You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226898805

Coating Weight Model for the Continuous Hot-Dip Galvanizing Process

Article  in  Metallurgical and Materials Transactions B · July 2007


DOI: 10.1007/s11663-007-9037-2

CITATIONS READS

39 2,757

6 authors, including:

Ehab Elsaadawy Joseph R. McDermid


Saudi Aramco McMaster University
25 PUBLICATIONS   208 CITATIONS    169 PUBLICATIONS   1,511 CITATIONS   

SEE PROFILE SEE PROFILE

Andrew N. Hrymak J. Fraser Forbes


The University of Western Ontario University of Alberta
190 PUBLICATIONS   3,261 CITATIONS    164 PUBLICATIONS   2,327 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Polymer Processing and Characcterization View project

HAZ Softening in DP and Martensitic Steels View project

All content following this page was uploaded by Ehab Elsaadawy on 20 November 2014.

The user has requested enhancement of the downloaded file.


Coating Weight Model for the Continuous Hot-Dip Galvanizing
Process
E.A. ELSAADAWY, G.S. HANUMANTH, A.K.S. BALTHAZAAR, J.R. McDERMID,
A.N. HRYMAK, and J.F. FORBES

A coating weight model was developed to describe the pressure and wall shear stress distribu-
tions as functions of slot gap (d) and impingement distance (Z), for the air knife wiping of the
liquid zinc coatings in continuous hot dip galvanizing at ratios of Z/d £ 8. This model was then
used in validation studies in order to predict the coating weight as a function of the process
parameters. The model was based on improved correlations for pressure and shear stress
developed by a combination of experimental and computation techniques, which has resulted in
more accurate predictions of coating weight validated using industrial coil average coating
weight data, particularly for coating weights of up to 75 g/m2. For this region, the maximum
deviation between the predicted and measured coating weights was 8 pct. The coating weight
model was further developed by incorporating a lumped heat-transfer analysis to predict the
solidification ‘‘dry line’’ of the coating. For a typical continuous galvanizing process, the model
predicts an 80 pct coating solid fraction for a coating weight of 130 g/m2 to occur at 15 m from
the air knives, which agrees qualitatively with visual observations in continuous galvanizing
lines.

DOI: 10.1007/s11663-007-9037-2
 The Minerals, Metals & Materials Society and ASM International 2007

I. INTRODUCTION grated into the air-knife wiping process. Essential to this


control system is the inclusion of a mathematical model
JET finishing, jet stripping, or gas jet wiping are all for predicting the coating weight as a function of the
synonyms used to identify a process in which the various process parameters.
thickness of a liquid coating is controlled by the In spite of the importance of the continuous hot-dip
stripping action of a gas jet. One of the more widely galvanizing process, relatively few studies on the gas jet
practiced applications of this technique is in the contin- wiping of liquid metal coatings have appeared in the
uous hot-dip galvanizing (CGL) process. In the CGL literature. Thorton and Graff[1] provided a theoretical
process, a metallurgically clean steel strip is continu- coating weight model based on the ‘‘maximum flux’’
ously immersed in a molten bath of the desired metallic theory for jet stripping in which the jet effect is
coating. Gas jet knives located just above the bath, represented as a body force, i.e., the gravity force is
which typically operate using either air or N2 as the gas, supplemented by a pressure gradient created by the jet.
control the thickness of the liquid metal coating. These In their study, Thorton and Graff[1] ignored the effects
are typically referred to as ‘‘air knives’’ in the contin- due to the wall shear stress of the wiping jet. Ellen and
uous galvanizing industry. The continuous hot-dip Tu,[2] Tu,[3] and Tu and Wood[4] refined the work of
coating process provides a highly adherent coating due Thorton and Graff[1] by considering the shear stress
to the chemical reactions between the molten metal bath imposed on the surface of the liquid coating by the
and the substrate and is significantly less expensive than wiping jet. The addition of the shear stress to the coating
alternative coating methods such as electroplating. In weight model led to better agreement with industrial
order to enhance both line productivity and uniformity coating weight data. However, for relatively small strip-
of coating thickness, a control system is always inte- to-knife distances, Z, the coating weights predicted by
these models, as compared to industrial data, called for
further improvements to the models. Furthermore,
E.A. ELSAADAWY, Postdoctoral Fellow, and A.N. HRYMAK, lower strip-to-knife distances, corresponding to values
Professor, Department of Chemical Engineering, and G.S. HANUMANTH, of Z/d £ 8 (where d is the air-knife slot width or gap),
Research Associate, and J.R. McDERMID, Associate Professor, Depart- are important in practice because they are employed for
ment of Mechanical Engineering, are with the McMaster University, the high quality/lighter coating weight galvanized prod-
Hamilton, ON, Canada L8S 4L7. Contact e-mail: mcdermid@mcmaster.ca
A.K.S. BALTHAZAAR, Applications Engineer, formerly Undergrad-
ucts used in automotive applications. A recent publica-
uate Student, Department of Chemical Engineering, McMaster Uni- tion by Naphade et al.[5] describes a model for the jet
versity, is with Imperial Oil Limited - Nanticoke Refinery, Nanticoke, finishing process for hot-dip coatings on steel strip.
Ontario, Canada. J.F. FORBES, Professor, is with the Department of While the authors have treated pressure as a function of
Chemical and Materials Engineering, University of Alberta, Edmon- the distance along the strip, it is not clear how the shear
ton, AB, Canada T6G 2G6.
Manuscript submitted October 30, 2005. stress is treated. It is essential to express shear stress as a
Article published online May 1, 2007. function of distance along the strip for accurate mod-

