EDITOR
P o . D . Raul D.S.G. CAMPILHO
ADVANCES IN MATERIALS,
STRUCTURES, SENSING, AND
DATA-DRIVEN ENGINEERING
APPLICATIONS
Published by
BZT TURAN PUBLISHING HOUSE
Ce i ica e Numbe : 202401
Delawa e, Uni ed S a es
www.bz u anpublishinghouse.com
in o@bz u anpublishinghouse.com
EDITOR: P o . D . Raul D.S.G. CAMPILHO
ADVANCES IN MATERIALS, STRUCTURES, SENSING, AND DATA-DRIVEN
ENGINEERING APPLICATIONS
OPEN ACCESS
Sugges ed Ci a ion:
Campilho, R. (2025). ADVANCES IN MATERIALS, STRUCTURES, SENSING, AND DATA-DRIVEN ENGINEERING
APPLICATIONS. BZT TURAN PUBLISHING HOUSE. DOI: h ps://doi.o g/10.30546/19023.978-9952-8610-7-
5.2025.0057
Language: English
Publica ion Da e: Oc obe 2025
Co e Design By Mehme ÇAKIR
P in and digi al e sions ypese by BZT TURAN Media Co. L d.
E-ISBN: 9 7 8 - 9 9 5 2 - 8 6 1 0 - 7 - 5
DOI: h ps://doi.o g/10.30546/19023.978-9952-8610-7-5.2025.0057
iii
PREFACE
Enginee ing esea ch oday anges om he design o high-pe o mance
ma e ials and s uc u es o he de elopmen o in elligen sys ems and human-
cen ed solu ions. The challenges aced by indus y, de ence, spo , and
socie y equi e specialized app oaches and an in eg a ion o pe spec i es om
di e se domains. This book e lec s such di e si y, by ga he ing con ibu ions
ha combine ad anced modelling, nume ical analysis, sensing echnologies,
and da a-d i en me hods o add ess p oblems o bo h echnical and social
impo ance. The i s pa o he olume ocuses on ad anced ma e ials and
s uc u al applica ions. I begins wi h a nume ical s udy o ballis ic p o ec ion
using ul a-high-molecula -weigh polye hylene (UHMWPE) and Ke la
pla es, p o iding insigh s in o hyb id a mou sys ems. This is ollowed by
a se o chap e s dedica ed o adhesi e bonding in s uc u al enginee ing,
whe e cohesi e zone modelling and ex ended ini e elemen echniques
a e employed o e alua e impac pe o mance, sca geome ies, and no el
solu ions in he canoeing boa indus y. These wo ks ein o ce he impo ance
o adhesi es in mode n s uc u al design and p opose new guidelines
and imp o emen s o hei applica ion in demanding en i onmen s. The
second pa o he book ansi ions in o sensing echnologies wi h a s udy
on d one-based g ound pene a ing ada o sub e anean objec de ec ion.
This con ibu ion highligh s he g owing ole o unmanned ae ial sys ems in
de ence, a chaeology, and en i onmen al moni o ing, showing how classical
ada me hods can be adap ed o mode n mobili y pla o ms. The hi d
and inal pa u ns o da a-d i en and human-cen ed enginee ing. He e,
one chap e in oduces a sen imen il e ing and ecommenda ion sys em
ailo ed o Tu kish e-comme ce, illus a ing he impac o na u al language
p ocessing and us -weigh ed heu is ics on consume -o ien ed pla o ms.
The inal chap e add esses spa ial quali y in s uden do mi o ies, using a
sys ema ic li e a u e e iew o compile and analyse p oblems a ec ing use
well-being. Taken oge he , hese se en chap e s emphasize he inno a ion
o con empo a y enginee ing esea ch, demons a ing how undamen al
ad ances in ma e ials, mechanics, sensing, and da a science can con e ge
i
owa d p ac ical applica ions. I is my hope ha his olume will se e as
a aluable esou ce o esea che s, p ac i ione s, and s uden s, and inspi e
u he explo a ion o hese opics.
Raul Dua e Salguei al Gomes Campilho has
a PhD in Mechanical Enginee ing (2009) and a
Habili a ion in Mechanical Enginee ing, bo h om
he Facul y o Enginee ing o he Uni e si y o
Po o (Po ugal). He is an associa e p o esso o
he Mechanical Enginee ing depa men o ISEP –
School o Enginee ing, Poly echnic Uni e si y o
Po o (Po ugal), whe e he eaches se e al cou ses
o he Bachelo and Mas e deg ee in Mechanical
Enginee ing. Since 2023 he is he ice-di ec o o
CIDEM – Cen e o Resea ch & De elopmen in Mechanical Enginee ing.
He de elops his esea ch ac i i ies in ISEP and INEGI – Ins i u e o
Science and Inno a ion in Mechanical and Indus ial Enginee ing, which
is a Resea ch and Technology O ganiza ion (RTO), and is a ilia ed o he
LAETA – Associa e Labo a o y o Ene gy, T anspo s and Ae ospace. His
esea ch mainly ocuses on adhesi e join s; s uc u al adhesi es; design o
bonded join s; expe imen al es ing; Fini e Elemen Me hod; Ex ended Fini e
Elemen Me hod; Meshless Me hods; Cohesi e Zone Models; composi e
ma e ials; nume ical modelling o composi e ma e ials; mic omechanics;
mac omechanics; design o mechanical s uc u es; lexible p oduc ion;
au oma ion; obo ics; and ac ua o sys ems. He was he p incipal in es iga o
o 1 na ional R&D p ojec and pa icipa ed/pa icipa es as a esea che in
6 na ional R&D p ojec s, all unded h ough compe i i e calls. He also
pa icipa ed in 1 R&D p ojec in collabo a ion wi h he Po uguese indus y
( unded h ough compe i i e calls). He is/was he (co)supe iso o 6 PhD
(1 concluded and 5 ongoing) and 258 MSc hesis. His publica ion eco d
con ains 340 a icles & e iews published in Web o Science-indexed jou nals
and 154 book chap e s. He also coedi ed 3 books and was he (co)au ho o
o e 400 communica ions ha we e p esen ed in in e na ional con e ences
and 18 communica ions p esen ed in na ional con e ences.
ii
CONTENTS
PREFACE .................................................................................................................... iii
CHAPTER 1
In es iga ion o Ballis ic Beha io o UHMWPE and Ke la Pla es ....... 1
Ahme Mu a Asan
CHAPTER 2
P ocedu es o he impac analysis o adhesi ely-bonded
s uc u es ........................................................................................................... 9
P.D.A. Da Sil a
R.D.S.G. Campilho
CHAPTER 3
Nume ical e alua ion o sca geome y adhesi ely-bonded join s
by XFEM modelling .............................................................................................21
I.R.S. A aújo
R.D.S.G. Campilho
CHAPTER 4
Nume ical CZM e alua ion o adhesi ely-bonding solu ions o
canoeing boa ab ica ion .................................................................................. 39
João C.M. San os
Raul D.S.G. Campilho
CHAPTER 5
Feasibili y o D one-Based G ound Pene a ing Rada o
Sub e anean Fo eign Objec De ec ion/Mapping ....................................59
Celile Nu Yalçın
INVESTIGATION OF BALLISTIC BEHAVIOR OF UHMWPE AND KEVLAR PLATES
6
Re e ences
Ab ew, M. A., Boussu, F., B uniaux, P., Loghin, C., & C is ian, I. (2019). Ballis ic impac
mechanisms–A e iew on ex iles and ib e- ein o ced composi es impac esponses.
Composi e S uc u es, 223, 110966. h ps://doi.o g/10.1016/j.comps uc .2019.110966
ANSYS. (2025). GRANTA Ma e ials Da a o Simula ion (Sample). h ps://www.ansys.com/
p oduc s/ma e ials
Ba os, D., Mo a, C., Bessa, J., Cunha, F., Rosa, P., & Fanguei o, R. (2023). Blas agmen
impac o angle-ply composi e s uc u es o buildings wall p o ec ion. Buildings,
13(8). h ps://doi.o g/10.3390/buildings13081959
Bajya, M., Majumda , A., Bu ola, B. S., A o a, S., & Bha acha jee, D. (2021). Ballis ic pe o -
mance and ailu e modes o wo en and unidi ec ional ab ic based so a mou panels.
Composi e S uc u es, 255, 112941. h ps://doi.o g/10.1016/j.comps uc .2020.112941
Ca , D. (1999). Failu e mechanisms o ya ns subjec ed o ballis ic impac . Jou nal o Ma e-
ials Science Le e s, 18(7), 585–588.
Edidin, A. A., & Ku z, S. M. (2000). In luence o mechanical beha io on he wea o 4 clini-
cally ele an polyme ic bioma e ials in a hip simula o . The Jou nal o A h oplas y,
15(3), 321–331.
Hea le, J. W. S. (2001). High-pe o mance ib es. Else ie .
Hu, P., Yang, H., Zhang, P., Wang, W., Liu, J., & Cheng, Y. (2022). Expe imen al and nume-
ical in es iga ions in o he ballis ic pe o mance o ul a-high molecula weigh pol-
ye hylene ibe - ein o ced lamina es. Composi e S uc u es, 290, 115499. h ps://doi.
o g/10.1016/j.comps uc .2022.115499
Ka hikeyan, K., Russell, B. P., Fleck, N. A., Wadley, H. N. G., & Deshpande, V. S. (2013a).
The e ec o shea s eng h on he ballis ic esponse o lamina ed composi e pla-
es. Eu opean Jou nal o Mechanics - A/Solids, 42, 35–53. h ps://doi.o g/10.1016/j.
eu omechsol.2013.04.002
Ka hikeyan, K., Russell, B. P., Fleck, N. A., O’Mas a, M., Wadley, H. N. G., & Deshpande,
V. S. (2013b). The so impac esponse o composi e lamina e beams. In e na ional
Jou nal o Impac Enginee ing, 60, 24–36.
Li, X., Zhang, X., Guo, Y., Shim, V., Yang, J., & Chai, G. B. (2018). In luence o ibe ype on
he impac esponse o i anium-based ibe -me al lamina es. In e na ional Jou nal o
Impac Enginee ing, 114, 32–42.
Ma, Y., Wang, J., Zhao, G., & Liu, Y. (2023). New insigh s in o he damage assessmen and
ene gy dissipa ion weigh mechanisms o ce amic/ ibe lamina ed composi es un-
de ballis ic impac . Ce amics In e na ional, 49(13), 21966–21977. h ps://doi.o -
g/10.1016/j.ce amin .2023.04.021
Peng, L., Zhou, J., Wang, Q., Zhang, X., & Guan, Z. (2024). Nume ical modelling o he
ballis ic impac esponse o hyb id composi e s uc u es. Composi es Pa C: Open
Access, 14, 100474. h ps://doi.o g/10.1016/j.jcomc.2024.100474
AHMET MURAT ASAN
7
Ramadhan, A., Talib, A. A., Ra ie, A. S. M., & Zaha i, R. (2012). Expe imen al and nume ical
simula ion o ene gy abso p ion on composi e Ke la 29/Polyes e unde high eloci y
impac . Jou nal o Ad anced Science and Enginee ing Resea ch, 2(1). h p://psasi .
upm.edu.my/id/ep in /23243
Reddy, T. S., Reddy, P. R. S., & Madhu, V. (2017). Response o E-glass/epoxy and Dyneema®
composi e lamina es subjec ed o low and high eloci y impac . P ocedia Enginee-
ing, 173, 278–285. h ps://doi.o g/10.1016/j.p oeng.2016.12.014
Tan, V. B. C., & Khoo, K. J. L. (2005). Pe o a ion o lexible lamina es by p ojec iles o di -
e en geome y. In e na ional Jou nal o Impac Enginee ing, 31(7), 793–810.
Vi ue. (2024). UHMWPE. h ps://www. i ue ex ile.com/
Wang, H., Wee asinghe, D., Moho i, D., Hazell, P. J., Shim, V., Shanka , K., & Mo ozo ,
E. V. (2021). On he impac esponse o UHMWPE wo en ab ics: Expe imen s and
simula ions. In e na ional Jou nal o Mechanical Sciences, 204, 106574. h ps://doi.
o g/10.1016/j.ijmecsci.2021.106574
Wikipedia. (2021). 9x19mm Pa abellum. Re ie ed Decembe 2, 2021, om h ps:// .wikipe-
dia.o g/wiki/9x19mm_Pa abellum
Zhu, W., Huang, G. Y., Feng, S. S., & S onge, W. J. (2018). Conical nosed p ojec ile pe o-
a ion o polye hylene ein o ced c oss-ply lamina es: E ec o ibe la e al displace-
men . In e na ional Jou nal o Impac Enginee ing, 118, 39–49.
Con ibu o s
Ahme Mu a AŞAN g adua ed om Fı a Uni e si y in mechanical
enginee ing and elec ical and elec onics enginee ing. He has a mas e ’s
and doc o a e deg ee in mechanical enginee ing. He is cu en ly wo king
as a esea che a Dicle Uni e si y. His esea ch in e es s include composi e
ma e ials, ac u e mechanics, he ini e elemen me hod, ib a ion, ballis ic,
ensile, comp ession, shea , and a igue es s.
9
CHAPTER 2
P ocedu es o he impac analysis o
adhesi ely-bonded s uc u es
P.D.A. da Sil a1
R.D.S.G. Campilho2
Abs ac
The applica ion o s uc u al adhesi es has been inc easing in he
indus y. Among he ields o applica ion ha mos use his ype o join s, he
ae onau ical and au omo i e indus ies s and ou . Cohesi e zone modelling
(CZM) is widesp ead o s a ic analysis o adhesi e join s. Howe e , in many
applica ions, impac analyses a e undamen al o assess he s uc u al sa e y
o adhesi ely bonded s uc u es. This wo k add esses he nume ical analysis
o adhesi e ubula join s subjec ed o impac loads, conside ing di e en
adhesi es. The nume ical app oach o model impac damage consis ed o an
adap ion o he CZM echnique using he Abaqus® so wa e. A pa ame ic
nume ical s udy ca ied ou on he in luence o he o e lap leng h (LO)
and adhe ends hickness ( p), on he s eng h o he adhesi e join s. Load-
displacemen (P-δ) cu es, abso bed ene gy (Ea) and maximum load (Pm)
a e p esen ed o all join s es ed. I was concluded ha he s eng h o he
1 CIDEM, ISEP – School o Enginee ing, Poly echnic o Po o, R. D . An ónio Be na dino
de Almeida, 431, 4200-072 Po o, Po ugal.
