Note: Descriptions are shown in the official language in which they were submitted.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
1
Method of monitoring individual anode currents in an electrolytic cell
suitable for
the Hall-Fleroult electrolysis process
Technical field of the invention
The invention relates to the field of fused salt electrolysis, and more
precisely to the
monitoring of the Hall-Heroult process for making aluminium by fused salt
electrolysis. In
particular, the invention relates to a particular approach determination and
monitoring of
individual anode currents of the Hall-Heroult type using simple equipment and
digital
communication for easy wiring, installation and maintenance.
Prior art
The Hall-Heroult process is the only continuous industrial process for
producing metallic
aluminium from aluminium oxide. Aluminium oxide (A1203) is dissolved in molten
cryolite
(Na3AIF6), and the resulting mixture (typically at a temperature comprised
between 940 C
and 970 C) acts as a liquid electrolyte in an electrolytic cell. An
electrolytic cell (also called
"pot") used for the Hall-Heroult process typically comprises a steel shell, a
lining usually
made from refractory bricks, a cathode usually covering the whole bottom of
the pot (and
which is usually made from graphite, anthracite or a mixture of both), and a
plurality of
anodes (usually made from carbon) that plunge into the liquid electrolyte.
Anodes and
cathodes are connected to external busbars. An electrical current is passed
through the
cell (typically at a voltage between 3.7 V to 5 V) which splits the aluminium
oxide in
aluminium ions and oxygen ions. The oxide ions are reduced to oxygen at the
anode, said
oxygen reacting with the carbon of the anode. The aluminium ions move to the
cathode
where they accept electrons supplied by the cathode; the resulting metallic
aluminium is
not miscible with the liquid electrolyte, has a higher density than the liquid
electrolyte and
will thus accumulate as a liquid metal pad on the cathode surface from where
it needs to
be removed from time to time, usually by suction.
Typical Hall-Heroult cells comprise tens of individual anode assemblies, each
anode
assembly comprising one (or two) anodes connected to an anode rod, said anode
rod
being mounted on the anode busbar (so-called "anode beam"). Anodes made from
carbon
are consumed during cell operation and need to be replaced. As their thickness
is
decreasing due to burn-off, their height needs to be adjusted regularly. In
order to improve
the monitoring and control of cell operation, it would be desirable to be able
to individual
anode currents and their distribution, as an indication of spatial variations
of cell operation
conditions. However, the currently available cell monitoring systems are
largely dependent
on the cell resistance. The cell resistance represents a combination of the
average anode-
cathode distance (ACD), global bath composition and bath properties in the
cell, which is
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
2
not able to reflect the spatial variations in the cell. As such they do not
help detect
localized abnormalities, such as so-called anode effects, until they become
severe and
apparent on the cell resistance.
To obtain more information of the process, prior art has proposed several
methods in view
of measuring the current through each anode. According to a first known
solution, anode
current is calculated from Hall Effect sensors. Such a method is described for
example by
J.W. Evans and N. Urata ("Wireless and Non-Contacting Measurement of
Individual
Anode Currents in Hall-Heroult Pots; Experience and Benefits", Light Metals
2012, TMS,
p. 939-942). This first proposition requires however significant modelling
effort for each
smelter cell design and may incur significant costs in deployment and
maintenance. It is
limited to low sampling rates (typically 1 Hz) for acquisition of anode
current data.
Other prior solutions have proposed to calculate anode current from voltage
drop
measurement across a length on the anode rod. To this end, individual sensors,
which are
located on respective anode rods, transmit individually current levels to a
remote
computer. This solution is for example described in US 4,786,379 and WO
94/02859. This
second method has however specific drawbacks. Thus, it can often lead to poor
signal
quality due to the poor contact of signal pickup points, as the electrical
contact to the rod
need to be disconnected each time the anode is replaced. Moreover, this may
cause
damage to wiring.
To eliminate the above drawbacks, it has been suggested to measure voltage
points on
the anode busbar, the number of voltage measurement points being one more than
the
number of anodes. The current specific to each rod is then determined from
difference
calculation (e.g., J. Keniry and E. Shaidulin, "Anode signal analysis: the
next generation in
reduction cells", Proceedings of TMS Light Metals, New Orleans (2008) p.329-
331). This
alternative method may however be complex, as it needs to solve a system of
simultaneous equations of Kirchhoff circuit laws based on the model of the
entire
superstructure.
Another approach is based on the calculation of the currents that pass the
anode bus bar,
at the left and the right of each rod. To this end, the electric potential
drops in a specific
length are first acquired at the same time. The values of currents in the beam
are then
obtained by the Ohm's law and the rod current is obtained according to
Kirchhoff's law.
(e.g. Li et al., "Experiments on measurement of online anode currents at anode
beam in
aluminium reduction cells", Light Metals 2015, TMS, p. 741-745).
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
3
The methods based on the measure of voltage points on the busbar have several
advantages, by comparison with methods based on the measure of voltage points
on the
anode rod. In particular, they lead to more reliable systems, which require
less
maintenance. However, they imply specific drawbacks, especially for what
concerns
global structure of the apparatus. Furthermore, the signals transmitted can be
susceptible
to noises.
The problem that the present invention endeavours to resolve is therefore to
propose a
method of anode current monitoring in an electrolytic cell, which can be
carried out with a
simpler apparatus than in prior art. Moreover, the invention wishes to provide
such a
method, which makes it possible to obtain reliable results, in particular for
what concerns
the noise issue in signal transmission.
Object of the invention
The inventors have identified that the drawbacks of prior art are mainly
linked to the way
signals are transmitted from the individual sensors to the central unit. Thus,
in known
methods, these signals are forwarded in an analog form. Therefore, this
implies the need
for numerous and long signal wires, to carry the raw voltage signals of a
fraction of
millivolts to the central unit, for signal amplification and data acquisition.
Hence it leads to
high wiring and significant maintenance costs. Furthermore the analogue
signals
transmitted, according to prior art, are highly susceptible to noises,
particularly when
signal/noise ratio is low.
