background image

Using ANSYS and CFX to Model Aluminum Reduction Cell since 1984 and Beyond
Page7
It is quite straightforward to mirror a full cell quarter model in order to produce a full cell half model (see
Figure 15). Yet, there is no need to solve such a model unless we apply asymmetric boundary conditions on
it. As the addition of the external busbars network will automatically introduce such an asymmetry between
the positive and the negative side, adding the external busbars network to the model is the best thing to do
at this point (see Figure 16).
Unfortunately, it was clear that trying to solve that 211,648 elements full half cell and external busbars
thermo-electric model on a PIII 800 MHz computer while the 105,096 elements full cell quarter thermo-
electric model took 52.48 CPU hours and 75.68 wall clock hours to compute would require a lot of time,
too much time!
2004: 3D half cell and external busbars thermo-electric model, part2
Figure 15 presents the mesh of a 3D half-cell model. This mesh is made of 290410 finite elements. That
model is the biggest model with constraint equations that could be solved. The P4 took 100 CPU hours to
converge with the assumed ledge profile. Figure 17 shows the thermal solution obtained. Of course, with
that much CPU time required for an assumed ledge profile, it was not practical to repeat this 5 ­ 10 times,
the number of iterations needed for the calculation of the ledge profile (Reference 27)!
2005: 3D full cell and external busbars thermo-electric model
The mesh of a 3D full cell and external bus-bar thermo-electric model of a 300 kA cell has been presented
in 2002 (Reference 26). That 423,296 elements model could not be solved on the PIII computer available to
the author at the time.
A bit later, the mesh of a 3D full cell and external bus-bar thermo-electric model of a 500 kA cell has been
presented (Reference27). That 585,016 elements model could not be solved either, even on a P4 3.2 GHz
computer with 2 GB of RAM.
Figure 18 presents the coarser 249,322 elements that could finally actually be solved. The P4 3.2 GHz
computer took 63.7 CPU hours to solve the model once (see the thermal solution in Figure 19). Since it
took so long to solve, no attempt to converge the ledge profile geometry has been performed (Reference
28).
2005: Weakly Coupled Thermo-Electric and MHD Mathematical
Models
As described in Reference 26, the velocity dependent local heat transfer coefficients at the bath/ledge and
the metal/ledge interfaces are at the heart of the coupling between the thermo-electric model and the MHD
model. The principal goal of the present work is to develop the convergence strategy of the weakly coupled
thermo-electric and MHD models, so it is important to get a significant coupling feedback loop between the
two models. The thermo-electric model is of course ANSYS based but since ANSYS-CFX cannot solve the
MHD problem in an efficient way, the specialized commercial code MHD-Valdis was used to solve the
MHD problem (Reference 28). The obtained non uniform ledge profile is presented in Figure 22.