International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
1
DM Make up Water Reduction in Power Plants Using
DMAIC Methodology a Six Sigma approach
Himanshu Kumar*, Anurag Singh#
*
Deputy Manager ( O&M – OPN)
#
Manager ( P&S)
Abstract- DM water is life line of a power plant. Presently DM
make up for all NTPC running plants is around 0.84% of BMCR,
however percentage figure seems misnomer if we observe the
absolute values. Overall in NTPC for the year 2010-11,11-12 &
12-13 DM make up has been 55 Lacs MT, 57 Lacs MT & 63
Lacs MT respectively and this DM Make up has financial
implication on cost in terms of crores of Rupees. However
significance of every additional DM Make up in present cost
competitive scenario is such that it needs special focus. Stringent
norms of CERC recently has given a red wake up call to run our
plants at maximum efficiency level and keep our performance at
better heat rate possible . On this motivation we have made our
research that will focus on how efficiently we are using this
scarce and precious commodity and how we can further improve
and increase overall profitability of our stations . Main aim of
this paper was to find statistical as well as subjective solutions
for minimizing DM water wastages or leakages. To initiate our
research , we firstly found loss potential of DM water
statistically. How much loss we find due to increased
consumption and how much it affects our performance level.
DMAIC approach is used to find out at what sigma level our
plants are performing. We have taken data of Dadri ( Thermal)
stage- 1 unit -3 three months data from feb 2013 to Aug 2013.
The process capability analysis of these data says that DM
consumption rate was less than 95 % conformal and the
consumption rate is +0.16σ. ( To be very much particular
regarding this sigma value two things are most important. Sigma
value is a variable term . In a layman's language it can be said as
if any company is running under six sigma levels, it doesn't mean
It will always remain at that value. It may increase or decrease
any time depending upon its performance at any point of time.)
For finding monetary loss analysis we have used heat transfer
method to find out the percentage loss in heat rate by additional
DM make up . If we include cost of heat loss due to loss of
enthalpy net additional make up of DM costs us Rs 1225 per
metric ton. This figure itself shows that if we reduce DM
leakages even by 0.1 % we can save so much of money. For
coming to solution we have drafted a thirty question audit based
online survey and on the base of that analysis a bar chart and a
paretochart has also prepared. This water audit helped us to
develop a framework not only for finding the root cause for poor
sigma level of DM make up pattern but also gives some
suggestive solutions that will help plants in finding solutions and
comprehensive plans to plug them. Lastly we have formulated a
0.1% reduction strategy on the basis of various inputs collected
through our research.
Index Terms- DM make up, DMAIC, Minitab 16.0 , process
capability , six sigma , 0.1% reduction strategy
I. INTRODUCTION
D
M water is life line of a power plant. Presently DM make up
for all NTPC running plants is around 0.84% of BMCR;
however percentage figure seems misnomer if we observe the
absolute values. Overall in NTPC for the year 2010-11,11-12 &
12-13 DM make up has been 55 Lacs MT, 57 Lacs MT & 63
Lacs MT respectively and this DM Make up has financial
implication on cost in terms of crores of Rupees. However
significance of every additional DM Make up in present cost
competitive scenario is such that it needs special focus. Stringent
norms of CERC recently has given a red wake up call to run our
plants at maximum efficiency level and keep our performance at
better heat rate possible . On this motivation we have made our
research that will focus on how efficiently we are using this
scarce and precious commodity and how we can further improve
and increase overall profitability of our stations. Main aim of this
paper was to find statistical as well as subjective solutions for
minimizing DM water wastages or leakages. To initiate our
research, we firstly found loss potential of DM water statistically.
