) Cross-sectional schematic view of counterflow fluidized beds arranged at angle to incident solar flux for effective flux spreading to maintain wall temperatures T w
below material limits with efficient heat transfer into the fluidized particle bed. b) Predicted average T w;ext (solid curves) and solar-to-particle thermal efficiency gsolar (dashed
curves) for a range of wall-to-particle hw as a function of solar concentration for receiver configuration in part a). The following parameters were used for the graph: average
T b ¼ 700  C, h ¼ 11:4 ; T amb ¼ 22  C, w ¼ 0:9; kw ¼ 25:0 W m1 K1 ; Dxw ¼ 0:005 m, hw;c;amb ¼ 10 W m2 K1. Solar concentrations based on 1 sun = 1000 W m2.
respect to decreasing T w;ext for improved receiver wall durability.
As such this preliminary analysis provides valuable insight for
designers as to preferred operating conditions for a solar particle
receiver based upon the narrow-channel, counterflow fluidizing
bed concept. The results indicate that the particularly high values
of hw demonstrated in this study could be noteworthy for inform-
ing full-scale receiver design and enabling sustainable receiver
operation. Full-scale receiver design will need to be informed by
further work using high-fidelity CFD models as cited earlier [57–
59] that instruct more fully how the variation in hw over a full-
scale receiver design impacts solar efficiency at different operating
conditions. The current results provides a basis for calibrating such
models and a motivation for further investment in such computa-
tional design tools.
5. Conclusions
This study has demonstrated that narrow-channel, counterflow
fluidized beds can achieve wall-to-particle total heat transfer coef-
ficients hw up to 1000 W m2 K1 for mean particle diameters of
260 lm. Such high heat transfer coefficients were achieved at ele-
vated particle bed temperatures (up to 650  C in the tests in this
study) and at superficial gas velocities no more than twice the min-
imum fluidization (U mf  0:12 m s1 at particle temperatures of
600  C). The wide range of hw measurements over a broad distribu-
tion of temperatures and gas velocities were performed both in
continuous particle flow and transient particle batch mode, and
fluidizing conditions fell within the so-called viscous fluidization
regime of Molerus [1]. The measured total hw results minus the
radiative wall-to-particle contribution were captured very well
by a modified version of Molerus Nusselt number correlation based
on laminar Archimedes number Ar lam [1,2]. The correlation was
modified by a function of a non-dimensional excess superficial
gas velocity and provided a means for estimating how hw increases
with smaller particle sizes or higher bed temperatures.
Analysis of the modified Molerus correlation suggests that hw
up to 1200 W m2 K1 would be readily achievable at realistic
receiver operating conditions. The benefits of higher hw were
explored in a simple analysis for a concentrating solar particle
receiver based on angled cavity surfaces for solar flux spreading
of 5.0. At such flux-spreading ratios, hw ¼ 1000 W m2 K1 of the
narrow-channel fluidized bed can enable solar aperture fluxes up
to 1500 suns with solar receiver efficiencies of very close to 90%
743
and external receiver wall temperatures (based on HaynesÒ 230Ò
properties) of less than 1030  C.
The robust operation of the counterflow fluidized bed
arrangement over a broad range of flow conditions with very low
gas-to-particle mass ratios suggests the value of this particle
flow configuration for wall-to-particle heat transfer at high-
temperatures for solar receivers and other high-temperature,
particle heat exchange applications. The range of operability and
very high hw within the viscous fluidization regime offers the
potential for very effective heat exchange to/from the particles
under flow conditions that may support very long-term operation.
Average hw of 1000 W m2 K1 or higher can be reliably predicted
from the modified Molerus correlation that fit the extensive data of
this study, but more extensive heat transfer studies over a range of
particle sizes will provide greater confidence in large-scale design
of solar particle receivers or particle heat exchangers that operate
in this promising bubbling fluidized bed configuration.
Conflict of Interest
Authors declare that there is no conflict of interest.
