1
E alua ion o Powe Sa ing Achie ed by Ro o Sail Sys ems o Ve y
La ge O e Ca ie s
Seungchul Shin1,*, Sang-Yeob Kim1, Min-Su Kim1, Joongyu Kim1, Cheolho Kim2
1 Ko ean Regis e , Busan, Republic o Ko ea
2 Siemens Indus y So wa e L d., Republic o Ko ea
Abs ac . This s udy p esen s a comp ehensi e compu a ional luid dynamics (CFD) analysis o e alua e he
ae odynamic pe o mance and powe -sa ing po en ial o a o o sail sys em ins alled on a ull-scale Ve y La ge
O e Ca ie (VLOC). The simula ion amewo k inco po a es ealis ic a mosphe ic bounda y laye (ABL) wind
p o iles, appa en wind condi ions, o o ope a ional cons ain s, and ae odynamic in e ac ions be ween he
o o sails and he ship hull. Con e gence es ing and model alida ion we e conduc ed o ensu e he accu acy
and eliabili y o he CFD esul s. The analysis showed ha o o -gene a ed ne h us is maximized when he
appa en wind angle app oaches 90°, and scales p opo ionally wi h he squa e o he ue wind speed. Howe e ,
o o RPM limi a ions a high wind speeds cons ain u he pe o mance gains. The s udy also e ealed ha
hull-induced low dis o ions in luence o o e ec i eness based on loca ion and wind di ec ion. Addi ionally,
la e al o ces we e ound o inc ease signi ican ly wi h wind speed, aising po en ial conce ns ega ding essel
maneu e abili y. Rou e-based powe -sa ing assessmen s, using wind da a om he ERA5 hindcas model,
yielded a e age ne powe sa ings o 283 kW and 309 kW o he B azil–Eas Asia and B azil–Eu ope ou es,
espec i ely—equi alen o 1.8–2.0% o calm wa e p opulsion powe . These simila esul s a e a ibu ed o
compa able appa en wind condi ions ac oss bo h ou es. As he es ima es a e based on hou ly ne powe a he
han annual sa ing, he indings o e ou e-speci ic, p ac ical insigh s o suppo shipowne s in making
in o med decisions ega ding o o sail implemen a ion. O e all, he s udy p o ides bo h echnical
unde s anding and ope a ional guidance o applying o o sail sys ems on la ge comme cial essels
Keywo ds: Ro o sail, Compu a ional Fluid Dynamics, Powe Sa ing.
1. Backg ound
The ma i ime indus y aces inc easing p essu e o imp o e uel e iciency and educe g eenhouse gas emissions
as in e na ional egula ions igh en. Wind-assis ed p opulsion echnologies ha e e-eme ged as p omising
solu ions, wi h o o sails (Fle ne o o s) gaining pa icula a en ion as an e ec i e means o ha nessing wind
powe h ough he Magnus e ec . A o o sail is a powe ed spinning cylinde ha gene a es li pe pendicula o
he wind low, augmen ing a essel’s h us . Ea ly ials o his concep da e back o he 1920s, bu mode n
ad ances and success ul ins alla ions (e.g., by No sepowe ) ha e demons a ed signi ican uel sa ings po en ial—
ypically anging om 5% o 20% unde a o able condi ions [1]— enewing in e es in o o sails as a iable
emissions- educ ion echnology. This backg ound unde sco es he g owing mo i a ion o quan i a i ely e alua e
o o sail pe o mance on con empo a y ca go ships.
Se e al nume ical s udies ha e ecen ly in es iga ed he ae odynamic pe o mance o o o sails using
Compu a ional Fluid Dynamics (CFD), examining how design and ope a ing pa ame e s in luence he gene a ed
o ces. De Ma co e al. (2016) [2] pe o med s eady-s a e Reynolds-A e aged Na ie –S okes (RANS) CFD
simula ions o sys ema ically a y key o o design pa ame e s—including spin a io, aspec a io, and he p esence
o end pla es—and cha ac e ized he esul ing li and d ag coe icien s. Thei s udy demons a ed ha inc easing
spin a ios and aspec a ios enhances li , and ha end pla es (Thom disks) imp o e ae odynamic e iciency by
mi iga ing ip o ices. Kwon e al. (2022) [3] ex ended hese analyses h ough addi ional pa ame ic CFD
in es iga ions o s andalone o o s, explo ing a wide ange o spin a ios and geome ic con igu a ions o assess
hei e ec s on ae odynamic o ces and o que equi emen s. These s udies p o ided aluable p elimina y
pe o mance p edic ions o o o sails unde idealized condi ions.