2 CIDEM, ISEP – School o Enginee ing, Poly echnic o Po o, R. D . An ónio Be na dino
de Almeida, 431, 4200-072 Po o, Po ugal.
Ins i u e o Science and Inno a ion in Mechanical and Indus ial Enginee ing, Rua D .
Robe o F ias, 400, 4200-465 Po o, Po ugal.
PROCEDURES FOR THE IMPACT ANALYSIS OF ADHESIVELY-BONDED STRUCTURES
10
adhesi e join s inc eases o s i adhesi es. Pm and E signi ican ly imp o ed
o highe LO. O e all, clea design p inciples a e p oposed o maximize he
ensile beha io o ubula adhesi e join s unde impac loads.
Keywo ds: Adhesi e join , S uc u al adhesi e, Impac analysis,
Cohesi e zone models.
1. In oduc ion
The e a e cu en ly many me hods o es ima e he s eng h o an adhesi e
join . Con en ional analy ical me hods p o ide he beha iou o adhesi e
join s a hei linea elas ic limi , ei he uling ou plas ic beha iou o
making he calcula ion qui e complex. Thus, when he complexi y o s udies
o de e mine he s a e o s ess in an adhesi e join unde ce ain condi ions
inc eases, he applica ion o analy ical me hods o s udy becomes un easible,
and nume ical me hods a e adop ed, in pa icula he ini e elemen me hod
(FEM). The FEM, in oduced by Ha is and Adams (1984) o adhesi e join s,
becomes he mos widely used me hod. The ela ed echniques inco po a e
ac o s such as join o a ion, he plas ici y o he adhe ends, he plas ici y o
he adhesi e, and he in luence o he ille . Opposed o s a ic loads, dynamic
loadings a y o e he ime. Wi hin his scope, a igue, modal analysis, and
a iable s ain a e and impac loadings can be p esen in a s uc u e. The
a iable s ain a e/impac scena io in pa icula in ol es dynamic beha iou
and equi es explici nume ical in eg a ion. Cu en ly, a iable s ain a e and
impac s udies a e di ided in o con inuum mechanics, damage mechanics and
CZM models (Ramalho, Sánchez-A ce e al. 2022). Con inuum mechanics is
used o e alua e s esses and s ains in adhesi e join s. This app oach is easy
o p og amme. Howe e , i should be conside ed ha , unde impac loading
condi ions, he s ess p opaga es as a wa e, which leads o se e al local s ess
peaks along he adhesi e. Damage mechanics makes i possible o simula e
he p og essi e deg ada ion o he ma e ial un il inal ailu e along an
a bi a y pa h. Di e en wo ks a e a ailable in he li e a u e unde dynamic
analysis o adhesi e join s. Rao and C ocke (1990) used a heo e ical model
o s udy he bending ib a ion o a sys em o o e lapping join s. Fi s , he
equa ions o mo ion in he join egion a e de i ed ma hema ically using a
di e en ial calculus app oach. The ans e se displacemen s o he uppe and
lowe beam a e conside ed o be di e en . The adhesi e we e conside ed o
be linea ly iscoelas ic and he Kel in-Voig model was used o ep esen
his beha iou . The model can be used o p edic he na u al equencies,
modal damping a ios and mode shapes o he sys em o ee ib a ion.
Esmaeili, Zehsaz e al. (2015) ca ied ou a a igue s udy on he e ec o he
P.D.A. DA SILVA • R.D.S.G. CAMPILHO
11
igh ening o que on bol ed and hyb id (bonded/bol ed) join s wi h di e en
cyclic longi udinal loads. The a igue o he specimens was p edic ed using
six di e en mul iaxial a igue c i e ia by means o he local s ess and s ain
dis ibu ion ob ained om FEM analyses. The hyb id join s showed be e
a igue li e compa ed o he bol ed join s. Inc easing he igh ening o que o
clamping o ce on he join leads o an inc ease in he a igue esis ance o
double o e lap bol ed join s.
This wo k add esses he nume ical analysis o adhesi e ubula join s
subjec ed o impac loads, conside ing di e en adhesi es. The nume ical
app oach o model impac damage consis ed o an adap ion o he CZM
echnique using he Abaqus® so wa e. A pa ame ic nume ical s udy ca ied
ou on he in luence o LO and p on he s eng h o he adhesi e join s. P-δ
cu es, E and Pm a e p esen ed o all join s es ed.
2. Expe imen al and nume ical de ails
2.1. Ma e ials
To ca y ou he nume ical analyses, i is impo an o de ine he mechanical
p ope ies o he ma e ials employed. The ma e ial adop ed o he adhe ends
is DIN 55 Si7 (Sil a, Pe es e al. 2022). This ma e ial was chosen o wo
easons: o a oid plas ic de o ma ion o he adhe ends du ing he nume ical
s udy and o ensu e ha he ailu e is cohesi e in adhesi e. Table 1 illus a es
he mechanical p ope ies o he ma e ial chosen o he adhe ends.
Table 1 – Mechanical p ope ies o DIN 55 Si7 (Sil a, Pe es e al. 2022).
P ope ies Value
Young’s modulus, E [GPa] 210
Tensile yield s ess, σy [MPa] 1078
Tensile s eng h, σ [MPa] 1600
Tensile ailu e s ain, εy [%] 6
Poisson’s a io, 0.3
Densi y, ρ [g/cm3]7.8
A b i le adhesi e (AV138) and wo adhesi es wi h high duc ili y (DP
8005 and XNR6852 E-2) we e used. The mechanical p ope ies we e
ob ained by ca ying ou di e en ypes o es s. Tensile es s p o ided he
Young Modulus (E) and ensile cohesi e s ess ( n
0), while he shea modulus
(G) and he shea cohesi e s ess ( s
0) we e ob ained h ough hick-adhe end
PROCEDURES FOR THE IMPACT ANALYSIS OF ADHESIVELY-BONDED STRUCTURES
12
shea es s (TAST). Rega ding he ac u e p ope ies, he double-can ile e
beam (DCB) es s p o ided he ensile ac u e oughness (GIC), while he
end-no ched lexu e (ENF) es s p o ided he shea ac u e oughness (GIIC).
The densi y (ρ) and Poisson’s a io (ν) a e also equi ed o p ocessing he
esul s. The inal p ope ies adjus ed o his s udy can be ound in Table 2
(Sil a, Pe es e al. 2022).
Table 2 – Adhesi es’ mechanical p ope ies (Sil a, Pe es e al. 2022).
Adhesi e DP 8005 AV138 XNR6852 E-2
Young’s modulus, E [MPa] 590 4890 1742
Shea modulus, G [MPa] 159 1560 645.2
Tensile cohesi e s eng h, n
0 [MPa] 27.5 70.2 53.7
Shea cohesi e s eng h, s
0 [MPa] 36.7 51.7 45.8
Tensile oughness, GIC [N/mm] 1.1 0.35 1.68
Shea oughness, GIIC [N/mm] 60.6 18
Densi y, ρ [g/cm3]1.06 1.7 1.5
Poisson’s coe icien , 0.3a0.35b0.4c
*a - es ima ed alue; b - alue supplied by he manu ac u e ; c - ypical alue o
epoxy adhesi es.
2.2. Geome ies
The Single Lap Join (SLJ) ha alida es he nume ical model wi h
expe imen s unde impac loads is ini ially de ined. SLJs we e manu ac u ed
o ca y ou he expe imen al es s, in acco dance wi h ASTM D1002 and
ISO 4587. Valen e, Campilho e al. (2019) used hese geome ies o alida e
a CZM-based nume ical model ha could accu a ely p edic he s eng h
o adhesi e join s submi ed o impac loads, by compa ing i wi h an
expe imen al model and analysing he esul s. The SLJ is composed o wo
adhe ends wi h a leng h o 120 mm, LO o 25 mm and an adhesi e hickness
( a) o 0.2 mm. The p employed is 2 mm. Figu e 1 p esen s he geome ic
pa ame e s ha de ine he adhesi e join s used in he expe imen al es s.
P.D.A. DA SILVA • R.D.S.G. CAMPILHO
13
Figu e 1 – Geome y o join used in he expe imen al es .
The ubula adhesi e join s used in he nume ical pa ame ic analysis a e
composed by wo o e lapping adhe ends wi h dissimila diame e s, bonded
by an adhesi e. In he case o ubula adhe ends, he alue o LO is 10 mm
and he leng h o he adhe ends is 60 mm, o a o al specimen leng h o
100 mm. The ou e diame e o he inne ube is ixed a 20 mm, p is 2 mm,
and a is 0.2 mm. Figu e 2 illus a es he ubula adhesi e join used in he
nume ical analysis.
Figu e 2 – Geome y o he ubula adhesi e join used in he nume ical analysis.
2.3. Nume ical modelling
A bidimensional nume ical analysis o he impac es s was pe o med
using Abaqus®. The ep esen a ion o he ubula na u e o he join
PROCEDURES FOR THE IMPACT ANALYSIS OF ADHESIVELY-BONDED STRUCTURES
14
was based on he axisymme ic pa ame e isa ion, simula ing he axis o
e olu ion o he join a ound i s neu al axis, which was de ined as being
o he de o mable ype and wi h he basic cha ac e is ic o he wi e ype.
Axisymme ic elemen s we e selec ed o model bo h adhe ends (CAX4) and
adhesi e (COHAX4R). Fo he adhe ends, he elemen s we e displayed in
a s uc u ed mesh, while he cohesi e elemen s we e assigned a sweep ype
mesh. Besides, o simula e he es ’s impac Ea, a solid homogeneous mass
was c ea ed. P elimina y calcula ions showed ha Ea o 40 J was equi ed
o sepa a e he adhe ends and, by combining his alue wi h he impac
eloci y p e iously es ima ed, i was possible o de ine he o al mass and i s
espec i e densi y. Conside ing bo h ma e ials p ope ies and na u e, elas ic-
plas ic beha iou was assumed o he adhe ends, and, o he adhesi e, a
quad a ic s ess c i e ion was employed o p edic ailu e ini ia ion, while
damage p opaga ion was uled by a linea ene ge ic c i e ion. Las ly, a
linea s ess-s ain beha iou was assigned o he mass, and i s elas ic
p ope ies we e de ined o ha e no in e e ence wi h he join ’s s eng h.
Rega ding bounda y condi ions (Figu e 3), hey we e applied as ollows:
1) nil longi udinal displacemen a he le specimen’s edge; 2) nil adial
displacemen and eloci y ype p ede ined ield o 1.75 m/s applied o he
mass a he o he edge, o assu e he equi ed Ea (40 J).
Figu e 3 – Se o bounda y condi ions o he nume ical simula ion.
3. Resul s
3.1. CZM alida ion
To alida e he pe o med nume ical analysis, expe imen al esul s and
nume ical e e ence alues ob ained by (Valen e, Campilho e al. 2019)
P.D.A. DA SILVA • R.D.S.G. CAMPILHO
15
a e aken as e e ence. Fo he adhesi es AV138, DP8005, and XNR6852
E-2, a compa ison be ween expe imen al and e e ence Pm alues wi h he
nume ical Pm ob ained in his wo k using he CZM model is p esen ed in
Figu e 4 and de ailed in Table 3 by he Pm ela i e di e ence (∆Pm).
11.75
14.43 13.61
16.79
19.93 19.90
24.00
26.92 28.39
0
10
20
30
40
Expe imen al Nume ical
e e ence
Cu en nume ical
P
m
[kN]
AV138 DP8005 XNR6852 E-2
Figu e 4 – Compa ison be ween expe imen al and nume ical Pm alues.
Table 3 – Compa ison be ween expe imen al and nume ical Pm alues.
Adhesi e AV138 DP8005 XNR6852 E-2
Nume ical Re e ence (1) 14.43 kN 19.93 kN 26.92 kN
Expe imen al Value (2) 11.75 kN 16.79 kN 24.00 kN
Nume ical Value (3) 13.61 kN 19.90 kN 28.39 kN
ΔPm (3) – (1) -5.68% -0.15% 5.46%
ΔPm (3) – (2) 12.89% 15.60% 16.31%
3.2. Pa ame ic analysis
3.2.1. O e lap leng h
Figu e 5 (a) shows he P-δ cu es o he join s bonded wi h he AV138
and LO=20 mm. The obse ed oscilla ions in P wi h δ a e ypical o impac
loadings due o he ine ial e ec s, leading o a P inc ease in s eps up o Pm
being eached. Be ween adhesi es, he AV138 has he highes Pm, ollowed
by he XNR6852 E-2 and inally he DP8005, ela ion ha was ound alid
o all LO. The maximum (o ailu e) δ was in e sely p opo ional o he
adhesi es’ s i ness. The Pm e olu ion wi h LO is shown in Figu e 5 (b) o
he h ee adhesi es. Conside ing he AV138 esul s, Pm inc eases wi h LO.
NUMERICAL EVALUATION OF SCARF GEOMETRY ADHESIVELY-BONDED JOINTS BY XFEM MODELLING
22
s udy o sca adhesi e join s in ension wi h di e en adhesi es (A aldi e®
AV138, A aldi e® 2015 and Sika o ce® 7752) and di e en sca angles o α
(3.43°, 10 °, 15°, 20°, 30° and 45°) by he eX ended Fini e Elemen Me hod
(XFEM). Ini ially, he expe imen al esul s ob ained in a p e ious wo k a e
desc ibed, o he pu pose o alida ing he ob ained nume ical esul s. The
de eloped nume ical wo k includes he dis ibu ion o he damage a iable
and join s eng h. Wi h he wo k ca ied ou , he cohe ence o he nume ical
esul s wi h he expe imen al ones was obse ed, wi h emphasis on he join
s eng h as a unc ion o α. I was ound ha join s wi h α=3.43° p esen he
bes esul s in e ms o ensile s eng h o he join s. The adhesi e A aldi e®
AV138 p esen s he bes ensile beha iou , ega dless o he α alue. Based
on he esul s ob ained, i was conside ed ha he XFEM is a ool ha can be
accu a ely used o design sca adhesi e join s.
Keywo ds: Adhesi e join , eX ended Fini e Elemen Me hod, C ack
p opaga ion me hod.