Acknowledgments
This work was funded in part through a grant from the Depart-
ment of Energy SunShot Initiative ELEMENTS program (award #
DE-EE0006537) and involved collaborations with the National
Renewable Energy Laboratory (NREL). The authors would like to
thank Luca Imponenti and Prof. Rob Braun from the Colorado
School of Mines and Dr. Kevin Albrecht now with Sandia National
Laboratories for helpful discussions on the design conditions
explored for this study. In addition, Drs. Zhiwen Ma and Janna Mar-
tinek from the National Renewable Energy Laboratory provided
helpful insight into the operations of fluidized beds. Finally, the
authors thank Claude Krause of CarboCeramics who generously
provided the particles used in this study.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.ijheatmasstransfer.
2018.05.165.
744 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745

References [29] T.D. Tan, Y.T. Chen, Review of study on solid particle solar receivers, Renew.
Sustain. Energy Rev. 14 (2010) 265–276.
[30] H. Zhang, H. Benoit, D. Gauthier, J. Degreve, J. Baeyens, I.P. Lopez, M.
[1] O. Molerus, Heat transfer in gas fluidized beds Part 1: Maximum heat transfer
Hemati, G. Flamant, Particle circulation loops in solar energy capture and
coefficients, Powder Technol. 70 (1992) 1–14.
storage: gas-solid flow and heat transfer considerations, Appl. Energy 161
[2] O. Molerus, Heat transfer in gas fluidized beds Part 2: Dependence of heat
(2016) 206–224.
transfer on gas velocity, Powder Technol. 70 (1992) 15–20.
[31] K. Kim, N. Siegel, G. Kolb, V. Rangaswamy, S.F. Moujaes, A study of solid
[3] P. Denholm, M. Hand, Grid flexibility and storage required to achieve very high
particle flow characterization in solar particle receiver, Sol. Energy 83 (2009)
penetration of variable renewable electricity, Energy Policy 39 (2011) 1817–
1784–1793.
1830.
[32] C.K. Ho, J.M. Christian, D. Romano, J. Yellowhair, N. Siegel, L. Savoldi, R. Zanino,
[4] A. Evans, V. Strezov, T.J. Evans, Assessment of utility energy storage options for
Characterization of particle flow in a free-falling solar particle receiver, J. Sol
increased renewable energy penetration, Renew. Sustain. Energy Rev. 16
Energy Eng. – Trans. ASME 139 (2017).
(2012) 4141–4147.
[33] M. Diago, A.C. Iniesta, Q. Falcoz, T. Shamim, N. Calvet, Energy and exergy
[5] O. Behar, A. Khellaf, K. Mohammedi, A review of studies on central receiver
analysis of a novel gravity-fed solid particle solar receiver, Energy Procedia 69
solar thermal power plants, Renew. Sustain. Energy Rev. 23 (2013) 12–39.
(2015) 812–821.
[6] M. Liu, N.H.S. Tay, S. Bell, M. Belusko, R. Jacob, G. Will, W. Saman, F. Bruno,
[34] A.C. Iniesta, M. Diago, T. Delclos, Q. Falcoz, T. Shamim, N. Calvet, Gravity-fed
Review on concentrating solar power plants and new developments in high
combined solar receiver/storage system using sand particles as heat collector,
temperature thermal energy storage technologies, Renew. Sustain. Energy Rev.
heat transfer and thermal energy storage media, Energy Procedia 69 (2015)
53 (2016) 1411–1432.
802–811.
[7] C.K. Ho, Advances in central receivers for concentrating solar applications, Sol.
