On he e ec o popula addi i e manu ac u ing echnologies on
he pe o mance and noise emission o UAV p opelle s
J. Ga cía-Tísca a,∗, P. Quin e oa, P. Va elaa, F. N. Ramí eza, A. C emadesb
aCMT – Clean Mobili y & The mo luids, Uni e si a Poli ècnica de València, Camino de Ve a, 46022 Valencia, Spain
bFLOW, Enginee ing Mechanics, KTH Royal Ins i u e o Technology, S ockholm, Sweden
Abs ac
Noise emains a majo conce n o he widesp ead employmen o Unmanned Ai Vehicles (UAVs) ope a ions, especially in
u ban en i onmen s. Mo eo e , no only mus o e all noise le els be conside ed when analyzing he acous ic impac o UAVs,
bu noise quali y based on he spec al con en o he acous ic signal mus also be conside ed due o he psycho-acous ic
e ec on he lis ene s. Addi ionally, as UAVs become widesp ead, he abili y o pe o m ield eplacemen o p opelle s
using addi i e manu ac u ing (AM) is o inc eased in e es o many ope a o s. Howe e , as di e en AM echniques
become popula , hei impac on pe o mance and noise mus be assessed. In his in es iga ion, he e ec o manu ac u ing
cha ac e is ics, such as su ace oughness and lexibili y, on he p opelle h us , o que, and acous ic signa u e is explo ed
h ough an expe imen al campaign, e ealing di e ences depending on he selec ed echnique. A nume ical expe imen is
hen pe o med o isola e he e ec o lexibili y and oughness on hese cha ac e is ics.
Keywo ds: P opelle s; UAV; Acous ics; FDM; SLS; SLA;
1. In oduc ion1
Unmanned Ai Vehicles’ (UAVs) popula i y as an ae ial
2
pla o m o a wide ange o ci il and mili a y applica-
3
ions con inues o inc ease. F om ec ea ional pu poses
4
o su eillance and i e sa e y, passing h ough ilmmaking,
5
in as uc u e inspec ion, oad a ic con ol, cell ne wo k
6
enhancemen , o name a ew, hese ai c a a e becoming
7
widesp ead in bo h he coun yside and u ban spaces [1].8
Howe e , e en hough he echnology in e ms o ligh
9
pe o mance is ma u e, i s ull po en ial, gi en all he de-
10
sc ibed applica ions, is no ye ully ealized, pa ly due o
11
a combina ion o egula o y, sa e y, and accep abili y con-
12
ce ns. On he i s wo on s, se e al ini ia i es in he US,
13
EU, UK, e c., aim o achie e a ha monized egula ion ha
14
allows he sa e ope a ion o UAVs [
2
]. Howe e , he e a e
15
s ill some conce ns abou he public accep abili y o UAV
16
ope a ions. Di e en public opinion su eys conduc ed by
17
a ia ion au ho i ies ha e ound ha UAV noise is one o he
18
leading ac o s among hese conce ns [3].19
When s udying UAV noise, wo aspec s should be con-
20
side ed. On he one hand, he e is he issue o o e all noise
21
le el, which is he added global le el ha con inued UAV
22
ope a ion could b ing in o homes, schools, hospi als, e c.
23
Acco ding o he indings o he Wo ld Heal h O ganiza ion
24
(WHO), noise le els a e he second mos ha m ul o m o
25
en i onmen al cause o heal h p oblems, jus a e chemical
26
pollu an s [
4
]. Thus, ca e should be aken ha he new
27
∗Co esponding au ho . jo ga i@mo .up .es
se ices p o ided by using UAVs do no inc ease hese noise
28
le els and hei heal h consequences o he popula ion. On
29
he o he hand, he psycho-acous ic impac o UAV noise
30
signa u es has been a es ed. Due o he pa icula ligh
31
dynamics o UAVs, based on mul i-p opelle speed con ol,
32
he equency con en o he noise p esen s a semi- andom
33
a iance a ound an a e age, o en pe cei ed as especially
34
annoying [
3
,
5
]. The e o e, bo h o e all noise le el and e-
35
quency con en mus be s udied o add ess hese conce ns
36
p ope ly. 37
Consequen ly, he amoun o esea ch in p opelle ae oa-
38
cous ics has inc eased conside ably in la e yea s. Recen ly,
39
a benchma k o low Reynolds numbe p opelle s has been
40
p oposed [
6
]. O he wo ks ha e ocused on e alua ing he
41
accu acy o di e en nume ical me hods compa ing di e -
42
en blade esol ed me hods wi h BEMT solu ions [
7
,
8
],
43
o wi h he ac ua o line me hod [
9
]. The in luence o
44
non-axial in low condi ions on he acous ic pe o mance
45
has been also explo ed bo h expe imen ally [
10
] and nu-
46
me ically [
11
]. In addi ion, ae odynamic and ae oacous ic
47
p opelle -p opelle and p opelle - uselage in e ac ions ha e
48
also been s udied o bo h mul icop e s [
12
] and ixed-wing
49
UAVs [13,14]. 50