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 38B, JUNE 2007—413


eling of the wiping phenomenon. A mechanistically
more detailed, but numerically tedious, approach
involving direct numerical simulation is provided by
Lacanette et al.[6] in their recent article on the simulation
of gas jet wiping in the continuous galvanizing process.
In order to improve on existing coating weight
models, the current study focuses on the development
of improved correlations that better represent the
pressure and wall shear stress distributions in the near-
field region of the jet, for the ratio of strip-to-knife
distance, Z, and nozzle gap, d, of Z/d £ 8, using a
combination of experimental and computational tech-
niques. It should be noted that for Z/d > 8, a Gaussian
distribution has proven to work well for the pressure
distributions on the strip due to the impingement of the
air knife.[2] However, for Z/d £ 8 the Gaussian distri-
bution is observed to be inadequate to describe the ‘‘flat
top’’ pressure distribution profile that is developed when
the potential core of the air-knife jet impinges on the Fig. 1—Schematic diagram showing the near-jet and far-jet regions
strip. This may be one cause for the relatively poor of the hot-dip galvanizing gas jet wiping process.
predictions of coating weight at low Z/d values, because
the pressure gradient is a key boundary condition of any With the preceding assumptions, the two-dimensional
mathematical model. Navier–Stokes equations for a thin film on a flat plate
The rate at which heat is lost from the coating, and reduce to
hence the rate at which the coating solidifies, is a critical  
d 2u dp
step in the hot-dip galvanizing process. This is because l 2  qg þ ¼0 ½1
the nucleation and growth of zinc grains during the dy dx
solidification process[7,8] largely determines the micro- The boundary
structure and surface texture of the coating, and hence  conditions are u = Vs, at y = 0 (no
the optical properties of the galvanized product. The slip), and l @u@y ¼ s, at y = w, where w is the film
latter is particularly important for the appearance of thickness; u and Vs are the velocity of the coating and
painted consumer goods. The proposed coating weight strip, respectively; q and l are the density and viscosity
model includes a lumped heat-transfer analysis that of zinc, respectively; g is gravitational acceleration; p is
allows computation of the solidification rate and, pressure along the strip; and s is the shear stress im-
thereby, the distance traveled by the strip before the posed on the film by the stripping jet. Integrating Eq.
liquid coating completely solidifies. This is a preliminary [1] and applying the preceding boundary conditions
step in the effort to model dendritic solidification, based leads to the following solution:
on phase-field models, to describe the evolution of the  
coating microstructure. y y y  GW 2
u ¼ Vs 1 þ SW  2 ½2
w w w 2

where qW ffiffiffiffiffi is the nondimensional film thickness,


II. COATING THICKNESS MODEL qg
W ¼ w lV s
; S is the nondimensional shear stress,
Figure 1 shows a schematic illustration of the air- ffiffiffiffiffiffiffiffi
s ffi
S ¼ qlVs g; and G is the effective gravitational accelera-
p
knife wiping system in the continuous hot-dip galvaniz- 1 dp
ing process and the coordinate system appropriate to tion, G ¼ 1 þ qg dx .
the coating thickness model. The principal step in the Conservation of mass requires that the net vertical
coating thickness model is the calculation of the steady- mass flow rate at any position is equal to the final
state flux, q, of the liquid, i.e., liquid zinc in the case of coating mass multiplied by the strip velocity. Thus, the
continuous galvanizing, adhering to the steel strip using molten zinc flux, q, is given by
a simplified form of the Navier–Stokes equation.[2] The Zw  
flow of liquid zinc in the coating is assumed to be steady, SW GW 2
q¼ udy ¼ Vs w 1 þ  ½3
isothermal, and incompressible. The coating thickness 2 3
varies along the plate, and fluid properties such as 0
viscosity, density, and thermal conductivity are assumed Introducing
constant. Furthermore, the pressure across the thin qffiffiffiffiffi the nondimensional withdrawal flux,
qg
coating layer is assumed to be constant, because the Q ¼ Vqs lV s
, Eq. [3] transforms to
velocity of the coating layer flow in the direction GW 3 SW 2
perpendicular to the strip surface is small compared to Q¼ þ þ W ½4
the flow velocity in the direction parallel to the strip. 3 2
Also, the effects of surface tension, oxidation, and Following Ellen and Tu,[2] the nondimensional
substrate surface roughness are neglected. film thickness W, corresponding to the maximum

414—VOLUME 38B, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS B


withdrawal flux, Qmax, can be determined by solving experimentally using the McMaster Pilot galvanizing
dQ/dW = 0, and is a function of S and G at any value simulator. The experimental air knife was a two-
of the coordinate x such that dimensional flat nozzle jet with a gap, d, of 1.52 mm
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi fed with air on one side by means of a Sonic SAS 700
S  S 2 þ 4G blower (Sonic Air Systems Inc., Los Angeles, CA) with a
W ¼ ½5
2G 7.5 kW motor. The air knife was set perpendicular to a
flat impingement plate, as shown in Figure 2, at a
Thus, for a given line speed, the maximum nondimen-
distance Z from the plate. The impingement plate was
sional withdrawal flux, Qmax, and the corresponding film
a 0.2 · 0.48 · 0.00635 m aluminum plate mounted on a
thickness, W, will be a function of dp/dx and s at any
manual traverse with x)y)z movement capability. The
coordinate of x. Based on mass continuity, the physically
air knife was kept stationary and the plate was moved
available withdrawal flux, Q, will be the minimum of the
for all tests. The plate could be moved with an accuracy
values of Qmax corresponding to every x value. Once the
of ±0.127 mm in all three directions.
liquid zinc is solidified, the coating has a uniform veloc-
It should be noted that the nozzle used in the current
ity equal to the strip velocity, Vs, and hence the resulting
study is geometrically similar to those used in the
final film or coating thickness, w¥, is given by
industrial continuous galvanizing coating lines. The
q ðQmax Þmin pressure profiles along the strip, p(x), were measured at
w1 ¼ ¼ qffiffiffiffiffi ½6 four different stations located across the strip at 0, 76.2,
Vs qg
lVs 152.4, and 228.6 mm from the center. The pressure
measurements were taken using Endevco 8510B-2 (End-
In order to determine the final film thickness, both the evco Corp., San Juan Capistrano, CA, 92675) pressure
pressure gradient and shear stress imposed by the transducers mounted behind a 0.5 mm pressure tap in
stripping jet on the strip are required. Ellen and Tu[2] the plate. The analog-to-digital conversion of the signal
and Beltao et al.[9] have developed correlations for the from the pressure transducer was carried out using a
pressure and shear stress imposed by the jet based on National Instruments data acquisition board (Austin,
experimental measurements such that TX 78759). All measurements were taken at a sampling
G2
p ¼ pm ecnp ½7
and
n   2o
cnG
s ¼ smax erf 0:833nG
p  0:2np e
p ½8

