You are on page 1of 10

A Comparative Experimental and Numerical Study to

Investigate the Relative Merits of Convectors and

‘‘C’’ Inserts in Cooling Cold-Rolled Coils

The coil cooling and storage unit (CCSU) is used to cool cold-rolled coils to the temper rolling
temperature after the annealing cycle is over at the batch annealing furnace (BAF) in a cold rolling
mill (CRM). In the CCSU, the coils are kept on the cooling bases for any fixed time irrespective of the
grade and tonnage. Therefore, the need for a mathematical model to accurately predict the cooling
time of the coils was felt. The current study involves experimental and numerical analysis of a stack of
coils with respect to heat transfer and fluid flow. A comparative study was carried out to ascertain the
relative merits of convectors and ‘‘C’’ inserts (CIs) in the cooling the coils. The air flow distribution
for the case of different convectors and CIs was measured by means of a full scale physical model.
Two different mathematical models were applied to model the fluid flow and flow distribution through
the stack of coils. The first flow model uses the hydraulic resistance concept for estimating the air flow
rate distribution, whereas the second flow model uses commercial computational fluid dynamics
(CFD) software and predicts the velocity distribution in the flow path between two coils in a stack.
The predictions from these two models compare well with the experimental data. The flow models
were used to calculate the average heat-transfer coefficient in different flow passages in a stack. The
heat-transfer coefficients thus obtained were used to tune and validate a two-dimensional transient
heat-transfer model of coils. The heat-transfer model predicts the cooling time of coils accurately and
also suggests a possible reduction of cooling time if CIs are used in place of convectors.

I. INTRODUCTION ent designs. For example, the base convectors have 16 short
and equal numbers of long fins on each side, whereas the
COLD-ROLLED coils are annealed in the batch anneal- intermediate convectors may have 20 short and long fins
ing furnace (BAF) for obtaining the desired properties and each on both sides. The top convectors, on the other hand,
mechanical strength to render the cold-rolled sheets ame- have 16 long fins only on one side with the fin thickness
nable to subsequent forming operations (e.g., deep drawing nearly double those of the previous two types. Figure 1(b)
of auto body parts). In the BAF, the coils are stacked inside shows the top view of a typical base convector plate.
a furnace and annealed for long hours (40 to 50 hours) in a The stacking as well as the destacking of coils is a time-
hydrogen atmosphere. There are three stages in the anneal- consuming process, which may take 30 to 45 minutes for a
ing cycle, namely, heating, soaking, and cooling. The coils single stack depending upon the number of coils. In order
are heated to a temperature of 720 °C and kept for soaking to reduce the time loss during stacking and destacking, the
at that temperature for a certain time before they are cooled intermediate convectors can be replaced by ‘‘C’’ inserts
at a slower rate first (up to 500 °C) and then at a faster rate (CIs). Figure 2(a) shows a CI with typical dimension. The
using a bypass cooling system until the hot spot (core) of arrangement of CIs is such that between two coils, there are
the coils reaches 160 °C. four CIs at diametrically opposite locations making a 90
It is the coil cooling and storage unit (CCSU) in which deg angular distance (refer to Figure 2(b) for arrangement).
the coils are made to pass through a final cooling step so The present work investigates the effect of using CIs on the
that the hot spot temperature comes down to around 50 °C. cooling time of coils in CCSU as compared to the practice
Thereafter, the coils may be taken to the skin pass mill with convectors by predicting the cooling time of coils.
(SPM) for final property and surface finish compliance. Cooling of the coils is done by passing dehumidified air
In the CCSU, identical coil stacks are built after finishing from the bottom of the stack through the eyes of the coils
the cooling cycle on the annealing bases at the BAF. Figure and the convector plates (or the CIs). After stacking, each
1(a) shows a typical stack configuration for a CCSU base stack is covered by a single hood in order to collect the
with four coils. The first coil of the stack is placed on a dehumidified air that has passed the convectors or CIs. To
bottom convector plate (CP). An intermediate convector avoid oxidation of the outer layers, the relative humidity of
plate is placed between each two coils of a stack. A top the recirculated air for cooling the coils has to be less than
convector plate is placed above the top coil of a stack. 29 pct.
These three types of convector plates normally have differ-

VIKAS SINGH, Researchers, are with Research and Development, Tata A. Overview
Steel Ltd., Jamshedpur, Jharkhand, PIN 831001, India. Contact e-mail: Extensive work[1–18] has been carried out on the process
Manuscript submitted September 2, 2005. of batch annealing furnaces, but there appears to be very