Because o he low Reynolds a which hese p opelle s
51
ope a e, ce ain phenomena like lamina sepa a ion bubbles
52
become impo an [
15
,
16
]. Mo eo e , b oadband noise be-
53
comes mo e ele an , so e o s a e being made o imp o e
54
he accu acy o nume ical me hods [
17
,
18
]. The in luence
55
o he Addi i e Manu ac u ing (AM) echnology used o
56
manu ac u e he p opelle has p o en o ha e an impac on
57
1
his b oadband noise [17,19].58
In pa allel wi h UAVs, ano he echnology ha is be-
59
coming widesp ead is Addi i e Manu ac u ing (AM). Since
60
UAVs a e o en subjec ed o c ashes o o ced landings,
61
which may imply a deg ee o damage, and gi en ha he
62
p opelle s a e o en he mos easily damaged pa , apid
63
in- ield eplacemen o hese componen s becomes an ex-
64
emely desi able capabili y o many kinds o ope a o s
65
[
20
]. Fo example, UAVs a e being used o deli e medical
66
supplies in emo e loca ions o e en emba ked on ocean-
67
going essels. Ob aining spa e pa s such as p opelle s om
68
he UAV manu ac u e is di icul o impossible in hese si -
69
ua ions, and hus, he abili y o apidly manu ac u e hem
70
om a supply o aw ma e ial becomes highly desi able.71
Howe e , eplacing a comme cial p opelle wi h a 3D-
72
p in ed one is no s aigh o wa d, e en i he geome y
73
is exac ly he same. As di e en AM echniques esul in
74
pieces wi h dis inc mechanical p ope ies such as lexibil-
75
i y, su ace inish, oughness, aniso opy, e c., he e is no
76
ce ain y ha he pe o mance o a ec ea ed p opelle , in
77
e ms o h us , o que, and noise emission, is equi alen o
78
ha o he o iginal. P e ious wo ks [
17
,
21
,
22
,
23
] ha e
79
p o en ha ho e pe o mance and noise emissions a e a -
80
ec ed by he AM echnology used, bu u he wo k needs
81
o be done o ully comp ehend hese e ec s unde di e en
82
ope a ing condi ions and o he en i e noise spec um. To
83
add ess his issue, in his in es iga ion, we conside one
84
wooden comme cial p opelle ypical o UAVs (XOAR 9x7)
85
and ec ea e i using he h ee mos widesp ead AM ech-
86
niques, expanding he esul s p esen ed by Ga cía-Tísca
87
e al. [
24
]. All he p oposed p opelle s ha e been es ed
88
a CMT’s anechoic chambe and wind unnel o assess he
89
impac o he di e en AM me hods on h us , o que, and
90
noise emission, including o e all le el, di ec i i y, and spec-
91
al con en . Since each echnique in oduces i s pa icula
92
combina ion o mechanical p ope ies, a nume ical simu-
93
la ion campaign is pe o med, in which each mechanical
94
p ope y o he ma e ial is modi ied sepa a ely, allowing
95
a be e unde s anding o hei impac on he p opelle
96
pe o mance.97
The p esen pape is s uc u ed as ollows: his sec ion
98
p o ides a b ie in oduc ion o he subjec and a con ex ual-
99
iza ion o he cu en s a e-o - he-a o he ield. Sec ion 2
100
p esen s he p opelle used, he addi i e manu ac u ing
101
echniques e alua ed, he expe imen al se up, and he nu-
102
me ical app oach. Sec ion 3shows and discusses he ex-
103
pe imen al esul s, and he conclusions ob ained om he
104
nume ical analysis. Finally, in Sec ion 4, he inal conclu-
105
sions eached a e summa ized.106
Scanning
P ocessing
+
Addi i e
manu ac u ing
Mesh de ail
Figu e 1: Wo k low o he scanning, meshing and manu ac u -
ing o he p opelle s used in his wo k. Top image cou esy o
UAV Wo ks G oup.
2. Me hodology 107
2.1. Selec ed p opelle 108
The comme cial wooden XOAR PJN 9x7" (22.86 cm)
109
p opelle was selec ed o his esea ch. This model o p o-
110
pelle is ypical o ixed-wing UAVs such as he VALAQ120
111
by he UAV Wo ks G oup. In o de o c ea e he addi i e-
112
manu ac u ed (AM) e sions o he p opelle , he geome y
113
Building pla o m
Ex ude head
Hea e
Suppo s
The moplas ic
ilamen spool
Figu e 2: Concep ual schema ic o he FDM echnology.
2
o he p opelle was ex ac ed by manually ine- uning a
114
s uc u ed-ligh 3D scanne digi al model. Figu e 1depic s
115
he a o emen ioned p ocess.116
2.2. Addi i e manu ac u ing117
The e m Addi i e Manu ac u ing (AM), also known
118
popula ly as 3D p in ing, encompasses a la ge a ie y o
119
e y di e en echnologies. Each echnology has i s own
120
s ong poin s and weaknesses, especially in e ms o me-
121
chanical p ope ies and su ace inish o he esul ing i ems.
