TYPE O iginal Resea ch
PUBLISHED 12 Ap il 2024
DOI 10.3389/ ma s.2024.1375200
OPEN ACCESS
EDITED BY
Na ayan L.,
Sa ee ha Uni e si y, India
REVIEWED BY
Mahesh Shewale,
ASML, Uni ed S a es
Na een Venka esh S.,
Vello e Ins i u e o Technology (VIT), India
Subash Thanappan,
KAAF Uni e si y College, Ghana
*CORRESPONDENCE
Sachin Saluankhe,
[email p o ec ed]
RECEIVED 23 Janua y 2024
ACCEPTED 21 Ma ch 2024
PUBLISHED 12 Ap il 2024
CITATION
Ja i VS, Sawan DA, Deshpande R,
Saluankhe S, Cep R, Nas EA and
Mahmoud HA (2024), T ibological analysis o
i anium alloy (Ti-6Al-4V) hyb id me al ma ix
composi e h ough he use o Taguchi’s
me hod and machine lea ning classi ie s.
F on . Ma e . 11:1375200.
doi: 10.3389/ ma s.2024.1375200
COPYRIGHT
© 2024 Ja i, Sawan , Deshpande, Saluankhe,
Cep, Nas and Mahmoud. This is an
open-access a icle dis ibu ed unde he
e ms o he C ea i e Commons A ibu ion
License (CC BY). The use, dis ibu ion o
ep oduc ion in o he o ums is pe mi ed,
p o ided he o iginal au ho (s) and he
copy igh owne (s) a e c edi ed and ha he
o iginal publica ion in his jou nal is ci ed, in
acco dance wi h accep ed academic p ac ice.
No use, dis ibu ion o ep oduc ion is
pe mi ed which does no comply wi h
hese e ms.
T ibological analysis o i anium
alloy (Ti-6Al-4V) hyb id me al
ma ix composi e h ough he
use o Taguchi’s me hod and
machine lea ning classi ie s
Vijaykuma S. Ja i1, Dh u A. Sawan 1, Rashmi Deshpande2,
Sachin Saluankhe3*, Robe Cep4, Emad Abouel Nas 5and
Hai ham A. Mahmoud5
1Symbiosis Ins i u e o Technology, Symbiosis In e na ional Deemed Uni e si y, Pune, India,
2Depa men o Ins umen a ion Enginee ing, D Y Pa il Ins i u e o Technology, Sa i ibai Phule Pune
Uni e si y, Pune, India, 3Depa men o Biosciences, Sa ee ha School o Enginee ing, Sa ee ha
Ins i u e o Medical and Technical Sciences, Gazi Uni e si y Facul y o Enginee ing, Depa men o
Mechanical Enginee ing, Mal epe, Tü kiye, 4Depa men o Machining, Assembly and Enginee ing
Me ology, Facul y o Mechanical Enginee ing, VSB-Technical Uni e si y o Os a a, Os a a, Czechia,
5Depa men o Indus ial Enginee ing, College o Enginee ing, King Saud Uni e si y, Riyadh, Saudi
A abia
The p epa a ion and ibological beha io o he i anium me al ma ix (Ti-
6Al-4V) composi e ein o ced wi h ungs en ca bide (WCp) and g aphi e (G p)
pa icles we e in es iga ed in his s udy. The s i cas ing p ocedu e was used o
ab ica e he i anium me al ma ix composi es (TMMCs), which had 8 weigh
pe cen o WCp and G p. The ibological s udies we e designed using Taguchi’s
L27 o hogonal a ay echnique and we e ca ied ou as wea es s using a pin-
on-disc de ice. Acco ding o Taguchi’s analysis and ANOVA, he mos signi ican
ac o s ha a ec wea a e a e load and dis ance, ollowed by eloci y. The
wea p ocess was asce ained by scanning elec on mic oscopy in es iga ion o
he wo n su aces o he composi e specimens. Pea son’s hea map and Fea u e
impo ance (F- es ) we e plo ed o da a analysis o s udy he signi icance o
inpu pa ame e s on wea . Machine lea ning classi ica ion algo i hms such as k-
nea es neighbo s, suppo ec o machine, and XGBoos algo i hms accu a ely
classi ied he wea a e da a, gi ing an accu acy alue o 71.25%, 65%, and
56.25%, espec i ely.
KEYWORDS
i anium me al ma ix composi e, K-nea es neighbo ing, suppo ec o machine,
XGBoos , wea a e, ibology
1 In oduc ion
Ti anium alloys a e widely used in many echnical applica ions due o hei excellen
combina ion o high ha dness, wea esis ance, s eng h, co osion esis ance, s i ness,
and low densi y (Jiaoe al., 2018;Su eshe al., 2018;Cao and Liang, 2020). Ti anium
alloys a e widely u ilized o hei excep ional s eng h- o-weigh a io o lowe ene gy
consump ion, inc ease p oduc i i y, and ex end p oduc li e in he au omo i e, ae ospace,
spo s, anspo a ion, and medical equipmen indus ies (A a e al., 2018;Chaoe al.,
F on ie s in Ma e ials 01 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
2019;Pe undyu aiThanga ele al., 2020). Ti anium me al ma ix
composi es (TMMCs) can be b oadly classi ied in o wo ypes based
on he shape and dis ibu ion o ein o cemen s: con inuously
ein o ced TMMCs and discon inuously ein o ced TMMCs
(Guoe al., 2012;Jiaoe al., 2018;Haya e al., 2019). WC, Al2O3,
TiB, CNTs, SiC, FE3O4, B4C, TiC, G , and o he ce amic
pa icles and whiske s commonly ein o ce TMMCs (Lie al., 2015;
Zhange al., 2016;Sune al., 2020).