Fig. 1—(a) Typical CCSU stack arrangement with convectors between the coils. CIs can also be kept in place of the intermediate convectors. (b) Top view of
the base convector of a CCSU stack. The dotted line shows the peripheral locations where velocity measurements were carried out in the physical model.

little literature available for CCSU, because the heat-trans-

fer process is somewhat similar to the BAF. The present
work involves experimental and numerical analysis of heat
transfer and fluid flow in a stack of coils. The analysis was
employed to compare the effectiveness of convectors and
CIs in cooling the coils. The heat-transfer model was used
to predict the cooling time of coils. The air flow distribution
for the case of different convectors and CIs was estimated
by means of a full scale physical model in which velocity
measurements were carried out. A flow model using the so-
called hydraulic resistance concept was developed for esti-
mating the flow distribution through convectors or CIs. A
commercial CFD package, Fluent,[19] was employed to
obtain the velocity distribution through the flow passages
of convectors and CIs. The actual velocity measurements
were compared with the CFD simulation. The information
from the flow models was used to calculate the average
heat-transfer coefficients for a coil. The heat-transfer coef-
ficients were used to tune and validate the heat-transfer
model. Plant trials were conducted to record the temper-
ature of a particular point of a coil with time when CIs were
used. The prediction of cooling time from the heat-transfer
model was compared with the measurements from the
plant trials.

B. Experimental Measurement of Flow Distribution

The estimation of the distribution of flow exiting each of
the convectors and CIs has been obtained from measure-
ments on a full scale physical model. The experimental
setup consists of four identical coils stacked on a bottom
convector. The coils have been made from plywood board
Fig. 2—(a) Different views of a CI used in CCSU. (b) The radial arrange-
ment of the CIs on the top surface of a coil. Another coil sits on top of
with scrap cold-rolled sheets as the inner and outer cylin-
these CIs, making a passage for the air in between two coils. All dimen- drical surfaces. The convector plates have also been fabri-
sions are in millimeters. cated from plywood board, and the fins have been made


from wood with very high surface finish so as to offer the The measurements were taken for three different flow rates
same flow resistance as that of the actual convectors made each for convectors and CIs. In the case of intermediate
up of weldable steel plates. The equivalent roughnesses of convectors, the velocities were measured for both sides of
galvanized sheet steel and plywood are comparable (0.12 to the convector plate along the periphery of the coils. The
0.15).[20] The fabrication using plywood and wood made average velocity was calculated at a particular convector or
the coils and the convectors lighter and easy to handle. CI, and the average flow was then calculated from the flow
The CIs are also fabricated from wood with dimensions area. For intermediate convectors, the average flow was
as shown in Figure 2(a). The experimental setup has been calculated by taking the mean of the average flows of both
shown in Figure 3(a) with actual dimensions. sides of the convector plate.
The air flow was supplied from a blower through pipe,
which was fitted at the bottom chamber of the stack. A C. Estimation of Flow Distribution Using
grate was placed at the eye of the bottom convector to make Hydraulic Resistance
the flow uniform when it passed through the first convector
and coil. The flow delivered by the blower was calculated A flow model was developed for predicting the flow
from the difference of water column in a ‘‘U’’-tube man- distribution through convector plates using the hydraulic
ometer. The cover on the stack is fitted with small windows resistance concept. The various resistances offered by var-
at different heights in order to access the convectors and the ious flow paths such as diverging slots between two radial
CIs during velocity measurements. At each level, there fins on the convector plate and the tube-like part at the inner
were six windows. Velocity measurements were carried surface of coils have been shown in Figure 3(b) using dif-
out through these windows using a vane-type anemometer ferent shades. The resistance network thus formed has been
at peripheral locations of a coil on the convector plate (refer depicted in Figure 3(c).
to Figure 1(b) for measurement locations) and also periph- The following relations have been used with some mod-
erally in the gap between two coils when CIs were used. ifications for the resistances[21] offered by different flow
(a) For flow through the tube region of the coil, i.e., for
resistances R2, R4, R6, and R8 in this case,
R ¼ ð128mLÞ ðpD4 Þ [1]
where m 5 viscosity of air flowing, L 5 length of flow path,
D 5 hydraulic diameter of flow path, Pin 5 pressure at inlet
to the flow path, and P 5 average pressure of the flow
path 5 [(Pin 1 Pout)/2].
(b) For flow on the convectors, the resistance equation
for the rectangular cross section has been used in approx-
Therefore, resistances R1, R3, R5, and R7 can be calcu-
lated from the equation
h  iP 
2 in
R ¼ a ðCmLÞ ðknA YÞ [2]
where a 5 a tuning parameter; C 5 a constant (=2824.859,
if all parameters are in SI units); m 5 viscosity of air
flowing; L 5 length of flow path; k 5 1 for base convector,
2 for intermediate convectors; n 5 number of long fins on
one side of convector plate; A 5 average cross sectional
area of the flow path; Y 5 correction factor for height-to-
width ratio of the flow path cross section, e.g., Y 5 1 for
square cross section;[21] Pin 5 pressure at inlet to the flow
path; and P 5 average pressure of the flow path 5 [(Pin 1
In reality, the flow path between two fins is a diverging
slot obstructed by a midbar (short fin), as shown in Figure
1(b). The tuning parameter a differs for base and intermedi-
ate convectors, and is used for tuning the model with data
sets obtained from the flow measurements in the physical
model since the original equations are ideal for laminar
Fig. 3—(a) Experimental setup of the physical model with convectors. CIs flow with regular geometries. The P for any flow path has
were also kept in place of the intermediate convectors. (b) The various
resistances offered by different flow paths in a stack with four coils. (c) been calculated by the iterative method using the correction
The flow resistance network to estimate flow through different convectors from the previous iteration. It may be noted that the resist-
for a four coil stack. All dimensions are in millimeters. ance of the cylindrical tube of the coils is far less than that


