relief along the length of the cell must be built-in the design.
Ramming paste small joints between cathode blocks will partially
play that role. Predicting the resulting effect is not trivial. Further,
to avoid leaks into the lining, the end walls must also provide an
adequate confinement.
A finite element model of a Hall-Héroult cell slice was built using
the in-house code FESh++. Key features of the model include the
baking of ramming paste, the quasi-brittle nature of carbon and
contact interfaces.
The transient fully coupled thermo-mechanical solution on an
Opteron 850 64-bit machine required approximately 2.6 GB of
RAM and 2.9 to 3.3 hours of wall clock time per time step. The
total wall clock time then varied between 35 and 120 hours.
Faster preheating rates mean larger thermal gradients in the
cathode blocks. For the modeled configuration, this leads to a
horizontal tensile stress zone in the cathode block. For the cases
studied these stresses were not sufficient to crack the block but
this highlights one of the possible cathode failure mechanisms.
From this work, it was shown that a slice model is not sufficient to
detect cracking of the cathode blocks for practical cell designs. At
least a quarter cell needs to be modeled to accurately represent the
confinement of the lining along the length of the cell. However, it
is possible to obtain valuable information with a slice model about
the extent of ramming paste baking and the opening of gaps that
would allow bath infiltration into the lining and possibly reduce
pot life.
To help determine the optimal final cathode surface temperature
and ramming paste baking profile, it is now necessary to study
what happens when the bath is poured in the cell. The timing of
the cathode block expansion and ramming paste shrinkage is
critical to ensure no gap will form which would allow liquid to
leak into the lining. Thermal shock of the cathode blocks and flash
pyrolysis of the ramming paste must also be avoided. The desired
conditions at the end of preheating could then be reverse
engineered.
From the work presented here, it can be seen that a numerical
model is an invaluable tool to gain insights into the complex
response of a cell during start-up. Once validated with
experimental measurements, it can be used in the optimisation of
the heat-up practices. The results of the model could also be used
to determine some control metrics in the operation, for example
by relating the evolution of cathode bar temperature to the risks of
cathode block cracking.
The authors wish to thank Jérôme Bédard from Laval University
for helping with the literature review, the mesh, the input files and
with running the simulations. The kindness of Professor Daniel
Marceau from Université du Québec à Chicoutimi is also
acknowledged, for providing us with an access to his numerical
computing facilities. Finally we thank the Natural Sciences and
Engineering Research Council of Canada (NSERC) and REGAL
for the granted partial financial assistance.
[1] M. Sørlie & H.A. Øye, "Cathodes in Aluminum Electrolysis",
2
Kvande, "Preheating Study of Smelting Cells", in Proceedings of
TMS Light Metals 2005, 333-336.
"Gas preheating and start of Søderberg cells", in Proceedings of
TMS Light Metals 1991, 741-747
Bake-out of Aluminium Reduction Cells, a Technology for the
Futur
Axial
"Development and Validation of a Thermo-Chemo-Mechanical
Model of the Baking of Ramming Paste
Evaluation of Ramming Paste for Aluminium Electrolysis Cells",
Heiling, "International Cooperation and High-Performance
Reduction in Siberia
Thermo-electro-mechanical Transient Simulation Applied to
Electrical Preheating of a Hall-Héroult Cell", in Proceedings of
mécanique des interfaces de contact d'une cuve de Hall-Héroult",
Thermal and Sodium Expansion in Prebaked Aluminium
Reduction Cells