Soft Sensing - Advanced Control and Supervision of Mineral Processing Plants - page 205

M

Sample Variance Template

Soft ore Template

Medium ore Template

Hard ore Template

in Euclidean space R M for operating condi-
tions corresponding to grindability indexes for hard, medium and soft ores. Also shown is a sample
variance template S

Figure 4.28 Pictorial representation of templates V ( Γ i )

for an unknown operating condition of the plant. Ellipses represent sets of
sample templates obtained from repeated on-line measurements for the three different kinds of
ores. The points representing the soft, medium and hard ore templates are calculated by averaging
similar sample templates for each case when the type of ore is known. The estimated grindability
index would be medium in this case

( Γ )

Results were obtained through simulation using a dynamic SAG mill circuit
model (able to handle two different kinds of ore) adjusted to data from a large grind-
ing plant with a layout shown in Figure 4.25 [22].
Comparisons of Figures 4.29 and 4.30 show the advantages obtained through
concatenation of the three measured variables. This advantage is reflected in Tables
4.9 and Table 4.10, which show that the improvement in the correct estimation of
the grindability index brought about by concatenation is quite remarkable.

Tabl e 4. 9 Percentages of hits (diagonal) and misses (off-diagonal) of actual grindability index
when only the template for return ore flow F r is used

Actual index Identified index
0.94 1.00 1.06

0.94 64.2 34.2 1.6

1.00 31.4 34.0 34.6

1.06 3.4 29.4 67.2