122
In his in es iga ion, h ee o he mos popula echnolo-
123
gies ha e been selec ed, which a e b ie ly desc ibed in his
124
subsec ion.125
2.2.1. Fused Deposi ion Modeling (FDM)126
Fused Deposi ion Modeling (FDM), also known as Fused
127
Filamen Fab ica ion (FFF), is p obably he mos well-known
128
AM echnology and can be conside ed as he one kicks a -
129
ing i s ecen popula i y. In i s essence, a con inuous ila-
130
men o he moplas ic ma e ial is ex uded h ough a hea ed
131
p in ing head, which causes he ma e ial o mel o so en
132
enough o use wi h he p e iously deposi ed ma e ial. The
133
ex ude head is con inuously mo ed as equi ed o o m he
134
geome y o he desi ed i em. Figu e 2shows his p ocess.135
The mo emen s o he ex uded heads a e p e-compu ed
136
h ough so wa e called slice s because he ypical app oach
137
is o “slice” he i em geome y in o a se ies o ho izon al
138
planes. Commonly, he head will comple e all he equi ed
139
XY mo emen s in each Z plane be o e mo ing o he nex
140
one, which is ypically accomplished by lowe ing he build-
141
ing pla o m i sel . This p ocess esul s in a manu ac u ed
142
i em ha clea ly shows he di e en Z laye s in i s su ace
143
inish, in addi ion o possible “Z seams” esul ing om he
144
ilamen ex usion’s s a ing and inishing poin s a each
145
laye . Mo eo e , he o e hanging pa s o he geome y
146
equi e suppo ing ma e ial ha needs o be emo ed in
147
pos p ocessing s eps.148
In addi ion, he p ocess o building he objec h ough
149
subsequen Z laye s o used o deposi ed co dons o he mo-
150
plas ic ilamen esul s in aniso opic mechanical p ope ies,
151
as he beha io in he Z di ec ion is usually qui e di e en 152
om ha o he X and Y di ec ions. Fu he mo e, objec s
153
a e no ully illed e en i 100% in ill is selec ed, as he 3D
154
space canno be illed by he cylind ical geome y o he
155
ilamen s.156
2.2.2. Selec i e Lase Sin e ing (SLS)157
SLS echnology sol es many o he common issues o
158
FDM. In his app oach, a high-powe , o ien able lase mel s
159
small pa icles o he building ma e ial ( ypically a polyamide
160
such as PA11 o PA12), using hem oge he o o m a
161
solid objec . Again, he p ocess in ol es slicing he objec
162
in o Z planes, as he lase can only a e se he XY plane.
163
Powde ed ma e ial om a ese oi is hinly sp ead o e a
164
mo able bed h ough a olle , whe e he lase sin e s he
165
equi ed slice. Then, he bed is lowe ed, a esh and un-
166
sin e ed powde laye is sp ead on op, and he p ocess
167
epea s un il he objec has been ully buil . The p ocess is
168
shown in Fig. 3.169
This echnique p esen s se e al ad an ages. Fi s , he
170
esul ing objec has p ac ically iso opic mechanical p ope -
171
ies, as he powde pa icles ha e used oge he . Also, he
172
su ace esolu ion is be e han ha o he FDM p ocess, as
173
i is no limi ed by he geome y o he deposi ed ilamen s
174
bu by he e ec i e size o he lase . Finally, as he sin e ed
175
ma e ial is suppo ed by he unsin e ed powde , no suppo
176
ma e ial is equi ed. In ac , e y complex and in e locking
177
geome ies can be p oduced by aking ad an age o his
178
ac . 179
Howe e , his me hod is no wi hou disad an ages.
180
Fi s and o emos , he equi emen o a high-powe ed lase
181
inc eases he complexi y and cos o he machine, es ic ing
182
i s use o mo e p o essional se ings a he han hobby o
183
household en i onmen s. Wo king wi h powde ed ma e ial
184
can also be a nuisance, and equi e p ocedu es and/o a
185
dedica ed wo kshop. The su ace inish is usually ough,
186
e en i i is ee o isible Z laye s o seams. 187
2.2.3. S e eoli hog aphy (SLA) 188
Finally, he SLA echnique elies on he pho opolyme -
189
iza ion o esin. When a UV ligh sou ce is ocused on he
190
bo om o a a con aining pho opolyme esin, he esin
191
solidi ies h ough a pho ochemical p ocess. The lase is
192
s ee ed o solidi y he Z slice as equi ed, and hen, he liq-
193
uid a is lowe ed and he lase d aws he nex slice, which
194
is used wi h he p e ious one, as shown in Fig. 4.195
Th ough his p ocess, laye ing o seam e ec s a e g ea ly
196
educed. Also, he esul ing pieces exhibi iso opic me-
197
chanical p ope ies. Nowadays, SLA machines a e mo e
198
accessible han SLS ones. Fu he mo e, a wide a ay o
199
esins wi h di e en p ope ies is a ailable, al hough he
200
ma e ial choice is no as a ied as wi h FDM. 201
A a ian o his echnique uses a p ojec o o shine
202
he comple e image o each slice in o he esin a , which
203
lowe s he cos o he machine e en mo e. These a e known
204
as Digi al Ligh P ocessing (DPL) machines. Howe e , he
205
p ojec s wo k by p ojec ing a as e image (made up o
206
Unsin e ed
powde
Rolle
Lase head
Building pla o m
Figu e 3: Concep ual schema ic o he SLS echnology.
3
Table 1: Mechanical p ope ies o he addi i e manu ac u ing ma e ials used in his s udy [25,26,27,28].
P ope y Uni FDM: ABS-M30 SLS: Nylon 12 SLA: G ey 4 (PC) SLA: Rigid (PC)
Tensile modulus XZ GPa 2.4 1.85 2.8 4.1
Tensile modulus ZX GPa 2.3
Flexu al modulus XZ GPa 2.22 1.6 2.2 3.4
Flexu al modulus ZX GPa 1.96
Ul ima e ensile s eng h XZ MPa 28.1 50 65 69
Ul ima e ensile s eng h ZX MPa 26.8
Flexu al s eng h XZ MPa - 66 - -
Flexu al s eng h ZX MPa 47.7
Tens. elonga ion a b eak X/Y % 8.1 11 6 5.3
Tens. elonga ion a b eak Z % 1.8 6
Hea De lec ion Temp 0.45 Mpa ºC 103.8 171 73 77
Hea De lec ion Temp 1.8 Mpa ºC 99.9 87 58 60
pixels) ins ead o he con inuous mo emen o he lase spo
207
o SLA machines. Thus, DPL esul s in pieces exhibi ing a
208
ce ain oxel e ec , as i cons uc ed o iny cubes.209
While SLA p esen s ce ain ad an ages, i also has some
210
d awbacks. Wo king wi h he pho opolyme can be di icul ,
211
and ce ain p ecau ions mus be aken o handle i . The
212
esul ing “g een” pieces mus be cleaned o liquid esin ha
213
has ailed o solidi y. Then, he pieces mus be cu ed wi h a
214
combina ion o empe a u e and UV ligh in o de o a ain
215
hei bes mechanical p ope ies. Also, simila ly o FDM,
216
he pieces o en equi e suppo as hey a e being buil .
