A METHOD TO IDENTIFY THE KEY CAUSES OF

DIFFERENCES IN ENERGY EFFICIENCY OF OPERATORS

Maryam Abdi OskoueiDr. Kwame Awuah-Offei

1

Energy Efficiency in Coal Mining

• Energy consumption in coal mines is estimated as 142 billion kWh per year (DOE, 2007)

• About 49% of this energy can be saved by improving energy efficiency and implementing best practices (Bonskowski et al. 2006)

• Dragline, as one of the main energy consumers in surface coal mines, consumes about 15-30% of total mine energy (Orica Mining Services 2010) 2

Energy Efficiency of Dragline Operation

3

Mine design & planning

Equipment characteristics

Operating condition

 Energy Efficiency

Operator practice

𝜂=𝑈𝑠𝑒𝑓𝑢𝑙𝑤𝑜𝑟𝑘𝐼𝑛𝑝𝑢𝑡𝑒𝑛𝑒𝑟𝑔𝑦

≅𝑃𝐸

Responsible Parameters

• Which parameter is causing the difference between energy efficiency of operators?

• Why identify responsible parameters?

4

The Flowchart

5

STARTCorrelation

analysisCreate pairs of

operatorsi=1

Select ith pair

Use linear regression of differences to find

significant parameters *

i= i+1

i <= No

YesUse the results to

determine significant variable across all

pairs

Repeat n times

Pearson Correlation

• Evaluate the relation between relevant parameters and energy efficiency

• Test the hypothesis of no correlation under significant level of α

6

,

cov( , )X Y

X Y

X Y

Structure of the Data (opr_i)

7

 Energy efficiency Correlated parameters

 

cycle 1   ηi1 pari11 pari12. . . pari1v

 

cycle 2   ηi2 pari21 pari22. . . pari2v

 .   . . . . . . .

 .   . . . . . . .

 .   . . . . . . .

 

cycle ci   ηic paric1 paric2. . . paricv

 

Issues with Missing Data   Opr i Opr j

Cycle 1 Xi1 Xj1

Cycle 2 Xi2 Xj2

  . .

  . .

  . .

Cycle cj Xi cj Xj cj

. .  

Cycle k Xi k  

. .  

Cycle ci Xi ci  

8

• Proposed method is based on pair-wise comparison of operators

• Select equal number of cycles

• Repeat the process n times to reduce the effect of random sampling error

Select cj cycles at random

Difference matrix (opr_i - opr_j)

  Δη Δpar  

cycle 1   ηi1 – ηj1 pari11 – parj11 pari12 – parj12. . . pari1v – parj1v

 

cycle 2   ηi2 – ηj2 pari21 – parj21 pari22 – parj22. . . pari2v – parj2v

 

.   . . . . . . .  

.   . . . . . . .  

.   . . . . . . .  

cycle c   ηic – ηjc paric1 – parjc1 paric2 – parjc2. . . paricv – parjcv

 

9Dependent variables

Independent variables

Linear Regression Analysis

• Test the significance of coefficients (ki) to identify responsible parameters (significance level of α)

• A parameter is a responsible parameter, if in (1-α)100 % of runs it is recognized as a responsible parameter.

10

ηi

par1i

par2i

parni

.

.

.

ηj

par1j

par2j

parnj

.

.

.

Operator i

Operator j

Δ η=𝑘0+𝑘1 Δ𝑝𝑎𝑟 1+𝑘2 Δ𝑝𝑎𝑟 2+⋯+𝑘𝑛 Δ𝑝𝑎𝑟𝑛

Case Study

• Dragline Bucyrus-Erie 1570w – 85 yd3 removes the blasted overburden

• Real time monitoring system by Drives & Controls Services, Inc.– Adjusted to record energy consumption of three sets

of motors in the free slots of the database– 34,327 cycles recorded in one month, 13 operators

