Thermal Analysis of The Hot Dip-Coating Process: Hui Zhang M. Karim Moallemi Sunil Kumar
Thermal Analysis of The Hot Dip-Coating Process: Hui Zhang M. Karim Moallemi Sunil Kumar
Thermal Analysis of The Hot Dip-Coating Process: Hui Zhang M. Karim Moallemi Sunil Kumar
Introduction
The hot-dip coating process has been utilized extensively sidered. Cheung (1985) analyzed freeze coating on a flat plate
during the last few decades in the manufacturing and cladding where constant plate temperature and saturated liquid tem-
of wire, and galvanizing and coating of metal strips, sheets, perature conditions were relaxed. Whereas Seeniraj and Bose
and wires (for example, see Prior and Tonini, 1984; Cook et (1981) predicted a monotonic increase of coating thickness over
al,, 1986). Applications of closed-loop computer controls have the plate, Cheung (1985) was able to predict initial deposition
made this process more attractive by improving product con- followed by remelting. However, both of these studies cannot
sistency and reducing excessive use of coating material (Town- be rigorously used for analyzing dip-coating since they assume
send and Bilski, 1988). Recently, fabrication of tapers, lenslike the existence of leading edges for the solidification front as
waveguides, and mono- and multilayer antireflection coatings well as for the melt flow. In doing so the effects of bath
has renewed interest in the dip-coating technology, where mi- geometry are ignored.
cro-controlled processes are envisioned that would permit pre- The modern applications of dip-coating require sophisti-
cise control of the complex spatial variations of coating cated modeling of heat and mass transfer in a finite molten
thickness (Herrmann and Wildmann, 1983; Tiefenthaler et al., bath where important effects such as axial diffusion, buoyancy,
1983). In spite of the broad range of applications, a funda- and shear-induced melt flow must be included. The present
mental understanding of the process heat and mass transfer paper describes a model developed to predict the transient
mechanisms involved during the solidification over a moving solidification process over a circular rod being pulled through
surface in a finite bath is not available. a melt pool at a constant speed. The temperature distributions
The dip-coating process has its roots in dip-forming, which in the solid and melt, as well as the flow field in the melt, are
is a continuous casting process (Carreker, 1963). The objective calculated for different initial temperatures of the rod and its
of the parent process (dip-forming) is to produce moldless Reynolds number. The influence of the natural convection in
castings, which is achieved by pulling a cold wire, rod, or strip the melt pool, subcooling of the solid entering the bath, and
through a molten bath of similar metal. Optimal process uti- the superheating of the melt entering the bath on the rod radius
lization is achieved when the overall solidification is maxi- profile are also investigated and reported. The role of the
mized, and sophisticated modeling of the heat transfer geometric parameters, namely, the radius and depth of the
mechanisms has therefore not been necessary. Horvay (1965) melt pool scaled by the rod radius, are also investigated. An
modeled this process by using simple thermal resistance net- approximate solution is also developed based on the energy
work analysis in a laterally unbounded bath and obtained integral method and presented in the appendix. The predictions
comparisons with experimental results to within 10 percent. of the numerical scheme are found to be in good agreement
In the dip-coating process the objective is to coat wires, with the experimental data. The results of the approximate
rods, or plates with layers of different materials of controlled solution, however, exhibit significant disagreement with the
thickness. Thus it is essential to have an in-depth understanding data, particularly at greater axial positions in the bath, which
of the heat transfer mechanisms involved in the process to is attributed to the simplifying assumptions used in its devel-
predict and control the coating thickness. However, there are opment.
no published works in the area of hot dip-coating that address
these issues, and a few relevant studies are based on oversim- Analysis
plified models. Seeniraj and Bose (1981) performed an analysis The dip-coating system is schematically depicted in Fig. 1.
of freeze-coating of polymeric materials over a moving metallic A metal rod of inlet temperature 7^, and radius rsi is pulled at
substrate by assuming the temperature of the metallic object speed Us through the bath of a molten metal. As the cold rod
to remain constant and uniform, and the molten bath to be enters the bath, a rapid solidification of metal on the rod
at its melting temperature. The first assumption is physically occurs, which may be followed by melt-out if the bath is of a
incorrect while the second limits the analysis to the one con- great height or if the velocity of the rod is small. A semi-
analytical seminumerical approximate solution for the problem
Contributed by the Heat Transfer Division and presented at the National Heat is included in the appendix, and here; after a short review of
Transfer Conference, Minneapolis, Minnesota, July 28-31 1991. Manuscript possible numerical approaches, the mathematical model used
received by the Heat Transfer Division, June 1992; revision received September and solution procedure adopted are explained.