217
Howe e , unlike wi h FDM in which an auxilia y ex ude
218
can p o ide a specialized, soluble suppo ma e ial, he
219
suppo s in SLA need o be made o he esin a ailable in
220
he a . When emo ing he suppo ing s u s, some ma ks
221
may emain on he su ace o he piece, equi ing manual
222
sanding.223
2.2.4. Facili ies and ma e ials224
Fo his esea ch, he FDM p opelle was manu ac u ed
225
in Ac yloni ile Bu adiene S y ene (ABS-M30) he moplas-
226
ic using a S a asys F170 machine. The Z laye esolu ion
227
was 125
µ
m, wi h he p opelle being p in ed in laye s pa -
228
allel o he o a ion plane. 100 % in ill was selec ed in he229
slice so wa e. A limonene-soluble suppo ma e ial was
230
also used o ensu e p ope a achmen , which was emo ed
231
in pos p ocessing.232
Suppo s
Liquid
pho opolyme
Building pla o m
Lase head
Figu e 4: Concep ual schema ic o he SLA echnology.
The SLS p opelle was manu ac u ed using a Fo mlabs
233
Fuse 1 sys em. I ea u es a 10 W y e bium lase and is
234
capable o a laye esolu ion o 110
µ
m, wi h a lase ocal
235
poin o 200
µ
m. The employed ma e ial was Nylon 12.
236
This is he only echnology ha equi es no suppo o
237
pos p ocessing o he han cleaning he unsin e ed powde
238
o he piece. 239
Finally, he SLA p opelle s we e c ea ed wi h a Fo mlabs
240
3L p in e . This machine ea u es a a ian o he SLA
241
me hod called Low Fo ce S e eoli hog aphy (LFS), which
242
uses a olle and a lexible ay loo in o de o ensu e ha
243
only a e y hin laye o esin is cu ed by he lase . Two
244
250 mW lase heads a e used o speed up he p ocess. The
245
selec ed laye heigh was 50
µ
m, and he XY esolu ion was
246
25
µ
m o he s anda d Fo mlabs G ey 4 p opelle . On he
247
o he hand, a laye heigh o 100
µ
m and an XY esolu ion
248
Figu e 5: The p opelle s es ed in his in es iga ion. F om
le o igh : FDM, SLS, SLA-g ey and SLA- igid e sions. A
zoomed iew highligh s he di e ences in su ace quali y.
4
o 25
µ
m was used o he Rigid 4000 esin p opelle . The
249
p opelle s we e hen cleaned wi h isop opyl alcohol and
250
cu ed wi h he empe a u e and UV ligh cycles p esc ibed
251
by he manu ac u e . The suppo ma ks we e manually
252
sanded o .253
In Fig. 5, he ou p opelle s manu ac u ed by each o
254
he selec ed echnologies a e compa ed agains each o he .
255
I can be seen how he FDM p opelle ea u es e y clea
256
laye s, along wi h he pa hs o he ilamen (no e especially
257
how he sha wall is c ea ed by a ci cula ilamen deposi-
258
ion). The SLS p opelle does no show any inhomogenei y,
259
bu a oughe su ace can be clea ly app ecia ed. Finally,
260
he SLA p opelle shows a smoo h su ace inish, appea ing
261
simila o a cas plas ic piece.262
Focusing on he ma e ial p ope ies, Table 1displays he
263
main mechanical p ope ies o he ma e ials used in his
264
s udy.265
2.3. Expe imen al se up266
In o de o measu e he noise emission o each p opelle ,
267
an elec ic mo o was ins alled in he anechoic chambe
268
a ailable a Labo a o y 5K o he CMT - Clean Mobili y &
269
The mo luids Ins i u e. This chambe ea u es a equency270
cu -o o 100 Hz, and an a ailable in e io space o 7.5
×271
6.5
×
6 m. The non-in luence o eci cula ion in his acili y
272
o 9-inch p opelle s measu emen s has been p e iously
273
demons a ed [
29
]. The engine is an A enge V3 2812-900
274
k and is go e ned by a V-GOOD 2-6S 40A Elec onic Speed
275
Con olle (ESC). An A duino UNO mic ocon olle , linked
276
o a compu e in he con ol oom, commands he ESC.277
To measu e he acous ic emission, six B üel & Kjæ Type
278
4190 ee- ield mic ophones a e moun ed in an a ch, a a
279
Mic ophones
90º
60º
-60º
30º
-30º
0º
Anechoic walls
Mo o
R = 8.75 Dp
ESC
RPM
PS
Con .
PC
DAQ
Figu e 6: Schema ic o he expe imen al se up ins alled in he
anechoic chambe a he CMT Ins i u e.
Fai ing
P opelle
6-axis balance
Mo o
ESC
Figu e 7: FDM-manu ac u ed p opelle ins alled in he «P o .
F ancisco Pay i» wind unnel.
dis ance om he p opelle hub o
8.75
p opelle diame e s
280
(Dp)
. The six mic ophones a e moun ed in inc emen s o
281
30
º
, s a ing om he e ical (ups eam) o he p opelle .
