Page 20

This clearly demonstrates that without major process perturbations, In Situ

control logic works exactly as designed. The dissolved alumina concentration is
effectively controlled with an impressive accuracy employing only standard line
amperage and cell voltage signals as feedback control inputs.

Yet, a 12 hours simulation period without process events like metal taping or

anode change is not enough to demonstrate that the logic is sufficiently robust to cope
with typical unavoidable process perturbations. For sure, the anode change event
affects the cell resistance and this effect must be accounted for in the In Situ logic.
Otherwise In Situ logic will assume the change is generated by a change of the
dissolved alumina concentration and it would very inappropriately change the feeding
regime accordingly.

It turns out that it is possible to account for the effect of the anode change on the

cell resistance by using a ramp the same way it is done in a standard cell controller.
Figure 21 presents the results of a 3 days simulation containing 3 metal tapping events
and 3 anode change events. The anode change event does perturb significantly the In
Situ

control logic despite the best possible compensation of the resistance change with

a ramp. Yet, when compared with a standard continuous tracking control logic, those
perturbations remain relatively very small.

Figure 21: 3 days Dyna/Marc simulation using the In Situ control logic