[35] J. Marti, A. Haselbacher, A. Steinfeld, A numerical investigation of gas-particle
Energy 152 (2017) 38–56.
suspensions as heat transfer media for high-temperature concentrated solar
[8] C. Prieto, R. Osuna, A.I. Fernandez, L.F. Cabeza, Molten salt facilities, lessons
power, Int. J. Heat Mass Transf. 90 (2015) 1056–1070.
learnt at pilot plant scale to guarantee commercial plants; heat losses
[36] H. Benoit, I.P. López, D. Gauthier, J.-L. Sans, G. Flamant, On-sun demonstration
evaluation and correction, Renewable Energy 94 (2016) 175–185.
of a 750 °C heat transfer fluid for concentrating solar systems: dense particle
[9] U. Pelay, L. Luo, Y. Fan, D. Stitou, M. Rood, Thermal energy storage systems for
suspension in tube, Sol. Energy 118 (2015) 622–633.
concentrated solar power plants, Renew. Sustain. Energy Rev. 79 (2017) 82–
[37] I. Perez Lopez, H. Benoit, D. Gauthier, J. Sans, E. Guillot, G. Mazza, G.
100.
Flamant, On-sun operation of a 150 kwth pilot solar receiver using
[10] C.S. Turchi, Z. Ma, T.W. Neises, M.J. Wagner, Thermodynamic study of
dense particle suspension as heat transfer fluid, Sol. Energy 137 (2016)
advanced supercritical carbon dioxide power cycles for concentrating solar
463–476.
power systems, J. Sol. Energy Eng. – Trans. ASME 135 (2013) 041007.
[38] Z. Ma, M. Mehos, G. Glatzmaier, B. Sakadjian, Development of a concentrating
[11] M. Mehos, C. Turchi, J. Jorgenson, P. Denholm, C. Ho, K. Armijo, On the Path to
solar power system using fluidized-bed technology for thermal energy
SunShot: Advancing Concentrating Solar Power Technology, Performance, and
conversion and solid particles for thermal energy storage, Energy Procedia
Dispatchability, Technical Report NREL/TP-5500-65688, National Renewable
69 (2015) 1349–1359.
Energy Laboratory, Golden, CO, 2016.
[39] A.B. Morris, S. Pannala, Z. Ma, C.M. Hrenya, A conductive heat transfer model
[12] W. Stein, R. Buck, Advanced power cycles for concentrated solar power, Sol.
for particle flows over immersed surfaces, Int. J. Heat Mass Transf. 89 (2015)
Energy 152 (2017) 91–105.
1277–1289.
[13] Y. Ting Wu, N. Ren, T. Wang, C. fang Ma, Experimental study on optimized
[40] N.S. Grewal, S.C. Saxena, Heat transfer between a horizontal tube and a gas-
composition of mixed carbonate salt for sensible heat storage in solar thermal
solid fluidized bed, Int. J. Heat Mass Transf. 23 (1980) 1505–1519.
power plant, Sol. Energy 85 (2011) 1957–1966.
[41] D. Subbarao, P. Basu, A model for heat-transfer in circulating fluidized beds,
[14] J.C. Gomez-Vidal, J. Noel, J. Weber, Corrosion evaluation of alloys and MCrAlX
Int. J. Heat Mass Transf. 29 (1986) 487–489.
coatings in molten carbonates for thermal solar applications, Sol. Energy
[42] S.W. Kim, J.Y. Ahn, S.D. Kim, D.H. Lee, Heat transfer and bubble characteristics
Mater. Sol. Cells 157 (2016) 517–525.
in a fluidized bed with immersed horizontal tube bundle, Int. J. Heat Mass
[15] K. Vignarooban, X. Xu, A. Arvay, K. Hsu, A. Kannan, Heat transfer fluids for
Transf. 46 (2003) 399–409.
concentrating solar power systems – a review, Appl. Energy 146 (2015) 383–
[43] S. Rasouli, M. Golriz, A. Hamidi, Effect of annular fins on heat transfer of a
396.
horizontal immersed tube in bubbling fluidized beds, Powder Technol. 154
[16] P.D. Myers, D.Y. Goswami, Thermal energy storage using chloride salts and
(2005) 9–13.
their eutectics, Appl. Therm. Eng. 109 (2016) 889–900.