* Co espondence o: [email p o ec ed]
16 h In e na ional Symposium on P ac ical Design o Ships and O he Floa ing S uc u es PRADS 2025
Ann A bo , MI, USA, Oc obe 19 h – 23 d 2025
2
Despi e hese ad ances, mos exis ing esea ch has been limi ed o simpli ied geome ies, small-scale models,
o gene alized wind condi ions [2]. Fo ins ance, p e ious s udies ypically conside ed isola ed o o s in uni o m
oncoming low, o en neglec ing he p esence o he ship’s hull and he a iabili y o a mosphe ic wind p o iles
[3]. While some s udies ha e a emp ed o apply highe - ideli y CFD app oaches o e ine o o sail pe o mance
p edic ions, such e o s o en emained con ined o simpli ied single- o o se ups and s eady-s a e assump ions
[4]. Simila ly, empi ical and semi-empi ical p edic ions equen ly elied on ixed appa en wind angles o
labo a o y-scale Reynolds numbe s [1]. Consequen ly, signi ican unce ain ies pe sis when ex apola ing hese
indings o ull-scale ships ope a ing in eal en i onmen al condi ions [5].
Recen e o s ha e sough o b idge his gap. Tillig and Ringsbe g (2020) [4] compa ed CFD p edic ions wi h
sea ial da a om ships equipped wi h o o sails, e ealing ha while nume ical models can cap u e gene al ends,
eal-wo ld de ia ions o en a ise due o hull– o o in e ac ions and luc ua ing a mosphe ic condi ions. Mo e
ecen ly, Sampaio e al. (2024) [5] ad anced he ield by conduc ing ull-scale CFD simula ions o a 325,000 DWT
e y la ge o e ca ie (VLOC) i ed wi h i e o o sails, inco po a ing 40 yea s o ERA5 hindcas wind da a o
p obabilis ically assess sys em pe o mance. Thei s udy emphasized he in luence o hull-induced low dis o ion,
o o – o o in e e ence, and deckhouse u bulence on h us a iabili y, and highligh ed he impo ance o
op imizing spin a io unde ealis ic condi ions. Howe e , hei app oach p ima ily ocused on long- e m
p obabilis ic ends a he han di ec ly quan i ying ime- esol ed ne powe sa ings o speci ic oyages.
Building upon hese insigh s and add essing he emaining gaps, he p esen s udy dis inguishes i sel by
cons uc ing de ailed ae odynamic pe o mance pola diag ams om ull-scale CFD simula ions and in eg a ing
hem in o a ship pe o mance model. This me hodology enables he di ec es ima ion o hou ly ne powe sa ings
ac oss di e en ade ou es, p o iding p ac ical and ope a ionally ele an me ics o shipowne s and e o i
designe s. By coupling ull-scale CFD-de i ed ae odynamic da a wi h ealis ic en i onmen al condi ions, his
wo k o e s a mo e accu a e and ac ionable e alua ion o he powe -sa ing po en ial o o o sail sys ems unde
eal se ice scena ios.
Speci ically, he p esen s udy conduc s high- ideli y uns eady RANS simula ions on a ull-scale VLOC
equipped wi h mul iple o o sails, de eloping h us and d ag pola diag ams ac oss a wide ange o appa en wind
angles and speeds. These pola esul s a e hen inco po a ed in o a ship pe o mance amewo k o simula e powe
sa ings o e ep esen a i e ade ou es using ERA5 wind da a. Fi s , he op imal o o ope a ing speed (RPM) is
iden i ied based on maximizing ae odynamic e iciency wi hou incu ing excessi e d ag o powe consump ion.