According to the invention, the above problem is solved by converting the
signal into a
digital form, in the sensing assembly itself. Therefore, the output of each
sensing
assembly, also called smart sensing assembly, delivers a digital signal, which
is
transmitted to the central unit. As only digital signals are transmitted, the
data
communication is virtually immune to electronic/magnetic interference (EMI)
and other
sources of noises. The adoption of the digital communication bus minimizes the
number of
signal wires. For example, only two wires are required for a local controller
communication
bus if the signals are transmitted electronically.
Typically, representative parameter of current is voltage drop and/or
temperature at said
location. The individual anode current is determined by measuring the voltage
drops and
temperatures at each location on the anode busbar. Any appropriate routine can
be
carried out, such as those described by Yao et al, as well as by Li et al. The
routine
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
4
according to Keniry and Shaidulin ("Anode signal analysis ¨ The next
generation in
reduction cell control", Light Metals, TMS 2008, p. 838-843) may also be used.
Advantageously, the number of sensing assemblies is equal to the number of
anode rods
plus one (additional assemblies are needed near risers). Typically, each
intermediate
sensing assembly is located on the anode busbar, between two adjacent anode
rods,
whereas each of the two end sensing assemblies is located in the vicinity of
the facing
anode rod, outside the latter.
Advantageously all sensing assemblies are connected with the central unit
through a local
controller network bus in a daisy chain configuration. The central unit
receives signals in
the digital form from the sensing assemblies and computes individual anode
currents.
Advantageously, this central unit delivers a cell operating information. To
this end, it
performs preliminary detection and diagnosis of abnormal conditions. This cell
operating
information is digitally transmitted to a signal receiver, for example a cell
controller or a
potroom computer server. Data of this cell operating information can be
transmitted
electronically (using appropriate signal wires) or optically (using a fiber-
optic cable).
A first object of the invention is a method of anode current monitoring in an
electrolytic cell
suitable for the Hall-Heroult electrolysis process, said cell comprising
- a cathode forming the bottom of said electrolytic cell and comprising a
plurality of
parallel cathode blocks, each cathode block carrying at least one current
collector bar
and two electrical connections points,
- a lateral lining defining together with the cathode a volume containing
the liquid
electrolyte and the liquid metal resulting from the Hall-Heroult electrolysis
process,
said cathode and lateral lining being contained in an outer metallic shell,
- a plurality of anode assemblies suspended above the cathode, each anode
assembly
comprising at least one anode and a metallic anode rod connected to an anode
busbar
(so-called "anode beam"),
- a plurality of alumina feeders by which alumina powder is fed into the
liquid bath,
said method comprising
- providing a plurality of sensing assemblies at a plurality of locations
along said anode
busbar, each sensing assembly comprising at least one sensing element and
converting means, for converting a measured analog signal into a digital
output,
- measuring with at least one of said sensing element(s) at least one set
of values of a
representative parameter of current at at least one sensing time,
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
- digitalizing said analog signals of values into digital outputs, using said
converting
means, said digital outputs representing the current flow in the anode beam in
the
vicinity of the sensing assembly having generated said digital output.
5 In an advantageous embodiment, said method further comprises:
transmitting said digital
outputs to a common central unit, and calculating from said digital outputs
the current flow
of each individual anode assembly.
In specific embodiments, said representative parameter of current can be the
voltage drop
at the vicinity of said location; each sensing assembly can be a voltage
sensing assembly;
.. said sensing assemblies can be all identical, except for their individual
address. Their
number should be sufficient to enable determination of anode current for each
anode
assembly. All sensing assemblies are connected with the central unit through a
local
controller network bus in a daisy chain configuration. A plurality of sets of
values of a
representative parameter of current, at a plurality of sensing times can be
measured by
said plurality of sensing assemblies. In an advantageous embodiment each
sensing
assembly samples at a programmed rate or predetermined multi-sampling rates
(including
periodically intermittent or on-demand fast sampling). The sampling rate is
preferably
between 0.01 Hz and 200 Hz.
Advantageously the temperature of the anode beam is measured simultaneously to
enable more precise determination of the anode current.
Said central unit delivers a cell operating information; this information can
be digitally
transmitted to a signal receiver. The process according to the invention can
comprise a
.. step of detecting abnormal conditions on the basis of said determined
individual anode
currents. In an advantageous embodiment this step comprises comparing the
oscillation
of at least one individual anode current with a predetermined threshold,
and/or if the FFT
(Fast Fourier Transform) spectrum shows a spike above a given intensity. The
method
according to this embodiment can further comprise taking samples of individual
anode
current to determine at least one parameter of bubble dynamics, if the
oscillation of at
least one individual anode current is superior to said predetermined threshold
and/or if the
FFT spectrum shows a spike above a given intensity. If said parameter of
bubble
dynamics decreases, low alumina concentration can be detected. This process of
monitoring individual anode current can be integrated into a process control
system, and
the method can further comprise moving upwards said anode if said parameter of
bubble
dynamics increases or does not substantially change.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
6
Another object of the invention is a sensing assembly configured to carry out
essential
processing steps of the process according to the invention. Each sensing
assembly
measures and samples a voltage drop on the anode busbar (left or right to an
anode
rod). It includes the functionalities of signal amplification, signal
filtering, data
acquisition, analogue to digital signal conversion, digital signal processing
and digital
signal (data) communication. These functionalities can be realized by using
integrated
circuits, typically including an amplifier, an analogue filter, a
microcontroller that
performs sampling, analogue to digital signal conversion, digital signal
processing and
a network controller for data communication.
The sensing assembly for use in the method according to the invention
comprises:
- a microcontroller configured to carry out essential process steps of the
method
according to the invention, said microcontroller comprising a CPU, a RAM and a
processor,
- at least one analog to digital converter capable of digitizing input
measured analog
signals into digital signals,
- an input channel for analog data.
All sensing assemblies are connected with a central unit through a local
controller network
bus in a daisy chain configuration. The central unit receives anode busbar
current signals
in the digital form the sensing assemblies, computes individual anode
currents, performs
data analysis and transmits the cell operating information digitally to an
appropriate signal
receiver (e.g., a cell controller or a potroom computer server). Data can be
transmitted
electronically (using appropriate signal wires) or optically (using a fiber-
optic cable). As
only digital signals are transmitted, the data communication is virtually
immune to
electronic/magnetic interference (EMI) and other sources of noises. The
adoption of the
digital communication bus minimizes the number of signal wires (e.g., only two
wires are
required the local controller communication bus if the signals are transmitted
electronically).
