Soft Sensing - Advanced Control and Supervision of Mineral Processing Plants - page 151

Tabl e 4. 3 ( continued )

Reference Model class Main features Applications, tests

Extraction of
features of
measured vari-
ables (images)
and genetic
classification

Estimation of ore lithology by
classifier using genetic algorithm.

Laboratory tests using monthly
rock samples from mine site.
Seven lithological classes are
considered and 130 features are
extracted from each rock sample
out of a database consisting of
760 digital images.

Perez et
al. , 1999
[25]

Clustering of
operating points

Clusters formed represent differ-
ent types of overload conditions.
Distances to cluster centers give
early warning of overload.

Clusters are formed using data
from an industrial ball mill grind-
ing circuit. If distance to an over-
load cluster is below a threshold
corrective actions are taken ac-
cording to the cluster characteris-
tics.

Espinoza,
et al. ,
1995 [26]

NN Industrial weightometer soft sen-

sor including monitoring sys-
tem for data validation, failure
detection, replacement of fail-
ing weightometer and alarming.
Among inputs are particle size
distribution and power draw.

Several models are developed and
tested in an industrial diamond
processing plant.

Pan et al. ,
2003 [27]

State model Kalman filter for combined state

and parameter estimation in SAG
and AG mill circuits. States are
total ore and water mill hold-ups,
ball hold-up, liner weight. Esti-
mated parameters are ore grind-
ability, liner wear rate and ball
wear rate.

Design and tests using milling
model. Estimated variables are
used in a model-based control
and expert control systems de-
signed using the models.

Herbst et
al. , 1989
[28]

State model SAG mill charge volumes estima-

tion using Kalman filter.

Installed in industrial grinding

circuits.

Herbst et
al. , 1996
[29]

Notation: LIP = Linear-in-parameters; NARMAX = Nonlinear ARMAX; NN = Neural Net-
work; SVM = Support Vector Machine SS = Soft Sensor; PCA = Principal Component Anal-
ysis; RMS = root mean square; SAG = Semi-autogenous.