where pm is the maximum static pressure at the plate,


c = ln(0.5)  )0.693; nG p ¼ x=bp is the dimensionless
distance; smax is the maximum shear stress at the plate;
and bp is the half width of the pressure distribution
(bp = x at which p = 0.5 pm), given by the expression:
 2  
bp Z Z
¼ 0:0019 þ 0:0551 þ 0:4035 ½9
d d d
where Z and d are the distance from the air knife to
the strip and the air-knife gap width, respectively.
With these correlations, the model has been found to
accurately predict coating thicknesses for high normal-
ized strip-to-knife distances, Z/d > 8, but the predic-
tions for low strip-to-knife distances, Z/d £ 8, deviate
from industrial line measurements by as much as
20 pct. Thus, in order to improve the model predic-
tions over a wider strip-to-knife distance (Z/d), the
current study focuses on the development of improved
correlations for the pressure and wall shear stress dis-
tributions in the near-field region of the jet, where the
strip-to-knife distance, Z, and nozzle gap, d, ratio are
less than or equal to eight (Z/d £ 8), using a combina-
tion of experimental and computational techniques.

III. PRESSURE DISTRIBUTION


MEASUREMENT
The pressure distribution imposed on the steel strip by Fig. 2—Schematic diagram of the air knife used in the pressure mea-
the impinging jet from the air knife was measured surements.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 38B, JUNE 2007—415


  h i2:5 Z
rate of 2 kHz. A LabView (National Instruments) d p
program was used to acquire and process the data, ¼ 3:6n3p 1 þ 0:6n4p ; 8 ½12
dnp pm d
which was stored for further processing off-line.
The pressure profiles, p(x), were measured for strip- The corresponding pressure gradient for the Gaussian
to-knife distances of Z/d = 2, 4, 6, and 8 and a plenum model (Eq. [7]) is
pressure of 13,790 Pa and are shown in Figure 3 in
normalized form. This plenum pressure corresponds to a dpG cnG
2

G
¼ 2cnG
p pm e
p ½13
jet Reynolds number (Rej) of 11,000, which is equal to dnp
the value used by Tu[3] and thus facilitates a direct
comparison of results. From the results of Figure 3, it where the superscript G denotes Gaussian model.
can be seen that when a turbulent jet is located less than From the results of Figures 4 and 5, it appears that
6 jet widths from the target surface, the jet potential core both correlations, Eqs. [7] and [10], are valid for the
will impinge upon the surface. Because the jet velocity region 6 £ Z/d £ 8.
profile is no longer self-similar due to the uniform Comparison of the experimental results with the
velocity of the potential core, the surface pressure preceding correlations showed that these two correla-
assumes a top-hat profile in the near-field region (i.e., tions fit the experimental results well, as can be seen in
Z/d £ 8) of the air knife.[10] Figures 4 and 5. The agreement appears to be better in
the case of pressure, whereas the calculated pressure
gradients show a larger deviation from the measured
values.
IV. PRESSURE CORRELATIONS
Based on the experimental and numerical results of
The experimental data and the computed pressure this study, the maximum pressure corresponding to Z/
were used to estimate the values of the half width of the d < 8 can be estimated from the following correlation:
pressure distribution, bp, for Z/d £ 8 by fitting the  2  
results to the following empirical formula using linear pm Z Z
¼ 0:0056 þ 0:0268 þ 1:0108 ½14
regression: po d d
 
bp Z Z Figures 6 and 7 show a comparison of the correlations
¼ 0:0453 þ 0:7921;  8 ½10 for the half width of the pressure profiles and the max-
d d d
imum pressure, Eqs. [10] and [14], respectively, with
The surface pressure data in the near-field region of the experimental and numerical results.
the impinging jet was fit to the following model form:
p h i1:5 Z
¼ 1 þ 0:6n4p ; 8 ½11 V. COMPUTATION OF SHEAR STRESS DIS-
pm d
TRIBUTION
where np is the normalized distance, x/bp, with bp cal-
culated from Eq. [10]. Equation [11] is a modified form Due to the lack of experimental equipment for wall
of the formula that was used by Brenhorst and shear stress measurements in our laboratory, a compu-
Harch[15] to initialize the computational domain for tational approach was used to model the wall shear
near-field measurements in a fully pulsed subsonic air stress imposed on the strip by the impinging jet. The
jet. Differentiation of Eq. [11] gives the corresponding computations were carried out using the FLUENT
correlation for the pressure gradient: 5.0[11] commercial software package.

Fig. 3—Experimental pressure distributions along the center of the Fig. 4—Comparison of the present and Gaussian pressure correla-
strip for low Z/d ratios (Rej = 11,000). tions to the experimental pressure data.