F a y e al. (F a ye al., 2003) epo ed ha he mechanical
and physical p ope ies o pu e i anium ein o ced wi h 10
weigh pe cen WP we e compa able o hose o he Ti-6Al-4V
(Ti64) alloy. A s udy by Choe e al. (Choee al., 2005a;Choee al.,
2005b) ound ha he size o he WP signi ican ly in luences
he mechanical cha ac e is ics o WP/Ti composi es. A ange o
mic os uc u es wi h good ensile s eng h and elonga ion we e
p oduced as a consequence o Wang e al.’s in es iga ions o he
a ied size dis ibu ion o TiC ein o cemen o i anium by DED
(Wange al., 2018). Using Ti/B4C composi e powde eeds ocks, Xia
e al. p oduced TiB + TiC ein o ced i anium in si u. They also
ho oughly examined he de elopmen o he in si u mic os uc u e
and he in e ac ion zone be ween he i anium and ce amic
ein o cemen (Xiae al., 2017).
Fu he mo e, in si u B4C/BN ein o ced Ti6Al4V composi es
we e s udied by Gup a e al. The main issue was wea pe o mance,
and s eng hening educed he sliding coe icien o ic ion (COF)
by hal compa ed o he Ti6Al4V ma ix (Gup ae al., 2018).
Using an in-si u app oach, Choi e al. p oduced hyb id composi es
o Ti6Al4V alloy wi h ein o cemen s TiB and TiC ha had a
cons an ein o cemen alloca ion. The esul s showed a signi ican
ela ionship be ween ein o cemen con en and mo e excellen
TMC wea esis ance, wi h mo e ein o cemen con en esul ing
in lowe wea loss (Kime al., 2011). A e elec oless pla ing o
manu ac u e coppe -coa ed CNTs, Wange al. (2017) used spa k
plasma sin e ing p ocedu es o c ea e coppe ma ix composi es.
The esul s showed ha elec oless pla ing signi ican ly imp o ed
he mechanical p ope ies by enhancing he elemen link be ween
coppe and ca bon nano ubes and enabling uni o m dispe sion
o CNTs.
By using elec oless nickel pla ing and SPS o c ea e composi es
o g aphi e lakes and coppe , Ren e al. d ama ically inc eased
he bonding a he g aphi e/coppe base con ac . Acco ding o
he indings, he bending cha ac e is ics and coe icien o he mal
expansion we e signi ican ly imp o ed by ins alling he NieP
ansi ion laye (Chene al., 2017). When 0.35w % o mul i-walled
ca bon nano ubes (MWCNTs) was added, Kondoh e al. disco e ed
ha he ensile pa ame e s, such as s eng h and yield, inc eased
by up o 27% and 42%, espec i ely, in con as o hose o pu e
i anium (Kondohe al., 2008). Wang e al. used an 823K sin e ing
empe a u e in conjunc ion wi h a spa k plasma sin e ing echnique
o c ea e a TMC composi e. The indings demons a ed ha when
he olume ac ion o MWCNTs eached 0.4 weigh pe cen , he
ma e ial’s comp essi e s eng h and yield s eng h bo h a ained
hei maximum alues. Then, when he MWCNT con en was aised
e en u he , he comp essi e s eng h d opped (Wange al., 2015).
Jin e al. s udied he p oduc ion o pu e i anium powde wi h
TiB2 ein o cing pa icles using a selec i e lase mel ing me hod.
All so s o wea cha ac e is ics, including adhesion, ab asion, and
oxida ion, we e imp o ed due o he Ti/TiB2 composi es (Jine al.,
2021). The i anium and Z O2 nanopa icle composi es ha Abd-
Elwahed e al. p oduced we e made ia powde me allu gy. The
esul s demons a ed ha aising he usual load imp o ed he
bonding, wea , and ic ion p ope ies, inc eased he o al amoun
o Z O2nanopa icles, and enhanced he sliding wea a e (Abd-
Elwahede al., 2020). An e al. assessed he ibological beha io s o
TMC coa ings a high empe a u es. The indings demons a e ha
delamina ion, plowing, and oxida ion wea p ocesses occu a 500C
oge he wi h inc eased wea a es o TMC coa ings (AnQie al.,
2019).
Rega ding ha dness and wea esis ance a oom empe a u e,
An e al.’s es s (Ane al., 2018;AnQ.e al., 2019) show ha he
hyb id TiBand TiC pa icles boos ed wi h i anium coa ing and
in e -g ow h ce amic s uc u es ou pe o m he TiB/Ti64 coa ing.
