You are on page 1of 4

Light Metals 2009 Edited by: Geoff Bearne

TMS (The Minerals, Metals & Materials Society), 2009

COMPARISON OF BUBBLE NOISE OF SDERBERG POTS AND PREBAKE POTS


Kjell Kalgraf1, Marianne Jensen2 Tarjei Nordb3, Tor Bjarne Pedersen2
1

Elkem Aluminium Research, P.O.Box 8040, 4675 Kristiansand, Elkem Aluminium Research P.O.Box 128, 4551 Farsund,
3
Elkem Research P.O. Box 8040, 4675 Kristiansand, Norway
Keywords: Sderberg bubble noise, prebake bubble noise
Abstract

The modeling work and the measurements have been done at


Elkem Aluminium Mosjen. Elkem Aluminium Mosjen is one of
two smelters owned by Elkem Aluminium ANS, which is a
partnership between Alcoa (50%) and Elkem (50%). The smelter
is located in Norway and is 100 % based on prebake technology.

In previous work [1] we analyzed the relationship between bubble


noise, bath height, and anode quality with data mostly from
Sderberg pots. We have now acquired additional data for both
Sderberg pots and for prebake pots. From an accidental current
increase from 127 to 145 kA over 2 days at a Sderberg plant, we
can see both the immediate increase of bubble noise proportional
to the square of the current, and the gradual decline of bubble
noise due to increasing crack area in the anode. For prebakes,
however, the response is different. In the short run the bubble
noise is increasing when bath height goes up, because of increased
adhesion when pressure goes up, but after 1-2 hours the bubble
noise has fallen to a value lower than before the bath height went
up. This happens because less current flows in the central regions
when there is more gas there, and more current flows further away
from the central region, giving a steeper slope of the anode closer
to the boundary. In fact, the long term decline of bubble noise for
prebake pots is very similar to the decline found for Sderberg
pots when bath height goes up. Because of the opposite short term
and long term response, the correlation between bath height and
bubble noise of prebake pots can be positive for rapid height
changes and negative for height changes with a bigger time
spacing.

In [1] we showed that the relationship between bubble noise Ubub


(V) and bath height at constant crack area for Sderberg pots can
be approximated by the formula (1)

I
U bub =

gas
A anode
H
ACD

bath

1 leak

0 . 3 ACD

1 . 14 10

CE

100

where =220 -1m-1 is the electrical conductivity of bath, gas


(kg/m3) is the density of the anode gas, I (A) is the line current,
Aanode =15 (m2) is the anode area, CE/100 is the fractional current
efficiency, and =1.5 s is the cycling time of the bubble noise. The
leakage will also depend on crack area, assumed constant here.

Sderberg anodes
Figure 1 shows the gas flows through anode and with bubbles.

The leakage factor, leak, must be estimated. Reference bath height


is set to 0.3 m here. The factor leak is the leakage fraction at the
reference bath height. The leak factor varies from pot to pot due to
variations in the number of cracks and their cross sections. A
bigger bath height immediately increases the gas flow through the
cracks, reducing the bubble noise. The increased gas flow will in
turn slowly widen the cross sections of the cracks, giving a long
term additional reduction of the bubble noise.
The density, gas, of the gas is given by the ideal gas law (2)

gas

mP
k T

(kg/m3)

where m = 44*1.67*10-27 (kg) is the mass of a CO2 molecule,


k=1.38*10-23 J/K is the Boltzmann constant, P (N/m2) is the
pressure, and T (K) is the bath temperature. The pressure (N/m2)
is given by (3)

Figure 1. The gas produced by the reduction process on a


Sderberg pot can accumulate in gas bubbles under the anode and
leave with the bubbles, or flow through cracks in the anode.

473

P = P
+
g H
ACD
top
bath
bath

Bubble voltage before/after bath filling. Sampled every 2.5 minutes.