282
They a e simul aneously acqui ed by a PULSE Type 3560-D
283
Da a Acquisi ion Sys em om B üel & Kjæ , also connec ed
284
o he con ol oom compu e . A schema ic o his se up is
285
shown in Fig. 6.286
A e calib a ing he mic ophones h ough a B üel &
287
Kjæ Type 3541 pis onphone, measu emen s we e made a
288
164.5 ps (9870 pm), cap u ing he acous ic signal o 2
289
seconds. 290
The ae odynamic pe o mance cha ac e iza ion o he
291
di e en p opelle s was pe o med in he «P o . F ancisco
292
Pay i» wind unnel a CMT. The es was ca ied ou in he
293
i s es sec ion o he wind unnel ese ed o low u bu-
294
lence ae odynamic es s, h ough a cus om 6-axis balance
295
as shown in Fig. 7. The same mo o and ESC as in he
296
anechoic chambe se up ha e been used. The balance has
297
been designed in-house and buil using 6 Tedea-Hun leigh
298
Model 614 load cells a 70 deg ees. Readings om he load
299
cells a e di ec ly acqui ed h ough a Na ional Ins umen s
300
DAQ sys em. 301
A e calib a ing he load cells using e e ence weigh s,
302
measu emen s we e made a di e en o a ional speeds,
303
bo h wi h he unnel s opped and a 10 m/s, o ob ain da a
304
in ho e and o wa d ligh condi ions. Ten epe i ions o
305
he measu emen s we e pe o med o each p opelle o
306
ensu e epea abili y. 307
2.4. Nume ical se up 308
In o de o be e unde s and he ole o su ace ough-
309
ness and lexing unde load on bo h ae odynamic and acous-
310
ic pe o mances, a nume ical expe imen was ca ied ou ,
311
whe e ad an age was aken o a p e iously alida ed nume -
312
ical se up o a igid, smoo h p opelle p e iously desc ibed
313
by Se ano e al. [
7
]. In his wo k, he simula ions we e
314
5
pe o med ollowing he same app oach bu using he com-
315
me cial so wa e SimCen e STAR-CCM+ 2210.316
As can be seen in Fig. 8, only he p opelle is modeled.
317
I is imme sed in a sphe ical luid domain wi h a adius
318
o
7.5Dp
. A Mo ing Re e ence F ame (MRF) app oach is
319
used, wi h he o a ing domain de ined as a small cylinde
320
a ound he p opelle . An addi ional ixed cylinde is used as
321
a e inemen olume in o de o imp o e he mesh quali y
322
in he p opelle wake, b inging he o al cell numbe o 2
323
million. An image o he compu a ional mesh is also shown
324
in Fig. 8.325
The i s main modi ica ion o he exis ing model o
326
his in es iga ion was he conside a ion o Fluid-S uc u e
327
In e ac ion (FSI), sol ing a Fini e Elemen Model (FEM) o
328
he p opelle in coupling wi h he ini e- olume luid sol e .
329
Gi en he na u e o he said coupling, bo h solid and
330
luid physics a e implemen ed, wi h a mesh o 400 hou-
331
sand cells o he solid egion. The simula ions a e ea ed
332
as pseudo-s a iona y. To achie e his, an implici uns eady
333
model is employed wi h a su icien ly la ge ime s ep, en-
334
su ing ha s eady-s a e condi ions a e eached a high sim-
335
ula ion imes.336
The s uc u al p oblem is sol ed using a nonlinea ge-
337
ome y model, which accoun s o la ge displacemen s and
338
o a ions by sa is ying he equilib ium equa ions in he
339
de o med con igu a ion and i e a i ely upda ing he s i -
340
ness ma ix ia New on’s me hod. A key implica ion is he341
ea men o ollowe o ces, whe e o ces, ac ions, and
342
p essu es adjus hei o ien a ion as he s uc u e de o ms:
343
o ce loads e ain hei ec o componen s, ac ion loads
344
p ese e hei di ec ion while hei esul an changes and
345
p essu e loads emain no mal o he de o med su ace. This
346
7.5Dp
Fluid domain
Mesh de ail
P opelle
Re inemen
olume
Ro a ing
olume
Figu e 8: Nume ical domain wi h de ail o he mesh.
app oach ensu es a mo e accu a e ep esen a ion o slende
347
s uc u es unde going la ge o a ions wi h ela i ely small
348
s ains [30,31]. 349
To ensu e ha he FSI simula ion con e ges o a s eady-
350
s a e solu ion, Rayleigh damping is in oduced o supp ess
351
s uc u al oscilla ions. The damping ma ix
C
is o mula ed
352
as a linea combina ion o he mass and s i ness ma ices
353
[32], Mand K espec i ely, as can be seen in Eq. 1.354
C=αM+βK,(1)
whe e
α
is he mass-p opo ional damping coe icien and
355
β
is he s i ness-p opo ional damping coe icien . In his
356
s udy, hese coe icien s a e se o
α
=0.001 Hz and
β
=0.001
357
s, calib a ed based on he imescales o he obse ed oscil-
358
la ions. This choice ensu es nume ical s abili y and accel-
359
e a es con e gence, allowing o an e icien e alua ion o
360
he s eady-s a e s uc u al esponse. Howe e , i inhe en ly
361
supp esses dynamic ins abili ies, which could be ele an
362
in cases o low s uc u al s i ness o high o a ional speeds.