Nex , he de i ed ae odynamic o ces a e mapped o powe -sa ing pola cha s, illus a ing he ne powe
con ibu ion achie able unde a ying en i onmen al condi ions. Finally, using hese cha s, p ojec ed powe
sa ings on di e en ou es a e compa ed, e ealing how ou e-speci ic wind pa e ns in luence o o sail
pe o mance.
O e all, he s udy demons a es ha by in eg a ing ull-scale CFD simula ions wi h oyage-speci ic
en i onmen al p o iles, i is possible o ob ain mo e accu a e and ope a ionally meaning ul p edic ions o wind-
assis ed p opulsion bene i s. The indings alida e he o o sail’s po en ial o educe powe consump ion on la ge
o e ca ie s and p o ide ac ionable insigh s o ship ope a o s and designe s, he eby b idging he gap be ween
concep ual es ima es and eal-wo ld ope a ional pe o mance. This wo k con ibu es o he ad ancemen o wind-
assis ed p opulsion echnologies and suppo s he b oade deca boniza ion objec i es o he ma i ime indus y.
2. Nume ical me hodology
1. Compu a ional domain and bounda y condi ions
The compu a ional domain was cons uc ed a ound a DWT 324,000- on Ve y La ge O e Ca ie (VLOC) equipped
wi h i e Fle ne o o s ins alled along he ship's cen e line. Each o o has a diame e o 4 m and a heigh o 24
m. The endpla e o each o o has a diame e o 6 m and a hickness o 0.35 m. The o e all ship geome y and he
con igu a ion o he compu a ional domain a e illus a ed in Figu e 1 and 2. Veloci y inle s we e de ined a he
on , le , igh , and op bounda ies o he domain, while a p essu e ou le was applied a he ea bounda y. The
bo om bounda y was se as a symme y plane. A ully de eloped A mosphe ic Bounda y Laye (ABL) p o ile
was applied a he eloci y inle s. The ABL eloci y dis ibu ion is de ined as:
𝑈
=
𝑈
(
𝑧
𝑧
)
(1)
3
whe e he exponen 𝛼 is 1/7 as ecommended by he In e na ional Towing Tank Con e ence (2014) [6], he 𝑧
is he e e ence heigh se a 10 m and 𝑈 is he wind speed a he e e ence heigh .
Figu e 1. 3D model o he essel used in his s udy
Figu e 2. Compu a ional domain and bounda y condi ions
4
2. Simula ion se up
To e alua e he powe sa ing pe o mance o Fle ne o o s ins alled on he uppe deck o he ship, he Uns eady
Reynolds-A e aged Na ie -S okes (URANS) equa ion is sol ed using comme cial so wa e STAR-CCM+ 18.06.
The ini e olume me hod is used o disc e ize he URANS. The Semi-Implici Me hod o P essu e Linked Equa-
ion (SIMPLE) algo i hm is adop ed o eloci y and p essu e coupling [7]. A second o de upwind scheme is used
o he disc e iza ion o con ec ion e m and a i s o de o wa d Eule scheme o he empo al disc e iza ion is
used h oughou all simula ion. Fo he ull-scale simula ion ha co esponds o ully u bulen calcula ion, he
𝑘−𝜔 Shea -S ess T anspo (SST) u bulence model is used o close he URANS equa ion [8]. The compu a-
ional domain consis s o a backg ound egion and an a ea o in e es as shown in Figu e 3. The backg ound egion
was meshed using a hexahed al g id, which is ad an ageous o educing compu a ional ime and gene a ing low-
skewness g ids. In con as , he egion o in e es including he o o and he hull, whe e complex low phenomena
a e expec ed, was meshed using a polyhed al g id due o i s obus ness in cap u ing in ica e low ea u es and
p o iding s able solu ions in highly complex low en i onmen s. (STAR-CCM+ use guide, 2023) [9]. The non-
dimensional wall dis ance (y+) alue is se o less han 1 using he cus om p ism laye . I is impo an o ecognize
ha he simula ion ime s ep is in luenced by he RPM o he o o . To ensu e he s abili y o he nume ical me hod,
he ime s ep mus sa is y a cons ain on he maximum cell-based Cou an -F ied ichs-Lewy (CFL) numbe [10].