How much loss we find due to increased consumption and how
much it affects our performance level. DMAIC approach is used
to find out at what sigma level our plants are performing. We
have taken data of Dadri (Thermal) stage- 1 unit -3 six months
data from Feb 2013 to Aug 2013. The process capability analysis
of these data says that DM consumption rate was less than 95 %
conformal and the consumption rate is +0.16σ. (To be very much
particular regarding this sigma value two things are most
important. Sigma value is a variable term. In a layman's language
it can be said as if any company is running under six sigma
levels, it doesn't mean it will always remain at that value. It may
increase or decrease any time depending upon its performance at
any point of time.) In dadri presently DM make up in our plant
is around 0.55-0.56%. DM leakages can occur from 'n' number of
points. To initiate our work a DM water audit sheet (online
survey) has been designed to find out various aspects of DM
consumption in NTPC Dadri plant. We all know there are two
mode of research – one is application research and second is
developmental research. This audit cum survey*acts as a tool for
both type of research. This audit sheet (which can be viewed
online
at http://www.surveymonkey.com/s/ZVDCHDS, http://www.sur
veymonkey.com/s/JLSKPR8,http://www.surveymonkey.com/s/J
X5CW2D ) can be applied to any plant to find out the DM make
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
2
up trend in any utility or industrial boiler. Before starting
DMAIC firstly we see what the financial implications of this
mych hyped thing are.
1.a. financial implications
Our current process in DM cycle make up and its analysis
suggest that there is a considerable scope of improvement in
plugging leakages and reducing variability. Cost of every
additional DM makeup at any given plant depends on cost of raw
water, chemical consumption and heat energy loss (of course
additional makeup is required due to some sorts of steam
leakage). Considering case study of NTPC Dadri where cost
calculation for steam leakage works out to be in the range of Rs.
790 to Rs. 1225/MT depending upon where the leakage is
occurring in the cycle. Our calculations suggest that every 0.1%
reduction in DM Makeup at Dadri alone will result in saving of
Appx. 2 Cr every year. However at NTPC level these figures
become further staggering at any thing between 60Cr. to 90Cr.
depending on primarily heat rate calculations of other plants .To
reiterate once again, as on today our plants do not have system to
carryout regular flow audits to identify sources & quantify
leakages and objective of our paper is to sensitize people with
regards to DM makeup and its actual considerable financial
impact.
Fig.1 DM make up pattern in Dadri
II. DMAIC APPROACH IN NTPC DADRI -A CASE STUDY
To start the main content firstly it is necessary to know
about DMAIC process. The DMAIC (Define-Measure-AnalyzeImprove-Control) method is a formalized problem-solving
process of Six Sigma. It’s made-up of five steps to apply to any
procedure of a business to improve effectiveness. These five
phases are design, measure, analyze, improve and control. In
power plants we all know DM water is main working fluid in
generation cycle and its process parameters are flow, pH,
conductivity, silica, COC, total hardness , TDS, langelier index
etc. The core philosophy of this paper is how to increase our
performance level in DM consumption and how can we attain
close to six sigma .If we increase sigma value by this approach
we will certainly improve our efficiency and will make our
process more sustainable.
Flow diagram of DMAIC :
Supplier
Input
DM plant
Make -up
Process
Opn practices
Steam is produced by DM water . So , It is main motive
element in generation cycle. DM water is of two kinds – cyclic
and non-cyclic DM. Cyclic dm water is that which is a part of
steam cycle and non cyclic dm water is used in various
applications like Stator water, SG ECW/ TG ECW etc. Here our
focus is on cyclic DM water consumption only. Calculations are
based on six months dm water trend of unit - 3 of stage- 1 dadri
plant. DM make up water enters in a condenser at atmospheric
temperature that is heated around 537°C for raising steam. DM
integrator is used to measure daily cycle make up water as
percentage of feed water flow. Presently, makeup water
consumption of Dadri is around 0.55-0.56% of MCR (Maximum
Continuous Rating) {while the best figure till date is 0.45%).
One of the best figures of DM make up is of Dhanu Plant
where make up was 0.37% (and this figure is of 2005, so
indeed we have much to do!)
Output
Customer
Reduction in m/u
NTPC Management
III. METHODOLOGY ADOPTED
To apply DMAIC approach in minitab 16.0 software six
month data from 1st Feb’13 to 31st July data of unit 3 has been
taken. (Just as random sample).
Cycle make up water
consumption has to be converted in terms of percentage of MCR
of feed water flow so that this methodology can be applied to
other power plants. As it is not possible to reduce cycle make up
water consumption to zero and minimum is the best, LSL (lower
specification limit) cannot be fixed for water consumption.