363
While such e ec s may play a ole in ce ain ope a ional
364
egimes, hei analysis alls beyond he scope o his s udy, 365
which ocuses on he s eady-s a e beha io o he coupled
366
sys em a he han ansien o ins abili y-d i en phenom-
367
ena, as s udied in p e ious wo ks [33,34,35,36]. 368
The o a ion is modeled using a Mo ing Re e ence F ame
369
(MRF), which imposes a cons an o a ional lux in he
370
conse a ion equa ions wi hou physically de o ming he
371
mesh. This app oach signi ican ly educes compu a ional
372
cos compa ed o a ully ansien o a ing mesh simula ion,
373
as can be seen in he wo k o Ga o ano-Soldado e al. [
37
]
374
o Liu e al.[38]. 375
The luid low is go e ned by he Reynolds-A e aged
376
Na ie -S okes (RANS) equa ions, disc e ized using second-
377
o de schemes o ad ec ion and di usion e ms. To model
378
u bulence, he
k−ω
Shea S ess T anspo (SST) model,
379
p oposed by Men e [
39
], is employed o compu e he
380
Reynolds s ess enso . This wo-equa ion u bulence model
381
is widely used in MRF and p opelle simula ions [37,38]. 382
Al hough he low emains essen ially incomp essible
383
ac oss all es ed ad ance a ios, he luid is modeled as an
384
ideal gas o enhance nume ical s abili y and ensu e sol e
385
con e gence. This app oach helps a oid po en ial issues
386
associa ed wi h sol ing he con inui y equa ion in s ic ly in-
387
comp essible o mula ions, whe e small nume ical p essu e
388
luc ua ions can a ec compu a ional s abili y. Addi ionally,
389
ea ing he luid as an ideal gas p o ides g ea e lexibil-
390
i y in sol ing he conse a ion equa ions, educing he isk
391
o di e gence wi hou comp omising accu acy wi hin he
392
e alua ed ange o condi ions [40,41]. 393
The second s udy in oduced a a iable wall oughness
394
ia a modi ica ion o he wall ea men o sweep his pa am-
395
e e and assess i s in luence on pe o mance. This allows
396
us o s udy he e ec o bo h pa ame e s in isola ion o
397
imp o e he unde s anding o he expe imen al esul s. 398
In o de o modi y he wall oughness, a oughness
399
unc ion is in oduced (Eq. 3) [
42
]. Said unc ion depends
400
6
on he oughness pa ame e R+, de ined as in Eq. 2.401
R+= ρu∗
µ(2)
=
1i R+≤R+
smoo h
hBR+−R+
smoo h
R+
ough−R+
smoo h +CR+ia
i R+
smoo h <R+<R+
ough
B+CR+i R+>R+
ough
(3)
Whe e
B,C,R+
smoo h,andR+
ough
a e model coe icien s (de-
402
ined in Table 2),
is he equi alen sand-g ain oughness,
403
ρ
he densi y,
u∗
he eloci y scale,
µ
he dynamic iscosi y,
404
and ais de ined as in Eq. 4.405
a=sin
π
2
logR+/R+
smoo h
logR+
ough/R+
smoo h
(4)
Finally, he Hanson acous ic model [
43
] was used o
406
es ima e he onal noise nume ically. This allows us o s udy
407
he in luence o he ma e ial and su ace oughness on he
408
onal noise wi hou ha ing o ca y ou cos ly ansien
409
simula ions.410
The Hanson model is a heo e ical p opelle noise o mu-
411
la ion ha akes in o accoun bo h he loading (Eq. 5) and
412
hickness (Eq. 6) noise sou ces and dis ibu es hem h ough
413
a helicoid. I has been ecen ly used o s udy p opelle s’
414
noise [
44
] and compa ed wi h o he acous ic analogies
415
[
45
,
46
]. In his wo k, he inpu o he Hanson model is
416
he h us and o que dis ibu ions along he blade span,
417
ex ac ed om he s eady CFD simula ions.418
PmL=mBM sin θ ex p [imB (ΩS /c+ (φ0−(π/2)))]
2p2πY (1−Mcos θ )×
Z ip
hub cosθ0
1−Mcosθ
dT
d −1
2M
dQ
d ex p(iφs)JmBΨLd
(5)
PmT=−ρc2Bsinθ ex p [imB (ΩS /c+ (φ0−(π/2)))]
4p2π(Y/D)(1−Mcosθ )×
Z ip
hub
M2
s(h/b)ex p(iφs)JmB k2
xΨVd
(6)
Loading and hickness componen s a e compu ed using
419
equa ions 5and 6, whe e
PmL
and
PmT
a e espec i ely he
420
loading and hickness componen s o he noise,
dT
and
421
B C R+
smoo h R+
ough
0 0.253 2.25 90
Table 2: Model coe icien s o oughness unc ion.
