Group 2: Christina Graziose Dave Lund Milan Nguyen

1Where Innovation Is TraditionWhere Innovation Is Tradition

Group 2:Christina GrazioseDave Lund Milan Nguyen

Determining the Efficacy of Modifications to T-AGS 60 Ships (DEMoTAGS)

Sponsor: Mr. Gregory Opas, Merrill-Dean Consulting

2

Agenda• Background• Problem Statement and Scope• Assumptions• Bottom Line Up Front• System• Approach• Model Overview• Data Analysis• Identification of Modifications Effects• Recommendations• Conclusion

3

Background• US Navy operates a fleet of 6 T-AGS Class Oceanographic Survey vessels

• Powered by 2 Z-drives: provide propulsion and directional control of the vessel• Recent ship modifications were made to enlarge the skeg• Towing tank and computational fluid dynamics analyses performed prior to mods

• Analyses suggested a level of fuel savings would occur• No comprehensive analysis of performance improvements done after the mods• T-AGS vessels operate in one of three modes:

• Underway (UW): vessel is moving and producing its own power• Not-underway (NUW): vessel is anchored and producing its own power• Cold iron: vessel is docked and receives power from outside generators

4

• Problem• Determine if skeg mods improved fuel consumption• Develop mathematical model

• Calculate propulsion fuel consumption and determine skeg mod effects on fuel efficiency based on ship speed and sea state

• Scope• Only UW and NUW will be analyzed

• NUW data will identify the hotel load power requirements• Overall, determine how skeg mods affected ship fuel consumption when UW

Problem Statement and Scope

5

Assumptions• When ship is not-underway, power generated solely supports

hotel load• Propulsion power can be sufficiently estimated by taking

underway power and subtracting not-underway power• Skeg mods do not affect the hotel load• No additional power is generated beyond what is needed to

support hotel load or propulsion power• Weight of diesel fuel is 7.2 lbs/gal• Weight of the vessel is constant• Ship speed and sea state are the primary variables that affect

fuel consumption

*All assumptions were approved by customer

6

Bottom Line Up Front (BLUF)• Fuel Consumption

• All vessels had fuel reduction post skeg modification• Reduced average yearly fuel consumption by 17%• Average yearly savings of ~$4.8 million

• Other modifications• Provided additional reductions in fuel consumption

• ANOVA to test if fuel consumption amongst vessels are the sameµ fuel consumption 1 = µ fuel consumption 2= … = µ fuel consumption 6

• Evidence of a difference between each vessel’s fuel consumption• Mathematical Model

• Calculated average fuel consumption based on speed and sea state

Model accurately represents actual data Skeg mods resulted in yearly savings of ~$4.8 million

7

• Multiple variables affect ship fuel consumption:• Ocean Current• Wind• Temperature• Speed• Sea State• Others

• Analyzed the effect of speed and sea state on the ship’s fuel consumption • Additive effect on the resistance acting on the ship

System

8

Approach• The study was completed through three tasks

• Task 1: Data Collection and Literature Research• Task 2: Data Analysis and Model Development• Task 3: Findings and Conclusions

9

Model Overview

• Goal of model to predict ship fuel consumption based on power consumption• Speed and sea state are major parameters used to calculate power

consumption• Hypothesis:

• Predicted fuel consumption will not be affected by skeg mods since it is computed from speed

• Actual fuel consumption will be affected by skeg mods• Predicted fuel consumption should start to deviate from actual fuel

consumption when skeg mods occurred

Regression Model for Speed Power

Relationship

Calculate Hourly Power in kW and HP

(qry-103)

Calculate Hourly Fuel Consumption

(qry-103)

Compute Monthly Fuel Consumption

Residuals(qry-105, qry-106)

Calculate Sea State Factor(qry-101)

Plot Residuals to Identify Fuel

Consumption Trends

Outlier Analysis Outlier Analysis

Model Baseline

Aggregate Hourly into Monthly Fuel

Consumption(qry-104)

Speed Power Data Hourly Ship Log Data

Monthly Fuel Data

Calculate Monthly Fuel Consumption

(qry-102)

= Input

= Process

= Output

11

Model Implementation

• Model was implemented using Microsoft Access• Three major data sets provided:

• Monthly Consumption and Op Hours• Ship Logs• Speed versus Power data

• Tables were created to store data • Queries were built to process the data

12

Tables

13

Queries

14

ShipLog Table• Contains ship log entries - recorded every few hours

Largest data table containing over 42,000 records

15

MonthlyConsumption Table• Stores monthly barrels of fuel consumed and hours of

operation while Underway and Not-underway

16

• Outlier Analysis:• Anderson-Darling normality test• Histograms • Boxplots (with fences)

• MonthlyConsumption Outlier Results:• Underway Fuel Consumption: 5.97% of data• Not-underway Fuel Consumption: 19.95% of data