All sensing assemblies are typically identical. They may be powered by a
common DC
supply which can be integrated into the central unit or alternatively be
powered from the
cell operating voltage (after DC-DC voltage boost and voltage regulation).
Each sensing assembly unit is identical but has a unique identifier number
stored in its
firmware. The information of identifier number is used during configuration of
the
distributed measurement system and is included in the transmission of anode
current
measurement. In this way, the sensing assemblies can be mass produced and
universally
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
7
installed to reduce their costs and the location of the measurement is also
obtained for
individual anode current monitoring. This scheme also enjoys easy maintenance,
as a
faulty unit can be easily identified and replaced.
The central unit processes anode current information received from all sensing
assemblies and calculates individual anode currents and performs preliminary
detection
and diagnosis of abnormal conditions. It transmits processed cell information
to a remote
computer server. The central unit is programmed such that it can decide the
level of
details of cell information depending on the process anomalies it detects, to
avoid
communication congestion.
.. The central unit also performs automatic calibration using an optimization
algorithm.
Figures
Figures 1 to 15 all relate to embodiments of the invention.
Figure la and lb each show a schematic view of a different embodiment of the
invention,
showing an anode busbar provided with sensing assemblies and central unit, for
carrying
a monitoring process according to the invention.
Figure 2 is a flowchart, showing the different steps of an embodiment of the
monitoring
process according to the invention.
Figure 3 is a schematic view, showing with more details the information
exchange
.. between the electronic components of the figure la.
Figures 4, 5 and 6 refer to Example 1.
Figure 4 shows schematically the anodes (Al to A20) and the feeders (F1 to F4)
in an
experiment in which feeder Fl was blocked.
Figure 5 shows the anode current (black line) determined for each of the four
anodes (Al,
A2, A3, A18) near feeder Fl, as well as the PCF concentration (in ppm)
measured in the
exhaust gas of this cell (grey).
Figure 6 shows the anode current (black line) determined for each of the four
anodes (A8,
A9, A10, All) near feeder F4, as well as the PCF concentration (in ppm)
measuredin the
exhaust gas of this cell (grey).
Figures 7 to 9 refer to Example 2.
Figure 7 shows the current determined over nine days for anode A20 (curve Cl)
and
anode A19 (curve 02) before and after the replacement of both anodes A19 and
A20. It
can be seen that the current of anode A20 had increased steadily over more
than one
week up to an abnormal level.
Figure 8 shows a Fast Fourier Transform (FFT) analysis of the current recorded
at anode
A19 in stages 1 to 4.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
8
Figure 9 shows the same analysis for anode A20.
Figures 10 to 12 refer to Example 3.
Figure 10 shows the current evolution (figure 10a) in Amperes and the current
pick-up rate
(in Amperes/hour) (figure 10b) of a new anode.
Figure 11 shows the FFT analysis for a new anode after 30 minutes of anode
setting.
Figure 12 shows the same analysis after 20 hours of anode setting. Figure 13
shows the
current of a new anode with abnormal anode current evolution due to low set
position.
Figure 14 shows a functional diagram for the central unit used for the
implantation of the
process according to an advantageous embodiment of the invention.
Figure 15 shows a functional diagram for the sensing assembly according to an
advantageous embodiment of the invention
Description
1. General presentation
An aluminium smelter comprises a plurality of electrolytic cells arranged the
one behind
the other (and side by side), typically along two parallel lines. These cells
are electrically
connected in series by means of conductors, so that electrolysis current
passes from one
cell to the next. The number of cells in a series is typically comprised
between 50 and
over 100, but this figure is not substantial for the present invention. The
cells are arranged
transversally in reference of main direction of the line they constitute. In
other words the
main dimension, or length, of each cell is substantially orthogonal to the
main direction of
a respective line, i.e. the circulation direction of current.
Each electrolytic cell substantially comprises a not shown potshell, a
superstructure and a
plurality of anodes, two of which 2A and 2B are illustrated on figure 1. This
superstructure
comprises a fixed frame (not shown on the figures) and a mobile metallic anode
beam 4,
hereafter called "anode busbar", which extends at the outer periphery of the
fixed frame.
Each anode 2A, 2B is provided with a respective metallic anode rod 6A, 6B for
mechanical attachment and electrical connection to the anode busbar. For
example,
anode busbar 4 may be provided with any appropriate known means, such as a
pair of
hooks (not shown on the figures), adapted to cooperate in a usual way with
anode rods,
for this attachment.
The Hall-Heroult process as such, the way to operate the latter, as well as
the cell
arrangement are known to a person skilled in the art and will not be described
here in
more detail. It is sufficient to explain that the current is fed into the
anode busbar, flows
from the anode busbar to the plurality of anode rods and to the anodes in
contact with the
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
9
liquid electrolyte where the electrolytic reaction takes place. Then the
current crosses the
liquid metal pad resulting from the process and eventually will be collected
at the cathode
block. In the present description, the terms "upper" and "lower" refer to
mechanical
elements in use, with respect to a horizontal ground surface. Moreover, unless
otherwise
specifically mentioned, "conductive" means "electrically conductive".
2. Sensing assemblies
As shown in two different embodiments on figure la and figure 1 b, the anode
busbar 4
supports the plurality of sensing assemblies 10. An electrolysis cell with n
anodes requires
n+1 sensing assemblies plus a number of extra units depending on the number
and
location of anodic risers 60 (shown on figure 1b). Seven of these sensing
assemblies 10A,
10B , 10C, 100, 10E, 10M, 10N are illustrated on figure 1 b; they are
connected to a
central unit 20. Each assembly is mounted on the anode busbar 4. By way of
example,
the assembly is housed in a protection case (not shown on the figures) with
gasket seals.
As shown in Figure la, each assembly has first sensing elements 101, 102, 121,
122,
141, 142 for voltage and temperature measurements, as analogue signal inputs.