of the resistances through the convectors. Therefore, the where Pk represents the shear production term:
minimum pressure drop occurs through the inner cylindri-  
cal core formed by the stack of coils, and it will be shown @U i @Uj @U i
Pk ¼ nt 1 [9]
in Section IV that the maximum flow leaves through the top @xj @xi @xj
convector since air flow takes the least resistant path. The
mathematical model for CIs can also be developed similarly. With the exception of the low-Reynolds number k-e
models, the damping functions f1, f2, and f3 are equal to
one. The model coefficients in the standard k-e model are
D. Estimation of Velocity Distribution using CFD given subsequently:

A commercial CFD software package (FLUENT)[19] has Cm ,C1 ,C2 ,sm ,se 5 ð0:09,1:44,1:92,1:0,1:3Þ [10]
been used to predict the velocity profiles inside the flow
passage through different convectors and CIs under steady- A detailed description of this model can be found in Launder
state conditions. This program is a finite volume code, which and Spalding.[23]
uses a cell-centered, nonstaggered grid. The Reynolds- Figures 4(a) and (b) show the computational domain for
averaged Navier–Stokes equations and the transport equa- convectors and CIs, respectively. The geometry was
tions of the turbulent quantities are solved by the pressure meshed using structured hexahedron cells. No slip boun-
correction algorithm SIMPLE[22] (semi-implicit method for dary conditions were used for the walls specified by the top
pressure-linked equation). In the current study, the fluid is and bottom coils and also for the side walls of convector
considered as incompressible. The fluid properties are fins and CIs. In the case of convectors, the outer surfaces of
assumed to be constant. The governing equations as given the included T-shaped short fins are also considered for no-
by Eqs. [3] and [4] describe the flow through the flow slip boundary conditions. The air flow rate was specified at
passage between two coils in a stack. the inlet and ambient pressure was applied as the outlet
Mass Conservation: boundary condition.

@U j E. Heat-Transfer Modeling and Temperature

50 [3] Measurement of Coils
The spatial and temporal thermal profile of any coil is
Momentum conservation: obtained through the heat balance of the control volume
method in the cylindrical coordinate. A computer program
  was written for the same. Owing to the cylindrical symme-
@U j U i 1 @p @ @U i
¼ 1 n  ui uj [4] try of the coils, only the rz plane is analyzed. It is assumed
@xj r @xi @xj @xj
that there is no temperature gradient in the u direction.
where ui uj is the unknown Reynolds stresses. Convective boundary conditions have been used at the four
The standard k-e model was used to model the turbulent
flow field. In this model, two entities that are used for
describing turbulence are introduced: the specific kinetic
turbulent energy k and its dissipation e. For both entities,
differential transport equations are introduced, which are
solved in addition to the other Eqs. [3] and [4]. Using the
Boussinesq hypothesis, the Reynolds stresses can be
described as follows:
@U i @U j 2
ui uj ¼ nt 1  kdij [5]
@xj @xi 3

the eddy (turbulent) viscosity vt is obtained from

n t ¼ Cm f m [6]
The kinetic turbulent energy k and its dissipation e are
determined using the following transport equations, respec-
@U j k @ nt @k
5 n1 1 Pk  e [7]
@xj @xj sk @xj