Consequen ly, as he angula eloci y inc eases, he allowable ime-s ep dec eases.
3. Ve i ica ion and alida ion
3.1 Con e gence es
To ensu e he eliabili y o he nume ical simula ions, he g id con e gence es was pe o med and e alua ed
using he G id Con e gence Index (GCI) me hod (Celik e al., 2008) [11], which is based on Richadson ex apo-
la ion (Richa dson, 1910; Richa dson and Gaun , 1927) [12][13]. The G id Con e gence Index p o ides a nume -
ical measu e o he unce ain y in he g id sys em. The con e gence es was conduc ed o a model-scale single
Fle ne Ro o a spin a io o 1 and an aspec a io o 5.1. Subsc ip s 1, 2, and 3 co espond o he ine, medium,
and coa se g ids, espec i ely. To ob ain he G id Con e gence Index (GCI), he o de o accu acy (𝑝) mus i s
be calcula ed. Which is exp essed as ollows:
𝑝
=
1
ln
(
𝑟
)
|
ln
|
𝜖
𝜖
⁄
|
+
𝑞
|
,
𝑞
=
ln
𝑟
−
𝑠
𝑟
−
𝑠
,
(1)
Figu e 3. Hyb id g id sys em
5
whe e 𝑟 is he g id e inemen ac o .
𝑆
=
𝑠𝑖𝑔𝑛
𝜖
𝜖
,
(2)
whe e 𝜖 =𝑆−𝑆, 𝜖 =𝑆−𝑆.
The G id Con e gence Index can be calcula ed as ollows:
Fo he coa se- o-medium g id:
𝐺𝐶𝐼
=
𝐹
𝑒
𝑟
−
1
,
(3)
Fo he medium- o- ine g id:
𝐺𝐶𝐼
=
𝐹
𝑒
𝑟
−
1
,
(4)
whe e 𝑒 =
, 𝑒 =
, 𝐹 is a ac o o sa e y o 1.25 which is applied by Roache (1998) [14].
Addi ionally, he con e gence a io (C𝑅) is a key pa ame e ha indica es whe he a solu ion is con e ging o
di e ging. I is de ined as:
𝐶𝑅
=
𝜖
𝜖
,
(5)
The Richa dson ex apola ed alue, which ep esen s he solu ion on an in ini ely ine mesh is calcula ed as ollows:
𝑆
=
𝑆
+
𝑆
−
𝑆
𝑟
−
1
,
(6)
The esul s o con e gence es s o h ee di e en g id sys ems a e shown in Figu e 4 and Table 1. The ime-
s ep o he g id con e gence es was se o 0.001s o achie e a o a ion o 4° pe ime-s ep. The con e gence a io
was ound o be be ween 0 and 1, indica ing mono onic con e gence. The li coe icien in he medium g id was
1.375, while in he ine g id, i is 1.36. Bo h esul s we e close o he ex apola ed alue o 1.34, wi h ela i e e o s
o 1.49% and 2.61%, espec i ely. The esul s a e also suppo ed by a low GCI alue wi hin 5%. Based on he
esul s o he g id con e gence es s, bo h he medium and he ine a e conside ed accep able. Howe e , as shown
in Figu e 5, when he compu a ional cos o he ine g id is app oxima ely 1.5 imes highe han ha o he medium
g id. The e o e, he medium g id was selec ed o u he s udy.
6
Figu e 4. Li coe icien a ia ion wi h espec o mesh e inemen
Table 1. Resul s o he g id con e gence es
While he g id con e gence es es ima es he dependency o he g id sys em, a ime-s ep con e gence es mus
be conduc ed o assess he dependency on ime s ep. The ime-s ep con e gence es s we e pe o med using he
medium g id sys em wi h h ee di e en ime s eps, ollowing he same GCI p ocedu e desc ibed abo e. As shown
in Figu e 6 and Table 2, he empo al disc e iza ion e o was ound o be qui e low, wi h a GCI alue wi hin 1%.
Based on hese esul s, a ime-s ep co esponding o a o a ion o 4° pe ime-s ep was conside ed he bes com-
p omise be ween compu a ional cos and he empo al disc e iza ion e o .