[44] F. Di Natale, P. Bareschino, R. Nigro, Heat transfer and void fraction profiles
[17] S. Khare, M. Dell’Amico, C. Knight, S. McGarry, Selection of materials for high
around a horizontal cylinder immersed in a bubbling fluidised bed, Int. J. Heat
temperature sensible energy storage, Sol. Energy Mater. Sol. Cells 115 (2013)
Mass Transf. 53 (2010) 3525–3532.
114–122.
[45] S.W. Kim, S.D. Kim, Heat transfer characteristics in a pressurized fluidized bed
[18] G. Flamant, D. Gauthier, H. Benoit, J.-L. Sans, R. Garcia, B. Boissiere, R. Ansart,
of fine particles with immersed horizontal tube bundle, Int. J. Heat Mass
M. Hemati, Dense suspension of solid particles as a new heat transfer fluid for
Transf. 64 (2013) 269–277.
concentrated solar thermal plants: on-sun proof of concept, Chem. Eng. Sci.
[46] A. Gupta, P. Nag, Bed-to-wall heat transfer behavior in a pressurized
102 (2013) 567–576.
circulating fluidized bed, Int. J. Heat Mass Transf. 45 (2002) 3429–
[19] N. Siegel, M. Gross, C. Ho, T. Phan, J. Yuan, Physical properties of solid particle
3436.
thermal energy storage media for concentrating solar power applications,
[47] A. Blaszczuk, W. Nowak, Bed-to-wall heat transfer coefficient in a supercritical
Energy Procedia 49 (2014) 1015–1023.
CFB boiler at different bed particle sizes, Int. J. Heat Mass Transf. 79 (2014)
[20] C.K. Ho, A review of high-temperature particle receivers for concentrating
736–749.
solar power, Appl. Therm. Eng. 109 (2016) 958–969.
[48] J. Krzywanski, W. Nowak, Modeling of bed-to-wall heat transfer coefficient in a
[21] A.J. Carrillo, J. Moya, A. Bayon, P. Jana, V.A. de la Pena O’Shea, M. Romero, J.
large-scale CFBC by fuzzy logic approach, Int. J. Heat Mass Transf. 94 (2016)
Gonzalez-Aguilar, D.P. Serrano, P. Pizarro, J.M. Coronado, Thermochemical
327–334.
energy storage at high temperature via redox cycles of Mn and Co oxides: Pure
[49] R. Yusuf, B. Halvorsen, M.C. Melaaen, An experimental and computational
oxides versus mixed ones, Sol. Energy Mater. Sol. Cells 123 (2014) 47–57.
study of wall to bed heat transfer in a bubbling gas-solid fluidized bed, Int. J.
[22] S.M. Babiniec, E.N. Coker, J.E. Miller, A. Ambrosini, Doped calcium manganites
Multiph. Flow 42 (2012) 9–23.
for advanced high-temperature thermochemical energy storage, Int. J. Energy
[50] G. Flamant, Y. Flitris, D. Gauthier, Heat transfer to walls in a high temperature
Res. 40 (2016) 280–284.
fluidized bed of group II particles, Chem. Eng. Process. 27 (1990) 175–184.
[23] K.J. Albrecht, G.S. Jackson, R.J. Braun, Thermodynamically consistent modeling
[51] G. Flamant, T. Menigault, Combined wall-to-fluidized bed heat transfer.
of redox-stable perovskite oxides for thermochemical energy conversion and
Bubbles and emulsion contributions at high temperature, Int. J. Heat Mass
storage, Appl. Energy 165 (2016) 285–296.
Transf. 30 (1987) 1803–1812.
[24] L. Imponenti, K.J. Albrecht, J.W. Wands, M.D. Sanders, G.S. Jackson,
[52] B.A. Abid, J.M. Ali, A.A. Alzubaidi, Heat transfer in gas–solid fluidized bed with
Thermochemical energy storage in strontium-doped calcium manganites for
various heater inclinations, Int. J. Heat Mass Transf. 54 (2011) 2228–2233.
concentrating solar power applications, Sol. Energy 151 (2017) 1–13.