1992. Keywords: Materials Processing and Manufacturing Processes, Moving
Boundaries, Phase-Change Phenomena. Associate Technical Editor: Y. Baya- The dip-coating process, like other solidification processes,
zitoglu. can be modeled via two different approaches. The first and
1.00
xl H
As mentioned earlier, although a steady-state solution was
Fig. 2 Comparison of experimental data with the predictions of the
sought, the problem was cast in time-dependent form for nu- numerical scheme (solid line) and the approximate solution (dashed line)
merical convenience. The initial conditions are arbitrary and
were typically taken to be fully developed flow with the melt
and the moving solid at the fusion temperature, and the sim- grid, the radial position of interface is evaluated by interpo-
ulation of the process started by imposing the thermal bound- lation on the radial side of the grid using the calculated value
ary conditions. The fully implicit form of discretized equations of the solid volume fraction, fs.
were solved iteratively. The time marching calculations were Figure 2 indicates that the numerical predictions compare
conducted over time steps which were initially small (At* = 0.1) well with the experimental data, thus establishing confidence
and were increased gradually (A/Jiax = 5.0) as the solution de- in the model. The approximate solution results are in agreement
veloped. At any time step the solution is considered converged with the numerical predictions and the experimental data at
if the early stages of the solid's travel in the bath (i.e., for small
+1 x*'s). The deviation of the approximate results from the nu-
\r v.j)-ra,j)\ < 10" (18) merical and experimental results increases with x*, which may
Maxl^"+1(/,y')l be explained in terms of the simplifying assumptions employed
where n is the iteration loop counter and 0 is u*, v*, or h*. in the development of the approximate solution. As stated in
The number of iterations required during the early stages of the appendix, the major assumptions of the approximate model
any simulation was about 80, which decreased as the solution are: (a) The axial heat conduction in the solid is negligible,
developed. When the criteria of Eq. (18) were satisfied, the and (b) the melt flow is induced solely by the moving solid
residual source of mass, momentum, and energy was less than (i.e., the effects of bath geometry, free surface, and thermal
10~7 for all the cases examined. The solution was considered buoyancy on the melt flow over the solid are neglected). From
the steady-state one when the change at any point on the solid- these two, the first assumption appears to be the source of the
melt interface was less than 0.1 percent over a time step. discrepancy observed in Fig. 2, since the extent of the dis-
agreement increases with x*. It must be noted, however, that
the degree of agreement or disagreement between the numerical
Results and Discussion results and the approximate solution was found to depend on
Before proceeding with a parametric study, the validity of the governing parameters of the problem, particularly, Re,
the numerical model is verified by comparing its predictions Stes, and Ste/. For example, for the conditions that resulted in
with the experimental data of Carreker (1963) who examined extensive remelting of the solidified layer (e.g., large Ste/), the
coating of pure copper on a copper rod. The limitation of an approximate model predicted smaller solid diameters when
approximate solution, which is presented as an appendix for compared to the numerical predictions. Also, the approximate
the sake of continuity in the text, is also examined by comparing solution is generally in better agreement with the numerical
its results with the experimental data. The representative ther- solutions for smaller Re's, indicating that the boundary layer
mophysical properties of the material used in the experiments approximation and the assumption regarding the existence of
and the operating conditions are presented in Table 1. a leading edge in the melt flow are valid for smaller Re's.