Figu e 5. Compu a ional ime o he di e en g id sys em
G id
𝐶
𝐷
/
∆
𝑥
𝑟
𝐺𝐶𝐼
𝐶𝑅
𝑝
𝐶
Fine
1.360
100
1.40
1.93%
0.6 1.6 1.34
Medium
1.375 71 1.38
3.27%
Coa se 1.400 50 - -
7
Figu e 6. Li coe icien a ia ion wi h espec o ime-s ep e inemen
Table 2. Resul o he ime-s ep con e gence es
3.2 Valida ion
To assess he accu acy o he p esen CFD model, he simula ion esul s we e compa ed o he expe imen al da a
o a cylinde wi hou end pla e which has an aspec a io o 5.1 (Badalamen i, 2008) [15]. The expe imen al
condi ion used o alida ion co esponds o 𝑅𝑒 =1.9 × 10. The li and d ag coe icien we e compa ed in a -
ious spinning a io. Figu e 7 shows he li and d ag coe icien s as he unc ion o he o o spin a io (SR).
Spin a io is exp essed as ollows:
𝑆𝑅
=
𝑛𝜋𝐷
𝑈
,
(5)
whe e 𝑛 is he RPM o he o o , 𝐷 is he diame e o he o o and he 𝑈 is he ees eam eloci y.
As shown in Figu e 7, he p esen CFD simula ion could p edic well o li coe icien in all spinning a io.
Con e sely, while he end o he d ag coe icien wi h espec spin a io is simila be ween he expe imen and
CFD esul s, a no able disc epancy has been obse ed. As discussed by Thouaul (2012) [16], he ully u bulen
calcula ions a e no he app op ia e app oach a low Reynolds numbe s. Howe e , since his s udy ocuses on high
Reynolds numbe s, u bulence modeling is essen ial. Ne e heless, due o he lack o expe imen al da a a high
Reynolds numbe s, he u bulen model was applied o lowe Reynolds numbe cases, which esul ed in he ob-
se ed disc epancies in d ag o ce.
G id
𝐶
𝐷𝑒𝑔
/
∆
𝑡
𝑟
𝐺𝐶𝐼
𝐶𝑅
𝑝
𝐶
Fine
1.379
2
1.93
0.27%
0.4
1.28
1.382
Medium
1.375
4 2.00 0.64%
Coa se 1.365
8 - -
8
(a) Li coe icien (b) D ag coe icien
Figu e 7. Valida ion o CFD esul s agains expe imen al da a
4. Resul
4.1 RPM Op imiza ion
To p ope ly e alua e he e iciency o he o o sail, i is essen ial o de e mine he op imal RPM. Fo his pu pose,
li and d ag coe icien s a e ob ained o spin a ios anging om 0 o 8 unde he condi ions whe e he o o sail
is expec ed o pe o m mos e icien ly, speci ically, a an appa en wind angle o 90°. Using he ex ac ed coe i-
cien s, he h us o ce in he di ec ion o mo ion is calcula ed o a ious spin a ios unde a bi a y appa en wind
speeds and angles. This h us o ce is hen con e ed in o powe , and he ne powe is de i ed by sub ac ing he
powe equi ed o o a e he o o om he powe gene a ed in he h us di ec ion. The RPM ha yields he highes
ne powe unde gi en condi ions is selec ed as he op imal RPM. Figu e 8 illus a es he p ocess o he RPM
op imaliza ion.
