Industrial Products for Advanced Control of Mineral Processing Plants - Advanced Control and Supervision of Mineral Processing Plants - page 299

In a further embodiment of the control system, several layers of the scheme de-
scribed above could be put in place to optimize the process control. In such a case
an ES is supervised by an OC/MPC-based controller, which in turn is supervised by
another ES, etc. This architecture allows optimal exploitation of the advantages of
each approach for each particular task.
Expert Optimizer has been applied in many cement plants. In flotation plants,
a strategy has been developed that incorporates first principle modeling, ES, fuzzy
logic and neural networks to generate the optimal grade-recovery curve, improve
concentrate quality, stabilize flows and minimize reagent use. The manipulated vari-
ables in this strategy are reagent flow rate, cell level set-point, and air flow rates for
cells, and wash water flow for columns [49].

ANDRITZ

Ideas Simulation & Control, a member of the Andritz group, markets three products
with potential for application to mineral processing [50]: Ideas dynamic process
simulator, ACE™ 8 (Andritz Control Expert) expert control system and BrainWave ®
predictive control software
Ideas is a dynamic simulation environment for use in process design, control
logic verification, operator training, and advanced control system design and opti-
mization. ACE is an ES that emulates the responses of an experienced operator, and
has been applied to crusher and dryer operations in combination with BrainWave ® .
BrainWave ® is a multivariable predictive controller based on Laguerre polyno-
mials, a technique developed at the University of British Columbia, see U.S. Patent
No. 5,335,164 [51]. It employs a model that adapts on-line and utilizes the prediction
horizon as a tuning parameter. The structure of the model, built around discrete-time
transfer functions (Z-transform), is shown in Figure 7.5. Once the model obtains the
state equation of the dynamic system, its parameters are then identified by a recur-
sive least-squares algorithm.

Figure 7.5 Model used by BrainWave ®

8

ACE is a registered trademark of IDEAS Simulation & Control Inc.