180
bath filling
ca 10 cm

160

Ptop= 10 (N/m ) is the atmospheric pressure on top of the bath,


bath =2100 (kg/m3) is the bath density, g=9.81 (m/s2) is the
acceleration of gravity, Hbath (m) is the bath height, and ACD (m)
is the anode cathode distance.

140

mV

120
100
80

Vbubble
Vtheory

60
40

Prebake anodes

20
050215:045330

050215:032330

050215:015330

050215:002330

050214:225330

050214:212330

050214:195330

050214:182330

050214:165330

050214:152330

050214:135330

050214:122330

050214:105330

050214:092330

050214:075330

050214:062330

050214:045330

050214:032330

050214:015330

050214:002330

050213:225330

050213:212330

050213:195330

050213:182330

Anode

050213:165330

050213:152330

Figure 3. Theory and observations of bubble noise at bath filling


on a Sderberg pot.

Bubble noise before and after current increase in room 2 EAL. Red curve is
moving average of mearements, and blue curve is calculated from current
change, change of gas channel area, and change of bath height
80

Gas bubble

Theory (blue)
Moving average (red)

70
60

Figure 2. Gas bubble underneath a prebake anode.


mV

50

Prebake anodes have no cracks and gas can not escape up through
the anode. That alters the mechanism of gas release between
Sderberg pots and prebake pots.

40
30
20
10

On a prebake pot gas can accumulate only at the flatter portion of


the anode. Outside a certain distance from the center, the anode
bottom surface inclines. At a critical inclination the bubbles start
to slide off the anode bottom. These regions are free of bubbles.
Gas disappears in the space between the anodes. When a center
bubble starts to approach the critical inclination, it will soon start
to slide up the inclined bottom. The steeper the bottom the smaller
the bubbles will be before they disappear. Close to the anode there
is an adhering layer giving zero velocity.

0
0

Time (days)

Figure 4. Development of bubble voltage before and after a step in


line current from 127 kA to 145 kA and back to 127 kA.
According to theory an increase of line current should give an
immediate increase of bubble noise due to increased gas
production as seen in Figure 4. In this case the line current
accidentally increased from 127 kA to 145 kA over 2 days. Over
days, however, the crack area will increase due to the greater gas
flow through the cracks, giving gradual decrease of bubble noise
as seen in Figure 4.

Practical determination of bubble voltage


Gas bubbles accumulate over 1-3 seconds and are then released.
May be the simplest way to determine bubble voltage is to
measure the peak to peak voltage difference over a 3 second
period in time series of voltages from a pot. When the resistance
is at its minimum, the conducting area under the anode is as big as
possible, and vice versa. Peak to peak give values comparable to
other ways of measuring bubble noise. We measure bubble noise
as 2.5 minute averages of 3 s peak to peak values of the voltage
and values on a prebake pot typically range from 30 to 50 mV and
between 50 to 130 mV on a Sderberg pot.

In Figure 4 we have also entered the calculations from our theory,


including a gradual change of the crack area due to an increase of
the gas flow. The initial rate of decline of bubble voltage is 0.3
mV/(kA*day), or about 0.5 % per kA and day, but the decline will
slow down over time as the anode approach a new equilibrium.
That takes many weeks.
The crack area changes according to (4)

Time variation of bubble noise for Sderberg pots


In Figure 3 the sampling time is 2.5 minutes and we see that the
bubble voltage drops at once when bath height goes up on a
Sderberg pot, the rate of decline being 3 mV/cm bath. Due to the
increased pressure difference more gas is flowing out through
cracks in the anode and less in the bubbles underneath the anode.
Over such a short time total crack area in the anode is constant.

dA
C
= gasflow ( A A0 )
dt
H
Where is a constant, and gas flow (kg/s) is the flow of gas
through the cracks, and C is anode consumption (m/s) and H (m)
is the height of the working zone of the anode, and A (m2) is the
average crack area in the working zone, and A0 (m2) is the crack

