Dedicated to Professor Dr.-Ing. Klaus Wiinnenberg on the occasion of his 651h birthday
Heat Flux Density and Heat Transfer Coefficient between Steel Melt and Metallic
Substrates
Michael Thiemann and Achim R. Buchner
The solidification of low carbon (LC) steels on metallic substrates was investigated with a twin roll caster and a model mould. The substrate
materials were steel, Ni, and two copper alloys, which are typical for the industrial construction of moulds. Heat flux density, heat transfer
coefficient, growth rate, and cooling rate were evaluated.
Keywords: solidification of steel, thin strip casting, heat flux density, heat transfer coefficient.
Experiments
Below the roll nip a compact strip left the caster with a 1.5 cm thickness. At the positions of the thermocouples
typical thickness of about 0.2 cm and 6.5 cm of width. The windows of 1.5 cm width were installed. (In the twin roll
strip surface temperature was continuously measured by a case this effect is not critical as the relative substrate-melt
pyrometer 13 cm below the roll nip. More details are given velocity there is about 5 times larger.)
in [l]. The temperatures in the walls were measured by four
In the steel- or CuNi3Si-rolls special information on heat equidistant NiCr-NiAl thermocouples in each wall covered
flux was to be obtained by a NiCr-NiAl thermocouple in with a sheathing ofNiCr16Fe. They were fixed at the meas-
one of the rolls 0.1 cm below the surface. Since the experi- urement position by a bajonet connection in combination
ments lasted less than 50 sec, the thermocouple cable could with an elastic spring, to ensure a proper contact pressure in
be coiled beside the casting roll, thus enabling a simple data spite of thermal expansion. The positions were 6 cm above
transfer to the registration device, which realises a data ac- the filter, 0.1 cm beyond the wall interface. The thermocou-
quisition rate of 500 per second. Due to the measurement ple contact devices were protected against overheating by
deficiencies discussed below, the real temperature curves four water cooled copper blocks. Figure 3 shows examples
are not obtained directly; nevertheless important informa- of measured curves. The large scatter within one wall is vis-
tion on heat transfer is estimated in an approximative way ible as well as the differences between the walls.
from the measured curves. The thermal resistance between the thermocouple tip and
the wall as well as the heat loss by the thermocouple sheath-
Model mould experiments ing and wires cause an inaccuracy in the measured temper-
ature signal. Following to [9-10] each thermocouple was
The bottom casting experiments were executed with a
calibrated, and the measured temperature signals were cor-
model mould, figure 2, in a vacuum furnace.
rected. Figure 4 exemplarily shows the measured and cor-
An induction melt of 17 kg steel was poured into a
rected temperatures in the steel wall.
tundish. The melt passed a ceramic filter and filled the
chamber (14 cm x 14 cm x 22 cm) between the walls. The
super heating was typically 100 °C, and the maximum pool 400
level was about 12 cm. The melt temperature was controlled steel
350
by two Pt-PtRh thermocouples in a position of 1 cm and 6 CuNi3Si - - - - - --
cm above the filter. The wall materials were the same as the 300
roll materials mentioned above. In contrast to the twin roll
250
experiments and in order to investigate the role of nickel
substrate over a longer time a wall made of pure nickel in- E 200
stead of copper covered with O. lcm nickel was used. The I-
150
wall thickness was 3.5 cm.
In principle the heat flux within the wall can be faster 100 ·.· ....: ...·
than the increase of the melt level, and then this problem .-- ·: ::: ..-- .,----:
.. :;::----· .: .. .-.... -~= :: ;; ::: .. :~: :. ~: ......... ::: -
..... :. ::..: ;~- ·: .
50 •.,..,,,;;::::::::::i::::;;;::::::;;;_,:·,... ,-
must be treated as a two-dimensional one; as the melt ve-
locity is only about 4 cm/sec, this effect was estimated to be 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
critical. By a simple measure the heat propagation problem
t [s)
can be reduced to one dimension. To avoid the wall internal
heat movement from the wall ranges below the thermocou- Figure 3. Temperature curves measured at four different positions
ples, the substrate walls were masked by ceramic plates of in the walls of steel and CuNi3Si; low carbon steel, air.
400
measured 0
350 corrected 0
300
250
e:200
I-
150
100
50
Figure 2. Construction of the model mould; 1: tundish, 2: Zr02-filter, Figure 4. Example for the effect of temperature curve correction;
3: NiCr-NiAI thermocouples, 4: Pt-PtRh thermocouples, 5: cooling steel wall; low carbon steel.
copper element, 6: ceramic plate, 7: substrate wall.
0
Steel Ni CuNi3Si CuNiBe
Hot strip just below the roll nip Thermal quantities from twin roll casting experiments
A very detailed study of heat transfer coefficients and
heat flux densities was possible by use of the thermocouple-
equipped rolls. As an example figure 7 shows a whole tem-
perature curve; eight peaks are visible corresponding to
eight passages of the thermocouple through the pool. Any
peak consists of an increase (starting with the contact to the
pool) and a decrease (during the heat removal into the roll
in the rest of roll rotation).
