Statistical analysis software article - jan 2016

24
T
he improvement in maritime
technology, such as sensors,
network connections and data
loggers, gives the chance to continuously
and automatically monitor a ship’s
performance. This means it is now
possible to evaluate and plan predictive
maintenance routines that will keep the
ship’s efficiency up.
One of the most significant benefits
of continuous monitoring is the
reduction of uncertainty in data
analysis due to the large amount of data
recorded and analysed.
The major operational cost for
the ship owner is related to the fuel
consumption. It is well known to ship
owners that the most important factor
influencing consumption is increased
ship resistance, in other words, the
appearance of speed loss.
According to one study published
by Casper Service, hull resistance may
increase from 12% in the first year after the
dry docking period to 40% by the end of the
fifth year. It is important to pay attention
to this in order to plan predictive action
before the efficiency levels become very
low, costing the operator a lot of money
due to extra fuel used and increasing
atmospheric emissions in consequence.
The objective of this document is to
explain a tool that will help ship owners
to operate their ships more efficiently. This
optimisation will focus on the evaluation
of the hydrodynamic performance,
studying the speed loss progression over
time by means of statistical analysis
software. A reliable prediction of ship
speed loss is essential from economic and
environmental perspectives.
The software will help owners to take
optimal decisions that could maintain or
raise their ships’ performance. Proactive
action, such as planning docking periods
for the right time (predicting when the
ship performance will reach the limits
of optimal efficiency, saving operational
costs) and evaluating the dry docking
periods (i.e. if the anti-fouling system
applied on the hull is effective or not), is
a powerful tool to operate the ships in the
most efficient way.
Statistical analysis (SA) software
In this document, I have used software
offered by the Norwegian company
Kyma as a reference; this company has 25
years of experience in the market and its
software is considered a good example
for understanding the usefulness of
SA software.
Statistical analysis definition
Satistical analysis refers to the setup
methods used to process large amounts of
data and report overall trends. Statistical
analysis is particularly useful when dealing
with noisy data because it provides ways
to objectively report on how unusual an
event is, based on historical data recorded.
Statistics are applied every day to
become more scientific about decisions
that need to be made.
Data analysed: performance
observations
The statistics will manage large amounts
of data in order to make a reliable long
trend analysis. These data are called
performance observations.
The SA software creates the performance
observations automatically. The software
collects data automatically from different
sensors, such us power meter, GPS, Speed
Log instrument, flow meters, etc.
The data from the sensors are not
always steady, so, in order to get sensible
and comprehensible instant values for
the operators, the software has a logging
period setup. During this logging
period, the software gathers the data
and calculates the average of the data.
Commonly accepted, the logging period
is 15 seconds. This method uses the
cumulative moving average (CMAn):
The brute-force method to calculate
this would be to store all of the data,
calculate the sum, and divide by the
number of datum points every time a
new datum point arrived. However, it is
possible to update a cumulative average
as a new value:
The above formula is used on the
SA software.
For the statistical analysis (long trend
evaluation) it is established as a reliable data
output frequency, with one performance
observation per day. Therefore, all the
averaged instant data calculated every
logging is recorded daily.
The purpose of the continuous
monitoring is to reduce the uncertainty of
the data analysed.
The average for the speed deviation
compared with the baseline (design data) will
givetheperformanceobservationthatitisused
in the statistical analysis. Each performance
observation is stored in a database for further
evaluationbySAsoftware.
Reference bases implemented
on the SA software
The performance observations are plotted
on graphs. However, these observations
are useless without reference bases to
compare with. The SA software uses two
reference levels.
One reference level is the design data
for each ship, obtained from the model
Feature 1 | Green Ships
The Naval Architect January 2016
In recent years, the main efforts within the maritime industry have been on
energy efficiency and regulations regarding safety and environment. Related
to that appear several new concepts such as “Green Shipping” and “Smart
Ships”, reports Carlos Gonzales, marine engineer
Statistical analysis software & speed loss
evaluation
n
XX
CMA n
n
++
=
...0
1
0
1
+
⋅+
=+
n
CMAnX
CMA n
n
NA Jan 16 -p24+26+27+28.indd 24 24/12/2015 10:35:13