Fa ias concen a ed on how he ibological cha ac e is ics o
he TiC ein o cing pa icles wi h open po osi y in TMCs we e
a ec ed by spa k plasma sin e ing. The s udy’s indings e ealed
ha adding TiC pa icles enhanced ibological cha ac e is ics
like wea esis ance, coe icien o ic ion, and nano ha dness
(Fa íase al., 2019). Insu icien bonding be ween he ein o cemen
pa icles and ma ix migh cause ce amic pa icles o unc ion
as an ab asi e ma e ial, as pe he indings o Ram P abh
e al. The dimensions, mass, and mo phology o he al e na i e
phase ein o ced pa icles, he equi alen ma e ial, he load, he
mic os uc u e, he en i onmen , and he humidi y can all impac
he wea esis ance o composi e ma e ials (Rame al., 2014). Room
empe a u e esea ch was done by Huang Xie, who also ca ied
ou sliding wea ials on TMCs wi h mild s eel g ade 35. I on
oxide addi i es, independen o load, a ely imp o e he TMC’s
wea pe o mance because hey equi e lub ica ing capabili ies. The
MLG/Fe2O3 nanocomposi e and MLG wi h Fe2O3 mechanical
combina ion may signi ican ly imp o e he wea pe o mance
(Xiee al., 2021).
Zhou e al. examined he e ec o g aphene/Fe2O3
nanocomposi es on he ibological pe o mance o he TC11
alloy. They ound ha on he inju ed su ace, a hin, s able double
ibolaye consis ing o laye s p ima ily composed o MLG and
Fe2O3 was c ea ed, signi ican ly educing wea and ic ion. This
esea ch shows ha adding speci ic nanopa icles o ma e ials can
imp o e hei ibological p ope ies (Zhoue al., 2017). Ti anium
hyb id composi es wi h single and mul iple ein o cemen s we e
c ea ed by Lixia Xi e al. and ab ica ed using SLM. The hyb id
composi e esul se be ween he ein o cemen s and ma ix
gene a ed he in e acial s uc u e. The wea esis ance and CoF
cha ac e is ics a e enhanced when TiC and TiN ein o cemen
pa icles a e in oduced (Xie al., 2021).
P akash conside ed emo ing his p ope y lag. Ti anium alloy
(Ti-6Al-4V) is used o o i y bo on ca bide (B4C) ce amic pa icles
using powde me allu gy (PM). As a esul o his esea ch, a
newe composi e was c ea ed and es ed, imp o ing ha dness,
co osion esis ance, and educed densi y. The wea pe o mance
o he composi e specimens is mo e a ec ed by he applied loads
han by he amoun o B4C added. Scanning elec on mic oscopy
esul s demons a e ha he B4C- ein o ced Ti-6Al-4V composi e
has be e wea esis ance han he un ein o ced Ti alloy and
shows signs o mildly wo n su aces (Soo yaP akashe al., 2016).
Hu in es iga ed and analyzed he quali y o he componen s, he
p ocesses by which mic os uc u es o m, and he e icacy o
F on ie s in Ma e ials 02 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
wo kpiece wea in TiB- ein o ced Ti ma ix composi es, which
we e p oduced using he LENS app oach. The esul s sugges
ha TiB-TMCs, wi h hei inno a i e mic os uc u es and TiB
ein o cemen , exhibi ed supe io wea pe o mance compa ed
o bulk componen s composed o comme cially pu e i anium.
Fu he mo e, by a ying he lase powe , he cha ac e is ics o he
p oduced componen s we e imp o ed wi h ewe in e nal de ec s,
leading o be e wea pe o mance (Hue al., 2017). Using i anium
alloy ein o cemen o imp o e he g apheme’s ul ima e comp essi e
s eng h, ensile s eng h, wea esis ance, he mal conduc i i y, and
di usi i y, Gu buz e al. in es iga ed he p ope ies o ibological,
mechanical, and he mal aspec s (Gü büze al., 2021).
Measu ing he pa icula wea a e is c ucial o he ma e ial
selec ion and op imiza ion p ocess. Enginee s can choose he bes
ma e ials o a gi en applica ion by compa ing he wea a es o
se e al magnesium o magnesium alloys wi h o he ma e ials.
This is especially c ucial in sec o s like manu ac u ing, ae ospace,
and au omo i e, whe e wea esis ance signi ican ly impac s he
longe i y and dependabili y o componen s. A i icial In elligence
(AI) has yielded many bene i s and made signi ican s ides
in se e al indus ies, including manu ac u ing and heal hca e.
A i icial in elligence (AI) has shown o be ex emely use ul
in he medical indus y o asks including disease diagnosis,
he apy planning, medica ion de elopmen , and pa ien moni o ing
(Ma hews, 2019;Buccinoe al., 2023a;Buccinoe al., 2023b;
Yange al., 2023). A ype o a i icial in elligence called machine
lea ning algo i hms has been widely used o e alua e complex
medical da a, spo ends, and gene a e p ecise o ecas s. As a esul ,
he e has been an imp o emen in he p ecision o diagnoses, he
c ea ion o cus omized ea men p og ams, and he gene al quali y
o pa ien ca e.