416—VOLUME 38B, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS B


wall functions, the latter was chosen for the current
work.
There are two appropriate wall treatment methods
provided by FLUENT 5.0: standard wall functions and
nonequilibrium wall functions.[11] Nonequilibrium wall
functions are recommended for use in complex flows
involving separation, reattachment, and impingement,
where the mean flow and turbulence are subjected to
severe pressure gradients and change rapidly. In such
flows, improvements can be obtained, particularly in the
prediction of wall shear stress and heat transfer.[11] The
nonequilibrium wall function employs the two-layer
concept in computing the budget of turbulent kinetic
energy at the wall-adjacent cells, which is needed to
solve the k equation at the wall-neighboring cells. The
Fig. 5—Comparison of calculated pressure gradients to the experi- key elements in the nonequilibrium wall functions that
mental data. make them different from the standard wall functions
are as follows.
(a) Launder and Spalding’s log law for mean velocity
is sensitized to pressure-gradient effects.[16]
(b) The two-layer-based concept is adopted to com-
pute the budget of turbulent kinetic energy in the
wall-neighboring cells.
Thus, a comparison with experimental data is crucial
to determine which wall treatment method is the best for
the present case. Available experimental data for com-
parison are those of Tu,[3] which were performed for a
Reynolds number of 11,000 and Z/d = 2. Both wall
treatment methods, i.e., standard wall functions and
nonequilibrium wall functions, were tested with both the
standard k-e and RNG k-e models in order to choose the
most convenient and accurate turbulence model and
wall treatment method combination to be used. Based
Fig. 6—Correlation of the half-width of pressure profiles with on this comparative study, it was found that no matter
impinging distance. which version of the model is used, standard or RNG k-
e models, the wall treatment method is the key element.
Using the turbulence kinetic energy, k, and the turbu-
lence dissipation rate, e, as boundary conditions for the
turbulence quantities of the inlet flow, along with the
standard k-e model with nonequilibrium wall functions,
gives a better shear stress approximation when com-
pared to the available experimental data, as shown in
Figure 8. This figure shows that the k-e turbulence
model correctly captured the linear relation between
wall shear stress and distance in the immediate vicinity
of the stagnation point, in agreement with Guo and
Wood.[14]
The mesh used in computations was refined for each
model in order to check the independence of the solution
from the grid. The full physical domain was solved, not
taking into account geometrical symmetry. Grid clus-
tering was used close to the wall and around the
Fig. 7—Correlation of the maximum pressure with impinging dis- centerline where there are large gradients. Three grids
tance. with different numbers of nodal points were tested to
ensure independence of the numerical results from the
grid density. Both, pressure distribution and wall shear
In the case of a plane jet impinging normally on a flat stress distribution were monitored, and as a result, a
plate, the results using either the standard k-e model or grid of 30,000 nodal points was chosen as it resulted in a
Reynolds stress turbulence model (RSTM) are almost grid-independent solution.
identical.[12] Since the RSTM requires more computa- For Z/d values of 2, 4, 6, and 8, computations
tion time, as compared to the standard k-e model with were performed using the standard k-e model with

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 38B, JUNE 2007—417


enhances the convergence rate to the steady-state solu-
tion. The computed pressure and wall shear stress
profiles along the strip are shown in Figures 9 and 10,
respectively. It should be noted that the pressure
distributions in Figure 9 are normalized by the maxi-
mum pressure (pm) value corresponding to each profile,
and the distance along the strip, x, is normalized by the
half width of the pressure distribution, bp. The shear
stress distributions are also normalized in the same way
for Figure 10 using smax. The half-width distance bp is
used to normalize the distance x in the case of the shear
stress distribution to allow comparison with Tu’s exper-
imental data[3] in Figure 10. A summary of the simula-
tion conditions and the results is shown in Table I.

Fig. 8—Comparison of computed wall shear stress with different


wall treatment methods and the experimental data of Tu and VI. WALL SHEAR STRESS CORRELATIONS
Wood[4].
Due to the lack of experimental data, the correlations
for shear stress for Z/d £ 8 were based solely on the
FLUENT computational results. The correlation for the
location of the maximum shear stress, bs, as a function
of the air-knife parameters has the following form:
 
bs Z Z
¼ 0:0443 þ 1:1687;  8 ½15
d d d
The nondimensional shear stress is correlated with
nondimensional distance, ns, by the following formula:
)
s 0:22n3s
smax ¼ erf ð 0:41n s Þ þ 0:54n s e ; for 0  n s  1:73
s
smax ¼ 1:115  0:24 lnðns Þ; for ns >1:73
½16
where ns = x/bs. The correlation, applied in the range
Fig. 9—Pressure distributions for Z/d £ 8; experimental and compu- 0 £ ns £ 1.73, is a modified form of that developed by
tational results. Beltao and Rajaratnam,[9] and the correlation applied
for ns > 1.73 was obtained by nonlinear regression.
A linear regression of the maximum shear stress
computed by Fluent resulted in the following correla-
tion:
 
smax Z Z
¼ 0:0001 þ 0:0035;  8 ½17
po d d

VII. HEAT-TRANSFER MODEL


The rate at which heat is lost from the coating, and
hence the rate at which the coating solidifies, is a
critical step in the galvanizing process due to the
constraints that this places on minimum cooling tower
heights and the placement of contacting rolls following
Fig. 10—Wall shear stress distributions for Z/d £ 8, computational
the coating bath. Furthermore, the nucleation and
results. growth of zinc grains during solidification[10,11] largely
determines the microstructure and appearance of the
coating, and hence the properties of the galvanized
nonequilibrium wall functions. The first 50 iterations of product. A model that computes the microstructure
the solution were performed with the laminar flow and surface texture of the coating would require a
option of the software enabled. This laminar solution rigorous microscopic analysis involving dendritic solid-
works as an initial solution for the turbulence model and ification, for example, phase-field modeling.[8] Rather,