• Missing ShipLog Data:• Excluded months with less than 75% of daily data

Data Analysis

Site NameTotal

MonthsMonths with

No DataMonths With < 75% Data

Usable Months

Percent Unusable Months

USNS Bowditch 96 30 33 33 66%USNS Heezen 96 14 38 44 54%USNS Henson 96 41 45 10 90%USNS Mary Sears 96 4 46 46 52%USNS Pathfinder 96 42 34 20 79%USNS Sumner 96 3 48 45 53%

Majority of outliers due to missing data

17

0

500000

1000000

1500000

2000000

2500000

3000000

65 75 85

The

Sum

of S

quar

es

Percentage of Monthly Data Required for Analysis

Data Variation - Sum of Squares for Recorded Propulsion Fuel Consumption

USNS Sumner

USNS Pathfinder

USNS Mary Sears

USNS Henson

USNS Heezen

USNS Bowditch

• Sensitivity analysis on monthly data• 65%, 75%, and 85% of monthly

data analyzed• Total variation (sum of squares)• Average variability (sample

variance)

Missing Ship Log Data Sensitivity

75% has low average variability

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

100000

65 75 85

The

Sum

of S

quar

es

Percentage of Monthly Data Required for Analysis

Data Variation - Sample Variance for Recorded Propulsion Fuel Consumption

USNS Sumner

USNS Pathfinder

USNS Mary Sears

USNS Henson

USNS Heezen

USNS Bowditch

Sam

ple

Varia

nce

18

Regression Model for Speed vs. Power• Relationship used for the mathematical model• R2 values used to determine correlation

• R2 value close to 1 indicates high correlation between curve and data points

Used polynomial equation in model implementation

0 2 4 6 8 10 12 14 16 18 200

100020003000400050006000700080009000

10000

f(x) = 2.88372411842636 x³ − 39.8892382137251 x² + 247.62591026939 x + 800R² = 0.994824444496485

f(x) = 728.856269704421 exp( 0.121870690566197 x )R² = 0.95954534829718

f(x) = 22.7902629099067 x^1.98739306345643R² = 0.962540655652399

Speed Power Curve

Power(kW)Polynomial (Power(kW))Exponential (Power(kW))Power (Power(kW))

Speed (kts)

Pow

er (k

W)

19

• Following formula was used for the conversion:• Fuel Consumption = (Specific Fuel Consumption * HP) / Fuel Weight

• Specific Fuel Consumption = 0.36 lbs/hp/hr• Fuel Weight (Diesel) = 7.2 lbs/gal

• Solved for HP and converted to kW by multiplying by 0.746• Histograms were developed for hotel loads

• Most frequent hotel load: ~800 kW range

Estimating Hotel Load

Site Name Mean Median Std Dev Confidence IntervalUSNS Bowditch 801.85 773.45 286.85 [857.79, 745.91]USNS Heezen 880.39 879.24 344.77 [950.84, 809.94]USNS Henson 747.64 704.97 329.73 [810.11, 685.16]USNS Mary Sears 759.08 783.30 122.66 [783.87, 734.28]USNS Pathfinder 871.33 792.55 340.46 [937.08, 805.58]USNS Sumner 831.04 783.30 378.31 [907.93, 754.15]Overall 814.18 783.30 300.46

Estimate of 800 kW for hotel load is reasonable

20

• Engine Fuel Consumption Estimate:• Caterpillar marine propulsion engine fuel consumption of 0.36 lb/hp-hr

• Engine HP is comparable to that of the T-AGS engines

Estimating Engine Fuel Consumption

Caterpillar C280-8 Marine Propulsion Engine (3,634 HP)Engine Speed

(rpm) Power (bhp)BSFC

(lbs/hp-hr)Fuel Rate (gal/hr)

500 386 0.39 21.5600 667 0.379 36630 773 0.376 41.4700 1,060 0.37 55.9750 1,303 0.364 67.7800 1,582 0.358 80.6850 1,897 0.352 95.1910 2,328 0.352 116.8950 2,649 0.355 133.9

1,000 3,090 0.351 154.8Average 0.36

BSFC: Brake Specific Fuel Consumption

21

• Used World Meteorological Organization (WMO) sea state codes• Sea state did not have an appreciable effect on fuel consumption• Sea state resistance curves were used to estimate Sea State Factor• Sea states 0 to 4 had a minimal impact on propulsion power• Sea states 5 to 9 had considerable impact on propulsion power

Calculate Sea State Factor

Sea State Wave Height (m) Wave Height (ft) Sea State Factor Description0 0 0 1 Calm (glassy)1 0.1 0.33 1 Calm (rippled)2 0.5 1.64 1 Smooth (wavelets)3 1.25 4.1 1 Slight4 2.5 8.2 1.016 Moderate5 4 13.12 1.094 Rough6 6 19.69 1.165 Very rough7 9 29.53 1.224 High8 14 45.93 1.271 Very high9 20 65.62 1.306 Phenomenal

22

Output Analysis (1 of 3)