In a
preferred embodiment, one sensing assembly is provided between two adjacent
anode
rods. By way of example, the distance between sensing elements of one single
same
assembly (e.g. 101 and 102 of 10A) is between about 70 mm and about 120 mm.
Moreover, the distance between facing sensing elements of adjacent different
assemblies
(e.g. 102 and 121 of 10A and 10B) is of the order of the distance between
anodes. Each
sensing assembly is linked to the adjacent assembly (assemblies) by a DC power
supply
cable 50 (shown on figure 1b).
As will be explained in more detail below, the sensing assembly 10 also
includes an
analogue low pass filter, a signal amplifier, as well as a microcontroller or
equivalent unit,
configured to execute the process steps necessary to carry out the process
according to
the invention. The microcontroller typically comprises a microprocessor with
analogue/digital converters, digital input/output channels, random access
memory,
EPROM, solid state storage and communication controller. The sensing assembly
10 is
electrically linked to a local controller area bus (typically a CAN bus 30)
transceiver for
digital communications. The local controller area bus transceiver implements
digital signal
by-passing to ensure uninterrupted transmission from normal smart sensing
assemblies to
central unit, should one sensing assembly be faulty.
Figure 3 illustrates data transfer means along the superstructure, in a more
detailed
manner. This figure schematically shows assemblies 10A, 10B and 10C of figure
la, as
well as assembly 10N which is located at the end of the busbar opposite to
assembly 10A.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
Each assembly is provided with a respective microcontroller 11A to 11N. Each
microcontroller has a respective input 12A to 12N, which receives data from
sensing
elements (shown on figure la but not on figure 3). Each microcontroller has
moreover a
respective output 13A to 13C, which delivers data in a digital form.
5 Each assembly is also provided with two connecting ports 14A to 14C, as
well as 15A to
15C which are connected in a daisy chain. The terminal sensing assembly 10N
has only
one connecting point 14N. First connecting port 14A of end assembly 10A is
linked to
central unit 20 via line 25 which is advantageously a CAN bus. Moreover,
second
connecting port 15A of end assembly 10A is linked to first connecting port 14B
of adjacent
10 assembly 11B via CAN bus 30A. Each intermediate assembly (like 10B),
i.e. each
assembly which is not provided at one end of the chain, is linked with a first
adjacent
assembly (like 10A) at its first port (like 14B) via a first CAN bus (like
30A), and is linked
with a second adjacent assembly (like 10C) at its second port (like 15B) via a
second
CAN bus (like 30B).
Figure 15 shows a functional diagram for the sensing assemblies 200 according
to an
advantageous embodiment of the invention. The large arrows indicate flux of
data.
Sensing assembly comprises a microcontroller 210. Said microcontroller 210
comprises a
CPU 211, a RAM 212 and a chip 213 with embedded firmware containing a computer
program that enables the microcontroller 210 to execute essential steps of the
process
according to the invention; said chip 213 may also include the identifier of
the sensing
assembly. Analog data enter the microcontroller 210 through analog input
channels 214
and are digitized by an analog-to-digital converter 215.
Anode beam voltage is measured at the anode beam by a voltage measurement
signal
pickup 250; this signal is amplified by a beam voltage amplifier 251, and
filtered by a low
pass filter 210 before entering the microcontroller 210 through one of the
analog input
channels 214. Anode beam temperature is measure at the anode beam by a beam
temperature sensor 253; this signal is amplified by a temperature signal
amplifier 254, and
filtered by a low pass filter 255 before entering the microcontroller through
one of the
analog input channels 214.
The microcontroller 210 receives electrical power 219 from the central unit
100 (figure 14)
through a power regulator and isolator 221 and an overvoltage protector 222.
Part of the
power 224 is derivated through a power converter 225 to supply appropriate
electrical
power to the beam voltage amplifier 251 and the temperature signal amplifier
254.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
11
Said microcontroller 210 exchanges digital data with the neighbouring sensing
assemblies
(and possibly with the central unit 100 if adjacent to the sensing assembly
under
consideration) through a CAN bus receiver 232, a digital signal isolator 231
and a CAN
bus connector 233 for a Daisy chain configuration.
3. Central unit
The central unit 20,100 typically comprises a microprocessor, a local
controller area bus
(CAN bus), transceivers for communications with all sensing assemblies and an
Ethernet
network adaptor or a fibre optic network adaptor for communication with at
least one
remote computer servers. The central unit is also equipped with a power
regulator with
overvoltage protection to produce regulated voltage from smelter cell voltage
and provide
power to all sensing assemblies through the same daisy chain connection as
used for
data communications.
Figure 14 shows a functional diagram for the central unit 120,100 according to
an
advantageous embodiment of the invention. The large arrows indicate flux of
data. The
central unit 100 comprises a microcontroller 110 and a power unit 120. The
microcontroller 110 comprises a CPU 111, a RAM 112 and a chip with embedded
firmware 113 containing a computer program that enables the microcontroller to
execute
essential steps of the process according to the invention. Since the process
generates
huge amount of data, these data may be stored on additional solid state memory
115,
possibly externally, and/or additional RAM chips 114 can be provided, possibly
externally.
The microcontroller 110 receives electrical power from a power unit 120. Said
power unit
120 receives electrical current 119 from the cell voltage; it comprises a
power regulator
and isolator 121 and an overvoltage protector 122. Said power unit 120 also
supplies
electrical current 123 to the sensing assemblies 200.
Said microcontroller 110 exchanges digital data with the first sensing
assembly through a
CAN bus connector 130, a digital signal isolator 131 and a CAN bus transceiver
100. Said
microcontroller 110 also exchanges digital data with the local area network
through an
Ethernet transceiver and network controller 140. If this data exchange is
carried out
through optical fibres, an Ethernet-to-optical fibre converter 141 is
necessary.
4. Use
In use, each sensing assembly samples, at a programmed rate or prescheduled
multi-
sampling rates (including periodically intermittent or on-demand fast (high-
frequency)
sampling). For example, the duration between two successive samplings may be
between
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
12
200 Hz) and about 0.01 Hz, and preferably between about 10 Hz and about 0.5
Hz. Each
pair of sensing elements permits acquisition of values of a representative
parameter of
anode current, such as voltage drop between the two locations of the elements
of said
pair.