Simila ly, AI has conside ably changed ope a ions in he
manu ac u ing sec o by acili a ing p edic i e main enance,
inc easing p oduc i i y, and s eamlining p ocedu es (Malekie al.,
2022;Malekie al., 2023). Demand o ecas ing, supply chain
op imiza ion, au oma ion, and quali y con ol ha e all bene i ed
om using machine lea ning algo i hms. These applica ions ha e
imp o ed ope a ional e ec i eness, lowe ed p oduc ion cos s,
and imp o ed p oduc quali y. The in eg a ion o E olu iona y
Compu ing wi h Machine Lea ning algo i hms has ecei ed e y
li le esea ch a en ion despi e he ema kable ad ancemen s in
AI and Machine Lea ning. E olu iona y compu ing is a sub ield
o a i icial in elligence ha uses me hods om na u al e olu ion,
including gene ic algo i hms, pa icle swa m op imiza ion, and
an colony op imiza ion, o ackle challenging op imiza ion
issues. Combining machine-lea ning echniques and e olu iona y
compu ing has excellen p omise in se e al ields. Resea che s
can mo e e ec i ely handle complex op imiza ion and p edic ion
asks by combining he adap i e sea ch capabili ies o E olu iona y
Compu ing wi h he lea ning and p edic i e powe s o Machine
Lea ning algo i hms. This in eg a ion can be u he implemen ed
in a ious ma e ial science and manu ac u ing domains
(Dhunganae al., 2019;Dhunganae al., 2021;Sadeke al., 2021;
Dhunganae al., 2022;Mish a and Ja i, 2023a;Dhunganae al.,
2023;G ecoe al., 2023). This hyb id echnique can help ind he
bes solu ions in complex p oblem spaces, inc ease he p ecision
and e ec i eness o op imiza ion algo i hms, and imp o e ea u e
selec ion in machine-lea ning models.
Es ima ing he p ecise wea a e o Hyb id Me al Ma ix
Ti anium alloy is essen ial in sec o s whe e wea esis ance is
a i al componen . While machine lea ning has demons a ed
po en ial in p edic i e modeling, e alua ing hei pe o mance
and choosing which algo i hm wo ks bes o his pa icula
use case is necessa y. Fu he mo e, he e needs o be a mo e
ho ough analysis and compa ison o a ious algo i hms and a
pauci y o esea ch in his a ea. Thus, compa ing he e ec i eness
o machine lea ning algo i hms, namely K-Nea es Neighbou ing
(KNN), Suppo Vec o Machine (SVM), and XGBoos classi ica ion
in o ecas ing he p ecise wea a e o he Hyb id Me al Ma ix
Ti anium alloy is he opic his esea ch s udy a emp s o add ess.
This s udy used he s i cas ing me hod o c ea e a i anium alloy,
Ti-6Al-4V, wi h ein o cemen s made o ungs en ca bide (WC) and
g aphi e (G ) hyb id me al ma ix composi e. P ocess a iables like
load, sliding eloci y, sliding dis ance, and ibological expe imen s
we e pe o med based on he Taguchi L27 o hogonal a ay.
2 Ma e ials and me hods
The p ima y ma ix ma e ial used was he i anium alloy Ti-
6Al-4V; Table1 shows he chemical composi ion o he ma ix
alues o his alloy. The pa icles ein o ced wi h WC and G ha e
been selec ed. The a e age size o he g aphi e pa icles was 25μm,
whe eas he WC pa icles we e 45μm. The equi ed amoun o Ti-
6Al-4V i anium alloy was mel ed in a g aphi e c ucible using an
elec ical u nace. The ein o cing pa icles we e hea ed o 500°C o
emo e he mois u e. A speci ic quan i y o ein o cing pa icles was
mixed wi h he i anium alloy. The hyb id composi e ma e ial was
egula ly blended. A e he hyb id composi e was inse ed in o he
p epa ed die a 800°C, i was le o solidi y a oom empe a u e.
The sample’s mic os uc u es and wo n su aces we e examined
using op ical and scanning elec on mic oscopy. Mic oha dness
es e o model-Mic oha dness es e om OMNI ech, MVH-1
au oma ic es load 10gm o 1,000gm employed o mic oha dness
measu emen . As indica ed in Figu e1, he composi e specimens’
d y sliding wea quali ies we e e alua ed using he DUCOM pin-
on-disc sliding wea es ing appa a us (Manu ac u e : DUCOM,
Bangalo e, India). The ASTM G99-95 ules we e ollowed when
conduc ing he d y sliding wea es ing. The pin was cleaned wi h
ace one, and i s ini ial mass was measu ed wi h a digi al elec onic
balance. A e ha , he pin was held up agains a e ol ing EN-32
s eel disk (coun e ace) wi h 65 HRC ha dness h oughou he es .
Figu e2A,B depic s he cas ed and machined samples, espec i ely.
