XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1775
the identification of model order or dead time ahead of
time. The modelling structure in the Laplace domain is
given in Figure 3. Given an input, U, multiple Laguerre
functions, L, are calculated, and then multiplied by coef-
ficients and summed up to approximate the response.
Once dynamic models are determined, the optimal
control actions are calculated based on the (Huzmezan et
al., 2001) cost function, Equation 1:
J Y S
i N
N
Q
i
Nu
R
2
1
2
1
2
DU
=-
+
=
=
/
/
t^ih ^DUh
^ih
^ih
(1)
In Equation 1, the error between SP and PV is mini-
mized by adjusting the control action (process inputs), U.
Specifically, Y(i), S(i), and U(i) stand for the predicted pro-
cess variable, the SP, and the control action at time instant,
Figure 1. APC, digital twin and real-time optimization with ANDRITZ’s platforms and software
Figure 2. Schematic diagram of CIL ACE APC solution contribution to reagent savings
the identification of model order or dead time ahead of
time. The modelling structure in the Laplace domain is
given in Figure 3. Given an input, U, multiple Laguerre
functions, L, are calculated, and then multiplied by coef-
ficients and summed up to approximate the response.
Once dynamic models are determined, the optimal
control actions are calculated based on the (Huzmezan et
al., 2001) cost function, Equation 1:
J Y S
i N
N
Q
i
Nu
R
2
1
2
1
2
DU
=-
+
=
=
/
/
t^ih ^DUh
^ih
^ih
(1)
In Equation 1, the error between SP and PV is mini-
mized by adjusting the control action (process inputs), U.
Specifically, Y(i), S(i), and U(i) stand for the predicted pro-
cess variable, the SP, and the control action at time instant,
Figure 1. APC, digital twin and real-time optimization with ANDRITZ’s platforms and software
Figure 2. Schematic diagram of CIL ACE APC solution contribution to reagent savings