418—VOLUME 38B, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS B


Table I. A Summary of the CFD Simulation and Results

Z/d Po Pm Pm/Po smax smax/Po bp bs bp/d bs/d


2 8100 8700 1.07 26.66 0.0033 1.35 1.95 0.89 1.28
4 8100 8199 1.01 26.14 0.0032 1.43 1.95 0.94 1.28
6 8100 8109 1.00 22.57 0.0028 1.45 2.25 0.95 1.48
8 8100 8043 0.99 22.26 0.0027 1.48 2.30 0.97 1.51

the emphasis in the present study is on a simple, ZwðxÞ 


macroscopic lumped analysis that yields an overall rate @ @T 
qcp uT ðxÞdy ¼ k ½21
of solidification and thereby an estimate of the place- @x @y y¼wðxÞ
a
ment of the ‘‘dry line,’’ which is nominally the total
distance the strip has to move from the air knives Now, applying boundary condition (3) yields the fol-
before the coating has largely solidified. This informa- lowing equation:
tion is valuable to the operators for maintaining quality
and ensuring safe operation. Z0 ZwðxÞ
@ @
A schematic representation of the hot-dip coating qcpuT ðxÞdy þ qcp uT ðxÞdy
process is shown in Figure 1. For steady, parallel flow of @x @x ½22
a 0
the coating liquid on the steel strip, neglecting viscous 
dissipation and assuming there is no heat generation, ¼ hi ðT  T1 Þ  ez cðTz4  T 4 Þ
1 y¼wðxÞ
that is, flow occurs in the absence of solidification, the
two-dimensional governing equation for energy conser- Equation [22], upon integration, results in the follow-
vation can be simplified as[17] ing expression:

@T @2T d

qcp u ¼k 2 ½18 qcp Vs T ðxÞ þ qcp T ðxÞq ¼ hi ðT  T1 Þ  ez cðTz4  T1


4
Þ
@x @y dx
½23
where u is velocity and T is temperature. The q, cp,
and k are density, specific heat, and thermal conductiv- The strip velocity Vs is constant, and assuming con-
ity, respectively. The following boundary conditions stant values of density (q), the specific heat (cp) of
apply: both the strip and zinc, and the liquid coating volume
flow rate, q, Eq. [23] can be simplified to the following
(1) ¼ 0 at y = )a (the strip centerline), where a is
@T
@y expression:[17]
the half thickness of the plate. This boundary con-
dition reflects either the symmetry of the tempera- dT hi ðT  T1 Þ  ez c ðT 4  T1
4
Þ
ture distribution in the strip, or, as pertains to the ¼ ½24
dx A
present model, a lumped analysis with uniform
temperature across the strip. where T is the lumped temperature along the strip, hi
(2) Ts = Tz at y = 0. This boundary condition states is the heat-transfer coefficient, ez is the emissivity of li-
that the temperature distribution at the interface quid zinc, c is the Stefan–Boltzmann constant, and A
between the strip and the liquid coating layer is is an effective thermal conductivity, equal to
continuous. (aqsCpsVs + qzCpzq), with qs, Cps, and qz, Cpz being
(3) At the free surface of the liquid coating, neglecting the density and specific heat of the strip and zinc,
solidification, the heat flux is given by the follow- respectively. The value of A will vary upstream and
ing expression: downstream from the air knife where the flow rates, q,
are different. The downstream value of q is calculated
@Tz using the maximum flux principle of Ellen and Tu,[2]
kz ¼ hi ðTz  T1 Þ  ez cðTz4  T1
4
Þ ½19 whereas the upstream value of q is calculated using the
@y expression derived by White and Tallmadge:[18]
where kz and ez are the thermal conductivity and emis- sffiffiffiffiffiffiffi
sivity of zinc, hi is the heat-transfer coefficient, c is the 2 lVs
q ¼ Vs ½25
Stefan–Boltzmann constant, and T¥ is the ambient 3 qg
temperature.
Integrating Eq. [18] with respect to y gives For calculating the heat-transfer coefficient, hi, it is
noted that the presence of the impinging jet divides the
ZwðxÞ  wðxÞ strip into two heat-transfer regions, as shown in Fig-
@ @T
qcp uT ðxÞdy ¼ k ½20 ure 1. In the near-jet region, the jet strongly influences
@x @y a the heat transfer, and in the far-jet region, the jet has
a
little or no influence on the heat transfer. For a single
Applying boundary condition[1], Eq. [20] transforms to slot jet impinging on a flat plate, the convective

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 38B, JUNE 2007—419


heat-transfer coefficient for the near-jet region, hi, can  0:5
NuF 1 + Pr
be computed from the empirical formula developed by ¼ 0:5642
Martin:[19] ðr Res Þ0:5 0.4621 + 0.1395Pr0:5 + Pr
½33
Nux 1.53
¼ x
Z
Rem
j ½26
0:42
þ and
Pr 2d 2d þ 1:39
 0:25
where NuN 1 þ Pr
¼ 0:502
1.0 ð / RaÞ0:25 0:492 þ 0:986 Pr0:5 þ Pr
m ¼ 0.695 

1:33 ½27
x Z ½34
2d þ 2d + 3.06
 For convective heat transfer in the near-jet region, Eq.
and 3000 £ Rej £ 90,000, 4 £ Zd £ 20, 4 £ dx  £ 50, and
V ð2dÞ [26] is used to compute the jet heat-transfer coefficient,
Rej = j m . The terms Nuj and Pr are the Nusselt and
hi, whereas in the far-jet region, Eq. [28] provides an
Prandtl numbers defined as Nuj = hi ð2dÞ and Pr = am;
ka estimate of the mixed heat-transfer coefficient, hi, in
ka, t, and a are the thermal conductivity, kinematic
terms of Nusselt number, Nm. Using the appropriate
viscosity, and thermal diffusivity, respectively.
calculated values of hi, Eq. [24] was solved numerically
An essential step in the heat-transfer model is to
for the lumped temperature, T, by the Runge–Kutta
define the boundary dividing the near-jet and far-jet
fourth-order method.
regions, because when the strip moves out of the jet
boundary, Eq. [26] for the heat-transfer coefficient loses
its validity and new correlations will have to be
formulated accounting for the fact that the strip is VIII. SOLIDIFICATION OF THE COATING
now moving in a stationary ambient atmosphere (Fig-
Typically, the temperature difference across the coat-
ure 1). In the far-jet region, the moving strip drags a
ing thickness is of the order of 0.01 K[8] and so can be
layer of air with it and the heat-transfer mechanism is
neglected. Further, assuming that perfect thermal con-
likely to be a combination of natural and forced
tact exists between the steel substrate and coating, the
convection. Correlations for the mixed heat-transfer
solidification of the coating can be described by the
coefficient for this type of flow are available in the
following lumped energy equation:
literature:[20]
"