The variation of the rod radius in the bath calculated by the For the parametric study, one set of baseline calculations is
numerical simulation of the experiments is shown in Fig. 2 and first performed and then compared with parametric variations
is compared with those of the original experiment. The pre- of the operating conditions, and bath geometry. The baseline
diction of the approximate solution procedure is also presented case considered corresponds to the following parameters: Re
in the figure for comparison. It should be noted that the ex- = 4xl0 4 , Gr=10 8 , Ste.^2.11, Ste, = 0.40, K = 1 , C,/CS=\,
periments by Carreker were conducted on baths of different rsi/H= 1/20, and R/H= 1. These are in the range of operating
melt depth for different solid (pull) speeds. The results were conditions as indicated by the experiments of Carreker (1963),
presented in terms of casting ratio (cast weight after emerging Table 1, with the exception of Ste., that the experimental con-
over input weight) versus immersion time (calculated from the ditions suggested 21.1. The high value of Ste^ used in the
bath depth and the pull speed). The experimental data pre- experiments is representative of the dip-forming process,
sented in Fig. 2 are all for the same solid velocity, but different whereas in the parametric study, smaller values of Ste^ typical
bath depths, and are plotted against the dimensionless position of the dip-coating process are used. The baseline case serves
in the bath, as presentation versus the immersion time would as a reference for comparison with the results of variations of
have confused the steady-state feature of the data and cal- the various dimensionless parameters.
culations. For the numerical results, at each of the 42 axial Figures 3(a) and 3(b) present the flow fields in the bath for
positions corresponding to the center of the computational Re = 4 x 104 (baseline case) and Re = 2 x 104, respectively, where
. . i | i . i
. ' ' ' ' I '
, | . . -
3.0 - -
F 0.0030 2.5 -
E 0.0018
2.11
D 0.0005
C -0.0007 > -
B -0.0020
A -0.0032 2.0 1.70 -
9 -0.0045
8 -0.0057
7 -0.0070
6 -0.0082 •
5 -0.0095 1.5
4 -0.0108 0.90 "
3 -0.0120
2 -0.0133
1 -0.0145 , . , , I . . . . i . . . i
-
xl H
(b)
Fig. 5 Effect of solid Stefan number on solidification thickness
Fig. 3 Stream function distribution in the bath: (a) Re = 4 x 1 0 4 , (b)
Re = 2 x 1 0 4
ciable increase in solidification after a certain limit. This limit
is a function of the solid and liquid Stefan numbers. Addi-
all other parameters remain at their baseline values. The left tionally, if the residence time were to increase indefinitely, the
boundary of the graphs is the centerline of the moving solid rod would start to remelt even at small levels of Ste/.
rod and the fresh melt is introduced from the upper right The effects of the two Stefan numbers Stes and Ste/ are
corner. It is seen that the flow field established is due to an presented in Figs. 5 and 6. Different values of Stes chosen are
interaction between the buoyancy in the melt and the shear 0.90, 1.70, 2.11, and 2.31 for Fig. 5, and those of Ste, are 0.0,
induced by the solid motion. By reducing the contribution of 0.10, 0.40, 1.00, and 1.60 for Fig. 6. The other parameters are
shear, i.e., decreasing the Reynolds number, a larger natural at their baseline values. The Stefan number of the solid is a
convection cell is generated, Fig. 3(6). The temperature dis- measure of its level of subcooling, and its influence on the
tributions in the bath represented typical characteristics (almost solidification thickness is shown in Fig. 5. The thickness de-
parallel temperature contours) of liquid metal flow, and thus creases with decreasing Stes as expected. The growth is defined
are not presented here. The shear-induced boundary layer over by the energy balance at the interface, and remelting of the
the rod is always present and isolates the rod from the buoy- solidified layer occurs if the subcooling is not sufficient to
ancy-induced motion in the bath. The corresponding variation compensate for the heat transferred to the solid from the mol-
of the solidification thickness is presented in Fig. 4. The Reyn- ten liquid. The liquid Stefan number indicates the level of
olds number is assigned the values 2 x 104, 4 x 104, 2 X 105, and superheating of the molten metal, and its increase reduces the
4x 105 while the other parameters are held to their baseline solidification rate. For large enough values of Ste/ the rod
values. Reducing the Reynolds number leads to increased res- radius may be reduced, even to values smaller than that at the
idence time of the rod in the bath, which results in more inlet.