   Fig. 4—(a) The geometry of the divergent flow passage between two long
@U j e @ nt @e e fins on the convector plate. The obstructing midbar and short fins (making
5 n1 1 ðC1 f 1 Pk  C2 f 2 eÞ [8] a T shape) could also be seen in the figure. (b) The geometry of the flow
@xj @xj se @xj k passage between two CIs separated radially by 90 deg on top of a coil.


boundaries, i.e., at the in, out, top, and bottom surfaces of where Nu is the Nusselt number, Kg is the thermal conduc-
the coil. The grid arrangement along with a typical heat tivity of fluid (i.e., air), and Dh is the hydraulic diameter of
balance for any internal node (control volume) has been the flow passage defined by 4(area/perimeter). The heat-
shown in Figure 5. It should be noted that there are nodes transfer coefficient at the inner surface has been obtained
at the surface (half node), corner (quarter node), and at the from the following Nusselt number correlation (due to
interior (full node). The alternating-direction implicit Dittus–Boelter) for flow through circular ducts:[25]
(ADI)[15,24] method was used for obtaining the sets of linear
algebraic equations. In ADI formulation, heat flows in the Nu 5 0:023Re0:8 Pr0:3 [12]
r direction for half of the time-step, and it flows in the where Pr 5 Prandl number for air, Re 5 Reynolds number 5
z direction for the other half. The resulting linear algebraic VDh/n, V 5 average velocity, and n 5 kinematic viscosity
equations are solved by the tri-diagonal matrix algorithm of air.
(TDMA). The following Nusselt number correlation (due to
A suitable Initial temperature profile (i.e., at time t 5 0) Gnielinski) for flow through cylindrical annuli is used for
of the coil has been used because the coils have a temper- calculating the heat-transfer coefficient at the outer surface
ature gradient owing to the thermal history at the BAF. Two of coil:[26]
thermal conductivities have been used, Kr and Kz, for radial
and axial directions, respectively. The main barrier to trans- ðf =2ÞðRe  1000ÞPr
Nu 5
forming heat from the coils is the low radial thermal con- 1 1 12:7ðf =2Þ0:5 ðPr2=3  1Þ [13]
ductivity, which depends upon the sheet thickness and air 2
gap between sheets. The radial conductivity varies from f 5 ð1:58 ln Re  3:28Þ
5 to 12 pct of the axial conductivity.[16] The low radial where Re and Pr are the same as mentioned earlier.
thermal conductivity decreases production rates and causes In the case of flow through the convectors (i.e., through
high thermal gradients in the steel coils. The simulations the diverging slot between two fins) or CIs, the correlations
were performed with a time-step of 100 seconds, and 51 for flow in rectangular ducts (due to Sleicher and Rouse)
nodes were used in both the r and z directions after check- may be used:[26,27]
ing for the grid and time-step independency.
Empirical correlations have been used to determine Nu 5 5 1 0:015Rea Prb
the heat-transfer coefficients at the inner, outer, top, and a 5 0:88  0:24=ð4 1 PrÞ [14]
bottom surfaces of the coils. The heat-transfer coefficient ð0:6PrÞ
is defined by b 5 0:333 1 0:5e
NuK g Because the diverging slot on the convector can be
h5 [11]
Dh thought of as made up of many rectangular cross sections
along the radial direction, the local Nusselt number and
hence the local heat-transfer coefficients were calculated
for the case of convector plates, and then they were aver-
aged to be used as the bottom and top heat-transfer coef-
ficients in the heat-transfer model.
The heat-transfer model was validated against actual
temperature measurements carried out in the plant for a
coil in which CIs were used for stacking. For this measure-
ment, a multipoint type hybrid chart recorder was used
along with some K-type contact thermocouples, and the
temperature data of some locations on the coil were
recorded for nearly 40 hours. The heat-transfer coefficients
were calculated for any coil of the stack for different input
flow rates and flow distribution. This was done for both the
convectors and CIs. In the case of the coil used for valida-
tion of the heat-transfer model, an input flow rate (less than
the maximum capacity of 10,100 m3/h per base in the plant)
and a flow distribution were assumed (which falls in the
range of the physical model findings for flow distribution
using CIs, refer to Table II). The calculated heat-transfer
coefficients for the coil under the aforementioned inputs
were tuned later to validate the model. This will again be
discussed in Section III.


