Environmental Engineering Reference
In-Depth Information
estimated average. This agrees with the fact that the optimal estimate of a random
variable is its expected value, when no other information is available.
The operational work index
W
iog
may be calculated with a fast sampling rate,
since it depends on power
P
T
(a continuous time measurement) and on
p
80
through
g
65
measuring +65# with a sampling period of 1 min. However, the plant data was
available with a sampling time of one month, which is the period in which the ore
work index is determined by laboratory analysis. This explains the time scale used
in Figures 4.22, 4.23 and 4.24.
Figure 4.22 shows the actual
W
iog
determined using
f
80
and
p
80
found by labo-
ratory measurements, and using instead the average value of
f
80
±
three standard
deviations.
16.5
16
15.5
15
14.5
14
13.5
13
12.5
0
10
20
30
40
50
months
Figure 4.22
Actual
W
iog
determined using
f
80
and
p
80
found by laboratory measurements is given
by the solid line. The
W
iog
determined using the average value of
f
80
±
three standard deviations
instead is given by the two dotted lines
The situation is different in the case of
p
80
, since if
p
80
changes by three standard
deviations around its average, the change in
W
iog
is 29
4%. This means that a soft
sensor is required for
p
80
. Only the two highly reliable measurements have been
used as secondary measurements: the feed ore flow
G
s
to the plant and the total
power
P
T
drawn by the mills. Then the following soft sensor was determined:
.
=
(
)+
(
−
)+
(
)+
(
),
p
80
θ
1
P
T
t
θ
2
P
T
t
1
θ
3
G
s
t
θ
4
l
t
(4.146)
using both stepwise regression and all possible subsets from a set of candidate bases
(components), which included
P
T
(
)
(
)
−
(
)
t
,
G
s
t
, their delays up to
t
1,and
l
t
to
accommodate a constant term.
Search WWH ::
Custom Search