1. In oduc ion
Cu en ly, adhesi ely bonded join s ha e wide applicabili y and a e used
in a ious indus ial sec o s. The ae onau ical indus y was he one behind
he widesp ead usage his joining me hod, which a he beginning o he las
cen u y applied adhesi es based on casein (a na u al polyme ic ma e ial) in
ai c a s uc u es. Du ing he 1950s, adhesi es used in ai c a s uc u es
we e capable o o e good s i ness and s eng h [1]. The use o adhesi e
join s in a ious applica ions p o es o be mo e ad an ageous compa ed o
mo e adi ional mechanical joining me hods, such as mechanical as ening,
welding, i e ing, among o he s. Adhesi ely bonded join s, when designed
and manu ac u ed co ec ly, p o ide signi ican ad an ages [2], such as
mo e uni o m s ess dis ibu ion h oughou he adhesi e laye , wi h educed
s ess concen a ions. This dis ibu ion allows o highe s i ness and load
ansmissions, p omo ing weigh educ ion and lowe cos . This bonding
me hodology also p o ides be e a igue esis ance and ib a ion damping,
as s esses a e pa ially abso bed. As any o he manu ac u ing p ocess,
se e al d awbacks can be iden i ied, e.g., he need o design join s in a
way ha minimizes peel and clea age loads as much as possible, a limi ed
esis ance o ex eme condi ions o empe a u e and humidi y due o he
polyme ic na u e o he adhesi es, he need o clamping jigs o hold pa s in
posi ion du ing he cu ing p ocess and also he necessi y o ca e ul su ace
p epa a ion p ocedu es. Se e al join a chi ec u es a e a ailable, allowing
he designe o selec he mos sui able one, conside ing he applica ion and
I.R.S. ARAÚJO • R.D.S.G. CAMPILHO
23
subjec ed loads. Fo example, he single-lap join (SLJ) is a widely used and
s udied design due o he ease o manu ac u e and p edominan shea loading.
One o he d awbacks o his con igu a ion is he non-collinea o ces ha
a e ansmi ed, which cause he adhesi e o be subjec o peel s esses a
he o e lapping ends. To educe his e ec , o he con igu a ions a e used,
such as double lap join s, s ep join s and sca join s, among o he s. Double
lap join s, when compa ed o SLJ, a e mo e complex and ime-consuming
o manu ac u e, howe e , he e ec s o bending a e subs an ially lowe [2].
Sca and s ep join s p esen high s eng h, since hese geome ies a ou he
educ ion o s ess g adien s along he adhesi e bondline. On he o he hand,
due o he need o machine he o e lapping a ea, hese join demands highe
cos s. In sca join s, he s eng h depends on he sca angle [3].
The e olu iona y p ocess o adhesi e join s is closely ela ed o he
de elopmen o eliable p edic ion me hodologies ha allow inc easing
e iciency in hei use, hus making i possible o o e come he pa adigm o
o e dimensioned adhesi e join s ha esul ed in mo e expensi e and hea ie
s uc u es, ela ed o he lack o p ecise ma e ial models and adequa e ailu e
c i e ia ha we e e iden a ew decades ago. The wo me hodologies ha
can be applied o he analysis o adhesi e join s a e analy ical and nume ical
me hods. The Fini e Elemen Me hod (FEM) is he mos commonly used
echnique o he analysis o adhesi e join s, ha ing been ini ially applied
by Ha is and Adams [4], who in oduced ac o s such as join o a ion,
adhe ends and adhesi es’ plas ici y and he in luence o ille s. Con inuum
mechanics was hen used o p edic he s eng h o adhesi e join s, which
equi es s ess dis ibu ion and an app op ia e ailu e c i e ion. FEM can
also be combined wi h ac u e mechanics echniques o p edic s eng h,
ei he by he s ess in ensi y ac o o by ene ge ic app oaches such as
he i ual c ack closu e echnique. Howe e , hese modelling echniques
make he p ocess o e alua ing c ack g ow h di icul due o he need o
ec ea e he mesh in he pa h o c ack p opaga ion, which has epe cussions
in e ms o compu a ional e o [5]. O e he las ew decades, nume ical
modelling has seen majo ad ances, one o which is he implemen a ion o
damage models using cohesi e zone models (CZM). This echnique couples
con en ional FEM models o egions whe e damage is no expec ed wi h
ac u e mechanics, h ough he use o cohesi e elemen s o p omo e c ack
p opaga ion. The concep o CZM began wi h s udies by Ba enbla [6]
and Dugdale [7], who desc ibed he damage in he ac u e p ocess zone in
on o he c ack unde he e ec o s a ic loads. CZM allows o cap u e he
beginning o a c ack and i s p opaga ion wi hin o a he in e ace o ma e ials,
o e en in he delamina ion o composi es. The implemen a ion o CZM can
NUMERICAL EVALUATION OF SCARF GEOMETRY ADHESIVELY-BONDED JOINTS BY XFEM MODELLING
24
be done in sp ing elemen s o , mo e con en ionally, in cohesi e elemen s [8].
The ex ended Fini e Elemen Me hod (XFEM) uses en iched shape unc ions
o ep esen a con inuous displacemen ield. XFEM is a ecen e olu ion o
CZM, which allows he analysis and modelling o damage g ow h o p edic
ac u e in s uc u es, based on he s eng h o ma e ials o damage ini ia ion
and de o ma ions o ailu e assessmen , ins ead o alues o n
0/ s
0 o δn
0/δs
0
(peak ac ions and displacemen s in ension and shea , espec i ely) used in
he CZM. Compa ing o CZM, in XFEM i is no longe necessa y ha he
c ack ollows a p e-de ined pa h, which is a signi ican ad an age. Thus, he
c ack can p opaga e eely wi hin he s uc u e wi hou he need o he mesh
o coincide wi h he geome y o he discon inui ies and wi hou he need o
apply he mesh close o he c ack [9]. Bely schko and Black [10], in he la e
90s, p esen ed he undamen al cha ac e is ics o his me hod, based on he
concep o pa i ion o uni y, and which can be implemen ed in he FEM by
in oducing local en ichmen unc ions o he displacemen s nea he edge
o he c ack, o allow g ow h and sepa a ion be ween he c ack aces.
Recen and ele an s udies ca ied ou o p edic he s eng h o sca
join s a e a ailable in he li e a u e. Al es e al. [11] p esen ed an expe imen al
and nume ical s udy o hyb id sca join s. Ca bon ib e ein o ced polyme
(CFRP) and aluminium adhe ends we e bonded wi h A aldi e® AV138
and A aldi e® 2015, b i le and duc ile adhesi e, espec i ely, conside ing
di e en sca angles (α). Using he FEM, he peel (σy) and shea s esses (τxy)
we e ob ained, while CZM was used o p edic join s eng h. The nume ical
esul s showed ha he magni ude o σy and τxy inc eases wi h α, al hough
his inc ease is mo e signi ican o σy. The damage a iable showed ha
ailu e o he adhesi e laye s a s a he adhesi e ends and g ows owa ds he
inne adhesi e un il ailu e. The expe imen al maximum load (Pm) inc eases
exponen ially wi h he educ ion o α o he wo es ed adhesi es, due o
he inc ease in adhesi e a ea and a mo e uni o m s ess dis ibu ion. The Pm
alues ob ained by CZM a e e y close o hose ob ained expe imen ally. In
he wo k o Sun e al. [12], an expe imen al and nume ical s udy was ca ied
ou on he ensile pe o mance o sca adhesi e join s. A duc ile adhesi e was
used and CFRP adhe ends we e conside ed. Expe imen ally, sca join s we e
es ed wi h di e en alues o α (3°, 5°, 10°, 15°, 20° and 30°). Nume ically,
o alida e he p edic ion accu acy o a use -de ined CZM, he nume ical
esul s we e compa ed wi h he expe imen al da a. A iangula damage law
was also used. Expe imen ally, i was ound ha Pm inc eases exponen ially
wi h α dec ease, excep o join s wi h adhe ends wi h he s acking sequence
[45/0/-45/90]3S. The s ess dis ibu ion in he adhesi e is no uni o m and
depends on α and he adhesi e s acking sequence. Compa ison o esul s
I.R.S. ARAÚJO • R.D.S.G. CAMPILHO
25
shows ha he use -de ined CZM was able o p edic join s eng h and
displacemen o ailu e wi h highe accu acy han he iangula damage law,
which unde es ima ed he expe imen ally ob ained alues.
The objec i e o his wo k is he pa ame ic nume ical s udy o sca
adhesi e join s in ension wi h di e en adhesi es (A aldi e® AV138,
A aldi e® 2015 and Sika o ce® 7752) and di e en sca angles o α (3.43°,
10°, 15°, 20°, 30° and 45°) by he XFEM. Ini ially, he expe imen al esul s
ob ained in a p e ious wo k a e desc ibed, o he pu pose o alida ing he
ob ained nume ical esul s. The de eloped nume ical wo k includes he
dis ibu ion o he damage a iable and join s eng h. Wi h he wo k ca ied
ou , he cohe ence o he nume ical esul s wi h he expe imen al ones was
obse ed, wi h emphasis on he join s eng h as a unc ion o α.
2. Ma e ials and me hods
2.1. Join geome y
Figu e 1 illus a es he geome y o he sca join . I s dimensional
speci ica ions a e as ollows (in mm): leng h L =170, adhe end hickness
P=3 and adhesi e hickness A=0.2. These pa ame e s emain cons an , wi h
only he sca angle α (3.43o, 10o, 15o, 20o, 30o and 45o) a ying, o s udy i s
in luence on he join ’s s eng h.
Figu e 1. Sca join geome y.
2.2. Ma e ials
The ma e ial used as adhe end was he aluminium alloy AW 6082-
T651. The selec ion o his ma e ial is no only due o i s good mechanical
p ope ies bu also o i s wide ange o s uc u al applica ions in ex uded
and olled o ms. This aluminium alloy was cha ac e ized in p e ious
wo ks [13], whe e he ollowing p ope ies we e de ined: ensile s eng h
(𝜎 ) o 324.00±0.16 MPa, Young’s modulus (E) o 70.07±0.83 GPa, ensile
NUMERICAL EVALUATION OF SCARF GEOMETRY ADHESIVELY-BONDED JOINTS BY XFEM MODELLING
26
yield s ess (𝜎y) o 261.67±7.65 MPa, and ensile ac u e s ain (ε ) o
21.70%±4.24%. The s ess-s ain cu es (σ-ε) o he aluminium adhe ends
we e ob ained expe imen ally, in acco dance wi h ASTM s anda d E8/E8M
[14], o be in oduced in he nume ical models. The adhesi es conside ed o
his wo k a e he A aldi e® AV138, a b i le epoxy adhesi e, he A aldi e®
2015, a duc ile epoxy adhesi e, and he Sika o ce® 7752, a highly duc ile
polyu e hane adhesi e. Table 1 p esen s all he ele an p ope ies o he
adhesi es, cha ac e ized in a p e ious wo k, along wi h hei espec i e
alues.
Table 1. P ope ies o he adhesi es A aldi e® AV138, A aldi e® 2015, and
Sika o ce® 7752 [15-17].
P ope y AV138 2015 7752
Young’s modulus, E [GPa] 4.89±0.81 1.85±0.81 0.493±0.0896
Poisson’s a io, 0.35a0.33a0.33a
Tensile yield s ess, 𝜎y [MPa] 36.49±2.47 12.3±0.61 3.24±0.5
Tensile ailu e s eng h, 𝜎 [MPa] 39.45±3.18 21.63±1.61 11.49±0.3
Tensile ailu e s ain, ε [%] 1.21±0.10 4.77±0.15 19.18±1.4
Shea modulus, G13 [GPa] 1.56±0.01 0.56±0.21 0.187±0.0164
Shea yield s ess, τy [MPa] 25.1±0.33 14.6±1.3 5.16±1.1
Shea ailu e s eng h, τ [MPa] 30.2±0.40 17.9±1.8 10.17±0.6
Shea ailu e s ain, γ [%] 7.8±0.7 43.9±3.4 58.42±6.4
Toughness in ension, GIc [N/mm] 0.2b0.43±0.02 2.36±0.2
Toughness in shea , GIIc [N/mm] 0.38b4.7±0.34 5.41±0.5
a Manu ac u e ’s da a
b Es ima ed in e e ence [18]
2.3. Nume ical modelling
The nume ical analysis o he sca join was ca ied ou using he FEM
so wa e Abaqus®, since i allows he use o he in eg a ed XFEM module
o p edic ing he s eng h o he sca join . The join s we e con igu ed in a
wo-dimensional o ma , employing solid plane s ain elemen s (speci ically
CPE4 and CPE3 in ABAQUS®) o model he adhe ends. The iangula
CPE3 elemen s we e used in he sca edge o enable he espec i e slope
wi hou inducing dis o ions in he mesh. Figu e 2 illus a es a de ail o mesh
e inemen o a model wi h α=45º.
I.R.S. ARAÚJO • R.D.S.G. CAMPILHO
27
Figu e 2. Rep esen a ion o he mesh cons i uen elemen s.
In he longi udinal di ec ion o he adhe ends, a selec i e mesh e inemen
was employed using he bias a io. The mesh has a highe le el o e inemen
nea he adhesi e laye . The numbe o elemen s and he e inemen a io o
each edge o he join we e chosen o ensu e g ea e e inemen in he c i ical
egions o he join . This a ia ion aims o educe compu a ional e o and
ime in ob aining esul s wi hou comp omising hei accu acy. Figu e 3
illus a es he e ec o he bias along he leng h o he sca join .
Figu e 3. Bias e ec along he leng h o he join .
To simula e eal expe imen al es condi ions, bounda y and loading
condi ions we e applied o he models in he ABAQUS® so wa e o emula e
eal es ing condi ions. The join was ixed a one end and cons ained in he
ans e se di ec ion a he opposi e end.