   # LDf s ðtÞ  Cpav DT ðtÞ ¼ hi ðT ðtÞ  T1 Þ þ ez cðT 4  T1


4
Þ Dt
3:2 3:2 NuF 3:2 NuN 3:2
Num ¼ k ð1  wÞ þ w ½35
kF kN
½28 where L is the latent heat of fusion, Df s(t) is the
change in solid fraction during the time-step Dt, Cpav is
where NuF, NuN, and Num are the local Nusselt num- the average specific heat for the steel strip and coating;
bers for forced convection, natural convection, and and T is the lumped temperature at time t. The solid
mixed convection, respectively, defined with respect to fraction, f s, in the coating as a function of tempera-
the distance along the strip, x, as hi x/ka, with ka equal ture is approximated by the Scheil equation[7]:
to the thermal conductivity of air, and kF, kN, k, and  
w are convection parameters given by the following (Tm  T (t)) 1=j1
f s (T ) ¼ 1   ½36
expressions: (Tm  Tliq )
kF ¼ (r Res þ w Re1 )0:5 ½29 where Tm is the melting point of pure zinc (419.58 C),
Tliq is the liquidus temperature of the Zn-0.2 wt pct Al
coating (418.29 C),[21] and j is the solute partition
kN ¼ (u Ra)0:25 ½30 coefficient, assumed to be 0.25.[7]
Equations [35] and [36] were solved iteratively for the
k ¼ kF þ kN ½31 lumped temperature T and solid fraction f s(T), using a
time-marching scheme with a time-step of 0.005 s. When
the strip entered the far-jet region, the average value of the
w ¼ kN =k ½32 heat-transfer coefficient, hi, for this region was calculated
The Reynolds number is given by Res = Vsx/m and using Eq. [28] and yielded a value of 5.05 W m)2 K)1.
Re¥ = V¥x/m. The term Ra is the Rayleigh number
defined as Ra = gb(Ts)T¥)x3/am. The terms r, x, and
u are functions of Pr, given by r = Pr2/1 + Pr, IX. RESULTS AND DISCUSSION
x = Pr/(1 + Pr)1/3, and u ¼ Pr =1 þ Pr, respectively.
The individual Nusselt numbers, NuF and NuN, for A. Coating Weight
forced and natural convection for the case of a vertically Incorporating the new pressure (Eqs. [10] through
moving plate in a quiescent ambient atmosphere are [14]) and shear stress correlations (Eqs. [15] through
given by the following expressions:[20] [17]) developed in the current work into the air-knife

420—VOLUME 38B, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS B


wiping model, the coating weights for typical industrial the pressure and shear stress correlations developed in
galvanizing process conditions were calculated and the current study. The significant improvement in the
compared to industrial data. Process data used in the predicted coating weights obtained with the new corre-
model, i.e., knife pressure, line speed, knife-to-strip lations becomes apparent by an examination of these
distance, and the corresponding coating weight data, figures. The predictions with the current model are
were obtained from an industrial continuous hot dip especially good for coating weights of less than 75 g/m2.
galvanizing production line. For coating weights up to 75 g/m2, the average deviation
A single average value for coating weight for each coil between the predicted and industrial coating weights is
was used as the experimental data for comparison with approximately 8 pct. However, the ability of the model
model results, as shown in Figures 11 and 12. A more to accurately predict the coating weight diminishes as the
thorough discussion of the sampling technique is avail- coating weight increases beyond 75 g/m2.
able in Balthazaar.[22] Figure 11 compares the industrial
data with model predictions using the pressure and shear
B. Heat Transfer
stress correlations of Ellen and Tu,[2] i.e. Equations [7]
and [8], respectively, whereas Figure 12 shows the same Figure 13 shows the variation of the jet heat-transfer
industrial data compared to predictions obtained using coefficient along the strip in the near-jet region, com-
puted using Eq. [26], for a typical galvanizing process.
The gradient is steep with hi varying from approximately
900 W m)2 K)1 to about 100 W m)2 K)1 over a distance
of 0.3 m from the impingement point. Because the
coating will cool rapidly in the near-jet region, it is
important to define the extent of this region. Above the
jet boundary, the convective heat-transfer rates are
significantly reduced with an average heat-transfer
coefficient of approximately 5.05 W m)2 K)1. In the
absence of any data, it is assumed that the jet boundary
is located at a distance equal to that covered by the
coating in the time it takes to cool to the liquidus
temperature, measured from the impingement point,
which for typical air-knife conditions is about 0.5 m
from the center of jet.
In Figure 14, the coating solid fraction is plotted as a
function of distance from the jet impingement point, for
a bath temperature of 733 K (460 C) and the typical
plant air-knife wiping conditions listed in Figure 14. The
distance from the impingement point the coating has to
travel to cool to the liquidus temperature (418.29 C)
Fig. 11—Comparison of coating weight predictions using Tu’s model was found to be equal to 0.44 m. Superimposed on the
(Eqs. [7] and [8]) and measured industrial line data. curve are the results of a sensitivity test on the jet

Fig. 12—Comparison of coating weight predictions using the current Fig. 13—Heat-transfer coefficients along the strip: Tbath = 460 C,
model (Eqs. [11] and [16]) and measured industrial line data. p0 = 20,685 Pa, Z = 18.24 mm, d = 1.52 mm, Vs = 2 m/s.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 38B, JUNE 2007—421