[53] R.S. Patil, M. Pandey, P. Mahanta, Parametric studies and effect of scale-up on
[25] C. Agrafiotis, M. Roeb, M. Schmuecker, C. Sattler, Exploitation of
wall-to-bed heat transfer characteristics of circulating fluidized bed risers,
thermochemical cycles based on solid oxide redox systems for
Exp. Thermal Fluid Sci. 35 (2011) 485–494.
thermochemical storage of solar heat. Part 2: Redox oxide-coated porous
[54] A. Reyes Urrutia, H. Benoit, M. Zambon, D. Gauthier, G. Flamant, G. Mazza,
ceramic structures as integrated thermochemical reactors/heat exchangers,
Simulation of the behavior of a dense SIC particle suspension as an energy
Sol. Energy 114 (2015) 440–458.
transporting vector using computational fluid dynamics (CFD), Chem. Eng. Res.
[26] R. Anderson, S. Shiri, H. Bindra, J.F. Morris, Experimental results and modeling
Des. 106 (2016) 141–154.
of energy storage and recovery in a packed bed of alumina particles, Appl.
[55] D.J. Patil, J. Smit, M. van Sint Annaland, J.A.M. Kuipers, Wall-to-bed heat
Energy 119 (2014) 521–529.
transfer in gas-solid bubbling fluidized beds, AIChE J. 52 (2006) 58–74.
[27] N.P. Siegel, C.K. Ho, S.S. Khalsa, G.J. Kolb, Development and evaluation of a
[56] L.M. Armstrong, S. Gu, K.H. Luo, Study of wall-to-bed heat transfer in a
prototype solid particle receiver: On-sun testing and model validation, J. Sol.
bubbling fluidised bed using the kinetic theory of granular flow, Int. J. Heat
Energy Eng. – Trans. ASME 132 (2010).
Mass Transf. 53 (2010) 4949–4959.
[28] T. Baumann, S. Zunft, Properties of granular materials as heat transfer and
[57] A.V. Patil, E.A.J.F. Peters, J.A.M. Kuipers, Comparison of cfd-dem heat transfer
storage medium in CSP application, Sol. Energy Mater. Sol. Cells 143 (2015)
simulations with infrared/visual measurements, Chem. Eng. J. 277 (2015)
38–47.
388–401.
 D.C. Miller et al. / International Journal of Heat and Mass Transfer 126 (2018) 730–745
[58] Z. Li, M.v.S. Annaland, J.A.M. Kuipers, N.G. Deen, Effect of superficial gas
velocity on the particle temperature distribution in a fluidized bed with heat
production, Chem. Eng. Sci. 140 (2016) 279–290.
[59] T. Lichtenegger, E.A.J.F. Peters, J.A.M. Kuipers, S. Pirker, A recurrence cfd study
of heat transfer in a fluidized bed, Chem. Eng. Sci. 172 (2017) 310–322.
[60] D.C. Miller, Heat Transfer Characteristics of a Novel Fluidized Bed for
Concentrating Solar with Thermal Energy Storage (Master’s thesis), Colorado
School of Mines, Golden, CO, 2017.
[61] G.G. Kuvshimov, Free flow of a granular material from an aperture in the
presence of a gas counterflow, J. Appl. Mech. Tech. Phys. 36 (1995) 869–876.
[62] A.S. Oles, G.S. Jackson, Modeling of a concentrated-solar, falling-particle
receiver for ceria reduction, Sol. Energy 122 (2015) 126–147.
[63] T. Aihara, S. Maruyama, K. Tanaka, J. Yamaguchi, Heat transfer and fluidization
characteristics of a high-temperature shallow fluidized bed, Exp. Thermal
Fluid Sci. 6 (1993) 282–291.