Th oughou he es ing, adjus men s we e made o he dis ance,
eloci y, and a e age load. A he end o each es , he pin’s ul ima e
mass was measu ed a e being cleaned wi h ace one. We calcula ed
he mass loss o he pin due o sliding wea by aking he di e ence
be ween i s ini ial and inal masses. The olume loss owing o wea
was calcula ed using he densi y alues linked wi h he pin. Nex , he
wea a e o he composi e pins was asce ained.
The ollowing is he p ocess o ca ying ou he wea es :
Fi s , he es sample is ca e ully weighed on a s a e-o - he-a
digi al balance, and i s o iginal mass is eco ded. The specimen is
hen secu ely secu ed using he no ch, and i s su ace is posi ioned so
ha i makes con ac wi h he disk. The ack adius is hen modi ied
o mee he unique equi emen s o he es . Following p ope
F on ie s in Ma e ials 03 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
TABLE 1 Chemical composi ion o Ti6Al4V.
Alloying elemen s Ti Al V Fe O N C
Chemical con en (w %) 85.096 7.75 6.5 0.34 0.02 0.04 0.05
FIGURE 1
Wea es on disc (Mish a and Ja i, 2023b;Mish ae al., 2023)
FIGURE 2
(A) Cas wea pins. (B) Machined wea pins.
specimen posi ioning speci ied no mal loads a e supplied, and he
sliding eloci y is se in compliance wi h he es pa ame e s. The
es is hen un o co e he speci ied dis ance o e a compu ed ime
in e al. Fo e e y es , he pin olume loss was de e mined using
he pin-heigh loss me hod. Each es was conduc ed h ee imes o
gua an ee epea abili y, and he a e age o he h ee es s was used
o de e mine he wea a e using Eq.(1). This p ocess is epea ed
o o he specimens wi h a ying olume pe cen ages and could be
F on ie s in Ma e ials 04 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
TABLE 2 Inpu p ocess pa ame e s.
Le el Sliding
speed (m/s)
Load (N) Sliding
dis ance (m)
1 2 20 500
2 4 30 1,000
3 6 40 1,500
es ed unde a ious condi ions. This me hod makes i possible o
compa e wea cha ac e is ics unde a ious ci cums ances.
The wea pa ame e s chosen o he es ing based on machine
capaci y, li e a u e analysis, and pilo ials a e shown in Table2.
Pilo expe imen s we e conduc ed o asce ain he p ac ical limi s o
he p e iously indica ed pa ame e s necessa y o he wea o occu
in a s eady s a e. Re e ing o ASTM G99-95, he pin used in he wea
es is 30mm long and has an 8mm diame e .
Wea a e(mm3/Nm)=(Volumeloss∗Ha dness)/
(No malLoad∗Slidingdis ance)(1)
The Taguchi echnique aims o minimize a ia ion in a p ocess
h ough obus expe imen design. The main objec i e o he
p ocedu e is o p o ide he make wi h high-quali y ou pu a a low
cos . D . Genichi Taguchi o Japan de eloped he Taguchi me hod
and has pe sis ed in using ha a ia ion. Thus, bo h he p oduce
and socie y a e impac ed by low p ocess quali y. He de eloped
a sys em o designing expe imen s o in es iga e how di e en
pa ame e s a ec he mean and a iance o a p ocess pe o mance
cha ac e is ic ha shows how well he p ocess is doing.
Taguchi’s expe imen al design ga he s he necessa y da a o
iden i y he a iables signi ican ly in luencing p oduc quali y
wi h mino expe imen a ion, sa ing ime and esou ces. This is
accomplished using o hogonal a ays o o ganize he a iables
in luencing he p ocedu e and he magni udes a which hey should
be shi ed. Key p ocess ac o s we e iden i ied using analysis o
a iance. An L27 o hogonal a ay was chosen o he cu en
expe imen , as indica ed in Table3.
This s udy classi ies and p edic s he wea a e o hyb id
me al ma ix composi e using KNN, SVM, and XG Boos machine
lea ning classi ica ion echniques. To build a con usion ma ix and
AUC-ROC cu es o u he in-dep h analysis and o p ecisely
assess he wea a e using classi ica ion, sample da a was cons uc ed
using a syn he ic da a gene a ion ool in MATLAB based on
he expe imen al da a. The classi ica ion me hod is a supe ised
lea ning echnique ha ca ego izes new obse a ions using aining
da a. Using he da ase o supplied obse a ions, a p og am lea ns
how o ca ego ize esh obse a ions in o a ious classes o g oups
in he classi ica ion p ocess. The da a in he cu en s udy is di ided
in o wo g oups acco ding o whe he he alue exceeds o alls sho
o he 0.084 a e age o all wea a e alues. The con usion ma ix
was plo ed using he Py hon me ics module om he sklea n
package. In o de o achie e p ecise ou comes o da a p edic ion,
he da ase was spli in o wo sec ions: 80 pe cen o aining da a and
20 pe cen o andom es da a. The con usion ma ix is composed
o h ee ypes o da a: False posi i es (FP) and False nega i es (FN),
which e lec w ongly o ecas ed da a o sugges ha he e was an
e o in he p edic ion p ocess; and T ue posi i es (TP) and T ue
nega i es (TN), which ep esen success ully an icipa ed da a. K-
old (k = 1) was conside ed o he numbe o neighbo s (k), and
KNN, SVM, and XGBoos classi ica ion was pe o med o ob ain
a wide ange o classi ica ion da a. Pea son’s hea map analysis and
ea u e impo ance plo (F- es ) we e plo ed u he o unde s and
he signi icance o ea u es on wea loss. A lowcha explaining
hea map analysis is depic ed in Figu e3.