A. The Flow Distribution
Fig. 5—(a) Cross section of a coil showing different nodes and the nomen-
clature adopted for the mathematical model. (b) A typical interior element Tables I and II show the percentage of flow distribution
(I,J) and the heat balance for the volume element. in the physical model through the convectors and the CIs,


Table I. Percentage Distribution of Air Flow through Different Convectors for Three Different Input Flow
Rates in the Physical Model

Percentage Distribution
Total Flow (m3/hr) Base Convector CP 2 CP 3 CP 4 Top Convector
1775 (200-mm WC) 10 10 18 20 42
2173 (300-mm WC) 10 11 18 20 41
2509 (400-mm WC) 13 17 23 21 26
Range (approximately) 10 to 13 10 to 17 18 to 23 20 to 21 26 to 42

Table II. Percentage Distribution of Air Flow Through Different CI Location for Three Different Input
Flow Rates in the Physical Model

Percentage Distribution
Total Flow (m3/hr) Base Convector CI 2 CI 3 CI 4 Top Convector
1775 (200-mm WC) 9 14 19 18 40
2173 (300-mm WC) 10 17 21 19 33
2509 (400-mm WC) 9 16 20 19 36
Range (approximately) 9 and 10 14 to 17 19 to 21 18 to 19 33 to 40

Fig. 6—The measured velocities at different peripheral locations when CIs

were used between coils 2 and 3 for a flow rate of 2173 m3/h (300-mm WC). Fig. 7—Comparison of percentage distribution of flow through different
convectors obtained by the flow model (hydraulic resistance concept) and
the physical model measurements for a flow rate of 2509 m3/h. Error bar is
8 pct of the experimental value.

respectively, for three different input flow rates. It is

obvious from these data that the maximum percentage of observed that the peripheral velocity distribution for CIs
flow leaves through the top convector and the reverse hap- was nearly uniform (refer to Figure 6).
pens through the bottom convector. This result is expected, Figures 7 and 8 compare the flow model (hydraulic
since the resistance to flow through the central core of the resistance) and the physical model results through each
stack is less due to its circular geometry, and the flow goes convector for two flow rates: 2509 and 2173 m3/h (blower
vertically upward taking the least resistant path. For this pressure difference of 400- and 300-mm water column
reason, the maximum of the flow leaves through the top (WC), respectively). There is good agreement between
convector (CP 5 here for a four coil stack). The bottom the observed and the calculated values of flow distribution
convector has only one side exposed to the flow and thereby at least for three convectors. Some discrepancies may be
offers less normal flow area compared to the other convec- attributed to the fact that the average pressure P may not be
tors. It should be noted from Tables I and II that the flow proper in the equation for hydraulic resistance (Eqs. [1] and
distributions are nearly reproducible, because the range of [2]). The tuning parameter a is 250 for the bottom CP and
distribution for a particular convector or CI is very close. 300 for intermediate CPs for all flow rates. Therefore, the
This is particularly evident in the case of CIs, where it was mathematical model can be used to get a rough idea regarding


the flow distribution for situations where we do not have that of a CI. Therefore, the velocity through the CI would
experimental data, for example, if more coils with different be higher as compared to the CP for the same flow rate.
dimensions are used in the stack and if convectors/CIs with This observation provides a clue to the better effective-
different dimensions are placed. ness of using the CIs compared to the convectors because
Figures 9 and 10 show the average velocities obtained for higher velocity means higher heat-transfer coefficient and,
three different flow rates at different CP and CI positions, hence, faster cooling. This issue is again discussed with an
respectively, in the physical model. The velocities at each example of cooling time of a particular coil in Section C.
CP/CI increase as the input flow rates increase. The com-
parisons of the average velocity of air passing through CPs B. The Velocity Distribution
and CIs have been shown in Figure 11 for a flow rate of
2509 m3/h. It is quite evident that the air passes through the The results obtained from the CFD model are discussed
top and bottom of the coils at higher velocities when the in this section. The velocity distribution obtained through
intermediate convectors are replaced by the CIs. The effec- the numerical simulation gives a better insight of the flow
tive available normal flow area at the outlet between two phenomena taking place between two coils in a stack.
coils (1500-mm o.d.) for a convector is nearly 1.78 times Figures 12(a) through (c) show the velocity distribution at
different planes in one of the flow passages of a convector
plate for a total stack input flow rate of 1775 m3/h (200-mm
WC). Similar contour plots are shown in Figures 13(a) and
(b) for a CI, located on top of the base coil, at the mid and

Fig. 8—Comparison of percentage distribution of flow through different

convectors obtained by the mathematical model (hydraulic resistance con-
cept) and the physical model measurements for a flow rate of 2173 m3/h.
Error bar is 8 pct of the experimental value. Fig. 10—Average velocities through different CIs for three flow rates in the
physical model.