[64] P. Basu, Heat transfer in high temperature fast fluidized beds, Chem. Eng. Sci.
45 (1990) 3123–3136.
[65] J.S.M. Botterill, Y. Teoman, Fluid-Bed Behaviour at Elevated Temperatures,
Springer, US, Boston, MA, 1980.
[66] M. Eriksson, M. Golriz, Radiation heat transfer in circulating fluidized bed
combustors, Int. J. Therm. Sci. 44 (2005) 399–409.
[67] A. Baskakov, B. Leckner, Radiative heat transfer in circulating fluidized bed
furnaces, Powder Technol. 90 (1997) 213–218.
[68] J. Yamada, Y. Kurosaki, T. Nagai, Radiation heat transfer between fluidizing
particles and a heat transfer surface in a fluidized bed, J. Heat Transf. – Trans.
ASME 123 (2001) 458–465.
[69] J. Yang, L. Wang, L. Tong, H. Li, Modeling of radiative heat transfer between
high-temperature fluidized beds and immersed walls, Chem. Eng. Sci. 59
(2004) 3195–3199.
[70] A. Baskakov, B. Berg, O. Vitt, N. Filippovsky, V. Kirakosyan, J. Goldobin, V.
Maskaev, Heat transfer to objects immersed in fluidized beds, Powder Technol.
8 (1973) 273–282.
745
[71] D. Kunii, Fluidization Engineering, Butterworth-Heinemann, Boston, 1991.
[72] J.C. Chen, J.R. Grace, M.R. Golriz, Heat transfer in fluidized beds: design
methods, Powder Technol. 150 (2005) 123–132.
[73] K. Pisters, A. Prakash, Investigations of axial and radial variations of heat
transfer coefficient in bubbling fluidized bed with fast response probe, Powder
Technol. 207 (2011) 224–231.
[74] J.A.M. Kuipers, W. Prins, W.P.M. Van Swaaij, Numerical calculation of wall-to-
bed heat-transfer coefficients in gas-fluidized beds, AIChE J. 38 (1992) 1079–
1091.
[75] M.A. Izquierdo-Barrientos, C. Sobrino, J.A. Almendros-Ibáñez, Modeling the
heat transfer coefficient between a surface and fixed and fluidized beds with
phase change material, J. Heat Transf – Trans. ASME 138 (2016) 072001.
[76] N. Gelperin, V. Einstein, Heat transfer in fluidized beds, in: J. Davidson, D.
Harrison (Eds.), Fluidization, Academic Press, London, 1971, pp. 471–540.
[77] O. Molerus, A. Burschka, S. Dietz, Particle migration at solid surfaces and heat
transfer in bubbling fluidized beds—II. Prediction of heat transfer in bubbling
fluidized beds, Chem. Eng. Sci. 50 (1995) 879–885.
[78] V. Borodulya, Y. Teplitsky, I. Markevich, A. Hassan, T. Yeryomenko, Heat
transfer between a surface and a fluidized bed: consideration of pressure and
temperature effects, Int. J. Heat Mass Transf. 34 (1991) 47–53.
[79] S. Zabrodsky, Hydrodynamics and Heat Transfer in Fluidized Beds, MIT,
Cambridge, Mass., 1966.
[80] O. Molerus, J. Schweinzer, Prediction of gas convective part of the heat transfer
to fluidized beds, Fluidization IV (1989) 685–693.
[81] O. Molerus, Arguments on heat transfer in gas fluidized beds, Chem. Eng. Sci.
48 (1993) 761–770.
[82] O. Molerus, A. Burschka, S. Dietz, Particle migration at solid surfaces and heat
transfer in bubbling fluidized beds—1. Particle migration measurement
system, Chem. Eng. Sci. 50 (1995) 871–877.
[83] Haynes International, HaynesÒ 230Ò alloy brochure, Technical Report, Haynes
International, 2017.