The cu en s udy uses a k-nea es neighbo (kNN) classi ica ion
me hod, which inds he closes Euclidean dis ance be ween he
a e age alue and each wea a e alue ( e e o Figu e4A).
Fo he kNN classi ica ion me hod, he e a e a ious ypes o
hype pa ame e op ions, such as:
a) The numbe o neighbo s decides he numbe o nea es
neighbo s o classi y each alue o poin in he a ge da ase .
b) Dis ance me ic, which is used o measu e dis ance be ween
2 poin s.
c) Dis ance weigh decides whe he he dis ance is equal o
in e se (1/dis ance).
SVMs we e c ea ed o sol e bina y classi ica ion issues. Howe e ,
when compu a ionally demanding mul iclass p oblems become
mo e common, se e al bina y classi ie s a e buil and coupled o
c ea e SVMs ha can ca y ou hese mul iclass classi ica ions using
bina y me hods. The SVM classi ie unc ion (SVC) is de ined
using inpu pa ame e s like he ype o ke nel used, he ma gin,
and he hype plane. Once he da a is ained using he SVM
classi ie unc ion, he es da a is p edic ed based on he ained and
alida ed da a. The wo k low o he SVM classi ica ion algo i hm is
depic ed in Figu e4B. XGBoos is a scalable and accu a e g adien -
boos ing solu ion ha pushes he compu a ional bounda ies o
boos ed ee algo i hms, p ima ily accele a ing compu a ional speed
and machine lea ning model pe o mance. As pa o an ensemble
app oach mean o p oduce supe io p edic ions wi h imbalanced-
class da a, Ex eme G adien Boos ing, o XGBoos , has become
inc easingly popula as a p edic ion algo i hm in ecen yea s.
XGBoos classi ica ion is simila o he F- es as i selec s he bes
ea u es o p edic he da a a e i ains and alida es he aining
da ase . I con inues un il he bes ea u e, which has he mos
signi icance on he da a, is selec ed, and no ea u es emain in
he da ase o e alua e. The wo k low o he XGBoos classi ica ion
algo i hm is depic ed in Figu e4C. The alues a e eco ded a e he
da ase is collec ed using wea loss expe imen s. Hea map analysis
and F- es a e done o iden i y he mos signi ican ea u es o he
da ase . Then, supe ised machine lea ning algo i hms, in his case,
classi ica ion algo i hms like kNN, SVM, and XGBoos , a e used o
classi y he da a based on he aining and alida ion da ase , and
he da a is p edic ed using he ained model. The esul s a e hen
analyzed o he bes wea loss a e based on he bes ea u es.
3 Resul s and discussions
In o de o es ima e he p ecise wea a e o hyb id composi es,
his sec ion p esen s he wea a e esul s oge he wi h s a is ical
analysis and a machine lea ning echnique. The expe imen al
pa ame e s used o he s udies a e shown in Table3, oge he wi h
he co esponding obse ed wea a e alues.
F on ie s in Ma e ials 05 on ie sin.o g

Ja i e al. 10.3389/ ma s.2024.1375200
TABLE 3 Expe imen al layou wi h obse ed alues.
S. No. Load (N) Veloci y (m/s) Dis ance (m) Wea a e (mm3/N-mm)
1 20 2 500 0.00021959
2 20 2 500 0.00026440
3 20 2 500 0.00044366
4 20 4 1,000 0.00018150
5 20 4 1,000 0.00009635
6 20 4 1,000 0.00015909
7 20 6 1,500 0.00056914
8 20 6 1,500 0.00007618
9 20 6 1,500 0.00004332
10 30 2 500 0.00014341
11 30 2 500 0.00028083
12 30 2 500 0.00024797
13 30 4 1,500 0.00007569
14 30 4 1,500 0.00006075
15 30 4 1,500 0.00006672
16 30 6 500 0.00017328
17 30 6 500 0.00024498
18 30 6 500 0.00023602
19 40 2 1,500 0.00006797
20 40 2 1,500 0.00005676
21 40 2 1,500 0.00007618
22 40 4 500 0.00023751
23 40 4 500 0.00011876
24 40 4 500 0.00014341
25 40 6 1,000 0.00010531
26 40 6 1,000 0.00009635
27 40 6 1,000 0.00004369
3.1 S a is ical analysis
The d y sliding wea es was pe o med using pin-on-disc
equipmen . The analysis o a iance and signal- o-noise (S/N) a io
echniques we e used o de e mine he signi ican pa ame e s. The
S/N a ios a e used o e alua e how noise ac o s a ec pe o mance
me ics. Th ee S/N a ios a e ypical and equen ly u ilized; hey
e alua e he deg ee o a ia ion in he answe da a and he deg ee
o which he a e age esponse esembles he a ge . Theo e ically,
highe , smalle , and be e a e he be e o hem. Smalle is
be e . The guideline was applied in his s udy o educe wea a e.