Figu e 8. Flow cha o RPM op imiza ion
9
4.2 E ec o hull
(a) AWA=20°
(b) AWA=41°
(c) AWA=63°
16
(a) Rou e 1 : B azil – Eas Asia (TWS) (b) Rou e 2 : B azil – Eu ope (TWS)
(c) Rou e 1 : B azil – Eas Asia (AWS) (d) Rou e 2 : B azil – Eu ope (AWS)
Figu e 16 P obabili y dis ibu ion o ue and appa en wind speed
The p e iously discussed ue wind condi ions (TWS and TWA) do no accoun o he essel's mo ion; in eali y,
he ship encoun e s appa en wind speed (AWS) and appa en wind angle (AWA), which a e in luenced by he
ec o sum o wind and essel eloci y. Figu e 16 (c), (d) and 17 (c), (d) illus a e hese cha ac e is ics, assuming
a cons an essel speed o 14.6 kno s. Unlike he ela i ely Gaussian dis ibu ion o TWS and TWA, he AWS and
AWA dis ibu ions a e no ably i egula and s ongly dependen on ou ing and encoun e geome y, o en di e g-
ing om ypical s a is ical models. In pa icula , Rou e 2 (B azil–Eu ope) shows mo e equen occu ences o
bo h high AWS condi ions exceeding 15 m/s and e y low AWS condi ions unde 5 m/s, indica ing a b oade
ange o encoun e scena ios. Rega ding AWA—an impo an ac o o o o sail pe o mance—bo h ou es p e-
dominan ly expe ience ollowing winds be ween 135° and 225°, bu wi h di e en cha ac e is ics: Rou e 1 (B azil–
Eas Asia) shows a ela i ely uni o m sp ead ac oss his ange, allowing o mo e s able h us gene a ion, whe eas
Rou e 2 exhibi s a concen a ed peak nea 180°, sugges ing mo e equen di ec ollowing wind, which may en-
hance ins an aneous e iciency bu could lead o g ea e a iabili y in o o -assis ed p opulsion due o i s na ow
di ec ional dis ibu ion.
17
(a) Rou e 1 : B azil – Eas Asia (TWA) (b) Rou e 2 : B azil – Eu ope (TWA)
(c) Rou e 1 : B azil – Eas Asia (AWA) (d) Rou e 2 : B azil – Eu ope (AWA)
Figu e 17 P obabili y dis ibu ion o ue and appa en wind angle
18
2. Discussion on ne powe sa ing o o o sail sys em
A ou e-based assessmen o wind-assis ed p opulsion pe o mance was ca ied ou o es ima e he ne powe
sa ings achie able using a o o sail sys em ins alled on a Ve y La ge O e Ca ie (VLOC). The analysis inco po-
a es AIS-de i ed essel posi ions and headings unde he assump ion o a ixed se ice speed o 14.6 kno s, and
accoun s o ins an aneous wind-ship in e ac ions o calcula e hou ly powe sa ings in kWh. While he e e ence
essel is a VLOC equipped wi h i e o o sails, he AIS da a includes a b oade popula ion o o e ca ie s, and
he s udy does no ollow speci ic oyages be ween de ined po s. As such, he esul s e lec gene al ope a ional
pa e ns along wo majo ou es—B azil–Eas Asia and B azil–Eu ope—and p o ide insigh in o ypical sa ings
on an hou ly basis, a he han o al oyage consump ion.
The p obabili y dis ibu ions o hese hou ly ne powe sa ings a e depic ed in Figu e 18, showing a e age alues
o app oxima ely 283 kWh o he B azil–Eas Asia ou e and 309 kWh o he B azil–Eu ope ou e. The modes
di e ence be ween he wo is consis en wi h he compa able AWS and AWA p o iles iden i ied ea lie . Howe e ,
he a ia ion in each dis ibu ion also emphasizes he impac o ou e-speci ic wind condi ions and ship-wind en-
coun e angles on o o sail e ec i eness. A supplemen a y able summa izes key s a is ical me ics, including he
a io o ne powe sa ings o calm-wa e p opulsion demand. Al hough his a io depends hea ily on essel size,
engine powe , and hull esis ance, i p o ides a ela i e benchma k o e alua ing o o sail con ibu ions unde
ealis ic oyage condi ions.
(a) Rou e 1 : B azil – Eas Asia (b) Rou e 2 : B azil – Eu ope
Figu e 18 P obabili y dis ibu ion ne powe sa ing o o o sail sys em
Table 4 S a is ical alue o ne powe sa ing
Rou e 1 Rou e 2
Region B azil - Eas Asia B azil - Eu ope
Powe sa ing (kWh) Mean 282.9 309.3
Median 47.2 76.7
S d. 638.6 662.7
Mean powe sa ing a io (% o calm sea p opulsion) 1.8% 2.0%
19
5. Conclusion
This s udy conduc ed a comp ehensi e compu a ional luid dynamics (CFD) analysis o e alua e he ae odynamic
pe o mance and powe -sa ing po en ial o a o o sail sys em ins alled on a ull-scale Ve y La ge O e Ca ie
(VLOC). The simula ion amewo k inco po a ed a ealis ic a mosphe ic bounda y laye (ABL) p o ile, appa en
wind condi ions, o o mechanical cons ain s, and ae odynamic in e ac ions be ween he o o sails and he ship
hull. To ensu e he eliabili y o he nume ical esul s, bo h con e gence es ing and model alida ion agains
e e ence da a we e pe o med and con i med p io o he main simula ions.