2.4. XFEM o mula ion
The XFEM se es as an enhancemen o he con en ional FEM. The
XFEM in eg a es en ichmen unc ions in o he FEM o mula ion, p ima ily
designed o ep esen displacemen jumps be ween c ack aces du ing
c ack p opaga ion [19]. When simula ing damage wi hin Abaqus®, damage
ini ia ion and p opaga ion a e igge ed in egions whe e he s esses and/o
NUMERICAL EVALUATION OF SCARF GEOMETRY ADHESIVELY-BONDED JOINTS BY XFEM MODELLING
28
s ains exceed p ede e mined h esholds. Abaqus® o e s a choice o six c ack
ini ia ion c i e ia. Among hese, he MAXPS (maximum p incipal s ess) and
MAXPE (maximum p incipal s ain) c i e ia ely on speci ic unc ions, as
desc ibed in he espec i e o de
max max
oo
max max
o
σε
σε
= =
(1)
σmax and σo
max ep esen he cu en and pe missible maximum p incipal
s ess. The use o Macaulay b acke s signi ies ha a pu ely comp essi e s ess
s a e does no lead o damage ini ia ion. Simila ly, εmax and εo
max ep esen
he cu en and allowable maximum p incipal s ain. The MAXS (maximum
nominal s ess) and MAXE (maximum nominal s ain) c i e ia a e exp essed
using he ollowing ma hema ical unc ions
nn
ss
00 0 0
ns n s
max , o max ,
εε
εε
= =
(2)
n and s a e he cu en no mal and shea ac ion componen s o he
c acked su ace. The s ain pa ame e s ha e iden ical signi icance. The
quad a ic nominal s ess (QUADS) and quad a ic nominal s ain (QUADE)
c i e ia a e based on he in oduc ion o he ollowing unc ions, espec i ely
22
22
nn
ss
00 0 0
ns n s
o
εε
εε
=+=+
(3)
All c i e ia a e ul illed, and damage ini ia es, when eaches uni y. Fo
damage g ow h, he undamen al exp ession o he displacemen ec o u,
including he displacemen s en ichmen , is w i en as [20]
( ) ( )
1
N
i
i
N x Hx
=
= +
∑
u ua
ii
. (4)
Ni(x) and ui ela e o he con en ional Fini e Elemen o mula ion.
Ni(x) ep esen s he nodal shape unc ions, while ui s ands o he nodal
displacemen ec o associa ed wi h he con inuous pa o he o mula ion.
I.R.S. ARAÚJO • R.D.S.G. CAMPILHO
29
The second e m enclosed in b acke s, H(x)ai, is only ac i e in he nodes
o which any ela ing shape unc ion is cu by he c ack and can be exp essed
by he p oduc o he nodal en iched deg ee o eedom ec o including he
men ioned nodes, ai, wi h he associa ed discon inuous shape unc ion, H(x),
ac oss he c ack su aces.
This me hod is buil upon he concep o in oducing phan om nodes
ha subdi ide elemen s in e sec ed by a c ack, e ec i ely simula ing he
sepa a ion be ween newly c ea ed sub-elemen s. The abili y o p opaga e a
c ack along an a bi a y pa h is acili a ed by hese phan om nodes, which
ini ially sha e he same coo dina es as he eal nodes. These phan om nodes
emain en i ely cons ained o he eal nodes un il damage ini ia ion occu s.
Once c ossed by a c ack, he elemen ge s di ided in o wo sub-domains.
The discon inui y in displacemen s is achie ed by adding phan om nodes
a op he o iginal nodes. When an elemen unde goes c acking, each o he
wo esul ing sub-elemen s consis s o eal nodes ( hose co esponding o
he c acked po ion) and phan om nodes ( hose no longe belonging o he
espec i e pa o he o iginal elemen ). These wo elemen s exhibi ully
independen displacemen ields and eplace he o iginal one. F om his poin
onwa d, each pai o eal/phan om nodes in he c acked elemen can sepa a e
ollowing an app op ia e cohesi e law un il ailu e. A his s age, he eal and
phan om nodes a e ee o mo e wi hou cons ain s, e ec i ely simula ing
c ack g ow h. A so ening XFEM law is conside ed, employing an ene ge ic
ailu e powe law c i e ion o a speci ic ype
I II
IC IIC
1.
GG
GG
αα
+=
(5)
whe e GI and GII ep esen he cu en ac u e ene gies in ension and
shea , espec i ely. The analysis in his wo k ocused on he QUADS
ini ia ion c i e ion and a linea damage law wi h a powe pa ame e o α=1.
3. Resul s
3.1. Expe imen al esul s
Figu e 4 p esen s he expe imen al Pm a e age alues and hei espec i e
s anda d de ia ion (STDV) as a unc ion o α, o he h ee e alua ed adhesi es.
F om da a analysis, a simila s eng h e olu ion wi h α is no iceable among
hem. Pm o all adhesi es is ound o α=3.43o, while o he o he angles Pm
diminish and he lowes educ ion was a ained be ween α=3.43o and α=10o.
NUMERICAL EVALUATION OF SCARF GEOMETRY ADHESIVELY-BONDED JOINTS BY XFEM MODELLING
30
Compa ing he pe o mance o all adhesi es, he A aldi e® AV138 s ands
ou , wi h highe Pm o α=3.43o by 35.6 and 120.0% when compa ed wi h he
A aldi e® 2015 and he SikaFo ce® 7752, espec i ely.
0
10
20
30
40
010 20 30 40 50
P
m
[kN]
α[º]
AV138 2015 7752
Figu e 4. Expe imen al Pm a e age esul s and STDV as unc ion o α o he h ee
adhesi es.
Table 2 shows he a e age expe imen al Pm alues oge he wi h STDV
and coe icien s o a ia ion (CoV). A highe CoV dispe sion was ound o
he join s wi h α=45o bonded wi h he A aldi e® 2015 and A aldi e® AV138,
o 17.43 and 11.03%, espec i ely, which may esul om specimens’
manu ac u e o issues associa ed wi h es ing p ocedu es. The SikaFo ce®
7752 p esen ed highe de ia ions wi h he sca a chi ec u es o α=15o,
α=20o and α=30o. All emaining alues a e accep able.
Table 2. Pm a e age alues, STDV, and CoV o sca join s bonded wi h he
di e en adhesi es.
α [o] 3.43 10 15 20 30 45
A aldi e® AV138
STDV [N]
29700
1500
10648
346.63
7926
642.49
5715
310.02
4480
150.89
3325
366.66
CoV [%] 5.05 3.26 8.11 5.42 3.37 11.03
A aldi e® 2015
STDV [N]
21903
1600
7509
772.04
4858
141.92
3567
205.56
2832
101.03
1868
325.56
CoV [%] 7.30 10.28 2.92 5.76 3.57 17.43
Sika o ce® 7752
STDV [N]
13500
510.90
4677
195.40
3132
286.25
2378
229.38
1543
167.86
1142
63.51
CoV [%] 3.78 4.18 9.14 9.65 10.88 5.56
I.R.S. ARAÚJO • R.D.S.G. CAMPILHO
31
Da a om Table 2 shows ha , while α diminishes, Pm inc eases o he
di e en e alua ed sca angles and adhesi es. Pm a ia ions p og essi ely
inc ease o join con igu a ions o α=45o, 30o, 20o, 15o and 10o, whe eas
be ween α=10o and α=3.43o a mo e p onounced a ia ion occu s. Compa ing
all adhesi es, he A aldi e® AV138 was he bes pe o ming adhesi e since
i a ains he highes Pm o he di e en es ed α. Con a ily, he SikaFo ce®
7752 p esen ed he wo s beha iou . Taking he α=3.43o con igu a ion as
baseline, he b i le A aldi e® AV138 p esen ed a Pm o 29.7 kN, whe eas he
duc ile A aldi e® 2015 a ained 21.9 kN. As men ioned abo e, he SikaFo ce®
7752, wi h Pm o 13.5 kN, was he wo s pe o ming adhesi e. Thus,
conside ing he Pm e alua ion o he h ee adhesi es, he A aldi e® AV138
is he bes choice o he sca join s, ollowed by he A aldi e® 2015, which
p esen ed qui e easonable esul s. Mo eo e , he e ogenei y was ound in
he ob ained CoV. None heless, bo h A aldi e® AV138 and SikaFo ce® 7752
p esen ed a lowe dispe sion o he esul s ob ained in he es s.
3.2. Damage analysis
This sec ion p esen s he s udy o he s i ness deg ada ion (SDEG)
damage a iable o he a ious join con igu a ions. This a iable ep esen s
he pe cen ile deg ada ion o he c acked elemen s compa ed o a pu ely
linea beha iou . The s udy o he SDEG damage a iable o he elemen s
o he adhesi e laye , along x/LO, is one o he ools ha helps o compa e
he a ious join con igu a ions o he di e en adhesi es. This a iable has
alues be ween 0 (undamaged ma e ial, up o peak s ess) and 1 ( ailu e),
p o iding he deg ada ion o he s i ness o he damage law unde mixed-
mode condi ions. Figu e 5 shows he ex en o damage (SDEG) a he ins an
o Pm wi h he no malised leng h o he adhesi e laye (x/L), in sca join s wi h
di e en α bonded wi h he adhesi es A aldi e® AV138 (a), A aldi e® 2015
(b) and Sika o ce® 7752 (c). Plo analysis shows ha he g ea es incidence
o damage in his ype o join akes place a he adhesi e laye ends, hus in
ag eemen wi h he dis ibu ion o s esses p esen ed in p e ious wo ks [11].
A he middle o he adhesi e laye , damage is ypically ze o a he momen
Pm is eached. The compa ison be ween di e en α shows ha , as his
geome ical pa ame e inc eases, he e is less localised damage a he ends
and mo e uni o m damage h oughou he adhesi e laye , which also ag ees
wi h he documen ed a ia ion o s ess dis ibu ions as a unc ion o α [21],
since s esses also become mo e uni o m o highe alues o α. Be ween
adhesi es, i can be seen ha , as hei s i ness inc eases, he magni ude o
he damage in he a ious α inc eases and is concen a ed in smalle a eas
a he ends o he bond, showing mo e oscilla ions. On he o he hand, he
39
CHAPTER 4
Nume ical CZM e alua ion o adhesi ely-
bonding solu ions o canoeing boa
ab ica ion
João C.M. San os1
Raul D.S.G. Campilho2
Abs ac
Canoeing is a nau ical spo ha appea ed in his o y housands o yea s ago
simply as a means o su i al. Nowadays, his spo is p ac iced all o e he
wo ld as a hobby o as a means o compe i ion. Gi en he desi e o imp o e he
quali y o cons uc ion and pe o mance o boa s, cu en ly hei manu ac u e
is ocused on he use o composi e ma e ials, which can be con enien ly
bonded wi h adhesi es. On he o he hand, o a manu ac u ing company o
hese boa s o emain compe i i e in he ma ke , i is equi ed he con inuous
imp o emen o hese join s in e ms o s eng h and manu ac u ing cos . In
his wo k, an adhesi e join exis ing in a canoeing boa is nume ically s udied
by cohesi e zone modelling (CZM), mo e speci ically he join be ween he
hull and he deck o a kayak. To achie e he goal o his wo k, a nume ical
1 Depa amen o de Engenha ia Mecânica, Ins i u o Supe io de Engenha ia do Po o, Ins-
i u o Poli écnico do Po o, R. D . An ónio Be na dino de Almeida, 431, 4200-072 Po o,
Po ugal.
2 Depa amen o de Engenha ia Mecânica, Ins i u o Supe io de Engenha ia do Po o, Ins-
i u o Poli écnico do Po o, R. D . An ónio Be na dino de Almeida, 431, 4200-072 Po o,
Po ugal.
INEGI – Pólo FEUP, Rua D . Robe o F ias, 400, 4200-465 Po o, Po ugal.
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
40
analysis is pe o med, in which he exis ing join con igu a ion was es ed,
di e en geome ic changes we e analysed, and di e en ypes o adhesi es
we e conside ed. A p io alida ion o he CZM echnique was pe o med
wi h expe imen al da a ob ained in a p e ious wo k. Ini ially, CZM was
posi i ely alida ed, ollowed by he bes geome y-adhesi e combina ion,
which signi ican ly imp o ed on he cu en used join .
Keywo ds: Canoeing, Adhesi e join s, S uc u al adhesi e, Composi es,
Nume ical modelling, Fini e elemen me hod, Cohesi e zone model.
1. In oduc ion
Since ancien imes, man has been ying o explo e he aqua ic
en i onmen . Canoeing was one o he i s means o a elling on wa e .
Because o he need o su i e, human beings seek ou i e s and seas o
ish, hun o e en na iga e hem. The oldes e idence o his spo was ound
in he omb o a Sume ian king du ing a chaeological exca a ions nea he
i e Euph a es (Wes e n Asia) and da es back a ound six housand yea s
[1]. Th oughou his long his o y, wo ypes o essels ha e eme ged ha
gene ally cha ac e ise he spo h oughou he wo ld. One in a “closed”
s yle, p opelled by a double-bladed paddle known as a kayak; and he o he
in an “open” s yle, p opelled by a single-bladed paddle known as a canoe.
The Kayak o igina ed in G eenland, whe e he Eskimos used i as a means o
indi idual anspo and o hun ing and ishing ac i i ies [2].
Adhesi e bonding is inc easingly p esen in socie y and is seen as an
al e na i e o adi ional me hods. Conside ing ha adi ional me hods can
only be applied o ma e ials abo e ce ain minimum hickness alues o
a oid damaging he ma e ial du ing joining (e.g., ea ing, bending, bu ning),
adhesi es a e ad an ageous because hey can be applied o ma e ials o
any hickness, and elimina e he need o d illing, which esul s in no s ess
concen a ions a hese poin s. Wi h he mos inno a i e adhesi es, i is now
possible o join mos dissimila ma e ials, such as me als, plas ics, o glass.
In addi ion, adhesi e connec ions a e associa ed wi h lowe manu ac u ing
cos s, highe a igue esis ance and high ib a ion damping capabili y [3].
Howe e , adhesi e join s also p esen nume ous disad an ages, such as
he need o ca e ul p epa a ion o he adhe ends, a e y long cu ing ime,
epai di icul ies, among o he s [4]. Va ious join con igu a ions ha e been
p oposed and used by designe s. One o he mos widely used geome ies is
he single-lap join , due o he simplici y o i s manu ac u ing p ocess [5, 6].
This con igu a ion has a majo d awback: when subjec ed o ensile loads, he
asymme y o he s ess ans e lines causes de lec ion o he join , leading
JOÃO C.M. SANTOS • RAUL D.S.G. CAMPILHO
41
o peel s esses (σy) a he edges o he o e lap, which has a de imen al
impac on i s pe o mance [7]. Aiming o o e come he a o emen ioned
limi a ions, se e al adhesi e join con igu a ions ha e been sugges ed, as is
he case o he double-lap, sca , s epped, among o he s [8, 9]. O e he yea s,
a ious app oaches ha e been p oposed o p edic he s eng h o adhesi e
join s, such as analy ical me hods like Volke sen’s pionee ing o mula ions
[10, 11] o nume ical me hods like he ones based on he Fini e Elemen
Me hod (FEM) [12, 13]. Cu en ly, one o he mos ecognised and widely
adop ed me hods is Cohesi e Zone Modelling (CZM). In combina ion wi h
FEM and employing ac u e mechanics concep s, his me hod allows o an
accu a e p edic ion o he adhesi e join s’ s eng h. Using he CZM me hod
i is possible o simula e he onse and p opaga ion o c acks o delamina ion
in composi e ma e ials, o o analyse cohesi e and in e acial de ec s [14].