Oct-200

2

Feb-20

03

Jun-200

3

Oct-200

3

Feb-20

04

Jun-200

4

Oct-200

4

Feb-20

05

Jun-200

5

Oct-200

5

Feb-20

06

Jun-200

6

Oct-200

6

Feb-20

07

Jun-20

07

Oct-20

07

Feb-20

08

Jun-200

8

Oct-200

8

Feb-20

09

Jun-200

9

Oct-200

9

Feb-20

10

Jun-201

0

Oct-201

0

Feb-20

11

Jun-201

1

Oct-201

1

Feb-20

12

Jun-201

2

Oct-201

20

50

100

150

200

250

300

Sumner - Propulsion Fuel Consumption

Predicted Prop FCRecorded Prop FC

Skeg Mod &Other Mods

• Model calculations vs. recorded data

Model underestimated FC prior to mod and was more accurate post mod

23

• Analysis of Mathematical Model Data• Analyzed ratio of the predicted to recorded fuel consumption

• 90% of the calculated UW data was within +/- 30% of the recorded UW data• ANOVA to test average fuel consumption amongst vessels

Output Analysis (2 of 3)

Model sufficiently represents real-life data

24

• Skeg modification data identified dates of “other” modifications• Analyzed effect of other modifications on fuel consumption

• Between modifications• After all modifications

Output Analysis (3 of 3)

VesselAverage Fuel Consumption

Post- Skeg Mod

Average Fuel Consumption

Post- Other ModDifference Percent Savings

USNS Heezen 157.81 gal/hr 136.67 gal/hr 21.14 gal/hr 13.4%

Other modifications resulted in fuel consumption reductions

Other Mods: Gondola, Bubble Fence, and Bilge Keel Skeg Extension

25

Site NameAvg Yearly FC Before

Mod (gal/hr)Avg Yearly FC After

Mod (gal/hr)Avg Yearly

Savings (gal/hr)Pct

SavingsUSNS Bowditch 157.12 129.59 27.53 17.5%USNS Heezen 150.54 147.67 2.87 1.9%USNS Henson 168.88 146.87 22.01 13.0%USNS Mary Sears 185.78 171.63 14.16 7.6%USNS Pathfinder 234.66 155.11 79.55 33.9%USNS Sumner 216.33 162.80 53.53 24.7%Overall 185.42 153.78 31.64 17.1%

Skeg Mod Effects on Fuel Consumption• Skeg mod effect on UW fuel consumption

Overall reduction in average fuel consumption

26

Skeg Mod Effects on Cost• Cost savings

• Used diesel fuel costs of $3.86 (current cost as of 15 April)• Cost Savings based on recorded average UW fuel consumption

Total expected monetary savings per year of ~$4.8 million

Avg Yearly Fuel (gal)

Avg Yearly Fuel Cost

Avg Yearly Fuel (gal)

Avg Yearly Fuel Cost

Avg Yearly Fuel Savings

(gal)Avg Yearly

Cost SavingsUSNS Bowditch 834,636 3,221,695$ 664,677 2,565,654$ 169,959 656,040$ USNS Heezen 932,700 3,600,222$ 921,992 3,558,887$ 10,708 41,335$ USNS Henson 1,016,513 3,923,742$ 856,718 3,306,932$ 159,795 616,810$ USNS Mary Sears 1,105,907 4,268,799$ 1,024,580 3,954,879$ 81,327 313,920$ USNS Pathfinder 1,340,815 5,175,545$ 905,664 3,495,864$ 435,151 1,679,681$ USNS Sumner 1,316,621 5,082,159$ 937,870 3,620,176$ 378,752 1,461,982$ Total 6,547,192 25,272,162$ 5,311,501 20,502,393$ 1,235,691 4,769,769$

Site Name

Before Skeg Mod After Skeg Mod Savings

27

Conclusions• Fuel Consumption

• All vessels had fuel reduction post skeg modification• Reduced average yearly fuel consumption by 17%• Average yearly savings of ~$4.8 million

• Other modifications• Provided additional reductions in fuel consumption

• ANOVA to test if fuel consumption amongst vessels are the sameµ fuel consumption 1 = µ fuel consumption 2= … = µ fuel consumption 6

• Evidence of a difference between each vessel’s fuel consumption• Mathematical Model

• Calculated average fuel consumption based on speed and sea state

Model accurately represents actual data Skeg mods resulted in yearly savings of ~$4.8 million

28

Recommendations• Further analysis on sea state effects on fuel consumption

• Perform sensitivity analysis on sea state factors• Perform study to determine exact sea state factors for a T-AGS vessel

• Improve recorded data quality• Daily or weekly data validity checks to capture outliers• Research methods for automatic data recording

• Mathematical model improvements• Incorporate additional variables that affect fuel consumption

• Wind speed/direction• Water Temperature• Variable total fuel weight during mission

• Would require refueling information• Vary BSFC based on vessel speed

29Where Innovation Is TraditionWhere Innovation Is Tradition

Questions?

https://sites.google.com/site/TAGSFuelStudy