25The Naval Architect January 2016
Feature1
tank test. This level will be kept constant
for all the ship.
Another reference level is the
benchmark level, which is created with
data collected after delivery and/or after
any major repair on the ships. Therefore,
this reference level will be dynamic and it
will change after any major event.
Constant reference base on SA software
The data from the model tank test will
show the relationship between the power
delivered to the propeller and the ship
speed through the water for ballast and
design draft conditions. The daily average
of speed through the water (performance
observation) is compared to the baseline,
giving a deviation in percentage, which is
used on the SA software.
As a result, the speed deviation is
corrected on the software for the ship’s
cargo condition (“power vs. speed”
relation is not the same at ballast and
laden condition). Following this premise,
the speed deviation value per day is
calculated taking into account the vessel
mean draft, ship speed through the water
and the power delivered to propeller.
This baseline designates the “zero level”
on the statistical analysis. This level is
constant for the ship’s life.
Dynamic reference base on SA software
The dynamic reference base corresponds
with the benchmark level. A benchmark
is a standard set by a number or several
numbers to estimate the basis of
something to compare with.
The benchmarking automatically updates
after any major event. The benchmark is
created by taking the average of a number
of daily observations (i.e. 120 performance
observations) starting after a major
event. The observations are only valid for
benchmarking if data is on the wind limits
and the ship is not on manoeuvring or in an
abnormal sailing mode (main engine load
above 35 % MCR).
It is very important to reset
benchmarking after any major event
because the ship will have a “new ship
status” to compare with. This will be a
new benchmark level. Therefore, the
benchmark designates the “new ship
condition” to be used as a new reference
level to compare the ship performance.
Filtering applied on the SA software
The software automatically creates
performance observations day after day
for this reason; it is required to set some
filtering on the SA software to avoid
useless data that could invalidate the long
trend analysis.
The vessels sail around the world
finding diverse weather conditions,
sea currents, sailing modes (normal
navigations, manoeuvring, etc.)
and cargo conditions (ballast/laden
conditions).
The software works continuously,
which makes it very important to apply
filtering to discard data in case of bad
weather conditions (using the Beaufort
scale as a reference: if the wind force
is above BF6 then the performance
observation must not be used on the
statistical analysis) and/or if the ships
are under abnormal sailing modes (if
ships are manoeuvring or the MCR is
less than 35%).
In addition, the speed deviation is
correctedbasedontheship’scargocondition.
Thisismadebycorrectingthedesignbaseline
with the mean draft value (mean draft will
indicate the ship’s cargo condition).
Analysis done by the SA software
After the filtering, the software has a
new set of data on which it performs
the statistical analysis. It creates a new
trend function in the form of f(x) = ax +
b based on only the included data points.
This line is plotted along with all the
included data points as a solid line.
The trend function is calculated
from the data points using the linear
regression model of least square fit.
This is a common method in statistics to
find a linear relationship between a set
of data points. The least square method
creates the following equations for
calculating “a” and “b” in the formula:
Using this trend line, the users can find
out how the ship is currently performing
(compared to benchmark and baseline).
By means of a simple colour coding,
it is possible to determine the ship
performance status referent to the speed
loss. The colour-coding (see figure 3)
included in the software is as below:
- Blue: the ship is in the benchmark
period
- Red: The performance status for the
ship is “not ok”
- Yellow: The performance status for the
ship “should be under observation”
- Green: The performance status for the
ship is “ok”
∑ ∑
∑ ∑ ∑
−
⋅−⋅⋅
= 22
)( xxn
yxyxn
a
bxaxf +⋅=)(
Figure 1: Trend line for ship performance comparison
NA Jan 16 -p24+25+26+27+28.indd 25 05/01/2016 10:11:46

26 The Naval Architect January 2016
Feature 1 | Green Ships
The trend line calculated with the
performance observations will show
the speed loss, and can, in consequence,
estimate the fuel impact due to this loss.
Such a calculation is made looking into
the baseline “FO consumption vs. Ship
speed”, and this baseline is made looking
at the same relation of “Power vs Speed”
and “SFOC vs. Power”.
The effect of this speed loss on fuel
consumption can be seen in Figure 1.
In the scenario that the ship desires to
reach 19 knots, the design baseline and the
actual speed loss due to the hull resistance
has increased. The ship’s 16.9 knots speed
(11.1% speed loss as the trend line shows
on the next section on this document)
can be extrapolated into extra fuel usage
that results in raising the operational costs
and increasing the atmospheric emissions
in consequence. The extra fuel used due
to the speed loss is 1,000 kg/hr if it is
converted to tonne/day:
The estimated fuel impact is related to fuel oil
based on ISO corrected calorific value (HCV
orLCV).
Hullstatus&speedloss
To make this evaluation, the main inputs to be
consideredare:
- Shaftpower
- Shipspeedthroughwater
The speed loss is caused by increasing the
ship’s advance resistance. Knowing that, it is
straightforward to analyse the hull status by
meansofaspeedlossstudy.
Itisimportanttoremarkthatthespeedinput
used is the speed through the water because it
takesseacurrentsintoconsideration.
The hull resistance is influenced by several
factors.Themostrepresentativefactorsforship
operationsare:
 Stillwaterresistance
 Windandwaveresistance
 Resistanceduetoverticalshipmotions
 Resistanceduetosteering
 Resistanceduetofouling
One of the filters applied to the SA
software is for weather conditions. Hence, if
the weather is bad, the data is automatically
deleted. In this way, the influence of
Figure 2: Long trend representation for Speed loss
Figure 3: Speed
Log deviation from
baseline. The SA
software shows the
results for the last
period (after last
drydock)