Nowadays, new nume ical me hods ha e been de eloped ha can be used o
p edic he s eng h o adhesi e join s, as is he case o he eX ended ini e
elemen me hod (XFEM) [15, 16] and meshless me hods [17, 18].
Se e al esea che s ha e s udied adhesi e join s applied in he con ex o
na al s uc u es. Alde ucci e al. [19] expe imen ally analysed he e ec s o
di e en su ace pa e ns p oduced by mechanical eng a ing on composi e
o aluminium adhesi e join s. Fou ypes o pa e ns we e conside ed o
he o e lap join s (no eng a ing, pa e n 0°, pa e n 45° and pa e n 45°X),
which we e subjec ed o ensile es s. In iew o he esul s ound, he Al/
Al join s showed highe s eng h in he 45° and 45°X pa e ns, while he
specimen wi h he 0° pa e n showed less s eng h han he “no pa e n”
specimen, i.e., when compa ing hese wo he p oduc ion o a pa e n
does no b ing any ad an ages. The Al/composi e join s showed highe
s eng h when manu ac u ed wi h some kind o pa e n. In conclusion,
in na al applica ions whe e ligh weigh ma e ials need o be joined (e.g.,
aluminium and/o composi es) he exis ence o g oo es can help imp o e
he mechanical s eng h o he adhesi e join . Ulus e al. [20] pe o med a
s udy o he ac u e and dynamic mechanical beha iou o hyb id adhesi e
join s a e long- e m ageing in seawa e . The aim was o de elop mode I
and mode II delamina ion esis ance and glass ansi ion empe a u e (Tg)
da a as a comp ehensi e design guideline o modi ied basal ib e ein o ced
polyme s (BFRP)-aluminium hyb id join s subjec ed o seawa e ageing (six
mon hs’ exposu e). In addi ion, he adhesi e was ein o ced wi h Halloysi e
nano ubes o enhance ac u e oughness and o slow down wa e abso p ion.
A e exposu e, he adhesi ely bonded ein o ced join s exhibi ed ∼36%
highe ac u e esis ance han he non-exposed bonded join s. Da la e al.
[21] p esen ed a s udy e alua ing he s eng h o he connec ion be ween
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
42
an aluminium hub and a ca bon ib e ein o ced polyme (CFRP) p opelle
blade p o ile o ma ine applica ions. The A aldi e® 2011 adhesi e was used
o bond he componen s, and he assembly was subjec ed o axial h us
condi ions. The au ho s also ca ied ou a nume ical s udy gi en he di icul y
in assessing he s eng h o he expe imen al join due o he complex
geome y, and concluded ha he bonded join specimens be ween he
aluminium hub and he CFRP p opelle blade p o ile we e in luenced by he
p olonged du a ion o ageing a he in e ace be ween he aluminium hub and
he adhesi e, esul ing in join ailu e a lowe load. In he wo k p esen ed in
e e ence [22], he au ho s p oposed a global app oach o assess he in eg i y
o a ull-scale adhesi ely bonded bi-ma e ial join o ma ine applica ions.
The join ep esen s a c oss-sec ion o he adhesi e bond be ween a s eel hull
and a sandwich composi e supe s uc u e used in he ma i ime indus y. The
join was subjec ed o a ensile quasi-s a ic load p o ile including six load
cycles ill he inal collapse o he specimen. Du ing he es s, he au ho s used
h ee s uc u al in eg i y moni o ing echniques, namely acous ic emission,
ib e op ic senso and digi al image co ela ion, o es ablish he s a e o
damage. In addi ion, a FEM model was de eloped o ep oduce he damage
mechanisms and compa e he nume ical esul s wi h he expe imen al ones.
The combina ion o all used echniques could de ec he onse o damage,
e alua e he ex en o damage, iden i y he c i ical egions and di e en ia e
he di e en damage mechanisms.
In his wo k, an adhesi e join exis ing in a canoeing boa is nume ically
s udied by CZM, mo e speci ically he join be ween he hull and he deck o
a kayak. To achie e he goal o his wo k, a nume ical analysis is pe o med,
in which he exis ing join con igu a ion was es ed, di e en geome ic
changes we e analysed, and di e en ypes o adhesi es we e conside ed.
A p io alida ion o he CZM echnique was pe o med wi h expe imen al
da a ob ained in a p e ious wo k.
2. Ma e ials and me hods
2.1. Join geome ies
In his wo k he goal is o op imize a bonded join employed in kayak
manu ac u ing. Gi en he complexi y o he s uc u e, i was decided o
educe he s udy o he adhesi e join o jus one sec ion o he essel. Fo
his pu pose, he on a ea o a single-sea e kayak’s (K1) was conside ed,
which ex e nally is subjec ed o small impac s caused by paddling o o he
impac s due o a collision be ween boa s and, in e nally, is an a achmen
JOÃO C.M. SANTOS • RAUL D.S.G. CAMPILHO
43
poin o he oo es . In ac , his sec ion is an essen ial poin o suppo ing
he use ’s ee and, consequen ly, a la ge pa o he o ce applied o he
wa e o he mo emen o he essel du ing paddling will be unloaded in
his a ea (Fig. 1 a). Fig. 1 (b) shows he sec ional iew o he sec ion ha
will be analysed. The ed line e e s o he hull while he blue line e e s o
he deck o he kayak. Cu en ly he wo pa s a e bonded conside ing a bu
join con igu a ion (Fig. 1 c).
a)
b) c)
Fig. 1. Zone o he Kayak unde analysis a), sec ional iew o he kayak b) and
de ail o he sec ional iew c).
In addi ion o he cu en bu join geome y (Fig. 2 a), wo mo e di e en
geome ies (Fig. 2 b and c) a e p oposed and analysed. Bo h solu ions
a e cham e - ype join s conside ing a cham e o sca angle o 45°, bu
symme ically a anged and an adhesi e hickness o 0.2 mm. Al hough hese
geome ies ha e simila cha ac e is ics, he cu a u e o he kayak and he
di e en hickness o he pa s o be joined will p oduce di e en beha iou s
in he adhesi e laye . Fig. 2 (b) shows he i s ype o he cham e ed
geome y (cham e 1). In his case, seen om he ou side, he deck will
o e lap he hull. The second ype (cham e 2) is p esen ed in Fig. 2 (c), in
which he hull o e laps he deck. These con igu a ions will be subjec ed o
ou di e en ypes o loading, namely ac ion, comp ession, bending and
shea , o de e mine he bes con igu a ion and imp o e he cu en assembly.
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
44
a) b) c)
Fig. 2. Schema ic ep esen a ion o he bu join (a), cham e 1 (b) and cham e 2
(c).
Fo nume ical model alida ion, he single-lap join geome y was
conside ed (Fig. 3). As bounda y condi ions, he join was clamped a one o
he edges, and he opposi e edge was subjec ed o a ho izon al displacemen
wi h e ical mo emen es ic ion. The main dimensions o he analysed
specimen a e he o e lap leng h (LO)=12.5, 25, 37.5, and 50 mm, adhe ends’
hickness ( P)=3 mm, adhesi e hickness ( A)=0.2 mm, join o al leng h
(LT)=170 mm, and wid h (B)=25 mm.
Fig. 3. Schema ic ep esen a ion o he single-lap join geome y and bounda y
condi ions.
2.2. Ma e ials
To p o ide a wide ange o esul s, in addi ion o he main adhesi e,
h ee o he adhesi es we e selec ed o e alua e he di e en adhesi e join
geome ies o bond he hull and deck o a kayak. The Sika Adeki ® A 140-1
JOÃO C.M. SANTOS • RAUL D.S.G. CAMPILHO
45
[23] has excellen mechanical pe o mance and esis ance o dynamic loads
( ib a ions and impac s). In addi ion, i is e sa ile enough o be applied bo h
e ically and ho izon ally, being sui able o illing i egula join s and being
used in agg essi e en i onmen s. To ensu e a wide ange o duc ili ies, he
b i le epoxy A aldi e® AV138 [24], he mode a ely duc ile epoxy A aldi e®
2015 [25, 26], and he duc ile polyu e hane Sika o ce® 7752 [27] we e also
conside ed.
Table 1. Mechanical and ac u e p ope ies o he ou adhesi es.
P ope y Adeki ® A
140-1
A aldi e®
AV138
A aldi e®
2015
Sika o ce®
7752
Young’s modulus, E [GPa] 2.57 4.89±0.81 1.85±0.81 0.493±0.0896
Poisson coe icien , - 0.35a0.33a0.33a
Tensile yield s ess, σy
[MPa]
- 36.49±2.47 12.3±0.61 3.24±0.5
Tensile s eng h, σ [MPa] 30 39.45±3.18 21.63±1.61 11.49±0.3
Tensile ailu e s ain, ε [%] 4 1.21±0.10 4.77±0.15 19.18±1.4
Shea modulus, G [GPa] 0.99 1.56±0.01 0.56±0.21 0.187±0.0164
Shea yield s ess, τy [MPa] - 25.1±0.33 14.6±1.3 5.16±1.1
Shea s eng h, τ [MPa] 20 30.2±0.40 17.9±1.8 10.17±0.6
Shea ailu e s ain, γ [%] - 7.8±0.7 43.9±3.4 58.42±6.4
Tensile oughness, GIc [N/
mm]
0.5 0.2b0.43±0.02 2.36±0.2
Shea oughness, GIIc [N/
mm]
5.56 0.38b4.7±0.34 5.41±0.5
a Manu ac u e ’s alue
b Ob ained in e e ence [28]
The geome y o he hull and deck ha e di e en con igu a ions. As he
hull is mo e p one o impac s and o he s esses, mo e laye s o ein o cemen s
a e applied o make he hull mo e esis an . In he a ea o highe s ess ( he
cen al pa o he boa , whe e he use will be), he hull is made up o i e
laye s, while he deck is only made up o h ee laye s in all. Fig. 4 shows
he co ec s acking o de o he deck adhe end (le ) and he hull adhe end
( igh ).
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
46
Fig. 4. S acking o de o he deck adhe end (le ) and he hull adhe end ( igh ).
Table 2 shows he p ope ies o he ma e ials ha cons i u e he sandwich
s uc u e o a boa . The (A) ca bon ib e [29] and (B) glass ib e [30] come
in he o m o ab ic and he co e o he kayak is (C) PVC oam [31]. To
connec hese ma e ials, he ma ix Resolcoa 1400-1407 epoxy esin was
used [32].
Table 2. P ope ies o he boa ma e ials.
P ope y (A) Ca bon
ib e
(B) Glass
ib e
(C) PVC
oam
Resolcoa
Young’s modulus, E [MPa] 240000 33000 95 3100-3300
Yield s ess, σ [MPa] 4200 626 2.5 -
Fib e densi y, ρ [g/cm3]1.78 2.6 0.08 1.15
A ea weigh , G [g/m2] 160 220 - -
Thickness, [mm] 0.30 0.25 3 -
Shea modulus, G [MPa] - - 27 -
Poisson coe icien , ν- - 0.4 0.35
Shea yield s ess, τ [MPa] - - 1.15 -
2.3. Nume ical modelling
The nume ical models we e de eloped in so wa e Abaqus®. To simpli y
he p ocedu e while gua an eeing accu acy o he esul s, a 2D ep esen a ion
o he geome y was chosen. The join o be modelled is a bu join wi h
cu ed adhe ends, o eplica e he join be ween he hull and deck o a
kayak. A iangula CZM is used o simula e ailu e in he adhesi e laye
connec ing he hull o he deck. The ma e ials, bo h he hull and he deck,
a e linea elas ic and iso opic. A hickness o 0.2 mm was conside ed o he
adhesi e. The adhe ends we e dimensioned o a leng h o 50 mm. In e ms
o hickness, he deck adhe end is 3.6 mm hick, while he hull adhe end
JOÃO C.M. SANTOS • RAUL D.S.G. CAMPILHO
47
is 4.2 mm hick, cen ed on each o he , i.e., wi h 0.3 mm o se on each
side. The join was subdi ided in o sec ions o allow he di e en ia ion o
each ma e ial and he subsequen a ibu ion o hei dis inc p ope ies. Fo
he PVC co e, a hickness o 3 mm was assumed, while o he GFRP and
CFRP, a hickness o 0.3 mm was conside ed o each laye . Fo each sec ion,
he mechanical p ope ies and hei espec i e beha iou we e de ined om
he da a o Table 1 and Table 2. PVC, CFRP and GFRP we e de ined as
homogeneous solid ma e ials, while he adhesi e was de ined as a cohesi e
ma e ial. The so wa e Abaqus® pe o ms he simula ion on an inc emen al
basis. To acili a e con e gence du ing he damage p opaga ion phase, he
minimum allowed inc emen size (in % o he applied loading) is 1×10-20, and
he maximum inc emen size is equal o he ini ial alue (0.5%). Bounda y
condi ions we e es ablished o emula e di e en possible loadings applied
o he join . To his end, a geome ic es ic ion was applied o he lowe
adhe end, while a displacemen condi ion was applied o he uppe adhe end.
Fo he adhe end laye s, a s uc u ed elemen de ini ion was selec ed, while a
sweep elemen de ini ion was used o he adhesi e. None heless, o analyse
s ess dis ibu ions, a s uc u ed elemen de ini ion was conside ed o he
adhesi e laye as well, which makes i possible o ex ac s esses wi h
p ecision a he adhesi e mid- hickness. CPE4 elemen s we e used o he
adhe end laye s, while COH2D4 elemen s we e used o he adhesi e. To
ob ain he s ess dis ibu ions in he adhesi e, CPE4 elemen s we e also used
in he co esponding pa i ion. In all cases, a iscosi y o 1×10-5 Pa.s. o
he elemen s we e conside ed. Bias (size g ading e ec s) in he cons uc ed
mesh ensu ed good e inemen a he mos c i ical loca ions o ex ac he
s esses a hese loca ions mo e accu a ely. In he CZM damage p opaga ion
models, in which he s eng h o he join is s udied, he es s a e ca ied ou
un il a aining he ini ially imposed displacemen , so ha he join ailu e is
achie ed.
