background image
TESTING IN SITU CONTROL WITH DYNA/MARC CELL SIMULATOR
For the purpose of the Dyna/Marc test, the simplest version of the In Situ control
logic has been coded in Fortran, the language in which Dyna/Marc kernel was
developed. In the simplest version, only the linear curve fitting option is available.

The simplest version of the In Situ controller subroutine has been added to
Dyna/Marc and tested. In the test runs, the CO
2
gas release noise generation is
neglected so the selected data sampling time step is 5 seconds. The In Situ calculation
step is 5 minutes which produces 60 normalized cell voltage data points. In normal
operation after the observation period, the estimation of the noise free normalized cell
voltage is used to calculate the estimate of the concentration of dissolved alumina, the
PID controller error and finally the new feeding regime.

The Dyna/Marc simulator runs start in perfect steady-state condition as any
Dyna/Marc run does at the target ACD and target alumina concentration of 2.3%. A
no feed observation period is started immediately. As can be seen in Figure 20, the
perturbation created by the observation last about an hour, after which the cell is back
in perfect steady-state conditions. The ACD is adjusted every 3 hours which create
very little perturbations on the feeding rate.

Figure 20: 12 hours Dyna/Marc simulation using In Situ control logic