The esul s indica ed ha he ne powe con ibu ion o he o o sails was maximized be ween TWA=90° and
TWA=120° and he pe o mance scaled app oxima ely wi h he squa e o he ue wind speed. Howe e , a highe
wind speeds, mechanical limi a ions on o o RPM imposed cons ain s on he e ec i e powe gene a ion. Addi-
ionally, he p esence o he ship hull was ound o locally dis o he in low eloci y ield, a ec ing he h us
gene a ed by each o o depending on i s posi ion and wind di ec ion. In he p esen essel con igu a ion, he
p esence o he ship hull was ound o induce a local accele a ion o he in low, esul ing in a posi i e impac on
o o h us gene a ion and, consequen ly, on ne powe sa ings. Howe e , i should be no ed ha di e en hull
geome ies may p oduce a ying e ec s on he in low ield, po en ially leading o ad e se ou comes. The e o e,
ca e ul conside a ion o hull-induced low modi ica ions is essen ial when de e mining he placemen and ins al-
la ion o o o sails o ensu e op imal pe o mance.
La e al o ces ac ing on he essel inc eased signi ican ly wi h wind speed, aising po en ial conce ns o maneu-
e abili y and udde load. These indings unde sco e he impo ance o balancing powe sa ing pe o mance o
RTS wi h na iga ional s abili y when designing o o sail sys ems.
Rou e-based ene gy sa ing assessmen s, conduc ed using wind da a de i ed om he ERA5 hindcas model,
e ealed a e age ne powe sa ings o app oxima ely 283 kWh and 309 kWh o he B azil–Eas Asia and B azil–
Eu ope ou es, espec i ely. These co espond o app oxima ely 1.8–2.0% o he essel’s calm wa e p opulsion
powe o VLOC, sugges ing ha o o sails o e modes bu measu able gains in ene gy e iciency unde ealis ic
ope a ing condi ions. The small di e ence in ne powe sa ings be ween he wo ou es is p ima ily due o he
simila i y in he dis ibu ions o appa en wind angle (AWA) and appa en wind speed along bo h shipping ou es.
I should be no ed, howe e , ha he o e all powe sa ing pe o mance o he o o sail sys em can a y signi i-
can ly depending on he size o he essel, he scale o i s engine powe , and i s ope a ional speed p o ile. When
same o o sail sys em is applied o smalle essels, he ela i e powe sa ing pe cen age may appea la ge due o
he educed calm wa e esis ance. The e o e, cau ion mus be exe cised when in e p e ing powe sa ing a ios
exp essed as pe cen ages. As his s udy es ima es powe sa ing e ec s based on hou ly ne powe con ibu ion o
speci ic ou es— a he han long- e m o annualized a e ages—i p o ides shipowne s wi h mo e immedia e and
ou e- ele an pe o mance insigh s o suppo p ac ical decision-making on o o sail ins alla ion.
In summa y, his esea ch no only quan i ied he p ac ical bene i s and limi a ions o o o sail sys ems unde
ull-scale, ou e-speci ic scena ios, bu also p o ided aluable insigh s in o hei ae odynamic beha io , design
conside a ions, and ope a ional implica ions o la ge comme cial essels.
20
Re e ences
[1] J. T au , P. Gilbe , R. Walsh, and A. Bows, “P opulsi e powe con ibu ion o a ki e and a Fle ne o o on
selec ed shipping ou es,” Applied Ene gy, ol. 113, pp. 362–372, 2014.