Fig. 9—Average velocities at the outlet of flow passages (at coil OD)
through different convector positions (base to top) for three flow rates in Fig. 11—Comparison of average velocity of air passing through convectors
the physical model. and CIs for a flow rate of 2509 m3/h in the physical model.


exit planes of the flow passage. It is clearly visible from the C. The Heat-Transfer Model
preceding plots that the maximum velocity at the outlet is
As outlined earlier in Section II–E, two sets of temper-
nearly 10 pct higher for CIs as compared to convectors.
ature measurement data were used to validate the mathe-
Table III compares velocities at the outlet of the flow pas-
matical model for heat transfer in the coil. The initial
sage for experimental measurements in the physical model temperature profile has been generated by running the
and the prediction from the CFD model. There is good heat-transfer code and setting the initial temperatures of
agreement between the actual measurements and prediction all nodes to a high value (as if, it is simulating the cooling
at lower flow rates. Although the prediction values for aver- cycle after soaking in BAF) with high heat-transfer coeffi-
age exit air velocity are always below the measured values, cients, and then interrupting the program after a suitable
we must mention that the agreement between these two temperature condition has been reached (e.g., when the hot
become poorer as we increase the air flow rate. For lower spot has reached 160 °C after a specified time as obtained
flow rates, the difference is as low as 5 pct from the mea- from the annealing cycle curves for the BAF). The initial
surements; however, for higher flow rates, this discrepancy heat-transfer coefficients obtained assuming a total input
is nearly 30 pct. The reason for this is perhaps that at higher flow and a flow distribution are tuned to validate one set
velocities, the degree of turbulence may not be captured of temperature profile data. This tuned model is then
properly by the present turbulence model. This validated applied to predict the temperature profile of another set of
CFD model would enable the parametric analysis for selec- data. Figure 14 shows the model predictions and actual
tion of a suitable design of a convector/CI in future. measurements along with the temperature profile of the

Fig. 12—Velocity (m/s) distribution at a plane (a) located 24 mm from bottom, (b) located 14 mm from bottom (i.e., mid plane), and (c) at the outlet of the
divergent flow passage between two long fins in the case of CP 1. The total input flow rate entering the stack was assumed to be 1775 m3/h (200-mm WC) in
the simulation.

Fig. 13—Velocity (m/s) distribution (a) on the midplane and (b) at the outlet of the flow passage between two CIs located on top of the base coil. The total
input flow rate entering the stack was assumed to be 1775 m3/h (200-mm WC) in the simulation.