2.4. CZM desc ip ion
Cohesi e Zone Models (CZM) es ablish a damage law be ween pai ed
nodes wi hin cohesi e elemen s ha ela es s esses and displacemen s,
enabling modelling he s uc u es’ beha iou up o hei mechanical limi s
[33]. O e ime, nume ous models ha e been de eloped, each ailo ed
o di e en ma e ial and beha iou s, o e ing enhanced accu acy, as
demons a ed by [34]. In his s udy, he iangula shaped law was chosen
o bo h adhesi e ma e ials. I s simplici y in con igu a ion and usabili y
wi hin nume ical analysis so wa e makes i an a ac i e choice. No ably,
i deli e s p ecise esul s in a ema kably e icien manne . This model’s
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
54
• Fo he adhesi e cu en ly used in he manu ac u e o boa s, he Sika
Adeki ® A140-1 adhesi e, he mos sui able geome y is cham e 2,
ollowed by he bu geome y and inally he cham e 1 geome y.
The cham e 2 geome y ob ains he highes Pm o ensile and shea
s esses, and he second highes alue o he o he wo loadings. The
bes geome y unde comp ession is bu , and unde bending is cham-
e 1. Howe e , by compa ing he a e age Pm o he cham e ed and
bu geome ies and o he ou loadings, he cham e 1 geome y is
ul ima ely he bes geome y o his adhesi e;
• Fo he A aldi e® AV138, he geome y ha gi es i he bes p ope ies
compa ed o he o he adhesi es is he bu geome y, ollowed by he
cham e 2 geome y and inally he cham e 1 geome y. By compa-
ing he Pm alues o his adhesi e, he cham e 2 geome y p o ides
he bes s eng h unde bending and shea , and he second-bes s en-
g h unde axial ension. The bu geome y once again gi es he bes
Pm o his adhesi e in comp ession, and he bes ensile s eng h is
gi en by he cham e 1 geome y. In sho , he a e age alues ob ained
o he di e en loadings show ha he cham e 2 geome y p o ides
he bes cha ac e is ics o e all;
• Analysing he A aldi e® 2015 shows ha he geome y ha gi es he
bes cha ac e is ics, compa ed o he o he adhesi es, is he cham e
1 geome y, ollowed by he cham e 2 geome y and inally he bu
geome y. The cham e 1 geome y has he bes esis ance o ensi-
le, bending and shea loadings. The bu geome y has he bes pe -
o mance unde comp essi e loadings in his adhesi e, bu he wo s
s eng h o o he loadings. As shown in Table 3, i can be concluded
ha he cham e 1 geome y is he bes o he h ee s udied geome ies
i his adhesi e is used. Compa ing he a e age Pm ob ained o each
geome y, he cham e 1 geome y once again s ands ou ;
• The Sika o ce® 7752 gi es he bes beha iou o he cham e 1 ge-
ome y, ollowed by he bu geome y and, inally, he cham e 2 ge-
ome y. Acco ding o he Pm alues, he cham e 1 geome y p o ides
he bes s eng h o his adhesi e unde ensile, bending and shea
loadings. On he o he hand, he bu geome y once again p o ides he
bes s eng h unde a comp essi e loading. BY compa ing he gene al
beha iou o he cham e geome ies in ela ion o he bu geome y,
i can be concluded ha he cham e 1 geome y gua an ees he bes
cha ac e is ics o his adhesi e.
JOÃO C.M. SANTOS • RAUL D.S.G. CAMPILHO
55
Analysing he ab ica ion p ocedu e o he di e en geome ies shows
ha he bu geome y has he sho es p oduc ion ime and consequen ly
he lowes p oduc ion cos s, since no su ace p epa a ion is ca ied ou and
he adhesi e is placed di ec ly on he join . Fo he cham e ed geome y,
whe he ype 1 o ype 2, i is necessa y o p epa e he join and ab ica e
he cham e ed su aces, which will inc ease manu ac u ing imes and
consequen ly p oduc ion cos s. Thus, his ea u e should be conside ed in he
design p ocess.
4. Conclusions
This wo k aimed o op imize he s eng h o an adhesi e join o med
be ween he hull and deck o a boa used in canoeing, and subsequen ly
p opose geome ic solu ions and ypes o adhesi e ha could imp o e he
cu en con igu a ion o he adhesi e join . The CZM echnique was ini ially
alida ed wi h s anda d SLJ geome ies and h ee adhesi es. The obse ed
p edic i e capaci y was qui e sa is ac o y o he wo A aldi e® adhesi es,
while mode a e unde p edic ions we e ound o he 7752, al hough his
does no in alida e he me hod. τxy/τa g and σy/τa g s esses showed s ess
concen a ions a he ends o he adhesi e o all loading ypes. By compa ing
he di e en geome ies, he bu geome y has smalles peak s esses unde
a shea loading, he cham e 1 geome y unde ensile and comp essi e
loadings, and he cham e 2 geome y unde bending loadings. Be ween
adhesi es, he AV138 had he highes peaks on accoun o i s s i ness. The
s eng h analysis showed ha , o he bu geome y, he AV138 maximizes
Pm. Fo he cham e 1 geome y, he 7752 is he bes solu ion, and o he
cham e 2 geome y, he AV138 pe o ms bes . In iew o all ob ained esul s,
inc easing he boa s eng h wi hou signi ican ly inc easing p oduc ion
ime is accomplished by applying he AV138 in a bu geome y. Fo he
bes possible con igu a ion dis ega ding cos s and ab ica ion ime, he e a e
wo possible con igu a ions: he 7752 applied o he cham e 1 geome y,
and he AV138 applied o he cham e 2 geome y. Abou he la e wo
con igu a ions, he i s excels in smalle τxy and σy s esses. Fu he mo e,
since a mo e duc ile adhesi e is used in he i s con igu a ion, i will be e
esis he s esses expe ienced.
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
56
Re e ences
1. S a , C.G., A his o y o he ancien wo ld. 1991: Ox o d Uni e si y P ess, USA.
2. Ve non, J., The Lincoln Kayak. The Ma ine ’s Mi o , 1984. 70(4): p. 415-426.
3. Pe ie, E.M., Handbook o adhesi es and sealan s. 2007: McG aw-Hill Educa ion.
4. Romano, M.G., M. Guida, F. Ma ulo, M. Giugliano Au icchio, and S. Russo, Cha a-
c e iza ion o Adhesi es Bonding in Ai c a S uc u es. Ma e ials, 2020. 13(21): p.
4816.
5. Tsai, M.Y. and J. Mo on, An e alua ion o analy ical and nume ical solu ions o
he single-lap join . In e na ional Jou nal o Solids and S uc u es, 1994. 31(18): p.
2537-2563.
6. Öz op ak, N. and G.M. Gençe , Load-bea ing capaci y o polyamide 6 (PA6) compo-
si e o 7075-O ae ospace Al-alloy single-lap join s: in luence o a ious lase ex u ed
pa e ns on ho p ess bonding. Jou nal o Adhesion Science and Technology, 2023: p.
1-19.
7. Ha -Smi h, L.J., Adhesi e-bonded single-lap join s. 1973, NASA Con ac Repo ,
NASA CR-112236.
8. Pe ie, E.M., The undamen als o adhesi e join design and cons uc ion: Func i-
on-speci ic cons uc ion is he key o p ope adhesion and load-bea ing capabili ies.
Me al Finishing, 2008. 106(11): p. 55-57.
9. Adams, R.D., J. Comyn, and W.C. Wake, S uc u al adhesi e join s in enginee ing.
2nd ed. 1997, London, Uni ed Kingdom: Chapman & Hall.
10. Volke sen, O., Die nie k a oe eilung in zubeansp uch en nie e bindungen mi kons-
an en loschonque schni en. Lu ah o schung, 1938. 15 p. 41-47.
11. Tse pes, K., A. Ba oso-Ca o, P.A. Ca a o, V.C. Bebe , I. Flo os, W. Gamon, M.
Kozłowski, F. San and ea, M. Shah e di, D. Skejić, C. Bedon, and V. Rajčić, A e iew
on ailu e heo ies and simula ion models o adhesi e join s. The Jou nal o Adhesi-
on, 2021: p. 1-61.
12. He, X., A e iew o ini e elemen analysis o adhesi ely bonded join s. In e na ional
Jou nal o Adhesion & Adhesi es, 2011. 31(4): p. 248-264.
13. de Sousa, C.C.R.G., R.D.S.G. Campilho, E.A.S. Ma ques, M. Cos a, and L.F.M. da
Sil a, O e iew o di e en s eng h p edic ion echniques o single-lap bonded jo-
in s. Jou nal o Ma e ials: Design and Applica ion - Pa L, 2017. 231: p. 210-223.
14. Saeedi a , M., M. Ahmadi Naja abadi, J. Youse i, R. Mohammadi, H. Hosseini Tou-
deshky, and G. Minak, Delamina ion analysis in composi e lamina es by means o
Acous ic Emission and bi-linea / i-linea Cohesi e Zone Modeling. Composi e S u-
c u es, 2017. 161: p. 505-512.
15. Bely schko, T. and T. Black, Elas ic c ack g ow h in ini e elemen s wi h minimal e-
meshing. In e na ional Jou nal o Nume ical Me hods in Enginee ing, 1999. 45(5): p.
601-620.
JOÃO C.M. SANTOS • RAUL D.S.G. CAMPILHO
57
16. Xa á, J.T.S. and R.D.S.G. Campilho, S eng h es ima ion o hyb id single-L bonded
join s by he eX ended Fini e Elemen Me hod. Composi e S uc u es, 2018. 183: p.
397-406.
17. Chen, J.-S., C. Pan, C.-T. Wu, and W.K. Liu, Rep oducing Ke nel Pa icle Me hods
o la ge de o ma ion analysis o non-linea s uc u es. Compu e Me hods in App-
lied Mechanics and Enginee ing, 1996. 139(1): p. 195-227.
18. Ramalho, L.D.C., R.D.S.G. Campilho, and J. Belinha, P edic ing single-lap join
s eng h using he na u al neighbou adial poin in e pola ion me hod. Jou nal o he
B azilian Socie y o Mechanical Sciences and Enginee ing, 2019. 41(9): p. 362.
19. Alde ucci, T., C. Bo sellino, and G. Di Bella, E ec o su ace pa e n on s eng h o
s uc u al ligh weigh bonded join s o ma ine applica ions. In e na ional Jou nal o
Adhesion and Adhesi es, 2022. 117: p. 103005.
20. Ulus, H., H.B. Kaybal, F. Cacık, V. Eskizeybek, and A. A cı, F ac u e and dynamic
mechanical analysis o seawa e aged aluminum-BFRP hyb id adhesi e join s. Engi-
nee ing F ac u e Mechanics, 2022. 268: p. 108507.
21. Da la, V., B. Sa ish Ben, and K.V. Sai S inadh, In es iga ion on in e acial bonding
s eng h be ween aluminum hub and CFRP bonded join s exposed o accele a ed
aging condi ions. Jou nal o Adhesion Science and Technology: p. 1-13.
22. Saeedi a , M., M.N. Saleh, A. K ai i, S.T. de F ei as, and D. Za ouchas, S uc u al
in eg i y assessmen o a ull-scale adhesi ely-bonded bi-ma e ial join o ma i ime
applica ions. Thin-Walled S uc u es, 2023. 184: p. 110487.
23. Sika, P oduc da a shee ADEKIT A140-1 / H9940-1. 2020.
24. Ne o, J.A.B.P., R.D.S.G. Campilho, and L.F.M. da Sil a, Pa ame ic s udy o adhesi-
e join s wi h composi es. In e na ional Jou nal o Adhesion and Adhesi es, 2012. 37:
p. 96-101.
25. Campilho, R.D.S.G., M.D. Banea, J.A.B.P. Ne o, and L.F.M. da Sil a, Modelling ad-
hesi e join s wi h cohesi e zone models: e ec o he cohesi e law shape o he adhe-
si e laye . In e na ional Jou nal o Adhesion & Adhesi es, 2013. 44: p. 48-56.
26. Campilho, R.D.S.G., A.M.G. Pin o, M.D. Banea, and L.F.M. da Sil a, Op imiza ion
s udy o hyb id spo -welded/bonded single-lap join s. In e na ional Jou nal o Adhesi-
on and Adhesi es, 2012. 37: p. 86-95.
27. Faneco, T.M.S., Ca ac e ização das p op iedades mecânicas de um adesi o es u u-
al de al a duc ilidade. 2014, Tese de Mes ado em Engenha ia Mecânica – Ramo de
Ma e iais e Tecnologias de Fab ico. Ins i u o Supe io de Engenha ia do Po o: Po o.
28. Campilho, R., Adhesi e, welded and weld-bonded single-lap join s: Nume ical ech-
nique o s eng h p edic ion. Vol. 24. 2012. 35-42.
29. HAUFLER COMPOSITES Techinacal da a shee Wo en Ca bon Fib e Fab ic 160
g/m², Plain 2017.
30. In e glas Po che Indus ies Glass Filamen Fab ics o Plas ics Rein o cemen - P o-
duc Speci ica ion. 2019.
NUMERICAL CZM EVALUATION OF ADHESIVELY-BONDING SOLUTIONS FOR CANOEING BOAT
58
31. Diab G oup Techincal da a Di inycell H. 2021.
32. Résol ech, Da a shee RESOLCOAT 1400-1407 and Accele a o AC140.
33. Al ano, M., F. Fu giuele, A. Leona di, C. Male a, and G.H. Paulino, Cohesi e Zone
Modeling o Mode I F ac u e in Adhesi e Bonded Join s. Key Enginee ing Ma e ials,
2007. 348-349: p. 13-16.
34. Campilho, R.D., M.D. Banea, J. Ne o, and L.F. da Sil a, Modelling adhesi e join s
wi h cohesi e zone models: e ec o he cohesi e law shape o he adhesi e laye . In-
e na ional Jou nal o Adhesion & Adhesi es, 2013. 44: p. 48-56.
35. Valen e, J.P.A., R.D.S.G. Campilho, E.A.S. Ma ques, J.J.M. Machado, and L.F.M. da
Sil a, Geome ical op imiza ion o adhesi e join s unde ensile impac loads using
cohesi e zone modelling. In e na ional Jou nal o Adhesion & Adhesi es, 2020. 97: p.
102492.