Sampling rate may be constant or variable. In an embodiment the sampling rate
is
constant, for instance between 0.1 and 0.01 Hz. The inventors have found that
this
sampling rate represents a sensible compromise between the depth of
investigation
(which would render it desirable to monitor the process continuously at high
sampling
frequency) and the constraints posed by the transmission and handling of such
an
.. enormous amount of data.
In another embodiment the sampling rate is scheduled in advance, and there is
a base
sampling rate (as in the first embodiment) on which short periods of faster
sampling rate
are superimposed at intervals that are regular or irregular, and for durations
that are
constant or variable. Such periods of faster sampling rate may be of the order
of 1 to 100
Hz. In an advantageous embodiment the interval between two periods of fast
sampling
rate are of the order of 20 minutes to 60 minutes. The sampling can also be
modified as a
result of the monitoring, in order to cope with specific conditions of the
pot.
The need and sampling frequency of high frequency sampling depends on the
events and
specific conditions that are supposed to be monitored. FFT analysis can be
applied on
any kind of periodically sampled data, but the sampling frequency needs to be
commensurate with the time scale of the specific event to be monitored. As an
example,
bubble dynamics typically shows up on a FFT spectrum between 0.5 and 1 Hz,
which
requires a sampling rate in the order of 10Hz. (Theoretically 2 Hz would be
sufficient, but
practically 10 Hz are preferred here due to imperfect low pass anti-aliasing
filters).
.. In one embodiment there is a constant sampling rate at a frequency f1
(typically
comprised between 0.01 and 0.1 Hz on which are superposed bursts of high
frequency
sampling at a frequency f2 comprised between 0.1 and 10 Hz; the duration d2 of
such a
high frequency sampling burst is typically between 0.2 and 3 minutes
(preferably between
0.5 and 2 minutes), and the spacing between two of such high frequency
sampling bursts
is typically between 0.2 h and 2 h (preferably between 0.5 h and 1 h).
Each anode current, i.e. the current value in each anode rod, is then
calculated upon the
basis of the above values of the representative parameter. This calculation is
carried out
according to any appropriate known method. For example, considering anode rod
6A, the
sensing members of 101/102 and 121/122 of the respective assemblies 10A and
10B
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
13
calculate the currents that pass the anode busbar, at the left and the right
of said rod. To
this end, the electric potential drops in a specific length are first acquired
at the same time.
The values of currents in the beam are then obtained by the Ohm's law and the
rod
current is obtained according to Kirchhoff's law, as described by Li et al.
("Experiments on
measurement on online anode currents at anode beam in aluminium reduction
cells",
Light Metals 2015, TMS, p. 741-745).
The input analogue signal of each sensing assembly is of the order of a few
millivolts,
depending on the distance between the voltage measurement points and the
busbar
materials. This input signal is filtered and amplified and converted to
digital form by the
microcontroller. The latter also calculates current on the busbar, using the
Ohm's law and
calibration algorithm. The sensing assembly digitally transmits the busbar
current values
to the central unit via a controller area network bus (e.g., CAN bus)
electronically, or via
an optic fiber (e.g., using a CAN bus-to-fiber converter).
The central unit collects all busbar current values and calculates (by
Kirchhoff's current
law) and analyses all anode current measured simultaneously. Current values on
the
anode beam can be calculated by sensing assemblies. The anode current values
can be
calculated by the central unit. In an embodiment, the central unit receives
anode busbar
current signals in the digital form from the smart sensing assemblies,
computes individual
anode currents, performs data analysis and transmits the cell operating
information
digitally to an appropriate signal receiver (e.g., a cell controller or a
potroom computer
server). Data can be transmitted electronically (typically by using
appropriate signal wires
or wireless transmission means) or optically (using a fiber-optic cable).
In a more detailed manner, referring to figure 3, input 12A to 12N of each
microcontroller
11A to 11C receives data from sensing elements, in analog form. The output 13A
to 13N
of each microcontroller delivers data in a digital form.
We describe here the process stages carried out with in a typical embodiment
of the
invention.
All sensing assemblies are installed on the anode busbar. An electrolysis cell
with n
anodes requires n+1 sensing assemblies plus extra units depending on the
number and
location of anodic risers. One sensing assembly unit is placed on each side
(left and right)
of every anode rod.
(i) Analogue voltage signals across a distance on the anode busbar at the
locations left
and right of anode rods and the temperature at the above locations are
acquired using
analog input channels 12A to 12N of the microcontroller. For example, the
voltage drop
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
14
between 101 and 102 and the local temperature are measured using analogue
input
channels 13A and 12A of sensing assembly 10A.
(ii) The voltage and temperature signals are then converted into a digital
form using the
Analogue to Digital Converter (typically built in the microcontroller), 11A in
this example.
Using the resistivity of the anode busbar corrected with temperature, the
current on the
anode busbar i is calculated from the above voltage drop. The beam current at
the
location left to anode 2A is then calculated using the following formula:
V
/A -
(a + bT)-1
A
where V is the voltage drop measured, / is distance on the anode busbar
between the
voltage measurement points, A is the cross sectional area of the anode busbar,
and T is
the temperature measurement. This temperature correction is necessary because
the
impact of small variations of the busbar temperature on its electrical
conductivity is
sufficient to perturbate the busbar resistance and thus the determination of
the beam
current.
Similarly, the beam current at the location right to anode 2A can be
determined using
sensing assembly 10B, denoted as /B.
(iii) All the data representative of current at different locations on the
busbar is transmitted
to the central unit through all sensing assemblies (connected in a daisy chain-
tyoe
arrangement) using a communication hardware and software protocol (typically
CAN bus)
in every sampling period. The central unit processes these data to determine
anode
currents values. Each anode current is determined from the difference between
the beam
current left and right to the anode rod. For example, the current of anode 2A,
denoted as
/24 can then be calculated from the following equation:
/2A = - /B.
The central unit also communicates with a cell controller or a computer server
via a local
area network (e.g., Ethernet) using, e.g. the deterministic TCP/IP protocol.