hot spot of the same coil. There was good agreement mind that the predicted values of the heat-transfer coeffi-
between the predicted and the measured temperatures. This cient are not exact.[28] The results obtained by various
validation of the plant data provides confidence in the use experimenters, even under carefully controlled conditions,
of the mathematical model. differ appreciably. In turbulent flow, the accuracy of a heat-
As has been discussed earlier, it is expected that CIs will transfer coefficient predicted from any available equation or
be very efficient in cooling the coils. The plot of local heat- graph is no better than 620 pct. In the transition region,
transfer coefficients along the radial direction has been where the experimental data are scant, the accuracy of the
shown in Figure 15, where the use of the CI has been Nusselt number predicted from available information may
compared with the use of the convector on top of the base be even lower.
coil of a stack for a total input flow rate of 10,100 m3/h For validation of the model in the present case, the tuned
(which is maximum in a base in the plant). The plot clearly heat-transfer coefficients are 22, 17, 38, and 20 W/(m2 K)
shows that the heat-transfer coefficient at any radial loca- for in, out, top, and bottom surfaces of the coil, respec-
tion is much higher if the CP is replaced by the CI. In the tively. Correspondingly, the computed values from empiri-
same plot, the isolated asterisk with a high value of heat- cal equations are 23.32, 1.83, 27.93, and 18.52 W/(m2 K),
transfer coefficient is for the presence of midbar (refer to respectively, assuming a flow rate of 10,100 m3/h and a
Figure 1(b)) on the convector plate, which tries to restrict suitable flow distribution for the CI. This shows that only
the flow by reducing the cross section of the flow path, thus the heat-transfer coefficient at the outer surface has been
increasing the velocity suddenly due to this sudden contrac- tuned extensively, whereas the other three agree reasonably
tion, and hence, the local heat-transfer coefficient shoots up well. Therefore, the computed values from the empirical
at that radial location. equations of the heat-transfer coefficient can be used to
In the application of any empirical equation for forced obtain a fair idea about the heat-transfer coefficients for
convection to practical problems, it is important to bear in any coil in the stack, at least, in the three surfaces. So, by
using them, the temperature profile of any coil can be rea-
sonably predicted by the heat-transfer model.
Table III. Actual and Predicted Velocities for a Particular
Convector and CI at the Outlet of the Flow
Figure 16 shows the temperature contours of the coil
Passage for Different Flow Rates after 36 hours. It is evident from Figures 14 and 16 that
the hot spot of the coil reaches 50 °C after 36 hours of
Average Velocity cooling. We can also compare the time for the hot spot to
(m/s) at the Outlet reach 50 °C if the convectors are used in the stack. From
the previous discussion, it is imperative to say that the hot
Flow Passage Total Flow (m3/h) Actual Predicted
spot of the coil will take up much more time if the con-
Base convector 1775 (200-mm WC) 0.461 0.42 vectors are used. With the use of the CI, the effective heat-
2173 (300-mm WC) 0.540 0.511 transfer area at the top or bottom of a typical coil goes up
2509 (400-mm WC) 0.874 0.589 from nearly 75 pct (for the CP) to 95 pct.
Cls on the top of 1775 (200-mm WC) 0.776 0.735 Therefore, there is a possibility of reduction of cooling
the base coil 2173 (300-mm WC) 1.161 0.90 time of coils if CIs are used in place of convectors. In the
2509 (400-mm WC) 1.254 1.041
absence of a proper predictive model, the coils were kept
for 40 to 50 long hours on the cooling bases irrespective of

Fig. 14—The validation of the mathematical model showing the predicted

and actual measurements of temperature for two different locations on the
coil. Location 1 is 480 mm below the top edge and location 2 is 280-mm- Fig. 15—Local heat-transfer coefficients along the radial direction on top
inside from outer edge on the top surface of the coil. The predicted temper- of a base coil for the case of convectors (CP) and CIs. The isolated asterisk
ature profile of the hot spot of the coil is also shown here (-m-). The hot is for the presence of the midbar on the convector plate, which reduces the
spot reaches 50 °C after nearly 36 h of cooling. flow area.


coefficient and makes the cooling process faster. More-
over, the effective heat-transfer area is higher in the case
of CIs.
3. The mathematical model can be used to predict the tem-
perature profile of any coil provided the heat-transfer
coefficients are properly tuned. The empirical equations
used for computation of the heat-transfer coefficient give
reasonably good values.
4. The cooling time is expected to be brought down by the
use of CIs and the actual cooling time can be accurately
predicted with the application of the mathematical