Fig. 15—Influence of coating emissivity on solid fraction of the coating
as a function of distance: Tbath = 460 C, p0 = 20,685 Pa, Z =
18.24 mm, d = 1.52 mm, Vs = 2 m/s, and w = 20 lm (130 g/m2).
Fig. 14—Solid fraction in the coating as a function of distance from
the air knife: Tbath = 460 C, p0 = 20,685 Pa, Z = 18.24 mm,
d = 1.52 mm, Vs = 2 m/s, w = 20 lm (130 g/m2), and 2z = 0.23: and, by reducing its value to 0.1 from 0.23, the heat flux
(a) jet boundary condition switched at T = Tliq (418.29 C) and (b) ratio drops to 0.62, indicating a linear dependence of the
jet boundary condition switched at x = 0.5 m.
heat flux on the emissivity in this range. The influence of
coating emissivity on the solid fraction profile is shown
in Figure 15, which indicates that for 80 pct of the
boundary. The dashed curve shows the results for the coating to solidify the required distances are approxi-
case where the jet boundary is arbitrarily shifted by a mately 15 and 13.5 m for ez values of 0.1 and 0.23,
small distance to 0.5 m from the impingement point; i.e., respectively. Hence, a doubling of emissivity reduces the
the jet heat-transfer boundary condition was switched at calculated dry line distance by only about 10 pct.
a distance of 0.5 m. The results suggest that small Coating thickness is a critical property of the galva-
perturbations in the boundary between the near-jet and nization product and it also plays a role in determining
far-jet regions have negligible impact on the solidifica- the microstructure and surface morphology of the
tion rate. For instance, the distance for 80 pct solid coating by influencing the solidification rate. Figure 16
fraction is only about 0.6 pct smaller in the case of the shows the dependence of the solidification rate on the
0.5 m jet boundary. This result is probably due to the coating weight for two different coating weights. It is
fact that the heat of fusion of zinc is sufficiently large observed that, as the coating weight is doubled from
(1.1 · 105 J kg)1) that a small increase in the jet 65 g/m2 (w¥ = 10 lm) to 130 g/m2 (w¥ = 20 lm), the
boundary does not noticeably affect the temperature
profile. The lag in solid fraction, observed in Figure 14,
is equal to the distance moved by the strip in the time it
takes to exhaust the coating sensible heat, i.e., the time
required to cool the coating to the liquidus temperature
(418.29 C) from its initial temperature of 460 C.
The model indicates that 80 to 90 pct of the coating is
solid by the time the strip covers a distance of 15 to 18 m
from the jet impingement point. This result agrees
qualitatively with observations in plant practice where
the freezing line was judged from a visual inspection of
the coating surface.
While convection dominates heat transfer in the near-
jet region, the situation is different in the far-jet region.
Preliminary computations done with a uniform emissiv-
ity of 0.23 demonstrate that, in the near-jet region,
although radiation increases steadily as the strip moves
away from the point of impingement, it is still less than
12 pct of the convective heat flux. This situation changes
significantly in the far-jet region. In this region, the ratio
of the radiative heat flux to convective heat flux (qr/qc)
Fig. 16—Solid faction in the coating as a function of distance from
has a value of approximately 1.4, indicating that the air knife for two different coating weights: Tbath = 460 C,
radiation is the dominant heat-transfer mechanism. As p0 = 20,685 Pa, Z = 18.24 mm, d = 1.52 mm, Vs = 2 m/s, and
expected, radiation heat flux is sensitive to emissivity 2z = 0.23.

422—VOLUME 38B, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS B


distance to 80 pct solidification increases by about ACKNOWLEDGMENTS
38 pct, from 9.3 to 15 m. There are no data available
in the literature on solidification rates of galvanized The authors thank the International Lead and Zinc
strips to compare with the present results, but plant Research Organization (ILZRO) for supporting this
operators suggest that the values predicted by the model study. In particular, the encouragement and advice of
are in the visually observed range. Experimental mea- Dr. Frank Goodwin are gratefully acknowledged. Pre-
surements of solidification rates are necessary in order to vious work that led to the model was supported by the
fully validate the current model. A more sophisticated Natural Sciences and Engineering Research Council of
model might include surface oxidation of the coating, a Canada (NSERC) and Dofasco Inc. We would like to
more precise description of the strip/coating interface, thank Mr. Norm Forrester and Dofasco Inc. for sup-
and multidimensional strip-width effects.
porting the validation studies. Thanks are also due to
Jennifer Knight for her contribution to the pressure
measurements.
X. CONCLUSIONS
A coating weight model was developed for air-knife
wiping of the liquid zinc coating in the continuous hot NOMENCLATURE
dip galvanizing process in order to predict the coating
weight as a function of the process parameters. The a half-thickness of the strip, m
model was based on improved correlations for pressure bp half-width of the pressure distribution, m
and shear stress developed by a combination of exper- bs location of the maximum shear stress, m
imental and computation techniques, which has resulted cp specific heat, J kg)1 K)1
in more accurate predictions of coating weight, partic- c constant, c = ln(0.5)
ularly in the low Z/d regions. In validation runs, the d air-knife (nozzle) gap

width,Rndp =2 1.52 mm
model showed good agreement with industrial coil erf error function, erf np ¼ p2ffiffip et dt
average coating weight data, particularly at lower fs solid fraction 0