ship per
is as bel
‐ Blue
‐ Red:
‐ Yello
‐ Gree
The tren
can, in c
looking
at same
The effe
In the sc
actual sp
speed lo
extra fu
emissio
converte
1
m
The esti
or LCV
Hull sta
To mak
rformance s
low:
: the ship is
The perfor
ow: The per
en: The perf
nd line calcu
consequenc
into the ba
the relation
ect of this sp
cenario that
peed loss du
oss as the tr
uel usage tha
ns in conse
ed to tonne/
310
24*000

E
imated fuel
V).
atus & spee
ke this evalu
status refere
s in the benc
rmance statu
rformance s
formance sta
ulated with
ce, estimate
seline “FO
n of “Power
peed loss on
t the ship de
ue to the hu
end line sho
at results in
quence. The
/day:
24
[tonne/
impact is re
ed loss
uation, the m
ent to the sp
chmark peri
us for the sh
status for th
atus for the
the perform
the fuel imp
consumptio
r vs Speed”
n fuel consu
esires to rea
ull resistance
ows on next
raising the
e extra fuel
day]
elated to fue
main inputs
4
eed loss. Th
iod
hip is “not o
he ship “shou
ship is “ok
mance obser
pact due to
on vs. Ship
and “SFOC
umption can
ach 19 knots
e has increa
t section on
operational
used due to
el oil based
to be consid
he colour-co
ok”
uld be unde
”
rvations wil
this loss. Su
speed”, and
C vs. Power
n be seen in
s, following
ased. The sh
n this docum
l costs and i
o the speed
d on ISO cor
dered are:
oding includ
er observatio
ll show the s
uch a calcul
d this baselin
r”.
n the graph b
g the design
hip’s 16.9 k
ment) can be
increasing t
loss is 1000
rrected calo
ded in the s
on”
speed loss,
lation is ma
ne is made
below:
baseline an
knots speed
e extrapolate
the atmosph
0 kg/hr if it
orific value (
oftware
and
ade
looking
nd the
(11.1%
ed into
heric
is
(HCV
NA Jan 16 -p24+26+27+28.indd 26 24/12/2015 10:35:16