474

When bath height increases on a prebake pot, adhesion and hence


bubble size increases. The critical inclination of bubble release
also incrases, and more gas can accumulate underneath the anode.
The bubble noise will increase. With more gas under the flatter
part of the anode, however, the anode consumption in the central
region decreases because less current is flowing there. More
current is flowing further out from the central region, and the
anode consumption there is going up, so the anode is gradually
becoming steeper. Because of this steeper inclination the gas
bubbles start to slide away more easily, and therefore the bubble
size start to decrease after a while, giving lower bubble noise.

area of the freshly baked anode entering the working zone. The
gas flow is proportional to the square of A for laminar flow and
proportional to A for turbulent flow.
With changing crack area the bubble noise formula has to include
leakage due to bath height Hbath and due to crack area A (5)

V bub

1 a

1 . 14 10

gas
H

A
anode

ACD
bath
0 . 3 ACD

CE

100

A
ref

The formulation of bubble noise on a prebake pot can be rather


similar to the formulation for Sderberg pots and be given as (6)

Vbub = frBuAr 1.14 10

where a and b are constants, and Aref is a reference crack area. Just
after the current increase, bubble voltage increases at the present
crack area. Because of the increased gas flow, the crack area starts
to increase, and more gas is leaving through the anode and less
with the bubbles accumulating under the anode, as can be seen in
the falling trend of the bubble voltage. When the current goes
back to 127 kA after 2 days, the crack area is greater than before
the current increase, and the bubble noise drops to a lower level
than before the current increase. However, the crack area is
starting on a slow decrease back toward the old value after the
current is back to normal because the gas flow decreases giving a
slowly rising bubble noise.

I CE

/ ( gas )
A 100

where frBuAr is fractional bubble area under the anode, I is


line current (A) , A is anode area (m2), CE is current
efficiency (%), is the average bubble noise cycling time
(s) , gas is density (kg/m3) of CO2 gas in bubbles, and =
220 electrical conductivity of bath (m)-1.
The fractional bubble area under the anodes, frBuAr, could either
be modeled using a detailed description of the physical
mechanisms of adhesion and change of anode bottom shape due to
consumption when currents flow in the anode, or by modeling the
net effect of the physics. The latter is chosen here. The fractional
bubble area is the product of two factors in the model, namely,
short term fraction, shortFrBuAr, which change rapidly, and long
term fraction, longFrBuAr, which change slowly, i.e. (7)

Time variation for prebake pots.

frBuAr = frBuAr Re f shortFrBuAr longFrBuAr

shortFrBuAr = shortFrBuAr +

dt

inShortFrBuAr shortFrBuAr

Tshort

Tshort = 450 s is chosen so that we get a peak consistent with observations

longFrBuAr = longFrBuAr +

dt
Tlong

inLongFrBuAr longFrBuAr

Tlong= 1200 s is chosen so that we get a decline consistent with


observations
The indicated short term frBuAr is an increasing function of bath height

Figure 5. Measured and simulated bubble noise after bath


filling on a prebake pot.

inShortFrBuAr =

Data shows that bubble noise is rising in the beginning when bath
is added on a prebake pot as shown in Figure 5. After a while the
bubble noise starts to fall, however, and after 1-2 hours the bubble
noise is lower than before bath was added. Thus we have a short
term increase and a long term decline of bubble noise on prebake
pots. The long term behavior resembles that of Sderberg pots,
but the short term behavior is different. Because of the opposite
direction of short term and long term effect, the correlation
between bubble noise and bath height for prebake pots can come
out as positive or negative depending on how fast bath height
changes.