dQ
a e he h us and o que o he di e en ial elemen
422
o leng h
d
,
m
is he ha monic numbe ,
B
he numbe o
423
blades,
M
is he ip Mach numbe ,
θ
he obse e angle
424
ela i e o ligh di ec ion,
θ
he obse e angle in e a ded
425
e e ence ame and
θ0
he same angle bu ela i e o he
426
p opelle sha axis,
Ω
he o a ional eloci y o he p opelle ,
427
S
he dis ance o he obse e om he p opelle hub and
428
S
he same dis ance in e a ded e e ence ame,
c
he
429
speed o sound,
φ
he angen ial angle and
φ0
he same
430
angle ela i e o he p opelle sha axis,
Y
he obse e
431
dis ance om p opelle axis,
he adius o he p opelle ,
M432
he ees eam Mach numbe ,
he non-dimensional adial
433
posi ion,
φs
he phase shi due o sweep,
ρ
he densi y,
D434
he diame e o he p opelle ,
Ms
he sec ion ela i e Mach
435
numbe ,
h
he maximum ai oil hickness, and
b
he cho d
436
o he blade a each sec ion. 437
Finally,
JmB
is de ined as in Eq. 7,
kx
is a wa e numbe
438
de ined by Equa ion 8, and
ΨV
and
ΨL
a e non-dimensional
439
sou ce ans o ms ep esen ing he e ec o cho d-wise non-
440
compac ness. As sugges ed in Re . [
47
] and used in Re . [
44
,
441
46
], a pa abolic hickness dis ibu ion and a uni o m li
442
dis ibu ion we e applied, de ined in Equa ions 9and 10.443
JmB =JmB mB M sinθ0
1−Mcosθ (7)
kx=2mBbM
Ms(1−Mcosθ )(8)
ΨV(kx) = ¨2/3kx=0
8
k2
x
[2
kxsin kx
2−cos kx
2]kx6=0(9)
ΨL(kx) = ¨1kx=0
2
kxsin kx
2kx6=0(10)
3. Resul s & discussion 444
In his sec ion, he expe imen al da a will be p esen ed
445
and discussed, and he esul s o he nume ical models will
446
be used o d aw u he conclusions on he ma e ial and
447
su ace oughness in luence as isola ed e ec s. 448
The pe o mance da a will be p esen ed as dimensional
449
o ces o he ho e esul s ( h us and o que), and as non-
450
dimensional coe icien s o he o wa d ligh da a, using
451
he de ini ions shown in Equa ion 11, whe e
ρ∞
ep esen s
452
he undis u bed ai densi y,
n
is he o a ional speed o he
453
blade in e /s, and Dis he diame e o he p opelle . 454
CT=T
ρ∞n2D4CP=P
ρ∞n3D5(11)
Fo he acous ic da a, he spec a a di e en angula
455
posi ions will be shown, and he SPL a he BPF and OASPL
456
(Eq. 12) di ec i i ies will also be compu ed. 457
OASP L =20 ·log10 PRMS
P0(12)
7
3.1. Expe imen al campaign458
A e manu ac u ing he di e en p opelle s using he
459
AM echnologies desc ibed in sec ion 2.2, hese we e in-
460
2000 4000 6000 8000 10000 12000
Ro a ional Veloci y [ pm]
0
2
4
6
8
10
12
Th us [N]
FDM
SLS
SLA
SLA Rigid
2000 4000 6000 8000 10000 12000
Ro a ional Veloci y [ pm]
0
0.05
0.1
0.15
0.2
To que [Nm]
FDM
SLS
SLA
SLA Rigid
Figu e 9: Th us ( op) and o que (bo om) s o a ional eloc-
i y in ho e o each p opelle measu ed in he wind unnel.
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55
Ad ance Ra io [-]
0.02
0.04
0.06
0.08
0.1
CT [-]
FDM
SLS
SLA
SLA Rigid
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55
Ad ance Ra io [-]
0.02
0.025
0.03
0.035
0.04
0.045
CP [-]
FDM
SLS
SLA
SLA Rigid
Figu e 10: CT( op) and CP(bo om) s J o each p opelle
measu ed in he wind unnel a a ees eam eloci y o V∞=
10 m/s.
s alled and es ed in he anechoic chambe and wind unnel
461
es benches, whe e he aw eadings om he mic ophones
462
and he load cells, espec i ely, we e eco ded. As s a ed
463
be o e, a leas 10 epe i ions o he o ce measu emen s
464
we e pe o med o each p opelle o ensu e epea abili y. 465
The ime e olu ion o h us and o que du ing he ex-
466
pe imen s was ob ained using he load cell calib a ions
467
pe o med be o e he measu emen . Measu emen s a se -
468
e al o a ional speeds we e ca ied ou wi h he wind unnel
469
s opped o he ho e measu emen s, and wi h he wind
470
unnel a
10
m/s o he ad ance ope a ional condi ions. Fo
471
each case, he a e age o he h us and o que signals was
472
compu ed, and he s anda d de ia ion o said mean was ob-
473
ained, wi h he esul s being shown in Fig. 9and 10. I can
474
be obse ed ha a ho e and low o a ional speeds, he e
475
is no signi ican di e ence in h us be ween p opelle s.