Hence, only USL (around 100 tons = 0.65%) and target value of
50 tons has been taken.
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
Define
The first phase defines the problem statement and goal
statement necessitating the team to identify the need of
implementation of six sigma and major challenges faced. Goal
statement – To minimise variability in DM consumption and
keeping consumption in a fixed band of percentage BMCR flow.
In define phase we define Critical success factors (CSF), which
are those factors which need to be worked upon for achieving
goal statement. Following are the critical success factors:
Reduction in DM make up and its inventories, Improvement in
heat rate due to reduction in APC and Secondary benefits and
periodically water audit for continuous improvement.
3
Measure
There is a very basic fact that we cannot control what we
don't measure. In cycle make up water consumption at TPP,
make up water flow is measured by an integrator. Although we
can use ultrasonic flow meter for finding velocity and flow from
any pipe easily. Every plant must have this apparatus as it is very
easy to handle and find out flow from any pipe easily. This phase
involves analyzing CSFs determined in measure phase. It also
identifies the Before Improvement Values (BIV) for each CSF
which will help in evaluation of monetary savings. Just before
going to six sigma tools a graphical summary is plotted for six
month data. It is a graphical summary of DM makeup.
Summary for DM
A nderson-Darling N ormality Test
150
300
450
600
750
900
A -S quared
P -V alue <
12.54
0.005
M ean
S tDev
V ariance
S kew ness
Kurtosis
N
162.81
106.88
11423.18
4.0299
21.7619
152
M inimum
1st Q uartile
M edian
3rd Q uartile
M aximum
41.00
115.00
136.50
178.00
883.00
95% C onfidence Interv al for M ean
145.68
179.94
95% C onfidence Interv al for M edian
130.00
148.17
95% C onfidence Interv al for S tDev
95% Confidence Intervals
96.06
120.46
Mean
Median
130
140
150
160
170
180
Fig.2 DM consumption stats in unit 3 ( Data : Feb-Aug 2013)
ANALYSE
Firstly when we take data of six months, minitab16.0 firstly
trims value of top 5 % and bottom 5% data. As these data are like
outliers. ( What is outlier – It is just like a boy coming school
daily on time but gets delayed on some day it doesn't mean that
he is a late comer. In a same way some max or min DM data may
be due to unit light up / shutdown or whatever). .We see A
squared value is 12.54 .Graph is skewed to higher side positively
skewed..
Runchart
Runchart was drawn for unit 3 from 1st Feb to 1st August
(six month timeframe). This is a DM water integrator reading. P-
values for clustering = (0.014), trend = 0.241, oscillation =
(0.798) and mixtures = (0.986). As the p-value for the cluster test
is less than the alpha value of 0.05, we can conclude that special
causes are affecting the process, and we should investigate
possible sources. Clusters can be even evidence of sampling or
measurement problems, but more to more data collection
removes this probability. p value for clustering is less than 0.05 it
means that with 95 % confidence we can say that pattern is not
normal. We have to look for the reasons for non normality. If p
value is more than 0.05 it means that distribution is normal.
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
4
.
Run Chart of C2
1000
800
C2
600
400
200
0
2
4
N u mb er o f ru n s ab o u t med ian :
E xp ec ted n u mb er o f ru n s:
Lo n g est ru n ab o u t med ian :
A p p ro x P -Valu e fo r C lu sterin g :
A p p ro x P -Valu e fo r M ixtu res:
6
7
12.0
6
0.014
0.986
8
10
12
Sample
14
N u mb er o f ru n s u p o r d o w n :
E xp ec ted n u mb er o f ru n s:
Lo n g est ru n u p o r d o w n :
A p p ro x P -Valu e fo r T ren d s:
A p p ro x P -Valu e fo r O sc illatio n :
16
18
20
22
13
14.3
3
0.241
0.759
Fig 3. Run chart
Process Capability Analysis
Process Capability of DM
Using Box-Cox Transformation With Lambda = -0.21
(using 95.0% confidence)
U S L*
t ransformed dat a
P rocess D ata
LS L
*
Target
50
USL
150
S ample M ean
162.809
S ample N
152
S tDev (Within)
101.486
S tDev (O v erall) 106.879
Target*
Within
O v erall
P otential (Within) C apability
Z.Bench
0.16
Low er C L 0.00
U pper C L 0.32
Z.LS L
*
Z.U S L
0.16
C pk
0.05
Low er C L 0.00
U pper C L 0.11
A fter Transformation
LS L*
Target*
U S L*
S ample M ean*
S tDev (Within)*
S tDev (O v erall)*
*
0.446754
0.356285
0.361257
0.0307728
0.0347414
O v erall C apability
0.24
O bserv ed P erformance
P P M < LS L
*
P P M > U S L 414473.68
P P M Total 414473.68
E xp.