At the locations where there are risers, additional sensing assemblies are
needed. For
example, as shown in Figure 1, due to the riser, an additional sensing
assembly 100 is
required to calculate the anode current for 20, which is the difference of the
beam current
from 100 and that from 10E.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
(iv) The above analog signals can be sampled at different sampling rates, as
scheduled or
on demand, which can be programmed in the software (in the firmware) of the
micro
controller.
For example, normal sampling rate can be 2 Hz to observe the trend of
individual anode
5 current. Periods of fast sampling can be scheduled, e.g., to have a 1
minute period for
every 30 minutes, in which a 10Hz sampling rate is implemented. This is useful
to obtain
fast current signal to reflect CO2 bubble dynamics. The Central unit sends
commands to
all sensing assemblies to apply the scheduled fast sampling rate (to all
sensing
assemblies) and return to normal sampling rate after 1 minute.
10 The central unit can also start a fast sampling period triggered by
abnormalities detected
from certain anode current signals (e.g., 2 Hz signals). In this case, the
fast sampling rate
(e.g., 10Hz) is applied to sensing assemblies located to the left and right of
the anode in
question. For example, if anode 2B is in question, a fast sampling rate (e.g.,
10 Hz) will be
applied to sensing assemblies 10B and 10C for a period of time (e.g., 1 to 5
minutes) to
15 collect more information on bubble dynamics. The beam current signal
from 10B will be
down-sampled, in this example, to 2Hz so that it can be used, together with
the beam
current signal from 10A (2Hz data) to calculate the anode current of 2A.
The beam current 10N and 10M are used to calculate the current of anode 2M.
All anode
beam current information is transmitted from the originating assemblies to the
central unit
through all intermediate assemblies. As an example, anode beam current
information is
transmitted from originating end assembly 10N to then central unit 20 via the
data
communication bus through all intermediate assemblies 10(N-1) to 10A (without
the
intervention of these assemblies). This allows the central unit to calculate
all anode
currents.
The central unit also decides the sampling rates depending on the anomalies it
detects to
obtain richer process data for diagnosis of abnormal operating conditions. For
example,
the central unit transmits processed anode current information to above
mentioned remote
computer server via a local area computer network (LAN). During normal
operations, less
detailed current distribution information is transmitted to avoid
communication channel
congestion. When certain abnormal condition is detected, significant details
of the process
data are communicated with the remote computer server to allow further
diagnosis. This
approach allows implementation of very high sampling rate to detect fast
process
dynamics (e.g., bubble dynamics) without causing the burden of large LAN
throughput in
large scale deployment. The central units also coordinate with each other to
give a high
priority to the faulty cell to transmit detailed process data.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
16
5. Detection of abnormal conditions
Abnormal conditions of cell operation can arise in many circumstances, but one
of the
most intricate abnormal condition is the so-called anode effect. Well-known to
persons
skilled in the art, this effect is related to the built-up of an insulating
gas layer under the
anode, leading to an increase in anode potential. Indeed, during the course of
the cell
operation the anode is consumed and its carbon reacts with the oxygen released
through
electrolysis to form carbon dioxide. These carbon dioxide bubbles that form
mainly
underneath the anode need to be released continuously. The dynamics of such
gas
bubbles leads to a perturbation of anode currents that can be detected as
electric noise
.. (oscillation) on the current signal.
Other abnormal conditions that can be monitored by the method according to the
invention are blocked alumina feeders or defective crust breakers, errors in
manual setting
of anodes, data input errors in the pot control system.
.. An example of a detection and diagnosis of abnormal conditions, carried out
according to
the invention, will now be described referring to the flowchart of figure 2.
All the steps of
this flowchart are carried out for at least one anode of the cell,
advantageously for the
majority of these anodes and, preferably, for all the anodes of the cell.
While the
detection/diagnosis method and the instrumentation scheme are independent to
each
.. other, the instrumentation scheme according to the invention does provide
an effective
way (requiring low maintenance and is less susceptible to electronic magnetic
interference) to collect information required for the detection/diagnosis
method.
a/ At step 100, the individual anode current (hereafter IAC) is calculated,
and its
.. oscillation is determined. Current oscillation can be determined by
performing a Fourier
transform on the current signal around a selected frequency. Then, the
determined
oscillation is compared with a predetermined threshold.
b1/ If the value of oscillation is superior to said predetermined threshold,
anode current
will be sampled at a higher rate (e.g. 10 Hz) over a period advantageously
comprised
between 10 sec and 100 sec (typically 1 minute) to capture bubble dynamics and
its
variation.
c1/ At downstream step 200, if a decrease in bubble dynamics has been
determined from
certain anode current signal, it implies reduced alumina concentration in the
vicinity of the
above anode, at stage 210.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
17
c2/ On the other hand, if an increase or a lack of change in bubble dynamics
has been
determined, the magnitude M of the considered IAC is used to determine if
there is anode
setting issues.
Let us consider that this magnitude fulfils at least one of the following
criteria:
- M is superior to x% of average IAC, where x is a predetermined value, and
"average IAC" is calculated upon the basis of the IAC of all the anodes of the
cell.
- M is superior to the sum of the current of multiple anodes:
If at least one criterion is met, this means that the considered anode is set
too low.
Therefore, at step 221, this anode is required to be moved upwards by a
certain distance.
If the magnitude M does not meet the criteria of step 221, no fault is
detected from anode
current signals.
b2/ If the value of oscillation is inferior to said predetermined threshold,
the magnitude M
of the considered IAC is used further to at step 300 to determine the root
cause.
c1/ Let us consider that this magnitude fulfils at least one of the following
criteria:
- M is superior to x% of average IAC, where x is a predetermined value, and
"average IAC" is calculated upon the basis of the IAC of all the anodes of the
cell.
- M is superior to the sum of the current of multiple anodes:
If at least one criterion is met, and IAC noise level increases (step 310),
this means that
the considered anode is set too low. Therefore, at step 311, this anode is
then moved
upwards by a certain distance.
If the magnitude M does not meet the criteria of step 310 and/or if IAC level
does not
increase at step 310, no abnormalities are detected from individual anode
current
measurement.
c2/ Let us consider that this magnitude fulfils at least one of the following
criteria:
- M is inferior to x% of average IAC, where x is a predetermined value and
"average
IAC" is calculated upon the basis of the IAC of all the anodes of the cell
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
18
- M is inferior to multiple anodes (see above explanation).