476
In con as , as he o a ional speed inc eases, he e is a
477
subs an ial inc ease in he pe o mance o he p opelle s
478
made by SLA ela i e o he o he wo, wi h he di e ence
479
eaching up o 9.5% be ween he s anda d SLA and he
480
FDM p opelle s a he highes o a ional speed. 481
On he o he hand, he p opelle manu ac u ed by SLS
482
gene a es he highes o que, which is app eciable a all
483
o a ional speeds. The o he h ee p opelle s gene a e a sim-
484
ila o que, being he FDM and SLA Rigid p opelle s sligh ly
485
highe han he SLA one a high o a ional speeds. The
486
di e ence be ween he SLA and he SLS p opelle o que
487
a he maximum o a ional speed is up o 15%. 488
The indings ega ding he SLA and SLS p opelle s a e 489
consis en wi h cu en li e a u e [
17
], whe e SLA p o-
490
pelle s we e ound o p oduce a highe h us and lowe
491
o que a ho e compa ed wi h SLS ones. 492
Rega ding he measu emen s wi h mean low, Fig. 10
493
shows he h us and powe coe icien s e olu ion wi h e-
494
spec o he ad ance a io. The ad ance a io can be de ined
495
as
J=Un/D
, whe e
U
is he low eloci y,
n
is he o a ional
496
speed and
D
is he p opelle diame e . I can be obse ed
497
ha he di e ences in
CT
a e highe han he ones seen in
498
ho e . Bo h SLA p opelle s p oduce a highe
CT
a low al-
499
ues o
J
, as expec ed based on he ho e esul s. Howe e ,
500
he slope o he SLA p opelle s is s eepe han he o he
501
wo. Because o his, he h us coe icien ob ained a high
502
alues o Jis highe o he SLS and FDM p opelle s. 503
In e ms o powe coe icien , he ho e ends seem o
504
con inue o all he
J
ange, wi h he SLS p opelle being
505
he one wi h highe o que and he SLA p opelle being he
506
one wi h a lowe
CP
. Ne e heless, he
CP
o he SLA Rigid
507
p opelle ends o ge close o he SLS one a high ad ance
508
a ios. 509
I can be seen ha he e a e no clea endencies ha can
510
be iden i ied, as he wo ac o s ha ha e he g ea es e ec
511
on pe o mance (ma e ial p ope ies and su ace inish)
512
ope a e simul aneously, hiding he indi idual e ec . Fo
513
his eason, a nume ical in es iga ion will be ca ied ou o
514
y o explain he indings ob ained and o acqui e a deepe
515
unde s anding o he unde lying phenomena. 516
As o he noise gene a ed by each o he p opelle s
517
8
manu ac u ed wi h AM, he SPL a he BPF and OASPL
518
di ec i i y ob ained a ound he p opelle a a adius o
519
2 me e s a e shown in Fig. 12. To expand he analysis,
520
in Fig. 11, he spec um o he sound signal eco ded a
521
−30deg and 30deg is p esen ed.522
I can be obse ed ha a he i s blade passing e-
523
quency, he sound le el achie ed is e y simila among he
524
ou p opelle s in he
−60deg
o
30deg
ange, being sligh ly
525
highe o he SLS p opelle han o he o he wo. Nea
526
he axis o o a ion, he di e ence inc eases, wi h he SLA
527
Rigid p opelle being he mos silen one, wi h a di e ence
528
o up o 8.75 dB.529
This sound inc ease is also obse ed, wi h a g ea e noise
530
inc emen , in he b oadband noise a highe equencies,
531
whe e i can be seen how he SLS p opelle gene a es a
532
conside able noise inc emen , p esumably due o he su ace
533
inish ha causes he gene a ion o a u bulen bounda y
534
laye . This phenomenon can be clea ly obse ed in Fig. 11
535
and explains he obse ed inc ease in OASPL di ec i i y o
536
he SLS p opelle in Fig. 12 (down), wi h a di e ence o up
537
o 5 dB depending on he angle.538
In addi ion, i can be obse ed in Fig. 11 how a he
539
sha equency (hal he i s blade passing equency) a
540
peak appea s. This peak is highe in he case o he p o-
541
pelle s manu ac u ed in SLA, which would indica e ha
542
hese a e he p opelle s wi h he g ea es geome ical di -
543
e ence be ween blades. This is p obably due o he e ec
544
o hand sanding on he geome y. In ac , he SLA Rigid
545
p opelle shows an inc ease o up o 10 dB in he noise a
546
0 1000 2000 3000 4000 5000
F equency [Hz]
20
30
40
50
60
SPL [dB]
FDM
SLS
SLA
SLA Rigid
0 1000 2000 3000 4000 5000
F equency [Hz]
20
30
40
50
60
SPL [dB]
FDM
SLS
SLA
SLA Rigid
Figu e 11: F equency spec a o he -30º( op) and 30º(bo -
om) mic ophone posi ions ho e ing a 9870 pm in he ane-
choic chambe .
0°
30°
60°
90°
120°
150°
180°
210°
240°
270°
300°
330°
30
40
50
60
70
FDM
SLS
SLA
SLA Rigid
0°
30°
60°
90°
120°
150°
180°
210°
240°
270°
300°
330°
60
70
80
90
FDM
SLS
SLA
SLA Rigid
Figu e 12: Sound P essu e Le el a he Blade Passing F e-
quency ( op) and O e all Sound P essu e Le el (bo om) di-
ec i i y o he p opelle s a 9870 pm.
he sha equency. This is consis en wi h he needed le el
547
o pos -p ocessing (highe in he Rigid case compa ed wi h
548
he s anda d SLA one). 549
3.2. Nume ical esul s 550
Rega ding he nume ical esul s, as s a ed be o e, s eady
551
simula ions ha e been ca ied ou , modi ying he ma e-
552
ial Young’s modulus (assuming iso opic ma e ials) and
553
modi ying he su ace oughness o he p opelle (using a
554
oughness unc ion and assuming iso opic oughness) 555
3.2.1. Ma e ial in luence 556
Rega ding p opelle pe o mance, Figu e 13 shows he
557
h us and powe coe icien s agains ad ance a io cu es
558
o di e en alues o Young’s moduli. I can be obse ed
559
ha a ho e and low alues o
J
he e is a small inc emen
560
in
CT
o Young’s moduli be ween 1 and 30 GPa. This
561
endency inc eases o lowe alues o E, wi h a di e ence
562
o
∼
22.5% be ween he 0.05 GPa case and he igid (30 GPa)
563
one. Ne e heless, all cu es con e ge o he same alue
564
when he ad ance a io is inc eased, educing he in luence
565
o he ma e ial. The same beha io can be obse ed o
566
he powe coe icien , wi h highe inc emen s a ho e (up
567
o
∼
58% o he wo s -case scena io). The gain in h us
568
p oduced by he in luence o he ma e ial is ound o be
569
smalle han he inc ease in o que, esul ing in educed
570
9