36. Sane, A.U., P.M. Padole, C.M. Manjuna ha, R.V. Uddanwadike , and P. Jhunjhunwa-
la, Mixed mode cohesi e zone modelling and analysis o adhesi ely bonded composi e
T-join unde pull-ou load. Jou nal o he B azilian Socie y o Mechanical Sciences
and Enginee ing, 2018. 40(3): p. 167.
37. Rocha, R.J.B. and R.D.S.G. Campilho, E alua ion o di e en modelling condi ions
in he cohesi e zone analysis o single-lap bonded join s. The Jou nal o Adhesion,
2018. 94(7): p. 562-582.
38. Dimi i, R., M. T ullo, L. De Lo enzis, and G. Za a ise, Coupled cohesi e zone mo-
dels o mixed-mode ac u e: A compa a i e s udy. Enginee ing F ac u e Mechanics,
2015. 148: p. 145-179.
59
CHAPTER 5
Feasibili y o D one-Based G ound
Pene a ing Rada o Sub e anean
Fo eign Objec De ec ion/Mapping
Celile Nu Yalçın1
Abs ac
This s udy e alua es he easibili y o using d one-based G ound
Pene a ing Rada (GPR) o de ec and map unde g ound o eign objec s a
high al i udes. In eg a ion o GPR echnology wi h d ones enables e icien
scanning o la ge a eas. Va ious pa ame e s such as ope a ing al i ude, d one
speed, equency, an enna gain, soil a enua ion, ield o iew, a ge size and
ma e ial ype we e aken in o conside a ion. The esea ch pa ame e s we e
kep wide o his ada sys em, which can be cus omized o be in eg a ed wi h
di e en unmanned ae ial ehicles. Simula ions we e un in MATLAB wi h
a cus om G aphical Use In e ace (GUI) using bo h pulsed and F equency
Modula ed Con inuous Wa e (FMCW) ada modula ion echniques. The
esul s show ha d one-based GPR sys ems can achie e su icien Signal- o-
Noise Ra io (SNR) and ange esolu ion o e ec i e unde g ound de ec ion
and show signi ican po en ial o a chaeology, de ense indus y, u ili y
mapping and en i onmen al moni o ing applica ions.
1 Anka a Yıldı ım Beyazı Uni e si y, Ins i u e o Science, Elec ical and Elec onics En-
ginee ing Mas e ’s Deg ee S uden , ORCID Code: h ps://o cid.o g/0009-0009-6402-
5083, [email p o ec ed]
FEASIBILITY OF DRONE-BASED GROUND PENETRATING RADAR FOR SUBTERRANEAN FOREIGN OBJECT
60
Keywo ds — D one, GPR, Unde g ound De ec ion, Fo eign Objec ,
Mapping.
1.In oduc ion
Subsu ace de ec ion holds c i ical signi icance ac oss a ious disciplines
such as ci il enginee ing, a chaeology, en i onmen al s udies, and secu i y.
G ound Pene a ing Rada (GPR) is a non-in asi e echnique ha employs
elec omagne ic wa es o image subsu ace ea u es. Con en ional g ound-
based GPR me hods a e o en ime-in ensi e and limi ed in e ms o scanning
a ea co e age. The ad ancemen o d one echnology o e s a solu ion by
enabling apid and ex ensi e ae ial su eys. The in eg a ion o d one and
GPR echnologies acili a es e icien and high- esolu ion mapping o la ge
a eas. This s udy e alua es he easibili y o d one-based GPR sys ems
ope a ing a high al i udes and in es iga es key sys em pa ame e s by
analyzing pe o mance me ics such as signal- o-noise a io (SNR) and ange
esolu ion h ough simula ion.
2. Sensing Techniques U ilized in GPR Technology
G ound Pene a ing Rada (GPR) is a geophysical imaging me hod ha
uses elec omagne ic wa es o de ec and map subsu ace s uc u es. GPR
sys ems can be op imized o di e en applica ions based on equency
and modula ion echniques. The choice o hese echniques is p ima ily
in luenced by he a ge ed dep h, esolu ion equi emen s, and en i onmen al
condi ions. The ollowing sec ions p o ide an o e iew o he p incipal
echniques commonly employed in GPR sys ems.
2.1 Pulsed GPR
Pulsed GPR sys ems emi high- equency elec omagne ic wa es in
sho pulses and analyze he signals e lec ed om subsu ace s uc u es.
The dep h o objec s is calcula ed using he ime delay o he e u n signal,
he p opaga ion eloci y o elec omagne ic wa es in he medium, and he
a eled dis ance eq (1)-(2).
2d
= (1)
whe e:
( ) = ime delay (s)
CELILE NUR YALÇIN
61
( d ) = objec dep h (m)
( ) = elec omagne ic wa e eloci y (m/s)
c
ε
= (2)
whe e:
( c ) = speed o ligh
𝜀𝑟 = ela i e pe mi i i y o he medium
Ad an ages:
• Enables apid da a acquisi ion.
• P o ides high pene a ion dep h.
• Simple signal analysis due o di ec ime-domain measu emen .
Disad an ages:
• Signal a enua ion inc eases a highe equencies.
• Highe noise le els can occu , leading o inc eased da a p ocessing
equi emen s.
2.2 Con inuous Wa e (CW) GPR
The Con inuous Wa e (CW) echnique ansmi s a con inuous
elec omagne ic signal and e alua es he signals e lec ed om benea h he
su ace. Unlike pulsed GPR, he signal du a ion is cons an , and dis ance
in o ma ion is ob ained h ough modula ion echniques.
Ad an ages:
• Deli e s a s ong and con inuous signal, esul ing in a high sig-
nal- o-noise a io (SNR).
• Mo e compa ible wi h mo ing pla o ms such as d one-based GPR
sys ems.
Disad an ages:
• Dep h measu emen accu acy a ies depending on he modula ion ap-
p oach.
• Requi es mo e complex da a p ocessing compa ed o pulsed sys ems.
FEASIBILITY OF DRONE-BASED GROUND PENETRATING RADAR FOR SUBTERRANEAN FOREIGN OBJECT
62
2.3 F equency Modula ed Con inuous Wa e (FMCW)
GPR
The FMCW echnique ob ains ange in o ma ion by con inuously a ying
he equency o he ansmi ed signal o e a speci ied ime in e al. The
posi ion and physical p ope ies o he objec s a e de e mined by analyzing
he equency di e ence be ween he ansmi ed and ecei ed signals eq (3).
2
es
c
R
B
= (3)
whe e:
es
R=
ange esolu ion (m)
( B ) = bandwid h (Hz)
( c ) = speed o ligh (m/s)
SNR calcula ion eq (4):
SNR =
n
P
P (4)
whe e:
P
= ecei ed powe (W)
n
P
= noise powe (W)
Ad an ages:
• P o ides highe esolu ion.
• Ene gy-e icien due o ope a ion wi h low powe signals.
• Enables mo e p ecise dep h es ima ion.
Disad an ages:
• Highe sys em cos due o inc eased complexi y o elec onic compo-
nen s.
• Da a p ocessing equi es mo e compu a ional esou ces compa ed o
pulsed GPR.
CELILE NUR YALÇIN
63
2.4 Time-Domain GPR
Time-domain GPR sys ems de ec subsu ace objec s by analyzing he
p opaga ion ime o elec omagne ic wa es. This echnique is capable o
dis inguishing media wi h di e en dielec ic p ope ies.
Ad an ages:
• Capable o high-speed da a acquisi ion o e la ge-scale a eas.
• O e s high dep h accu acy and is sui able o mul ilaye ed en i on-
men s.
Disad an ages:
• Requi es la ge s o age capaci y due o high da a olume.
• Analysis ime may be longe compa ed o o he echniques.
2.5 Mul i-F equency GPR
Mul i- equency GPR sys ems ansmi elec omagne ic wa es a mul iple
equencies simul aneously, allowing o p ecise de ec ion o objec s a
a ious dep hs.
Ad an ages:
High equencies enable de ec ion o su ace de ails, while low equencies
allow pene a ion o g ea e dep hs.
O e s a b oade ange o applica ions.
Disad an ages:
Complexi y in da a p ocessing due o mul i- equency inpu s.
Ha dwa e cos s a e highe ela i e o single- equency sys ems.
Conclusion and E alua ion
The echniques employed in GPR echnology a y based on ope a ional
p inciples and applica ion domains. Pulsed GPR o e s ad an ages such
as deep pene a ion and s aigh o wa d da a p ocessing, whe eas FMCW
sys ems p o ide supe io esolu ion. Time-domain and mul i- equency GPR
sys ems a e mo e sui able o ex ensi e su eys. Selec ion o he app op ia e
echnique should conside en i onmen al condi ions, equi ed dep h, desi ed
esolu ion, and ope a ional cos s. This summa y elabo a es on undamen al
GPR sensing echniques, he eby con ibu ing o in o med decision-making
ega ding me hod selec ion o di e se applica ions.
TRUST-WEIGHTED SENTIMENT FILTERING AND QUERY-BASED RECOMMENDATION FOR TURKISH
70
Yildiz e al. add essed hese challenges by in oducing a BERT-based
sen imen classi ica ion model ailo ed o Tu kish e-comme ce e iews,
ou pe o ming s anda d classi ie s and con i ming he alue o language-
speci ic op imiza ion17. Thei indings ein o ce he impo ance o oken-
le el handling, synonym expansion, and sen imen consis ency in Tu kish-
language ecommenda ion sys ems.
Finally, domain-speci ic esea ch has begun o emphasize aspec -based
sen imen analysis (ABSA) as a pi o al echnique o aligning use que ies
wi h ele an p oduc ea u es. Liu e al. p oposed an a en ion-based ABSA
amewo k o e-comme ce pla o ms18, allowing ecommenda ions o
espond no only o gene al sen imen pola i y bu also o con ex -speci ic
ea u es like “ba e y li e” o “sc een quali y,” which a e especially pe inen
in p oduc e iew il e ing .
In ligh o hese de elopmen s, he p esen s udy con ibu es by in eg a ing
high-p ecision egex-based e ie al wi h sen imen il e ing and e iew
eliabili y sco ing o p oduce a ligh weigh , anspa en , and use -in en -
aligned ecommenda ion pipeline. This app oach is no only explainable and
language-sensi i e bu also compu a ionally e icien —add essing he dual
challenge o anspa ency and scalabili y.
METHODOLOGY
This s udy adop s a design-o ien ed me hodological amewo k o
de elop a sen imen -awa e and eliabili y-weigh ed ecommenda ion sys em
ailo ed o Tu kish e-comme ce da a. The app oach in eg a es ule-based
sen imen labeling using s a a ings, linguis ic p ep ocessing ( okeniza ion,
lowe casing, and s opwo d elimina ion), and isual explo a ion ia wo d
clouds. Reliabili y is calcula ed h ough use engagemen me ics, speci ically
a no malized a io o posi i e (clap) o o al eedback (clap + humbsdown),
e lec ing ecen indings ha use us signals enhance ecommenda ion
ele ance and c edibili y⁷.
Ins ead o using p e- ained ans o me -based a chi ec u es, his wo k
emphasizes ligh weigh , in e p e able, and language-speci ic echniques,
which ha e shown e ec i eness in mo phologically complex languages like
Tu kish⁸. A que y-d i en e ie al mechanism is also implemen ed, allowing
use s o sea ch p oduc eedback wi h na u al ph ases such as “ ele on ek anı
güzel”. The ma ching logic inco po a es pa e n-based keywo d de ec ion and
il e s he esul s acco ding o sen imen class and a isual eliabili y sco e,
in line wi h ecen esea ch emphasizing human-cen e ed and anspa en
ecommenda ion pipelines⁹⁻¹¹.
MUSTAFA ERŞAHİN • UTKU GEZENSOY
71
Da ase Desc ip ion
The da ase u ilized in his s udy consis s o Tu kish-language e-comme ce
p oduc e iews, including s uc u ed me ada a such as p oduc iden i ie s,
s a a ings, e iew i les, ull e iew ex s, and use engagemen me ics
(clap and humbsdown coun s). These e iews ep esen eal-wo ld cus ome
opinions om a di e se ange o p oduc ca ego ies, allowing o sen imen
analysis g ounded in au hen ic linguis ic pa e ns and use p e e ences.
(Table 1: Each e iew en y con ains he ollowing key a ibu es).
Table 1: Each e iew en y con ains he ollowing key a ibu es
To enhance in e p e abili y and ensu e consis ency in sen imen
anno a ion, a sen imen pola i y label was c ea ed based on s a a ings:
• Re iews a ed 1 o 2 we e labeled as nega i e (-1)
• Ra ings o 3 we e conside ed neu al (0)
• Ra ings o 4 o 5 we e labeled as posi i e (+1)
In addi ion o sen imen labels, a eliabili y sco e was calcula ed o each
e iew. This sco e, no malized and con e ed in o pe cen age o ma , se es
as a p oxy o use us wo hiness and plays a i al ole in il e ing and
p io i izing e iews wi hin he ecommenda ion sys em.
The da ase includes ens o housands o e iews, making i a sui able
co pus o sen imen modeling, eliabili y analysis, and use -que y ma ching
in Tu kish-language con ex s. Fu he mo e, he p esence o bo h quan i a i e
and quali a i e eedback enables a hyb id app oach ha le e ages ule-based
classi ica ion alongside linguis ic ea u e explo a ion.
Da a P ep ocessing
P io o model aining and ecommenda ion gene a ion, he aw da ase
unde wen a se ies o p ep ocessing s eps designed o enhance da a quali y,
educe noise, and s anda dize ex ual ea u es o na u al language analysis.
TRUST-WEIGHTED SENTIMENT FILTERING AND QUERY-BASED RECOMMENDATION FOR TURKISH
72
Fi s , missing alues we e add essed. All missing en ies in he i le
column we e eplaced wi h he placeholde “Basliksiz”, while e iews
wi h null alues in he e iew ield we e emo ed o p e en downs eam
inconsis encies. Addi ionally, he display se ings o he da a ame we e
adjus ed o e eal comple e ex ual con en and ensu e isual cla i y du ing
analysis. Nex , a sen imen label was assigned o each e iew based on i s
co esponding s a a ing. Following es ablished con en ions in he li e a u e,
s a a ings o 1 o 2 we e ca ego ized as nega i e (-1), 3 as neu al (0), and 4
o 5 as posi i e (1). This ans o ma ion enabled supe ised lea ning models
o u ilize a s uc u ed a ge a iable o classi ica ion asks.