solidification and larger exit thickness. The variation of the A change of Ste, is accompanied by a change of Gr if the
coating growth with Re is quite nonlinear, and indicates the geometric parameters and phase-change material are kept the
significance of the convection effects. This may be observed same. The variations observed in Fig. 6 are essentially due to
by examining Fig. 4 and comparing the rod thicknesses for change in Ste/. This point is illustrated in Fig. 7, which shows
two values of Re at any given residence time. However, since the results for different values of Grashof number, Gr= 108
the sensible heat of the solid rod is limited by its inlet tem- and 1.6 x 109, the other parameters remaining at their baseline
perature, an increase of residence time will not lead to appre- values. The effect of Grashof number is seen to be minimal
*
•
2.S
« R / H = 2.0 ___- •
_:— —
^ ^ 1.0
^ ^ 0.5
2.0 ~
1.5 -/ -
-
1.0 . . , , i , . , . i . .
0.00 0.25 0.50 , .. 0.75 1.0<
0.00 0.25 0.50 0.75 1 00
xl H
Fig. 6 Effect of liquid Stefan number on solidification thickness x/ H
Fig. 8 Effect of bath radius on solidification thickness
1.00
xl H
xl H
Fig. 7 Effect of Grashof number on solidification thickness
Fig. 9 Effect of solid inlet radius on solidification thickness
and the solidification profiles are nearly the same. This is due
to the fact that the solidification is primarily influenced by the lower temperatures and thus rapid solidification occurs. All
shear-driven boundary layer on the solid rod, which tends to the previous results (Figs. 3-9) are for the case where the
isolate it from the buoyancy effects in the bath. The insignif- bottom wall is held constant at the same temperature Tw as
icant effect of the thermal buoyancy on the solidification rate that specified for the side walls of the molten bath.
justifies the use of forced convection correlation in the de-
velopment of the approximate solution. Conclusions
Figures 8 and 9 explore the effect of geometric parameters Although laminar, mixed convection flow in a finite cavity
on the solidification process. The results indicate that the lat- has been extensively studied in the literature, the related prob-
eral dimension R of the bath, as represented by R/H, is not lem of flow with solidification of a pure medium over a moving
an important variable for the set of parameters considered. surface in a finite bath has received comparatively little atten-
Increasing R results in more solidification because the hot walls tion. Such an analysis, required for the modeling of the dip-
are at a greater distance from the interface. The effect of the coating process, is presented in the present study. A continuum
inlet solid radius rsi is indicated in Fig. 9. The smaller radius model, which includes axial diffusion, thermal buoyancy, and
rod reaches thermal saturation faster due to its smaller thermal shear-induced flow, has been used to investigate the solidifi-
mass, and the ratio of local to inlet areas thus reaches an cation of metal during dip-coating. In general, in the range of
asymptotic value. normal operating parameters, the solidification thickness was
Figure 10 presents the effect of two different types of bound- found to depend strongly on the Reynolds and Stefan numbers
ary condition at the bottom boundary of the bath. The bottom and the ratio of inlet radius to height, and weakly on the
wall is assumed either to be adiabatic, or to be at the same Grashof number and the ratio of bath radius to height unless
temperature as the side walls. The constant parameters for the bath radius is much smaller than the height.
these two cases are those corresponding to the baseline case. An approximate solution for the problem was developed
The solidification thickness for the adiabatic condition is larger using the energy integral method, which is presented in the
because the overall liquid temperatures are smaller in mag- appendix. Its predictions of the variation of the solid radius
nitude than those for the constant wall temperature case. For in the bath did not compare well with the numerical predictions
the adiabatic case, the entering solid rod encounters liquid at or the experimental data. This was attributed to the simplifying
1.5 - I
APPENDIX
1 The purpose of this section is to develop an approximate
analysis of solidification on a moving substrate. As mentioned
1.0 » 1 1 — — 1 ' 1 1 1 1 1 -J 1 1 1 1 ' • • |
in the introduction, the available analyses of the problem are
0.00 0.25 0.50 , ., 0.75 1.00
X / H based on oversimplified models (Seeniraj and Bose, 1981;
Fig. 10 Effect of boundary conditions on solidification thickness Cheung, 1985). In this study, the isothermal conditions for the
solidified layer (Seeniraj and Bose, 1981) and the moving sub-
assumptions employed in the development of the approximate strate (Cheung, 1985) are relaxed. The saturated-melt tem-
solution. perature condition (Seeniraj and Bose, 1981) is also removed.