The development of the roH temperature is obtained from
the starting temperatures in any peak (in this case of non
water cooled rolls). The roll surface velocity was calculated
from the mean time interval between the starting points of
the peaks (to be used in the calculation of pool contact
Strip after cooling times).
Figure 8 shows details of a peak in magnification. From
Figure 5. Strip temperature distributions and strip surfaces under
various conditions; strip width 6.5 cm. a visually estimated starting time to the temperature in-
soo-,-~~~~~~~~~~---,-~~ --.----.----.---,--.--,
550
T,
500
450
400 h,
350
~300 F
t- 250
200
150 R.
100
50
~ ~ ~ ~ % w ~ ~ ~ ro
I [s]
Figure 7. Measured temperature curve during the casting experi- Figure 9. Sketch of the pool [3]; F = roll force, d = strip thickness,
ment; low carbon steel, steel rolls. hp = pool height, hip = liquid pool height, hrp = rigid pool height,
T, = starting roll temperature, R = roll radius.
280
260 .I, ._
------------
solid pool
contribution
variation of heat contact in the pool. The roll surface area
with the thermocouple below (area of about 1 mm2) meets
by chance areas with better or worse heat contact in a sta-
240 tistic manner. As the peak height follows from the heat
flown into the roll, it images the variable heat transfer coef-
~ 220 ficients between material and roll.
i=' liquid pool Now heat transfer coefficients in the small area of obser-
200 <S- contribution
vation are estimated, which refer to the liquid pool. For any
180 peakj a local value aj is estimated as follows:
The total heat qJ°1°1 flown over the total pool time shall
160
ti be proportional to the total peak height LlTJ°tal, figure 7;
the heat flown during the time t1 of the liquid pool shall be
M2 M4 M6 MB ~o ~2 ~4 ~6 ~8 ~o ~2
t[s]
a portion of this, which is characterized by the ratio "% liq-
uid pool" as to be deduced from figure 8; it is
Figure 8. Magnification of the third peak from figure 7; to: begin of t:...T/P = t:...Tr01 • (% liquid pool).
pool contact; t1: end of liquid pool = begin of rigid pool. It is assumed in the liquid pool of peak j:
creases due to heat gain from the pool. An important detail (1)
can be discussed at point t1, where the temperature curve
shows a break. As outlined in [3], the pool consists of two µ = unknown factor; t:,,TjP from figures 7 and 8.
main ranges, figure 9. The mean heat flux density is obtained after dividing by
In the upper part (over the height h1p) of the pool there is the liquid pool time t1 (from figure 8):
liquid melt between the growing half shells; in this part the
rolling force F cannot be submitted; thus the heat contact
between half shells and rolls is bad; from estimations in [3]
(2)
this part of the pool is nearly the whole pool, in real cases
e.g. 12 cm long. When solid bridges from the two half shells
begin to meet, the roll force F can be submitted, and the
pressure between material and rolls is growing thus strong- The local heat transfer coefficient is estimated from the
ly increasing the heat transfer. As estimated in [3], this heat flux density as usual by
range of length hrp is less than about 3 mm, i.e. only a very
small portion of the whole pool. It is concluded that t1 in
figure 8 marks a sudden heat transfer increase, i.e. the (3)
change from liquid to solid pool. This could be also con-
firmed by calculating pool-times after [3] and comparing
them with t1 from figure 8. Tuq (LC steel) ... 1530 °C; T,j = roll temperature before
In figure 7 it is visible that the height of the peak is vary- pool contact, see figure 7.
ing strongly. This can be understood by help of figure 5. {Jj and the denominator are assumed to be constant over
The temperature distribution in the strip corresponds to the short times as they are given in a twin roll pool.
The assembly of peaks can be understood as number of Table 2. Calculation of individual heat transfer coefficients ai and
arbitrary local measurements, and the mean value of all lo- heat flux densities for a cast of low carbon steel in CuNi3Si rolls.
cal aj shall be close to a mean a, measured over the whole
a; qj
pool area of about 160 cm 2 (in these experiments). a can Peakj
[W/cm2 K] [W/cm2]
be estimated from the strip thickness, some physical con- I 0.22 326
stants and the pool contact length, to be measured from the
2 0.20 299
coloured shape at the side dams after the experiment (de-
tails see [3]). It is 3 0.16 222
4 0.17 238
6 0.18 236
8 352 0.55 75 84 63 0.25 292 Figure 10. Calculated heat flux densities versus time for steel or
CuNi3Si walls; low carbon steel, air.
aoo~-------------------, 800
700
LC steel, air D 700
LC steel, air D
LC steel + Ti, air ~ LC steel + Ti, air ~
f0
600 LC steel, He - ~600
Ill
LC steel, He -
i500 ~500
Q)
iljj 400 1i,
~400
"O .s
~ 300
.:::: ~ 300
m200
I
200
100 100
0 0
Steel Ni CuNi3Si Cu Ni Be Steel Ni CuNi3Si CuNiBe
Figure 11. Mean heat flux density at the melt-wall interface at the Figure 14. Mean cooling rate of the solidification front.
beginning of contact; model mould experiments, variable alloys,
substrates, and atmospheres.