11

Operators with Sufficient Working Hours

• Mean Standard Error (SE) of energy efficiency for each operator

12

ii

i

SEnOc

A B C D E F G H I J K L M0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

05001,0001,5002,0002,5003,0003,5004,0004,500

Standard error Cycles

Operator

Mea

n st

anda

rd e

rror

Num

ber o

f cyc

les

• First cut-off standard error = 0.06

Operators with Sufficient Working Hours

13

A B C D E F G H0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

05001,0001,5002,0002,5003,0003,5004,0004,500

Standard Error Number of cycles

Operator

Mea

n st

anda

rd e

rror

Num

ber o

f cyc

les

Final cut-off standard error = 0.04

The Flowchart for the given data

14

STARTCorrelation

analysisCreate 10 pairs

of operatorsi=1

Select ith pair

Use linear regression of differences to find

significant parameters *

i= i+1

i <=10No

YesUse the results to

determine significant variable across all

pairs

Repeat 30 times

nop = 5 operatorsn = 30 runs for each pair95% significance level

Correlated Parameters# Parameter

PCP-value

1 Dump height-0.6560(<0.001)

2 Hoist energy-0.5857(<0.001)

3 Drag distance (vertical)-0.5089<0.001

4 Drag energy-0.4569<0.001

5 Drag distance (horizontal)-0.4807<0.001

6 Load bucket time-0.4548<0.001

7 Dump time-0.3050<0.001

# ParameterPC

P-value

8 Cycle time-0.3755(<0.001)

9 Swing energy-0.2724(<0.001)

10 Swing in time-0.3362(<0.001)

11 Spot time-0.1725<0.001

12 Angle swing out-0.1556(<0.001)

13 Swing out time0.0123

(0.0913)

14 Payload0.2429

(<0.001)

Results of Linear Regression Analysis (30 runs)

16

Correlated parameters D,B D,E D,A D,C B,E B,A B,C E,A E,C A,C

Dump height 30 30 30 30 30 30 30 30 30 30

Drag distance (vertical) 30 30 30 30 30 30 30 30 30 30

Drag distance (horizontal) 30 30 30 30 30 30 30 30 30 30

Load bucket time 4 30 26 30 0 8 20 5 22 30

Dump time 30 30 8 30 30 8 30 30 30 30

Cycle time 3 16 6 2 2 26 0 10 5 27

Swing in time 3 16 28 10 16 29 30 3 6 7

Spot time 30 30 30 30 30 30 30 21 30 30

Angle swing out 14 30 4 8 18 30 30 22 30 2

Payload 14 3 15 12 2 13 1 6 8 30

number of times of non-zero regression coefficient

Responsible at significance level 95%

17

Correlated parameters D,B D,E D,A D,C B,E B,A B,C E,A E,C A,C total

Dump height 1 1 1 1 1 1 1 1 1 1 10

Drag distance (vertical) 1 1 1 1 1 1 1 1 1 1 10

Drag distance (horizontal) 1 1 1 1 1 1 1 1 1 1 10

Load bucket time 0 1 0 1 0 0 0 0 0 1 9

Dump time 1 1 0 1 1 0 1 1 1 1 8

Cycle time 0 0 0 0 0 0 0 0 0 0 4

Swing in time 0 0 1 0 0 1 1 0 0 0 4

Spot time 1 1 1 1 1 1 1 0 1 1 2

Angle swing out 0 1 0 0 0 1 1 0 1 0 1

Payload 0 0 0 0 0 0 0 0 0 1 0

Responsible Parameters

18

Correlated parameters total probability

Dump height 10 100%

Drag distance (vertical) 10 100%

Drag distance (horizontal) 10 100%

Spot time 9 90%

Dump time 8 80%

Load bucket time 4 40%

Angle swing out 4 40%

Swing in time 2 20%

Payload 1 10%

Cycle time 0 0%

Discussion

19

Dump

heig

ht

Drag

dista

nce

(ver

tical

)

Drag

dista

nce

(hor

izont

al)

Spot ti

me

Dump

time

Load

buc

ket ti

me

Angle

swin

g ou

t

Swing

in tim

e

Paylo

ad

Cycle

time

0%

20%

40%

60%

80%

100%

Pro

ba

bili

ty o

f b

ein

g a

re

sp

on

sib

le p

ara

me

-te

r

In this case study:

• There is only a 40% probability for dig time (load bucket time) to be a responsible parameter

• High probability for engagement and disengagement position of bucket to affect dragline performance

• Payload and cycle time have a low chance of being a responsible parameter

Conclusion• A guideline for improving

operator performance can be established using the results

• The method suggested for identifying responsible parameter is a valid and robust approach and can be used for other mines and draglines

20

Question?

21