operator skill & strategy identification in process industry doc. student research seminar...
TRANSCRIPT
Operator Skill & Strategy Identification in Process Industry
Doc. student research seminar4.4.2011Janne Pietilä
Objective
• Despite a high level of automation, the human operator nevertheless has a significant role in controlling industrial processes
• The objective is to survey the performance and operating practices of different operators, using data-based analysis methods
• The industrial plant in case is a flotation process of the Pyhäsalmi mine in central Finland
• Results are useful in e.g. operator training or transfer of latent knowledge
The Process
• The copper flotation process of the Pyhäsalmi mine – a complicated process, whose state is difficult to measure– a relatively high level of automation– the operator’s expertise and insight significantly affect the efficiency of
the process• The control variables and setpoints
– the air feeds and froth thicknesses and the chemical addition rates are the most significant
• Measurements– levels of the slurry and the froth surface, concentrations, froth image
analysis
grinding
copper flotation
zinc flotation(pyrite flotation)
thickeningdewatering
The Operator
• The role of the operator– optimizing grade and recovery
– monitors the operation and reacts to emergencies, failures etc.
– coordinates maintenance and repair tasks during the shift
• There are 5 operators at the Pyhäsalmi mine– work group includes also
maintenance personnel
• The concentrator operates in three shifts
Performance
• The essential variables describing the process operating performance– recovery (index)
– concentrate grades (quality index)
– economic index
– tailings grades – fed to the zinc flotation circuit
• Other important variables– the ore feed properties
• grades • particle size distribution after
grinding
The data
• Gathered from the process automation system’s database
• The sampling time of the data is 1 minute, and from this data– the outliers and measurement errors are removed– hourly averages are calculated– the data is grouped according to the operating shifts
• The time span for the comparison analysis is e.g. 2-3 weeks
• The compared variables are the recovery, grades and production indices
Data preprocessing
CuV tal.ind. CuR Cu CuV Cu saanti
Kokonaissyotto
Malmi Cu
Malmi Zn
Malmi S
Syötteen raekoko -20um
Syötteen raekoko -74um
Syötteen raekoko +149um
Compensated Variables
Inpu
t var
iabl
es
CuV MLR Model Parameters: 01.10.2010 00:00 - 30.11.2010 23:59
-0.5
0
0.5
• Feed compensation– a fair comparison is sought– changes in the ore properties are
independent of the operator– an MLR model from the feed
properties to all comparison variables– estimated separately for each
comparison period
Testing and pairwisecomparisons• The pairwise comparisons
indicate those groups that differ statistically significantly from the others
• By combining the analysis results from different comparison variables, differences in process operating practices can be discovered
Analysis of the results
• Based on the comparison results, the following observations of the operating practices can be made:– Group A is ”evenly good”; the
recovery, concentrate grade and the economic index are all reasonably good
– Group B pursues a high recovery, even if the concentrate grade becomes lower
– Group C aims for a high quality concentrate, but at the expense of recovery
– Groups D and E seem to have some room to improve
Saanti-pitoisuus ja taloudellinen indeksi
27 28 2993
94
95
96
97
CuR Cu%
Saa
nti %
Ryhmä E
27 28 2993
94
95
96
97
CuR Cu%
Saa
nti %
Ryhmä D
27 28 2993
94
95
96
97
CuR Cu%
Saa
nti %
Ryhmä C
27 28 2993
94
95
96
97
CuR Cu%
Saa
nti %
Ryhmä A
27 28 2993
94
95
96
97
CuR Cu%
Saa
nti %
Ryhmä B
Tal
oude
lline
n in
deks
i
-1.5
-1
-0.5
0
0.5
1
Esimerkki 1: Pitkän aikavälin operaattorikohtainen vertailu• Kuparirikastuspiirin ohjaaminen on kesällä vaikeampaa
– Lietteen lämpötila vaikuttaa mineraalien käyttäytymiseen– Operaattorien väliset erot tulevat selvemmin esiin
01 02 03 04 05 06 07 08 09 10 11 12 01
94.5
95
95.5
96
96.5
97
Month
Rec
over
y (%
)
A
B
CD
E
Kuparin saanti 2010
Esimerkki 2: Syöttötason ylläpito
• Operaattori voi vaikuttaa syöttötasoon jauhatuksen aktiivisella valvonnalla
A B C D E
158
160
162
164
166
168
Med
ian
flota
tion
feed
(t/
h)
Operator
Syöttötaso
A B C D E0
20
40
60
Operator
% o
f ho
urs
incl
udin
g ch
ange
s
Operaattorin aktiivisuus
Vertailutyökalu
• Rikastamolle kehitetty automaattinen vuorojenanalysointi-työkalu– Operaattorien vertailu
halutulta ajanjaksolta– Datakompensointi– Suoritusindeksit– Jakaumat– Raportointi