PPM
PPM
PPM
0.28
Within P erformance
> LS L*
*
< U S L* 435824.42
Total
435824.42
0.32
0.36
0.40
E xp. O v erall P erformance
P P M > LS L*
*
P P M < U S L* 443102.20
P P M Total
443102.20
0.44
Z.Bench
0.14
Low er C L -0.02
U pper C L 0.30
Z.LS L
*
Z.U S L
0.14
P pk
0.05
Low er C L -0.01
U pper C L 0.10
C pm
0.33
Low er C L 0.31
Fig. 4. Process capability
Process capability analysis was performed using minitab to
draw curve for cycle make up water from TPP measured through
flow meter. In our case Z bench value comes 0.16σ presently dm
make up is +0.16 sigma level. If we limit upper specification
level around 150 ton (1% BMCR flow) and target value around
50 Ton
Fish Bone diagram
Using survey results and brainstorming fish bone diagram is
plotted to find different causes related to man,machine,method
and material .
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
5
Fig. 5. Fish bone diagram
Bar Chart
Actual DM water wastage from different points was possible. On the basis of audit sheet, bar chart was drawn.
Fig. 6. Barchart
Pareto Chart
Pareto chart is prepared by getting responses from dm water audit survey. The above barchart is replicated into paretochart
Pareto Chart for DM leakage contribution
Commulitive %
80
80
60
60
40
40
20
20
0
Commulitive R
Commulitive %
Percent
Cum %
37
52.86
51.2
51.2
49
20.34
19.7
70.9
59
14.29
13.8
84.8
67
11.43
11.1
95.8
Other
4.29
4.2
100.0
Percent
100
100
0
Fig. 7. Pareto Chart
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
6
for any SQC we require UCL /USL (upper critical limit or upper
specification limit) and LCL/LSL... OEM gives us relaxation of
DM up to 2% of BMCR.. We know our LCL/LSL can't be zero
as power generation does not work on isolated system concept.
The processes are practically open system (although it should be
closed one).So, our goal statement converges to a band of DM
(%BMCR) .Say like 0.3 to 0.5%. So for calculation of LCL, we
have used statistics.Firstly we see that our DM consumption is
following which type of distribution. To get more specific results
we find DM distribution pattern individually by minitab16.0.
Improve
From the above analysis we find that valve passings like
safety valve passing, drain valves passing are the biggest cause
of problem. High energy drains are also major culprit. SWAS
drains are very much neglected we can reduce DM leakages from
swas sample drains by periodic awareness and training of lab
analysts, communicationgap between operation and chemistry
staff and casual approaches can be removed to find out
solutoions. For NTPC this analysis has two -three things which
are very important: Firstly Lower limit definition – As we know
Probability Plot for DM
99.9
99
99
90
90
50
10
Logistic
A D = 3.797
P -V alue < 0.005
Loglogistic
A D = 0.894
P -V alue = 0.011
50
10
1
1
0.1
0
250
500
DM
0.1
10
750
99
99
90
90
50
10
100
DM
1000
3-P arameter Loglogistic
A D = 0.927
P -V alue = *
Johnson Transformation
A D = 0.345
P -V alue = 0.480
Normal - 95% C I
99.9
Percent
Percent
3-P arameter Loglogistic - 95% C I
99.9
50
10
1
0.1
Goodness of F it Test
Loglogistic - 95% C I
99.9
Percent
Percent
Logistic - 95% C I
1
10
100
DM - Threshold
1000
0.1
-4
-2
0
DM
2
A fter Johnson transformation
Fig. 8 Probabilistic distribution
We find lologistic normal distribution in dm pattern and p
value is more than 0.05 (with 95% conformality). The greater the
CpK value, the better. A CpK greater than 1.0 means that the 6σ
(±3σ) spread of the data falls completely within the specification
limits. A CpK of 1.0 means that one end of the 6μ spread falls on
a specification limit. A CpK between 0 and 1 means that part of
the 6σ spread falls outside the specification limits. A negative
CpK indicates that the mean of the data is not between the
specification limits. In layman's language roughly sigma level is
3 x cpk values...
