Environmental Engineering Reference
In-Depth Information
a multi-camera system and expert controllers are combined to improve flotation
performance. Forbes [20] shows that the Fourier ring and texture spectrum-based
features perform well in classifying new images at a relatively low computational
cost; results from three industrial case studies show that bubble size and texture
measures can be used to identify froth classes. As described in Chapter 3, Lin
et al.
[21] present a modified texture spectrum approach and a reliable method based on
binary images of bubble size estimation. Liu and MacGregor [22] use multiresolu-
tional multivariate image analysis for numerical estimation of froth status. Nunez
and Cipriano [23] introduce a new method for the characterization and recognition
of visual information using dynamic texture techniques for prediction of froth speed.
Outotec has evaluated the benefits of using visual information provided by the
froth characteristics sensor FrothMaster™ at Newcrest Mining's Cadia Valley Oper-
ations in New South Wales, one of Australia's largest gold producers. As Figure 7.1
shows, a first controller, using FrothMaster™ data, adjusts cell slurry level, aeration
rate and reagent addition to control the froth speed set-point. A second controller
uses the process data from the on-stream analyzer to manipulate the froth speed
set-point so as to achieve a grade set-point specified by the plant metallurgist. For
the purposes of the evaluation, three FrothMaster™ units were installed in the first
three cells of line 1, enabling grade and recovery calculations to be accurately com-
pared with their manually controlled counterparts in line 2. The results achieved in
line 1 showed significant improvements in both grade control and product recovery.
During a 44-day trial period the average grade error in line 1 as a percentage of
target remained within 0.03%, while the average grade error recorded in line 2 was
9.99%. Copper recovery from the line 1 first rougher circuit was 5.28% higher than
from the corresponding cells in line 2, while gold recovery was 5.14% higher. At
the conclusion of the trial period, average frother consumption in line 1 was reduced
by 7.1%, representing a significant saving in annual processing costs.
7.3 Concepts for Advanced Control
An analysis of the available information on advanced control in mineral processing
reveals that despite the wide variety of existing methodologies and techniques for
process control, only a handful have given rise to commercial products applied in
industrial settings. These techniques may be divided into two categories:
•
those we will call intelligent control, including ES and fuzzy logic;
•
model predictive control, using linear or non-linear models originating in phe-
nomenological or empirical models adjusted on the basis of operating data.
Both categories have their own underlying concepts and application methodolo-
gies, and attempts have been made to combine them into a single integrated solution.
This is the case, for example, with the algorithms known as fuzzy model predictive
control [24].
Search WWH ::
Custom Search