27The Naval Architect January 2016
Feature1
“wind and wave resistance” is discarded
from the statistical analysis.
Each performance observation
represents the average data per day,
deleting the momentary effect caused
by steering.
The resistance due to still water and
vertical ship motions have an important
influence when a vessel is sailing in rough
water. Normally, if there are strong winds,
there will also be rough seas. Therefore, by
discarding the performance observation
for windy conditions it is ensured that the
effects of still water and vertical motions on
the ship resistance have been neutralised.
By eliminating other factors, the speed
loss evaluation provides an excellent
indicator about the resistance added
due to fouling. If the hull is painted with
low quality paint, fitted with low quality
anti-fouling systems, or the ship is berthed
for long periods at ports, the performance
of the hull will deteriorate faster.
Deteriorating hull performance results
in speed loss, and means that more power
is required to reach the service speed and
fuel consumption is higher, producing
greater atmospheric emissions.
Above in Figure 2 is the long trend
representation available on SA software
offered by Kyma for one LNG carrier that
recently underwent a drydocking period.
In Figure 3, the black line at “zero level”
is the design level from the model tank test.
The blue dotted line is the benchmark
level, generated with the average values
of the performance observations (120
points) after the last drydock period,
during the “benchmark period”.
The red dots are the performance
observations generated daily within
acceptable weather conditions and normal
sailing status.
On the above example, the trend line
(speed loss) level on the end of that
period is 4.3% below the benchmark
level or 11.1% below the baseline [-4.3%
(compared with the benchmark) +
-6.8% (benchmark level compared with
baseline) = 11.1 %].
In conclusion, the speed loss is 11.1%
compared with the design status (baseline)
or 4.3% compared with the benchmark
generated after the docking period.
The SA software gives the speed loss
(in knots) for three MCR levels (50%,
75% and 90 %) with the actual ship
performance.
The effect of this speed loss converted
into fuel impact is estimated for the SA
software. For the actual example the
results are in Table 1.
In addition, the software predicts when
the ship will reach next colour area, giving
one visual alarm that shows “Observe”
when the trend line reaches the yellow area
and “Not OK” when the trend line reaches
the red area.
Analysing the trend line, the ship owners
can see when the speed loss will be below
the optimal level. If the speed loss shows
a deviation of more than 10% compared
with the benchmark level then the ship
performance is not optimal, and action
should be taken to raise the performance
(such as drydocking). Left unchecked, the
operational costs will increase.
Evaluation of the actions executed on the
ship by the SA software
Ships require different repairs and
modifications during their operational
life. These modifications aim to keep the
ship’s performance high; for instance,
applying new coatings, fitting the ship
with a new bulbous bow design or using a
new propeller design. These actions must
be taken into account for the statistical
analysis. The software helps to evaluate if
the modifications have had a good effect on
the ship’s performance or if they have failed
to meet expectations.
Related to the previous paragraph,
the SA must be interrupted if some
major event takes place, such as a
drydocking period. When the repair is
over, a new benchmarking period will
resume and a new trend line will start.
The new benchmark level for the actual
ship condition will be the reference for
further comparison of the performance
observations after the trend event.
See Figure 4 for an image of the Kyma
SA software, which shows a long-term
progression over 12 years of the speed
loss with major events set on the graphic.
Two trend events (green vertical lines)
corresponding with two drydockings can
be seen.
After the first drydocking, the statistical
analysis does not show a big improvement
on hull performance. However, after the
second drydock period, the trend line
jumped up to 8% better performance than
before that drydocking period. It shows that
the actions carried out on the ship during
Table 1: The effect of speed loss converted into fuel impact.
*Calculated with the conversion factor used by IMO for HFO, C=3.114 tonne FO/ton CO2
MCR (%) Speed loss (knots) Fuel oil impact
(tonnes/day, ISO
reference)
CO2 emissions
(estimated, tonne/
day)*
50 -0.75 + 8.5 26.47
75 -0.87 + 14.2 44.22
90 -0.91 + 19.9 61.97
Figure 4: Long-term progression of ship speed loss over 12 years
NA Jan 16 -p24+26+27+28.indd 27 24/12/2015 10:35:16

28 The Naval Architect January 2016
Feature 1 | GReen ShipS
that drydocking period have been efficient,
raising the hydrodynamic performance
about 8%, which in turn cuts fuel and
emissions as a consequence.
Future development
Some companies are already making
software that includes the automatic
evaluation of statistical analysis such as
Kyma or BMT Smart. The next step is to
enable the automatic and real time transfer
of data from onboard the ship to the ship
owner’s office onshore. That transmission
of data will allow owners to evaluate
performance and look at their vessels’
progression in order to be more effective
when making predictive maintenance. In
addition, this possibility will also create a
“Big Data” database for owners. They will
have a big database with all the parameters
included on the software from their ships
for further evaluation and comparison
between sister ships. To do that, they will
require a bigger server or cloud to manage
the information.
The company used in this document has
theautomatictransferoption,andaccording
to that company, they are improving
constantly to offer the best service possible
on the evaluation of the data on real time.
Conclusion
By using software with statistical analysis for
analysis of speed loss, it is easier to predict
the hydrodynamic performance as well as
to evaluate the jobs carried out on the ships,
evaluating, for example, how effective the
applied coating system is or how effective
the latest drydocking period was.
It will help owners to keep track of
the actual speed performance and keep
efficiency as high as possible, while
reducing operational costs and atmospheric
emissions.
The real examples used in this document
show how easy it is to quantify the speed
loss (hydrodynamic performance) and its
effects on the operational costs. In the end,
SA software is a useful tool for supporting
the owners in their decision-making.
References
www.kyma.no
www.bmtsmart.com/
www.epa.gov/oms/regs/nonroad/marine/
ci/fr/r98021.pdf
www.imo.com
AOGEXPO.COM.AU
SAVE 10%
ON ALL CONFERENCE PASSES
USING PROMO CODE: RINA
Proudly supported by
NA Jan 16 -p24+26+27+28.indd 28 24/12/2015 10:35:26

Statistical analysis software article - jan 2016

More Related Content

What's hot (19)

Similar to Statistical analysis software article - jan 2016 (20)

Recently uploaded (20)

Statistical analysis software article - jan 2016