and the cathode carbon blocks already present in the "almost empty
shell" potshell model type, the "half empty shell" potshell model
type also includes the geometry of the cathode blocks themselves.
As already briefly described in [5], the Dewing sodium expansion
behavior of the cathode blocks is treated in ANSYS® as a "creep-
that model in transient mode following the build-up of the sodium
concentration in the cathode blocks from start-up to around 1000
days of operation where the cathode blocks get fully saturated in
sodium. This way, the model computes the incremental build-up of
the strain-stress relationship due to the gradual and non-uniform
carbon swelling generated from the gradual increase in the sodium
concentration in the cathode blocks and also to the gradual and
non-uniform restraining effect of the potshell on the cathode blocks
sodium expansion. Of course, this must be done using relatively
small time steps hence the huge CPU time requirement.
effort, yet fortunately, the bulk of the R&D model development
work had already been done about 20 years ago! Reference [5]
reports 90 MFLOPS of sustained performance on solving such a
model using ANSYS® 4.4a on a CRAY X-MP/24. We can now
similar model using the SPARSE solver available in ANSYS® 11
M6300 portable computer. This huge increase in MFLOPS is due
to improvements in both the hardware and the software side. On
the software side, Reference [5] reports delays caused by huge I/O
activities and corresponding huge I/O wait time. With the use of
the SPARSE solver available in ANSYS® 11.0, it is possible to
RAM available on the computer. Furthermore, because it is a dual
core computer and ANSYS® solver is using both processors, the
shell" potshell model still requires a lot of computer power, clearly,
all the computing resources available on the Dell M6300 were
required. Solving the "half empty shell" demo potshell model in
elastic properties mode presented in Figure 8 took 25,335 CPU
seconds or 7.0 CPU hours which is 6.5 times more than what was
required to solve the "almost empty shell" demo potshell model in
elastic properties mode. Solving the "half empty shell" demo
potshell model in plastic properties mode presented in Figure 9
took 103842 CPU seconds or 1.2 CPU days which is 3.8 times
more than what was required to solve the "almost empty shell"
demo potshell model in plastic properties mode. It is interesting to
note that the CRAY X-MP/24 would have required about 80 CPU
days to solve the save model at 90 MFLOPS, which could explain
why Figure 1 of [5] presents model results after 60 days of cathode
life only!
the "almost empty shell" and the "half empty shell" potshell model
are quite similar, yet, in both cases, the "half empty shell" potshell
model predicts less side deflection because the "half empty shell"
model type also predicts a significant cathode panel upward
deflection accommodating part of that cathode panel expansion
which is of course not accounted for by the "almost empty shell"
model type. It is important to point out that that cathode panel
upward deflection has an impact on the MHD cell stability as
demonstrated in [9] and only the solution of an "half empty shell"
potshell model type generates the data required to account for that
cathode panel upward deflection in the MHD cell stability model.
have been presented and described in details. Nowadays, all three
types of models can be used as potshell design tools.