If at least one criterion is met, this means that the considered anode is set
too high.
Therefore, at step 320, this anode needs to be lowered by a certain distance.
If at step 300 none of the possibilities 310 and 320 are fulfilled, no
abnormalities are
detected from individual anode current measurements.
6. Calibration
According to an advantageous embodiment, the central unit 20 also performs
automatic
calibration using an optimization algorithm. It collects anode bus bar current
measurements during each time when an anode is removed and uses them to
progressively improve the calibration accuracy to capture the effects of the
time-varying
operating conditions and properties of bus bar materials, using a recursive
optimization
algorithm.
Due to the spatial variations in the properties of bus bar materials (e.g.,
resistivity) and
actual distance between signal pickup points, proper calibration is
advantageously
required. When an anode is removed, the anode rod current, i.e. the difference
between
the currents on the anode bus bar left and right to the anode rod, will be
zero. The data
collected each time when an anode is removed for a cycle of anode replacement
are
used, together with the line current, to determine the calibration factor for
each pickup
point using an appropriate optimization algorithm.
During the j-th anode setting, the difference of anode bus bar current left
and right to the
anode rod, denotes as /i+ and If are expected to be zero, with respect to a
reference
current direction. Assume that there are N number of anodes and there will be
N rounds of
anode setting practice. The objective of the calibration is to minimise each
individual
anode current during anode settings.
All the cells are connected in series therefore the total line current is
maintained. Hence,
the sum of all individual anodes current at any given time has to equal to the
total line
current, denoted as IL.
Denote 07 and ar the calibration factors for the anode busbar current
measurement left
and right to the j¨th anode rod respectively. The calibration algorithm can be
written as the
following optimization problem with all the calibration factors as the
decision variables:
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
19
min IVY ¨ )2 I ¨
\
where W is a weighting to penalize the error in individual anode current
measurement
during anode setting.
Constraints can be placed on the calibration factors to reflect the effect of
the anode
busbar structure on the resistivity at different locations.
The above optimization problem can be written in a recursive form so that the
changes in
the calibration factors, sa7 and AaT, are determined as the decision variables
to improve
the calibration results after N set of measurements during anode setting have
been
collected and used for initial calibration.
7. Additional advantages
The invention can be carried out by using a distributed instrumentation scheme
with smart
sensing assemblies with digital communication for real time continuous
individual anode
current measurement. The smart sensing assembly has its own processor which
takes the
local voltage measurements, converts the signal to digital via an A/D
converter and carries
out signal processing and communicates to other sensing assemblies and the
central unit
in digital form.
As explained above, the invention has many advantages for monitoring
electrolytic cells in
a Hall-Heroult electrolysis plant. By itself, such a structure of a
distributed instrumentation
scheme according to the invention provides additional advantages, such as;
(i) By using a digital communication daisy chainõ the number of wires needed
for
transmitting the measurement of local cell conditions is significantly reduced
to
theoretically two wires plus wires for power supply. This leads to significant
benefits of reduction of workload of installation and maintenance.
(ii) By using digital communication, the signal transmission is immune to
noises which
can be a significant problem in instrumentation for Hall-Heroult cells due to
the
small voltage signals (several milli-volts) and significant level of noise
caused by
strong electrical and magnetic fields surrounding the cells.
(iii) Smart sensing assemblies process signals and provide functionalities
including
calibration, preliminary detection of anomalies of local cell conditions.
(iv) Each unit is identical but has a unique identifier number stored in its
firmware. This
scheme also enjoys easy maintenance, as a faulty unit can be easily identified
and
replaced.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
Examples
Operating modern cells at low energy input and electrolyte volume can be
improved by
monitoring individual anode current signals. Several publications have shown
the impact
5 of increasing alumina concentration gradient while operating at a lower
electrolyte volume.
Sequentially, an increase in anode potential beyond the enabling limits of the
co-evolution
of fluorocarbon species permits the formation of bubble resistive film under
the anode and
reduces anode wettability.
10 Example 1: PFC detection
In this example, the impact of a blocked feeder on the spatial variations in
alumina
concentration in the electrolyte across a Hall-Heroult reduction cell was
reflected and
detected by individual anode current distribution. The effect of the resultant
non-uniform
15 current distribution on the initiation of PFC emissions can be analyzed
for corner and
central anodes.
Individual anode electrodes are designed to operate under similar conditions
of electrolyte
composition and electrode potential. However, because of non-uniform alumina
dispersion
20 in the electrolyte, the behavior of individual anodes will be
distinctive with a risk of
individual electrode potentials shifting to levels corresponding to the
discharge of fluoride
ions, with resultant PFC formation. Formation of PFCs is undesirable for
environmental
reasons and is indicative for the cell operating at lower current efficiency;
PFC
accumulation can lead eventually so-called anode effects.
In this example, an industrial Hall-Heroult cell comprising twenty anodes (A1-
A20) and
four feeders (F1-F4) was used. Figure 4 shows schematically the positioning of
the
feeders. One of the feeders (F1) was blocked, and this introduced an overall
alumina
concertation gradient in the cell. During this time, the total amount of
alumina supply to the
cell remained the same while the feeding rate proportionally was increased in
the other
three feeders. An increase in the co-evolution of fluorocarbon species was
triggered
shortly after blocking feeder F1 and this was associated with current
redistribution
predominately in the anodes around feeder F1. To facilitate the analysis, the
cell was
divided into four zones based on the feeder's location as illustrated by the
different grey
scheme in Figure 4.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
21
The steady depletion of alumina in the vicinity of the anodes near feeder Fl
(anodes Al,
A2, A3 and A18) has resulted in a lowering of current at these anodes as shown
in
Figure 5.
On the other hand, due to constant cell current condition and better alumina
concentration
in other zones of the cell, an increase in current for anodes adjacent to
feeder F4 (anodes
A8, A9, A10, All) occurred as shown in Figure 6.
The example shows quick response on the co-evolution of fluorocarbon species
when
introducing a spatial change in alumina concentration in a modern industrial
cell. This
demonstrates the capability of the process according to the invention to
detect at an early
stage abnormal operating conditions that may lead to background PFC emission
or even
anode effects.