To u he p epa e he ex o analysis, a cus om ex cleaning unc ion
was implemen ed. This unc ion in ol ed lowe casing all le e s, emo ing
punc ua ion and special cha ac e s, and elimina ing s opwo ds speci ic o he
Tu kish language. The s opwo d lis was sou ced om NLTK and manually
ex ended o be e sui colloquial usage pa e ns. The cleaned ex was s o ed
in a new column named cleaned_ e iew. To isualize he dis ibu ion o
high- equency e ms a e cleaning, a wo d cloud was gene a ed, o e ing
a compac summa y o dominan exp essions used ac oss p oduc e iews
(Figu e 1: Mos F equen ly Occu ing Wo ds).
Figu e 1: Mos F equen ly Occu ing Wo ds
Re iew Reliabili y Sco ing
To enhance he in e p e abili y and us wo hiness o use -gene a ed
con en , a eliabili y sco e was compu ed o each e iew based on
MUSTAFA ERŞAHİN • UTKU GEZENSOY
73
communi y engagemen me ics. Speci ically, he clap and humbsdown
coun s— ep esen ing posi i e and nega i e use eac ions, espec i ely—
we e combined using a a io-based o mula:
Figu e 2: Reliabili y Sco e Fo mula
The eliabili y sco e RRR, as shown in Figu e 2, was designed o
e lec he social us o each e iew based on posi i e e sus nega i e
use in e ac ions. This a io quan i ies he p opo ion o use s who ound a
pa icula e iew help ul o ag eeable. To ensu e consis en eadabili y, he
esul ing sco e was ounded o he nea es mul iple o 5 (e.g., 73% becomes
75%), p oducing eliabili y alues in inc emen s such as 60%, 70%, o 95%.
Unlike adi ional sen imen analysis echniques ha ely solely on
e iewe inpu , his addi ional me ic in oduces a communi y- alida ed
pe spec i e, en iching he ecommenda ion logic wi h a measu e o e iew
c edibili y. This laye o alida ion is pa icula ly bene icial in il e ing ou
po en ially misleading o low-quali y e iews, as i accoun s o collec i e
ag eemen o disag eemen among pla o m use s. The inal eliabili y alue
is included as a sepa a e a ibu e in he da ase and is displayed alongside
each ma ched e iew du ing keywo d-based que y esul s, p o iding use s
wi h con ex ual insigh in o he social us o each opinion.
Sea ch Que y Ma ching Mechanism
To enable e ec i e e ie al o ele an p oduc e iews, he sys em
inco po a es a ligh weigh , pa e n-based que y ma ching engine. Ins ead
o elying on embedding-based o ans o me -based seman ic simila i y
models, which a e compu a ionally in ensi e and o en language- esou ce
dependen , his s udy adop s a egula exp ession ( egex)-d i en app oach
op imized o Tu kish ex ual inpu .
Use que ies—ph ases such as “ek anı güzel” o “kali eli kumaş”—a e
i s p ep ocessed by con e ing all cha ac e s o lowe case and emo ing
non-alphanume ic symbols. These cleaned que ies a e hen okenized in o
keywo ds. A basic keywo d ex ac o isola es meaning ul e ms, which a e
TRUST-WEIGHTED SENTIMENT FILTERING AND QUERY-BASED RECOMMENDATION FOR TURKISH
74
ma ched agains p ep ocessed e iew ex s s o ed in he cleaned_ e iew
column. The e iew co pus is il e ed using he ollowing ma ching ule: a
e iew is conside ed ele an only i i con ains all he okens om he use
que y. This Boolean AND logic ensu es high-p ecision e ie al, emphasizing
explici seman ic o e lap. Addi ionally, a sen imen cons ain can be applied
(e.g., es ic ing o only posi i e sen imen e iews), as de ined by he ank
ou pu label.
To accoun o lexical a ia ion in Tu kish, egula exp essions a e
ex ended o include common synonyms and mo phological a ian s. Fo
ins ance, he sea ch oken “ek an” may also ma ch “gö ün ü” o “panel”
based on a manually de ined synonym map. This mechanism no only
allows o in e p e able and anspa en esul gene a ion, bu also p o ides
a ounda ion o aspec -based sea ch, whe e use s can e ie e e iews
ha exp ess sen imen owa d speci ic p oduc ea u es a he han gene al
sa is ac ion.
An example o success ul que y execu ion is illus a ed in Figu e 3,
whe e he sea ch o “kali eli kumaş” e u ns 152 ele an e iews anked
by eliabili y.
In con as , Figu e 4 demons a es he sys em’s g ace ul handling
o unma ched que ies—such as “uzay eknolojisi”—by p o iding clea
eedback and e u ning an emp y esul s able ins ead o ailing silen ly o
p oducing un ela ed ou pu .
Figu e 3: Example Que y 1
MUSTAFA ERŞAHİN • UTKU GEZENSOY
75
Figu e 4: Example Que y 2
Recommenda ion Ou pu and Ranking Logic
Following he que y-based e ie al o ele an e iews, he sys em
pe o ms an addi ional anking s ep o guide use s owa d he mos
us wo hy and ep esen a i e con en . This anking is p ima ily go e ned
by he eliabili y sco e, which quan i ies he collec i e use ag eemen on he
help ulness o a e iew based on posi i e (clap) and nega i e ( humbsdown)
eedback. Re iews wi h highe eliabili y a e p io i ized o appea ea lie in
he ou pu lis , enhancing he anspa ency and quali y o ecommenda ions.
Beyond me e p esen a ion, a seconda y op ional p oduc ecommenda ion
mechanism is igge ed when he use ’s que y includes a leas one e m
om a p ede ined p oduc ocabula y (e.g., “ ele on”, “pan olon”, “ able ”).
I such a e m is de ec ed, he sys em in e p e s he que y no only as a
eques o ea u e-aligned e iews, bu also as a p e e ence o disco e ing a
ele an p oduc . In such cases, he sys em iden i ies p oduc s associa ed wi h
high- eliabili y e iews (≥60%) and e u ns one op p oduc pe ma ching
i em, using g oupby logic o p e en epe i ion. This p ocess mi o s ecen
de elopmen s in explainable ecommende sys ems, whe e use que ies a e
mapped o opinion-en iched i ems h ough anspa en ules ins ead o la en
ac o embeddings.
The inal ou pu includes:
• Re iews il e ed by que y keywo d(s) and sen imen class
• O de ed esul s based on e iew eliabili y
• A dis inc ecommenda ion lis (i applicable), whe e one high-con i-
dence e iew pe p oduc is displayed as a sugges ion
This dual-s uc u ed a chi ec u e—combining exac -ma ch e iew
e ie al wi h agg ega ed, us -based ecommenda ion ou pu —enables use s
o ecei e con ex ual insigh s while p omo ing high-quali y use eedback.
Figu e 5 shows he sys em esponse o he que y “ek anı iyi”, whe e he
inpu exp esses a ea u e p e e ence bu lacks a speci ic p oduc class. As a
TRUST-WEIGHTED SENTIMENT FILTERING AND QUERY-BASED RECOMMENDATION FOR TURKISH
76
esul , only ele an e iews a e displayed, and no p oduc ecommenda ion
is made.
In con as , Figu e 6 p esen s he que y “ek anı iyi ele on”, whe e he
inclusion o a known p oduc e m (“ ele on”) igge s he ecommenda ion
logic. The sys em no only e ie es ele an e iews bu also sugges s
speci ic phones ha mee bo h he sen imen and eliabili y h esholds,
demons a ing he adap abili y o he hyb id a chi ec u e.
Figu e 5: Example Que y 3
MUSTAFA ERŞAHİN • UTKU GEZENSOY
77
Figu e 6: Example Que y 4
CONCLUSION AND DISCUSSION
This s udy success ully de eloped a sen imen analysis–based
ecommenda ion app oach ailo ed o Tu kish-language e-comme ce e iews.
The p ep ocessing pipeline included s opwo d emo al, no maliza ion, and
oken-le el cleaning o ensu e linguis ic consis ency. A sen imen label
was hen assigned o each e iew based on s a a ings, enabling sen imen
classi ica ion in o nega i e, neu al, and posi i e classes.
In addi ion o ex ual sen imen indica o s, a eliabili y sco e was
in oduced, calcula ed om he a io o posi i e o o al use eedback (clap
s. humbsdown), he eby inco po a ing communi y consensus in o he
ecommenda ion logic. This us me ic p o ided an addi ional dimension
o il e ing and in e p e ing use e iews, pa icula ly when p esen ing
sugges ions o use -en e ed que ies.
The co e e ie al sys em allowed use s o en e aspec -based sea ch e ms
such as “ek anı iyi ele on” (“a phone wi h a good sc een”), upon which he
algo i hm il e ed and lis ed e iews ha con ained bo h he ele an ea u e
(e.g., “ek an”) and a alid p oduc class (e.g., “ ele on”), anking hem based
on sen imen and eliabili y.
TRUST-WEIGHTED SENTIMENT FILTERING AND QUERY-BASED RECOMMENDATION FOR TURKISH
78
The hyb id use o egula exp ession–based pa e n ma ching and
eliabili y-weigh ed display logic enabled he sys em o mimic a ligh weigh
ecommenda ion engine wi hou elying on adi ional collabo a i e il e ing
o ma ix ac o iza ion me hods. Despi e he lack o a machine lea ning
classi ie a his s age, he p oposed model e ec i ely handled aspec -based
sen imen e ie al in Tu kish, which is pa icula ly aluable gi en he
sca ci y o high-quali y, labeled NLP esou ces o he language.
In conclusion, he p oposed a chi ec u e demons a es a unc ional and
scalable amewo k o ex ac ing ele an , sen imen - ich, and us wo hy
p oduc sugges ions om ee- o m Tu kish e iews. I lays ounda ional
g oundwo k o u u e in eg a ion o mo e ad anced NLP models, such as
ans o me s o supe ised classi ie s, while o e ing immedia e alue in
low- esou ce ecommenda ion scena ios.
he indings o his s udy p o ide meaning ul insigh s in o he p ac ical
deploymen o sen imen -based ecommenda ion sys ems in low- esou ce
languages such as Tu kish. By combining pa e n-based que y logic wi h
sen imen classi ica ion and use -based eliabili y sco ing, he p oposed sys em
b idges he gap be ween compu a ional e iciency and in e p e abili y—
wo o en con lic ing goals in mode n na u al language p ocessing (NLP)
applica ions.
Unlike neu al ne wo k-based models ha demand la ge anno a ed da ase s
and high compu a ional esou ces, ou app oach elies on simple ule-based
echniques—such as Boolean keywo d ma ching and sen imen labeling ia
s a a ings— ha a e bo h scalable and anspa en . This s a egy aligns wi h
ecen calls in he li e a u e ad oca ing o low-complexi y, explainable AI
me hods in eal-wo ld deploymen scena ios, pa icula ly in unde - esou ced
language con ex s .
The use o a de i ed eliabili y sco e based on use engagemen
(claps s. humbsdown) in oduces a no el me ic o quali y con ol in
ecommenda ion sys ems. T adi ional collabo a i e il e ing and ma ix
ac o iza ion app oaches o en lack such di ec quali y signals and depend
hea ily on la en use -i em in e ac ions . In con as , ou amewo k di ec ly
inco po a es pe cei ed use ulness in o anking logic, enhancing use us
and con en ele ance. Mo eo e , he in eg a ion o aspec -based il e ing—
allowing use s o sea ch o ea u e-speci ic exp essions like “ek anı iyi”
(good sc een)—demons a es he easibili y o deploying ine-g ained
sen imen analysis wi hou he need o dependency pa sing o p e ained
ans o me s. This is pa icula ly ele an in languages like Tu kish, whe e
mo phological ichness can hinde gene aliza ion in black-box models .
MUSTAFA ERŞAHİN • UTKU GEZENSOY
79
Despi e i s simplici y, he sys em also suppo s p oduc -le el
ecommenda ions by le e aging a hyb id o con en -based il e ing and ule-
d i en heu is ics. This laye ed ecommenda ion a chi ec u e is inspi ed by
ecen de elopmen s in in e p e able ecommenda ion amewo ks . No ably,
he dual-ou pu s uc u e— e ie ed e iews and agg ega ed p oduc
sugges ions—enhances he sys em’s p ac icali y in e-comme ce scena ios.
Howe e , he app oach is no wi hou limi a ions. The eliance on keywo d
o e lap may es ic ecall in cases o pa aph ased o implici sen imen
exp essions. Fu he mo e, he handc a ed synonym maps, while e ec i e,
a e no exhaus i e and may in oduce bias o omission e o s. Fu u e
i e a ions o he sys em could explo e ligh weigh wo d embedding models
(e.g., Fas Tex ) o complemen he ule-based engine wi hou sac i icing
in e p e abili y .
In conclusion, his s udy demons a es ha e ec i e sen imen -based
ecommenda ion is achie able in Tu kish using ule-based NLP pipelines
augmen ed by use eedback mechanisms. I con ibu es o he ongoing
discou se on building ligh weigh , anspa en , and cul u ally adap i e AI
sys ems o ecommenda ion in low- esou ce en i onmen s.
FUTURE WORK
While he cu en s udy demons a es he iabili y o ule-based sen imen -
d i en ecommenda ion sys ems in Tu kish-language e-comme ce con ex s,
se e al a enues emain open o u he enhancemen .
Fi s and o emos , he inclusion o embedding-based seman ic sea ch
me hods—such as hose u ilizing Fas Tex o BERTu k—could mi iga e
he igidi y o keywo d-based que y ma ching. These models a e capable
o cap u ing lexical and seman ic simila i y e en when use inpu does no
exac ly ma ch e iew ph asing, which could signi ican ly imp o e bo h
ecall and use expe ience in ee- ex que ies .
Second, expanding he synonym mapping s a egy o a da a-d i en,
co pus-based synonym disco e y me hod (e.g., PMI o co-occu ence
s a is ics) would allow o g ea e adap abili y ac oss p oduc domains.
Gi en Tu kish’s agglu ina i e mo phology, s a is ical expansion echniques
o mo phological analyze s like Zembe ek could be employed o mo e
accu a ely de ec s emmed and de i ed o ms . Fu he mo e, he eliabili y
sco e cu en ly uses only bina y eedback (clap/ humbsdown). Fu u e sys ems
could in eg a e empo al da a (e.g., ime-weigh ed eedback), e iewe
me ada a (e.g., e i ied pu chase s a us), o beha io al engagemen me ics