Future studies of the dip-coating process are required to In other words, the freeze coating of a superheated liquid on
evaluate the non-steady-state operation of the process where a nonisothermal moving substrate with a limited cooling ca-
the solid velocity is continually varied in a controlled manner pacity is investigated.
to obtain thicknesses of complex, albeit prescribed, variations. In addition to the basic assumption made in the general
The dip-coating technology is moving toward such applica- model, Eqs. (l)-(4), the following simplifications are made in
tions, which would be computer controlled with appropriate developing the approximate solution:
feedback mechanisms. In order to prescribe velocity changes 1 A steady-state freeze-coating process is considered. This
to induce exit thickness variations of the solidified metal, the implies that the solid velocity, the bath height, and the
time lag between the cause and effect has to be modeled and far-field (inlet) liquid temperature remain constant in
accurately predicted via thermal analysis of the type presented time. The transient growth of the solidified layer is then
in this study. In addition to the effects included in the present expressed as
analysis, secondary effects such as surface tension will also
have to be included. (A.I)
dt dx
2 The axial heat conduction in the solid is neglected. This
References may be justified because the solid radius (including the
Bennon, W. D., and Incropera, F. P., 1987, " A Continuum Model for Mo- solidified layer) is small compared to the height of the
mentum, Heat, and Species Transport in Binary Solid-Liquid Phase Change
Systems. I. Model Formulation," International Journal of Heat and Mass Trans- bath.
fer, Vol. 30, pp. 2161-2170. 3 The melt flow is assumed to be solely induced by the
Carreker, R. P., 1963, "Dip-Forming—A Continuous Casting Process," Jour- moving solid. The effects of bath walls and free surface,
nal of Metals, Vol. 15, pp. 774-780. as well as the effect of thermal buoyancy on the melt
Cheung, F. B,, 1985, "Analysis of Freeze Coating on a Nonisothermal Moving
Plate by a Perturbation Method," ASME JOURNAL OF HEAT TRANSFER, Vol.
flow, are neglected to enable the development of the
107, pp. 549-556. approximate solution.
Christenson, M. S., Bennon, W. D., and Incropera, F. P., 1989, "Solidifi- In view of the last assumption, only the equation for the
cation of an Aqueous Ammonium Chloride Solution in a Rectangular Cavity— conservation of energy in the solid is solved, and the local
II. Comparison of Predicted and Measured Results," International Journal of
Heat and Mass Transfer, Vol. 32, pp. 69-79. convective heat flux from the warm liquid to the moving solid
Cook, T. H., Mergen, D. E., and Clark, D. L., 1986, "Increasing Profits in is treated as an input parameter, which may be directly obtained
Hot Dip Galvanizing," Metal Finishing, Vol. 84, No. 5, pp. 23-27. from the conventional solution of forced convection over a
Eckert, E. R. G., and Drake, R. M., Jr., 1972, Analysis of Heat and Mass moving surface without phase change such as those given by
Transfer, McGraw-Hill, New York, pp. 439-445.
Eckert and Drake (1972). In this practice, negligible interaction
Epstein, M., 1976, "The Growth and Decay of a Frozen Layer in Forced
Flow," Int. J. Heat Transfer, Vol. 19, pp. 1281-1288. between the melt flow and the shape of the solid-liquid in-
Herrmann, P. P., and Wildmann, D., 1983, "Fabrication of Planar Dielectric terface is implied, which is understood to be only valid for the
Waveguide With High Optical Damage Threshold," IEEE Journal of Quantum growth of thin crusts. Similar arguments have been successfully
Electronics, Vol. QE-19, No. 12, pp. 1735-1738. employed by previous investigators, e.g., Epstein (1976) and
Horvay, G., 1965, "The Dip-Forming Process," ASME JOURNAL OF HEAT
TRANSFER, Vol. 87, pp. 1-16.
Cheung (1985).
Moallemi, M. K., and Viskanta, R., 1986, "Analysis of Melting Around a With the above assumptions, the equations governing the
Moving Heat Source," International Journal of Heat and Mass Transfer, Vol. conservation of energy in the solid in terms of its temperature
29, pp. 1271-1282.
Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, McGraw-
Ts (r, x), and the solid profile rs(x) can be written as follows:
Hill, New York.
Prandtl, V. C , and Dawson, P. R., 1983, "Application of Mixture Theory
to Continuous Casting," in: Transport Phenomenon in Materials Processing,
M. M. Chen, J. Mazumder, and C. L. Tucker, eds., ASME HTD-Vol. 29, pp.
47-54. pUsh/ =k r r hx (T T)/ (A-26)
Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T., 1989,
Numerical Recipes, Cambridge University Press, Cambridge, United Kingdom,
p. 554.
% '^£' = *~ "~
Prior, D. C , and Tonini, D. E., 1984, "Hot Dip Galvanizing of High Strength where Ta is the melt temperature far from the interface, equal
Low Alloy Steel," Metal Finishing, Vol. 82, No. 5, pp. 15-19. to Tw of the general model. The local coefficient hx of con-
Nu v = C, Rel /2 Pr (A.3)
" */"
where Rex = Usx/v is the local Reynolds number, and Cx is a
constant order of 0.5. The boundary conditions of the problem
are as follows:
r = rs: Ts=Tf, (A. 4a) h,(T„-T()
dTs
= 0: = 0, (AAb)
dr
d p 9,r* dr* =
39,
(A.5)
dx* J0 RePr a/ dr*
To simplify the solution of Eq. (A.6), different solid tem- r* = r*i - 2Ste,(/<9j; + C, V I (A. 11)
perature profiles are assumed to characterize the thermal de- 3RePr
velopment of the solid as it travels through the bath. The (Note that Bsi < 0 due to the choice of nondimensionahzation.)
temperature profiles are required to satisfy the integral energy The thermal penetration thickness and solid radius are cal-
equation Eq. (A.5) and the boundary conditions Eqs. (A.4a) culated from Eqs. (A.8) and (A.9) up to x* = x* where A(x*)
and (A.4b). As shown in Fig. A. 1, the first profile is consistent = r* (x*), and beyond that a quadratic temperature profile is
with the notion of a thermal penetration thickness A(x*), be- assumed,
yond which the effects of the interface conditions are negli-
gible, i.e., 95=9C 1 for 0 < r <rs (A. 12)
„*2
(U, for ()</•*</•;-A where Qc(x*) is the centerline temperature of the solid, which
2 2 is calculated along with the solid radius by substituting the
( A , [!-(/•*-/•,+A) /A ] for/v -A<r*<rs
above profile into Eqs. (A.5) and (A.6),
where 0 < A(x*) -<r*(x*). Substituting this temperature profile -4[>-;29j=^ (A. 13)
into Eqs. (A.5) and (A.6) results in dx RePr
-24r* drs /cSte, / - 2 9 , CiSte;
4Ars + A 1 = (A.8) (A. 14)
dP ^ " RePrA dx* ~ RePr
•\f~RePTX*
drs /cSte/ -29, CiSte ;
(A.9)
dx* RePr The system of Eqs. (A. 13) and (A. 14) is solved using a fourth-
VRePr? order Runge-Kutta scheme with the starting condition pro-
Equations (A.8) and (A.9) form a system of first-order non- vided from the endpoint solution of system of Eqs. (A.8) and
linear differential equations, which is solved for the unknown (A.9). This marching integration is carried on up to a point
functions r*(p) and A(x*) with Eq. (A.4c) and A(0) = 0 as x*2 where 9 C and thus the entire solid reached the uniform fusion
boundary conditions. The system is integrated using fourth- temperature. Beyond x*, therefore, the first term on the right-
order Runge-Kutta scheme with adaptive step-size control hand side of Eq. (A.7) vanishes, and the rest is integrated to
(Press et al., 1989). The first step of the marching integration yield
required special consideration to ensure the correct starting
behavior since both r* and A have infinitely large gradient at * 2C,Ste,
rs (x )=rs(x2) (A. 15)
x* = 0. The system of Eqs. (A.8) and (A.9), however, may be RePr
simplified using the following facts: The marching integration routines and Eq. (A. 15) are used
* with the bounding conditions
A « r*s and 4 « — (A. 10)
dx* /\>0andx*<1.0