Control
In this stage, new process considerations are documented
and frozen into systems so that the gains are permanent. All
possible related causes of specific measured or approximated
where no measurement was identified problem from analysis
phase were tackled and shut out in control phase. With mere
1ton addition of DM makeup in system doesn’t only add
monetary burden of Rs.1225/-(Ref. Appendix 2). Regular
checking of all sources to be done on fixed time basis. We can
also go for some retrofitments in flanges as we always see in DM
line there is very much flange leakage. Victaulic coupling can be
used in place of flanges which have best reliability. Ultrasonic
flow meter should be used frequently to find actual amount of
DM leakages. Deareator vent valve orifice to be checked during
overhauling for its erosion. Presently there is no forum like
bureau of water efficiency. On a pilot basis we can pitch this idea
.In our audit report we can give some rating analogous to star
rating given by BEE for every equipments. We can improve
some better operation practices likewise if there is unit light up
just after condenser flood test and if we get no abnormality we
can send that water to drum directly. (Generally after flood test
hotwell is drained). We can use AHU drains in chilling of swas
compressors. If ambient temperature is very less during winter
season we can even isolate one PHE also. (Turbine ECW has 3
PHE and boiler side have 2 PHE). Passing of some less using
valves like pegging steam valve or CRH extraction valves in
deaerator or CRH supply to seal steam should be done regularly.
Steam traps are very much neglected. Steam trap drains to be
collected at one vessel. CBD flow measurement and LPD valves
passing to be checked properly. One can say that by regulating
so much we can’t suffocate our plant operation system. It is a
wise idea but we should keep one thing always in mind. And this
thing is a Taguchi loss function. This function says even we are
under specification limit but we are not at our specified target
value we are making our losses in square term. If we are “x”
away from target it means our loss into the system is order of “x 2
” and this loss is transmitted to society itself. It can be easily
depicted by following curve.
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 4, Issue 2, February 2014
ISSN 2250-3153
7
Fig. 9. Taguchi loss function / Learning curve
IV. ANALYSIS AND CONCLUSIONS
The preferred Study proves that we can implement Six
Sigma methodology in even power plant to perform better.
Higher consumption of DM water is found to be a big problem in
a thermal power plant. The causes for more DM water
consumption are safety valve passing, swas drain leakage,
various pumps seal problems, vacuum pump overflow and of
course blow down and drain or vents. This paper is open ended
paper as actual quantification of different leakage points to be
noted down in plant by using ultrasonic flow meter. Our R&D
team can do a cost benefit analysis on use of Victaulic coupling.
After doing and implementing DMAIC study a post work
financial audit is to be done. We all know in order to minimise
our DM consumption we can't suffocate our plant on the basis of
equipment healthiness. So after analysing our study O&M
management team can see this scenario under acceptable risk,
tolerable risk or unacceptable risk. For this purpose a sensitivity
analysis is required to be carrying out.
ANNEXURE
[1]
[2]
Online survey webpage: (
In PDF format attached with sheet)
http://www.surveymonkey.com/s/ZVDCHDS, http://www.surveymonkey.c
om/s/JLSKPR8,http://www.surveymonkey.com/s/JX5CW2D
DM heat loss calculation excel sheet ( In word format attached with sheet )
[1]
www. minitab.com
REFERENCES
AUTHORS
First Author – Himanshu Kumar, Deputy Manager ( O&M –
OPN), Email: [email protected]
Second Author – Anurag Singh, Manager ( P&S)
www.ijsrp.org