The signal- o-noise a io, o S/N a io, gauges he suscep ibili y
o he quali y a ibu e unde s udy o expe imen ally induced
uncon ollable e en s.
F on ie s in Ma e ials 06 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
FIGURE 3
Hea map lowcha .
FIGURE 4
(A) kNN classi ica ion lowcha . (B) SVM classi ica ion lowcha . (C) XG-Boos classi ica ion lowcha .
TABLE 4 Analysis o a iance o wea a e.
Sou ce DF Seq SS Adj SS Adj MS F P
Load (N) 2 0.0000001 0.0000001 0.0000001 3.31 0.057
Veloci y (m/s) 2 0.0000001 0.0000001 0.0000001 0.49 0.622
Dis ance (m) 2 0.0000001 0.0000001 0.0000001 3.35 0.056
E o 20 0.0000002 0.0000002 0.0000002
To al 26 0.0000004
An analysis o a iance (ANOVA) was pe o med o in es iga e
he e ec s o load, eloci y, and dis ance on he wea a e;
he esul s a e shown in Table4. The load and he dis ance
subs an ially impac ed he wea a e, acco ding o he F- es and
p- alue esul s. Table5 and Table6 depic he esponse able
o signal- o-noise a ios and means, espec i ely. As pe he
esul s, he load ollowed by dis ance and eloci y a ec s he
wea a e.
Figu e5A displays he plo o he wea a e’s p ima y in luence. I
is e iden ha he wea a e ends o inc ease wi h highe eloci ies
and no mal loads. Fu he mo e, Figu e5B illus a es he wea a e
in e ac ion plo . The lines in e sec o e eal a s ong in e ac ion
e ec be ween he wea a e and he load and eloci y, load and
dis ance, and eloci y and dis ance.
Figu e5C displays a esidual plo o he pa icula wea a e.
A s anda d p obabili y plo deno es a no mal dis ibu ion o he
F on ie s in Ma e ials 07 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
TABLE 5 Response able o signal- o-noise a ios.
Le el Load (N) Veloci y (m/s) Dis ance (m)
1 71.94 75.31 72.70
2 76.40 78.33 78.88
3 79.96 74.66 78.78
Del a 8.02 3.67 6.18
Rank 1 3 2
TABLE 6 Response able o means.
Le el Load (N) Veloci y (m/s) Dis ance (m)
1 0.000228 0.000200 0.000229
2 0.000170 0.000127 0.000114
3 0.000105 0.000176 0.000116
Del a 0.000123 0.000073 0.000116
Rank 1 3 2
esiduals. The e was no e idence o da a skewness o ou lie s in
he his og am plo . The e we e no appa en pa e ns in he esidual
e sus o de ed plo o he esidual e sus i ed plo . These indings
imply ha di e ences in ime o en i onmen al condi ions did no
cause any inaccu acies in he da a ga he ing.
The wea a e indica es he amoun o ma e ial loss o
wea olume pe uni o sliding dis ance and uni load. Se e al
wea egimes we e ound in he wea map by analyzing he
wea a e alues. A di e en combina ions o sliding eloci ies
and no mal loads, hese wea egimes o e insigh s in o he
p edominan wea mechanisms, including adhesi e wea , ab asi e
wea , delamina ion, plas ic de o ma ion, oxida ion, and mel ing.
Changes in wea a e o e ela i ely modes di e ences in
pa ame e s like a e age load, sliding eloci y, empe a u e, and ime
a e e e ed o as wea ansi ions. Wea ansi ion cha s usually
iden i y and cha ac e ize a ious wea egimes o mechanisms. Low
wea a es and mild con ac condi ions wi h a mix o adhesi e
and oxida i e wea p edomina e a e cha ac e is ics o he mild
wea egime. On he o he hand, he se e e wea egime is
cha ac e ized by ele a ed wea a es and usually a ises om mo e
igo ous wo king ci cums ances. Nume ous mechanisms, including
ab asi e wea , plowing, and delamina ion, can con ibu e o
se e e wea .
Plowing emo es ma e ial due o he in e ac ion o mic oscopic
aspe i ies on one su ace wi h ano he , which causes ab asi e wea .