[2] A. De Ma co, S. Mancini, C. Pensa, G. Calise, and F. De Luca, “Fle ne o o concep o ma ine applica-
ions: a sys ema ic s udy,” In e na ional Jou nal o Ro a ing Machine y, ol. 2016, A icle ID 3458750,
pp. 1–12, 2016.
[3] C. S. Kwon, S. M. Yeon, Y. C. Kim, Y. G. Kim, Y. H. Kim, and H. J. Kang, “A pa ame ic s udy o a Fle -
ne o o in s andalone condi ion using CFD,” In e na ional Jou nal o Na al A chi ec u e and Ocean Engi-
nee ing, ol. 14, a . no. 100493, 2022.
[4] F. Tillig and J. W. Ringsbe g, “Design, ope a ion and analysis o wind-assis ed ca go ships,” Ocean Engi-
nee ing, ol. 211, p. 107603, 2020.
[5] C. M. P. Sampaio, K. Nishimo o, F. Rugge i, M. L. Pin o, and P. on P i zelwi z, “P edic ion o o o sails
pe o mance on la ge o e ca ie s,” in P oc. In . Symp. on Ma ine P opulso s, Be lin, Ge many, Ma . 2024.
[6] In e na ional Towing Tank Con e ence (ITTC), P ac ical Guidelines o Ship CFD Applica ions, ITTC Rec-
ommended P ocedu es and Guidelines 7.5‐03‐02‐03, 27 h ITTC, Copenhagen, Denma k, 2014.
[7] S. V. Pa anka and D. B. Spalding, “A calcula ion p ocedu e o hea , mass and momen um ans e in h ee-
dimensional pa abolic lows,” In e na ional Jou nal o Hea and Mass T ans e , ol. 15, no. 10, pp. 1787–
1806, 1972.
[8] F. R. Men e , “Two-equa ion eddy- iscosi y u bulence models o enginee ing applica ions,” AIAA Jou nal,
ol. 32, no. 8, pp. 1598–1605, 1994.
[9] Siemens Digi al Indus ies So wa e, STAR-CCM+ 18.06 Use Guide, 2023.
[10] R. Cou an , K. F ied ichs, and H. Lewy, “On he pa ial di e ence equa ions o ma hema ical physics,” IBM
Jou nal o Resea ch and De elopmen , ol. 11, no. 2, pp. 215–234, 1967.
[11] I. Celik, U. Ghia, P. J. Roache, C. J. F ei as, H. Coleman, and P. E. Raad, “P ocedu e o es ima ion and e-
po ing o unce ain y due o disc e iza ion in CFD applica ions,” Jou nal o Fluids Enginee ing, ol. 130,
no. 7, p. 07800, 2008.
[12] L. F. Richa dson, “The app oxima e a i hme ical solu ion by ini e di e ences o physical p oblems in ol -
ing di e en ial equa ions, wi h an applica ion o he s esses in a mason y dam,” Philosophical T ansac ions
o he Royal Socie y o London, Se ies A, ol. 210, pp. 307–357, 1910.
[13] L. F. Richa dson and J. A. Gaun , “The de e ed app oach o he limi . Pa s I–II,” Philosophical T ansac-
ions o he Royal Socie y o London, Se ies A, ol. 226, pp. 299–361, 1927.
[14] P. J. Roache, “Ve i ica ion o codes and calcula ions,” AIAA Jou nal, ol. 36, no. 5, pp. 696–702, 1998.
[15] C. Badalamen i and S. A. P ince, “The e ec s o endpla es on a o a ing cylinde in c oss low,” 26 h AIAA
Applied Ae odynamics Con e ence, Honolulu, HI, USA, Aug. 2008.
[16] N. Thouaul , C. B ei sam e , N. A. Adams, J. Sei e , C. Badalamen i, and S. A. P ince, “Nume ical analysis
o a o a ing cylinde wi h spanwise disks,” AIAA Jou nal, ol. 50, no. 2, pp. 271–283, 2012.
[17] H. He sbach, B. Bell, P. Be is o d, S. Hi aha a, A. Ho ányi, J. Muñoz‐Saba e , e al., “The ERA5 global
eanalysis,” Qua e ly Jou nal o he Royal Me eo ological Socie y, ol. 146, pp. 1999–2049, 2020.