Example 2: Spike detection
Irregular variation in individual anode current magnitude is suggested as an
effective tool
to detect local abnormalities in Hall-Heroult cells. An example supporting
this was the
formation of a spike in corner anode A20 of the same cell as illustrated in
Figure 4 which
was removed at 70% of its expected service life because its current has
increased
steadily to 50% in excess of the average. This is illustrated in Figure 7
where the curve Cl
represents the typical current determined for anode A20 and curve 02
represents the
typical current as determined at adjacent anode Al 9 that was changed at the
same time.
Time domain analysis was divided into four stages as summarized in Figure 7
and Table
1. As summarized in Table 1 below, the diversion was linked with other anode
changes
that also impacted superheat, bath flow and mixing.
Table 1:
Analysis of anodes Al 9 and A20 during spike formation process.
Anode A19 Anode A20
Stage current current Comment
magnitude magnitude
Normal Normal Both anodes operate normally (prior of
replacing
1 current current anodes 2 and 3).
(-1% from (+11% from
target) target) Age of both anodes is 13 days.
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
22
Normal Drawing A reduction in superheat under corner
anode A20
abnormal due to changing neighbour anodes.
current
2 higher
(+12% from
current by Presence of any undissolved material as
crust lump
target)
24% or carbon dust results in initiation of
anode spike.
N Drawing Anode A20 continues to draw high abnormal
ormal
abnormal current which suggests that spike is
already
current
3 higher formed and current is flowing through the lower
(-8% from
current by ohmic resistance path due to short-circuit
(localized
target)
27% low ACD).
Drawing
Normal abnormal As spike extends into metal pad, current
draw of
current hi her anode A20 further increases abnormally.
4
(-12% from g
current by
target) Age of both anodes is 21 days.
30%
The above analysis shows that the spike has formed as a consequence of the low
superheat zone due to the recent anode set combined with the possible presence
of
undissolved material or carbon dust under the corner anode which strongly
effects spike
formation in anode A20. Furthermore, activities related to anode setting, such
as cavity
cleaning, could physically direct undissolved material to accumulate under the
corner
anode.
Fast Fourier Transformation FFT analysis of bubble frequency and amplitude
components
of anode A19 (at 10 Hz) did not show significant dynamics during stage 1 since
the anode
was still partially slotted. Both anodes require an extra day to complete the
slot
consumption process. As the slots start to disappear in stage 2, however, a
bubble peak
in the frequency range of 0.7 to 1.0 Hz starts to appear as illustrated in
Figure 8. In stages
3 and 4, anode A19 has continued to show the normal bubble peak due to the
release of
cell gases.
Similarly, anode A20 shows the slight appearance of the same bubble peak at
1.0 Hz as
anode A19 in stage 1 (see Figure 9). However, a change in anode A20 frequency
response was observed for the first time in stage 2 where some low frequency
component
appears with a slight bubble peak shift to 1.5 Hz. When an anode is partially
shorting and
is therefore drawing a higher current the electrode potential is automatically
lowered, and
therefore the electrochemical current density will drop, thus lowering the
rate of gas
formation and changing its release frequency. The change could also be related
to the
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
23
presence of a low superheat zone with some undissolved material which hinders
bath
mixing and initiates electrode potential changes as well as localized MHD
instability
(Figure 9). It is noted that a new peak starts to appear in stages 3 and 4
where the bubble
components split into two peaks, consistent with a change in gas release
pattern due to
the presence of the spike.
Example 3: Monitoring the new anode current pick up curve
In the presented study involving thirty repeated experiments where anode
setting
reference was fixed, a rapid initial increase in current pick-up rate during
the first two
hours changed to a slower uniform rate for the next twelve hours as
illustrated in Figure
10. Occasionally small spikes occurred as highlighted after nine hours in
Figure 10b. This
type of spike is probably due to detachment of pieces of the frozen layer
followed by re-
freezing a new insulating coating. Removal of the coating may be aided by the
induced
bubble forces. The first two stages were confirmed by photographic pictures
taken of new
industrial anodes after anode setting for one, eight and twelve hours in three
neighboring
cells for the same stall number.
In the first one hour after anode setting, bath freeze has developed on the
bottom and
side vertical plane of the anode surface. The side freeze is noticeably
thinner than at the
bottom. On the bottom, its thickness was 35-40 mm which matches the predicted
inter-
electrode distance.
Fast Fourier Transform (FFT) analysis of the new anode current signal at 10 Hz
data
sampling frequency after 30 minutes of anode setting in Figure 11 shows a
noisy signal
which is attributed to MHD instability resulting from the anode bottom freeze.
The frequency response spectrum of anode current has shown a high bubble
dynamic at
low frequency range which perhaps due to energy input deficit at stable
magnetohydrodynamic conditions following anode replacement. The early current
pick is
expected to be on the lower side of the anodes adjacent to those unchanged
because of
the mixing and heat transfer enhanced by their gas release.
The early current carried (presumably by the side) steadily increases to about
10% of the
final current in the first two hours as seen in Figure 10 and then changes to
a slower more
uniform rate. During the second phase, the remainder of the freeze, which
presumably is
predominantly underneath the anode, dissolves slower and the rate of current
pick up
drops off. This is consistent with the second phase of Figure 10 and the
photographical
CA 03012163 2018-07-20
WO 2017/141135 PCT/IB2017/050666
24
observations reported above. The slow dissolution of freeze increases the
available
surface area.
After about twenty hours the process enters a third stage which exhibits low
frequency
dynamics for the same anode, imposing a clear path for bubbles to escape
through the
slotted anodes. In this stage the rate of current pickup drops off further,
perhaps due to
the continuing very low superheat of the adjacent anode and the balance in
competition
between the entropic energy" demand for the electrode reaction and for the
energy
transfer for heating the anode. This is confirmed by the absence of the bubble
peak at 1
Hz frequency in Figure 12, and the appearance of a peak at 0.5 Hz.
While monitoring the current pick up, in some instances a more rapid rate of
increase was
observed perhaps due to freeze detachment or very low setting height of the
anode as
illustrated in Figure 13.
