Multivariate Image Analysis in Mineral Processing - Advanced Control and Supervision of Mineral Processing Plants - page 123

3

2

19

630

1

245

0

511

-1

840

-2

413

-3

SS1
SS2
SS3
SS4
SS5
SS6

-4

-5

-6

-3 -2 -1 0 1 2 3 4

t 1

840 511 630

245 19 413

Figure 3.19 The t 1 t 2 score plot of MR-MIA II features for selected steady-states images (SS1-

SS6). Representative images for each steady-state are also shown at the bottom. From Liu et al.
[37]

−

The continuous progression from one steady-state to the other is shown in greater
detail in Figure 3.21. The transition from steady-states SS2 to SS3 is tracked using
the solid line. Some representative froth images are also provided in order to follow
this transition visually. This figure clearly demonstrates the ability of the proposed
approach to monitor froth health, both in steady-state as well as in transitions. Note
that the PCA model projection residuals ( i.e. , SPE or DMODX) should also be used
to monitor froth health. This plot would safeguard against operators reacting in a
normal way to abnormal images ( i.e. problem with the camera or lighting condi-
tions) or other extreme conditions [37]. This plot was not included here since all the
images were found to be well explained by the PCA model ( i.e. , DMODX values
were lower than limit), but is available in the paper by Liu et al. [37].