Rough su aces o he p esence o ha d pa icles a e linked o his
mechanism. Oxida i e wea is caused by he in e ac ion o he
alloy wi h ambien oxygen, p oducing oxide laye s on he alloy
su ace. Inc eased wea a e and su ace de e io a ion may esul
om his mechanism. The e m “delamina ion” desc ibes how laye s
FIGURE 5
(A) Main e ec s plo o wea a e. (B) In e ac ion plo o wea a e. (C)
Residual plo s o wea a e.
come away om he su ace o a subs ance. I equen ly happens
in a eas wi h a concen a ion o localized ension, which causes
wea deb is and su ace oughness o accumula e. This phenomenon
is known as plas ic de o ma ion, when a ma e ial de o ms and
lows due o an applied a e age load. A wo-dimensional g aph
o he wea a e alues is displayed in Figu es6A–C. The colo
ep esen ed he wea a e alues. This image made he analysis
o wea a e a ia ions unde a ious es se ings possible. These
wea maps help de e mine p e ailing wea egimes, comp ehend
he wea mechanisms, and choose he bes ma e ials and ope a ing
se ings o speci ic applica ions. This in es iga ion ound ha he
wea a e dec eased a lowe eloci y and load alues and inc eased
F on ie s in Ma e ials 08 on ie sin.o g
Ja i e al. 10.3389/ ma s.2024.1375200
FIGURE 6
(A) Con ou plo o wea a e e sus load, eloci y. (B) Con ou plo o
wea a e e sus load, dis ance. (C) Con ou plo o wea a e e sus
eloci y, dis ance.
a highe eloci y and load alues. The wea map’s a ied colo ed
zones co espond o dis inc wea a es a a ying eloci ies and
loads. The ansi ion lines we e used o de ine he zones by he
expe imen al se up.
The ideal eloci y and load combina ion ha educes wea a e
can be ound h ough wea map analysis, making i possible o
iden i y ope a ing si ua ions whe e he ma e ial exhibi s excellen
wea pe o mance. This knowledge is a e e ence o echnical
applica ions using he hyb id me al ma ix composi e alloy. This is
especially impo an in he au omo i e, ae ospace, and biomedical
indus ies, whe e wea esis ance is c i ical. Fo he p esen s udy, he
wea a e is lowe be ween 3.5 and 4.5m/s eloci y and 25N–35N
load; see Figu es6A,B shows ha he wea a e is lowe be ween
27.5 N and 40N load and 1,000m–1,500m dis ance. Figu e6C
shows ha he wea a e is lowe be ween 2 and 4.5m/s eloci y and
1,250m–1,500m dis ance.
3.2 Op ical and SEM analysis
The op ical mic og aph o he composi e is shown in Figu e7A.
The sample has excellen compac ness and is ee o mic o- issu es
and po es. The WCp-G p ein o cemen was e enly dis ibu ed
h oughou he composi es o pe mi i s p esence in he Ti-6Al-4V
ma ix. The Ti-6Al-4V ma ix has uni o mly dis ibu ed iny WCp-
G p ein o cemen s, e en hough Ti anium Composi es In e ac ion
does no obse e any eac ions. I is impo an o no e ha
WCp and G p do no combine o o m a s aigh o wa d bina y
combina ion bu p oduce a new in e ace s uc u e. The a e age
alue o Vicke s Mic oha dness o he hyb id composi e ob ained
using a mic oha dness es e is 892 VHN.
SEM analysis o he wea su aces de eloped in o d y sliding
wea in he s eady s a e egime p o ides a c ucial ool o accu a ely
cha ac e izing he wea beha io o he composi es. Figu e7B shows
he composi es’ wo n-ou su ace ollowing wea . The ex ao dina y
ha dness o he composi e means ha he wo n-ou su aces a e
ba ely pe cep ible. This example has a e y smoo h su ace because
he WCp and GRp ein o cemen pa icles a e secu ely bonded o
he ma ix phase a ha le el. I is also clea ha ein o cemen s
ha e no wo n down only a li le. The sel -lub ica ing ac ion o
he ibo su ace ein o cemen s causes his. The wo n su ace
o he composi e makes he p esence o lamina ed laye s qui e
e iden . In his pic u e, he laye has changed he easily obse able
sliding su ace. The su aces also appea smoo h because o he
ein o cing componen .
3.3 Wea a e p edic ion using machine
lea ning classi ie
Da a was ca ego ized using machine lea ning based on he
ape angle o he squa e slo s in he s ainless s eel pla e. Using
Py hon lib a ies, Pea son’s hea map and F- es plo s we e c ea ed
o de e mine he ea u e impo ance o inpu pa ame e s, including
load, eloci y, and dis ance. Pea son’s hea map analysis is plo ed as
shown in Figu e8A, which depic s he minimal signi icance o load
and eloci y on he wea a e.
As shown in Figu e8B, a ea u e is deemed insigni ican i i s
F- es alue is below he F-dis ibu ion alue. Howe e , any F- es
alue o an inpu pa ame e o e he c i ical F-dis ibu ion will be
ega ded as a signi ican ea u e o inpu pa ame e . The k-nea es
Neighbo s (kNN) me hod p edic s he label o alue o a new da a
poin by conside ing he labels o alues o i s k-nea es neighbo s
in he aining da ase . In he cu en analysis o wea loss, he
alues below he a e age alue o wea loss we e conside ed as 1,
and he alues abo e he a e age alue we e conside ed as 0, e e o
Figu e9A. Wea loss should be minimal as i will gi e be e esul s.
The p edic ion accu acy o he kNN classi ica ion o wea loss was
71.25%, e e o Table7.
F on ie s in Ma e ials 09 on ie sin.o g