Hbath ACD

Hbathref ACD

2.5

and the indicated long term frBuAR is a decreasing function of bath height

inLongFrBuAr

Hbathref ACD
=

Hbath ACD

3.15

The short term effect moves toward indicated short term


effect with a time constant of 450 s and the long term effect
moves toward indicated long term effect with a time
constant of 1200 s. The direction of the two effects is

475

the adhesion goes up when pressure goes up, and then


gradually increase the slope of the anode bottom so gas
bubbles can escape more easily in the longer run.

opposite, and both depend on the bath height. The long


term effect is stronger than the short term effect, so a
change of bath height that last for a long time will first give
a rise and then a decline of bubble noise to a lower value
than before the rise of bath height. The exponent 2.5 has
been assigned a value such that the short term rise
approaches the observed rise, and the exponent 3.15 is
assigned a value that gives a long term decline corresponding to the observed value of 0.8 mV/cm bath. That gives a
behavior as shown in Figure 5.

Montly averages of Bubble Noise versus Bathheight - ACD at EAM pots 456-482
19.00
y = -0.777x + 29.46
R2 = 0.687

Bubble noise (V)

18.50

From the model we are able to estimate the correlation between


bubble noise and bath height for rapid and or slow changes of bath
height. We alter the bath height with certain time intervals and
sample the bubble noise just before each new change. The results
are shown in Figure 6 and 7. The short term correlation comes out
as 0.6 mV/cm bath when bath height changes every 400 s and as 0.8 mV/cm bath when bath height changes every 5000 s, i.e. a rise
in the short run and an even bigger fall in the long run.

17.00

16.00
14.45

14.85

15.25

15.65

16.05

16.45

16.85

Bathheight - ACD (m)

Figure 8. Measured bubble noise and bath height-ACD


from monthly data for a prebake pot

17.0

Figure 8 shows an example of correlation between bubble noise


and bath height from monthly averages for a prebake pot. The
slope of the curves and the correlation coefficient will vary
depending on how rapid changes of bath height the pot has
experienced, just as expected from the simulation.

y = 31 - 0.766 *x
Bubble noise (mV)

17.50

16.50

Sampled and estimated bubble noise

16.5
16.0
15.5
15.0

Conclusions

14.5
14.0
0.180

0.185

0.190

0.195

0.200

0.205

0.210

0.215

The bubble noise for Sderberg pots and prebake pots has
different mechanisms. In Sderberg pots more gas will flow up
through the cracks in the anode when bath height goes up, and the
change is immediate. The increased flow of gas will start an
increased erosion of the cracks so there will be a long term decline
of the bubble noise to a level almost twice as big as the initial
drop. Prebake pots have no cracks in the anodes and gas is leaving
through the space between anodes. When bath height goes up in a
prebake pot the adhesive forces increase. More gas accumulates in
bubbles under the anode in the short run (less than 1 hour), but
this increase of gas reduce the electrical current in the central
region of the anode. The current increases outside of the central
region, and gradually the anode bottom becomes steeper so the
gas bubbles escape more easily. Therefore the bubble noise in
prebake pots increases in the short run, and decrease in the long
run when bath height goes up and stays up for a long enough time.
After 1-2 hours the bubble noise on a prebake pot is lower than
before the increase of bath height. For changes of bath height
with several hours between changes, both prebake pots and
Sderberg pots show a falling bubble noise when bath height
increases, Sderberg pots because the gas flow through cracks in
the anode increases, and prebake pots because the anode bottom
becomes steeper and gas bubbles escape more easily.

0.220

Bath height (m)

Figure 6. Simulated bubble noise vs. bath height for a pot


when bath height change every 5000 s
Bubble noise sampled just before bath height changes every 400 s
17.5
y = 16 + 0.6*x
17

16.5
Vbub (mV)

18.00

16

15.5

15

14.5
0.185

0.19

0.195

0.2

0.205

0.21

0.215

Hbath (m)

Figure 7. Simulated bubble noise vs. bath height for a pot when
bath height changes every 400 s.

The short term rise for prebake pots is not seen for
Sderberg pots because more gas can escape through
cracks in the anode when the pressure goes up. That is not
possible for prebake pots because there are no cracks. So
the only option is to collect more gas in the bubbles since

References

Kalgraf K.et al, Theory of Bubble Noise, Bath Height, and


Anode Quality, Light Metals 2007, pp 357-361 .

476

You might also like