background image
subsequent model validation exercises, is a serious weakness. The
internal forces that will develop in a new potshell and lining design
will depend on that new potshell structural rigidity, on the sodium
swelling characteristics of the grade of cathode blocks selected, and
on the materials between the blocks and the potshell as previously
discussed.
It may be a serious disadvantage, but the corresponding advantage
is the fact that this type of model can be solved much more rapidly
than the other two types of potshell mechanical models available.
Per example, the "empty shell" demo potshell model using only
elastic steel properties presented in Figure 4 took only 150 CPU
seconds to solve on a 64 bits dual core Intel Centrino T9300 Dell
Precision M6300 portable computer running ANSYS® 11.0
version.
Figure 4: Elastic mode "empty shell" demo potshell model
The "empty shell" demo potshell model using elasto-plastic steel
properties presented in Figure 5 took 11 times more CPU to solve,
but still this is only 1706 CPU seconds or 28 CPU minutes.
Figure 5: Plastic mode "empty shell" demo potshell model
"Almost empty shell" potshell model
Clearly, a 30 minutes response time is not a problem, so the "empty
shell" potshell model would be the perfect design tool if we could
trust that the appropriate internal loads have been applied and
hence that the correct potshell deflection has been calculated.
This relatively small turnaround time is opening the door to the
option of spending more CPU time in order to increase the
accuracy of the model prediction and still having an efficient
design tool. In the "almost empty shell" model type, this is
achieved by adding the lining geometry between the potshell walls
and the cathode blocks all around the potshell and by applying a
pressure loading as boundary condition at the carbon block/side
lining interface that is lying on the Dewing strain-stress
relationship (i.e. that represents the carbon block equilibrium
condition). In order to be able to do so, a new convergence
numerical scheme external to the ANSYS solver must be setup [1].
Starting from an assumed initial internal load, the task of that
external convergence loop is to converge toward that cathode block
equilibrium condition pressure loading for each element face of the
carbon block/side lining interface.
It turns out that that can be achieved by using an under-relaxation
numerical scheme in about 20 iterations adding about the same
factor to the "almost empty shell" type model solution time
compared to the "empty shell" type model. Per example, the
"almost empty shell" demo potshell model solved in elastic
properties mode presented in Figure 6 took 3898 seconds CPU or
65 minutes CPU to solve while the "almost empty shell" demo
potshell model solved in plastic properties mode presented in
Figure 7 took 27524 seconds CPU or 7.6 hours CPU to solve.
Total displacement (m)
von Mises Stress (MPa)
von Mises Stress
Total displacement (m)
von Mises Stress