1. S.S. Sahay: J. Phys. IV, 2004, vol. 140, pp. 809-17.
2. S.S. Sahay, A.M. Kumar, and A. Chatterjee: Ironmaking and Steel-
Fig. 16—Different contour plots showing spatial temperature distribution making, 2004, vol. 31, pp. 144-52.
after different hours of cooling of the coil obtained from the heat-transfer 3. B. Sarkar, B.K. Jha, and A. Deva: J. Mater. Eng. Performance, 2004,
model. vol. 13, pp. 361-65.
4. J. Shi and D. Cui: Mater. Manuf. Processes, 2003, vol. 18, pp. 51-66.
5. S.S. Sahay and A.M. Kumar: Mater. Manuf. Processes, 2002, vol. 17,
the coil tonnage and grade. For example, the coil for which pp. 439-53.
the mathematical model was validated was kept for 50 6. R.D.F. Antunes, M. Das Gracas, and M.M. Cesar: Metal. Mater. ABM,
2002, vol. 57, pp. 18-23.
hours on the cooling base before it was sent to storage or 7. P.K. Roy, S.M. Merchant, and S. Kaushal: J. Electron. Mater., 2001,
SPM. Therefore, we can clearly appreciate the fact that vol. 30, pp. 1578-83.
there could have been a savings of nearly 14 hours if the 8. S.S. Sahay, A.M. Kumar, S.B. Singh, A.N. Bhagat, and M.S.S. Sharma:
cooling time was optimized with a mathematical model. Tata Search, 2001, pp. 39-46.
For a typical four coil stack (strip width 1485 mm), cooling 9. X. Peng, X. Wang, Z. Ru, and Q. Cao: J. Univ. Sci. Technol. Beijing,
2000, vol. 22, pp. 223-26.
down from the 160 °C hot spot temperature to the 50 °C hot 10. A. Buckley, A.J. Moses, and L. Trollope: Ironmaking and Steelmaking,
spot temperature in the CCSU takes nearly 47 hours if con- 1999, vol. 26, pp. 477-82.
vectors are used for stacking.[29] For the same strip width, 11. C.J. Kang and C.D. Huang: Technol. Training (Taiwan), 1999, vol. 24,
the present mathematical model provides nearly 45 hours to pp. 1-10.
12. A. Fouarge, L. Chefneux, M. Cambier, and H. Meunier: Cahiers d’In-
cool if CIs are used. So, with the use of CIs, the cooling time formations Techn., 1995, vol. 92, pp. 945-51.
is expected to be reduced. This can only be realized with the 13. K. Steck, K.F. Huttebraucker, and H. Fischer: METEC Congr. 94, 2nd
proper application of the mathematical model to accurately Eur. Continuous Casting Conf., 6th Int. Rolling Conf., Dusseldorf,
predict the cooling time while using CIs. Germany, 1994, vol. 2, pp. 319-22.
14. W. Scheuermann, P. Wittler, and T. Bovalina: Metallurgical Plant and
Technology International, 1993, vol. 16, pp. 140-42, 144, 146, 148,
D. Application of the Heat-Transfer Model 150.
15. R.I.L. Guthrie: Engineering in Process Metallurgy, Oxford University
A lookup table has been constructed with the off-line Press, New York, NY, 1989.
data produced for the cooling time of several thousand coils 16. A.R. Perrin, R.I.L. Guthrie, and B.C. Stonehill: Iron Steelmaker, 1988,
having different dimensions. The standard cooling time vol. 15, pp. 27-33.
data for a particular stack could be viewed through the 17. Y. Jaluria and K.E. Torran Computational Heat Transfer, 1st ed.,
Hemisphere Publishing Corporation, Washington, DC, 1986.
look-up table as soon as a stack of coils was built. This 18. K.V.V. Iyer: Steel India, 1984, vol. 7, pp. 43-53.
suggested cooling time would be used by the operator to 19. FLUENTÒ 6.1 User’s Guide, Fluent Inc., Lebanon, NH, 2003.
determine the duration of cooling for a particular stack. The 20. I.E. Idelchik: Handbook of Hydraulic Resistance, 3rd ed., CRC Press,
undercooling or overcooling of coils, thus, can be mini- Boca Raton, FL, 1994.
mized by reducing the human intervention. 21. A. Roth: Vacuum Technology, 3rd ed., Elsevier, Amsterdam, 1990.
22. S.V. Patankar: Numerical Heat Transfer and Fluid Flow, Hemisphere,
New York, NY, 1980.
23. B.E. Launder and D.B. Spalding: Comput. Meth. Appl. Mech. Eng.,
IV. CONCLUSIONS 1974, vol. 3, pp. 269-89.
24. M.N. Ozisik: Finite Difference Methods in Heat Transfer, 2nd ed.,
The following conclusions can be drawn in view of the CRC Press, Boca Raton, FL, 1995.
observations made in the preceding sections. 25. F.P. Incropera and D.P. DeWitt: Fundamentals of Heat and Mass
Transfer, 3rd ed., John Wiley & Sons, New York, NY, 1990.
1. The air flow distribution through different convectors or 26. S. Kakac and Y. Yener: Convective Heat Transfer, 2nd ed., CRC Press,
CIs in a stack falls within a close range for different Boca Raton, FL, 1995.
input flow rates. The distribution is therefore independ- 27. L.C. Burmeister: Convective Heat Transfer, 2nd ed., John Wiley &
ent of the flow rate entering a base. Sons, New York, NY, 1993.
28. F. Kreith and M.S. Bohn: Principles of Heat Transfer, 6th ed., Harper
2. The velocity of air on the top or bottom of any coil and Row, New York, NY, 2000.
increases if intermediate convectors are replaced by 29. Performance Guarantee Test Results for Tata Steel: Technical Manual
CIs. Therefore, it results in an increased heat-transfer for Batch Annealing Furnace, LOI Thermoprocess Gmbh, Germany, 2000.