coating weights. For coating weights up to 75 g/m2, G dimensionless effective gravitational acceleration
the maximum deviation between the predicted and g gravitational acceleration, g = 9.81 m s)2
measured coating weights was 8 pct. hi heat-transfer coefficient, W m)2 C)1
The coating weight model was further developed by k thermal conductivity, W m)1 K)1
incorporating a lumped heat-transfer analysis to predict Nu local Nusselt number, Nu = hi x/ka
the solidification ‘‘dry line’’ of the coating. For a typical p static pressure on impingement plate, Pa
continuous galvanizing process, the model predicts a dry pm maximum static pressure on impingement plate,
line distance for a coating weight of 130 g/m2 of 15 m Pa
for 80 pct coating solid fraction, which agrees qualita- p0 plenum pressure, pressure inside the plenum
tively with visual observations of continuous galvanizing chamber of the air knife, Pa
lines. Pr Prandtl number, Pr = m/a
The modeling approach in this study uses a proven Q dimensionless withdrawal flux
analytical solution for the coating film flow, with the jet q volumetric flow rate per unit of film width, m2 s)1
pressure and shear stress distribution terms described by Ra Rayleigh number, Ra = gb(Ts )T¥)x3/am
correlations developed using experiments and CFD Re Reynolds number
simulations. This semianalytical approach has allowed S dimensionless wall shear stress
the implementation of the model in an Excel-VBA based W dimensionless film thickness
spreadsheet application, which can be readily incorpo- T temperature, K
rated in process control strategies to control the coating t time, s
thickness in galvanizing lines. A more rigorous model u velocity component in the x direction, along the
that can, for example, cover a wider range of strip-to- strip, m s)1
knife distance (Z/d > 8) would require the inclusion of Vs strip velocity, m s)1
additional physical factors: a detailed description of the Vj jet velocity at the exit of the air knife, m s)1
jet fluid dynamics that includes the transition from w final coating film thickness, lm
potential flow to jet flow, substrate roughness, morpho- x horizontal Cartesian coordinate direction
logical description of the strip/coating interface, and y vertical Cartesian coordinate direction
multidimensional strip-width effects. A comprehensive Z impingement distance, distance from the nozzle
model should also have the capacity to predict the exit plane to the plate, m
coating properties, which largely determine the quality
of the galvanized product. This would require a descrip- Greek symbols
tion of microstructure evolution during dendritic solid-
ification of the coating, using techniques such as phase- a thermal diffusivity, m2 s)1
field modeling. In our ongoing effort to further develop b thermal expansion coefficient, K)1
the model, work is currently underway to quantify the j solute partition coefficient
influence of substrate roughness on the coating thick- c Stefan–Boltzmann constant,
ness. c = 5.67 · 10)8 W m)2 K)1

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 38B, JUNE 2007—423


l coefficient of viscosity, PaÆs 6. D. Lacanette, S. Vincent, E. Arquis, and P. Gardin: ISIJ Int.,
m kinematic viscosity,m = l/q, m2 s)1 2005, vol. 45, pp. 214–20.
7. A. Semoroz, L. Strezov, and M. Rappaz: Metall. Mater. Trans. A,
q mass density, kg m)3 2002, vol. 33A, pp. 2685–94.
s shear stress at the impingement plate, Pa 8. A. Semoroz, S. Henry, and M. Rappaz: Metall. Mater. Trans. A,
smax maximum shear stress at the impingement plate, 2000, vol. 31A, pp. 487–95.
Pa 9. S. Beltao and N. Rajaratnam: J. Hydraul. Res., 1973, vol. 11, pp.
29–59.
np dimensionless distance, np = x/bp 10. D.J. Phares, G.T. Smedley, and R.C. Flagan: Phys. Fluids, 2000,
ns dimensionless distance, ns = x/bs vol. 12, pp. 2046–55.
11. FLUENT 5.0 User’s Guide, Fluent Inc., Lebanon, NH, 1983, vol.
II.
12. M. Bouainouche, N. Bourabaa, and B. Desmet: Int. J. Numer.
Subscripts Methods Heat Fluid Flow, 1997, vol. 7, pp. 548–64.
13. S.E. Kim and D. Choudhury: American Society of Mechanical
a air Engineers, Fluids Engineering Division (Publication) FED, New
j jet York, NY, 1995, vol. 217, pp. 273–280.
s strip 14. Y. Guo and D.H. Wood: Exper. Therm. Fluid Sci., 2002, vol. 25,
z zinc pp. 605–14.
15. K. Brenhorst and W.H. Harch: Turbulent Shear Flows I, Springer-
¥ ambient Verlag,New York, NY, 1977, pp. 37–54.
16. B.E. Launder and D.B. Spalding: Comp. Meth. Appl. Mech. Eng.,
1974, vol. 3, pp. 269–89.
17. A.N. Hrymak, J.F. Forbes, E. Elsaadawy, A. Balthazaar, Q. Hu,
REFERENCES and J. Knight: ‘‘Improving Coating Weight Consistency,’’ Final
1. J.A. Thorton and H.F. Graff: Metall. Trans B, 1976, vol. 7B, pp. Report prepared for the International Lead and Zinc Research
607–18. Organization, Inc, Durham, NC 27713, USA, July 2004.
2. C.H. Ellen and C.V. Tu: J. Fluids Eng., 1984, vol. 106, pp. 399– 18. D.A. White and J.A. Tallmadge: Chem. Eng. Sci., 1964, vol. 20,
404. pp. 33–44.
3. C.V. Tu: Ph.D. Thesis, University of Newcastle, Newcastle, New 19. H. Martin: Adv. Heat Transfer, 1977, vol. 13, pp. 1–60.
South Wales, 1995. 20. H.T. Lin and H.L. Hoh: Heat Mass Transfer, 1997, vol. 32, pp.
4. C.V. Tu and D.H. Wood: Exp. Therm. Fluid Sci., 1996, vol. 13, pp. 441–45.
364–73. 21. S. an Mey: Z. Metallkd., 1993, vol. 84, pp. 451–55.
5. P. Naphade, A. Mukhopadhyay, and S. Chakrabarti: ISIJ Int., 22. A.K.S. Balthazaar: Report No. 4Y04, McMaster University,
2005, vol. 45, pp. 209–13. Hamilton, ON, Canada, 2004.

424—VOLUME 38B, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS B

View publication stats

You might also like