G aphical Abs ac
A Hyb id WAAM Moni o ing Sys em o Co ela e Elec ical Sig-
nals wi h he Con ac -Tip-Wo king-Dis ance
Paul D. Rose o-Mon al o, Ma in Ma inez-Bal a , Roi M´endez-Rial, F´elix
Vidal-Vila i˜no
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
A Hyb id WAAM Moni o ing Sys em o Co ela e
Elec ical Signals wi h he
Con ac -Tip-Wo king-Dis ance
Paul D. Rose o-Mon al oa, Ma in Ma inez-Bal a a, Roi M´endez-Riala,
F´elix Vidal-Vila i˜noa
aSma Sys ems and Sma Manu ac u ing G oup, AIMEN Technological
Cen e , Pol´ıgono Indus ial SUR-PPI-2, Po i˜no, 36418, Pon e ed a, Spain
Abs ac
Wi e a c addi i e manu ac u ing u ilizes an elec ic a c as a hea sou ce o
deposi me al wi e laye by laye , enabling he p oduc ion o la ge and com-
plex me al componen s wi h educed ma e ial was e and lead ime. Howe e ,
his p ocess needs high ol age and cu en low o keep he a c on mel ing
he elec ode, which is uns able since i injec s noise in o he elec ical sig-
nal. Con e sely, he wi e a c sp eads ays in se e al wa eleng hs, blinding
mos came as. The e o e, his wo k aims, on one side, an elec ical signal
analysis mus be done o de ec ou lie s ha could a ec he weld pool a
his heigh o wid h. On he o he side, a welding came a could de ec he
con ac - ip-wo king-dis ance o he wi e o co ela e wi h he elec ic pulses,
alida ing ou lie s. As a main esul , he hyb id moni o ing sys ems could
check whe e he welding machine ails when i is welding he ma e ial.
Keywo ds: waam, moni o ing sys em, addi i e manu ac u ing, senso da a,
ML models
1. In oduc ion
Pe sonaliza ion o p oduc s wi h changing cus ome equi emen s a ec s
p oduc li ecycles, o cing manu ac u ing companies o sho en p oduc de-
elopmen amp-up and indus ializa ion pe iods. Mo eo e , Manu ac u -
ing companies ha e shi ed o a ci cula manu ac u ing s a egy o in e-
g a e p oduc s’ li ecycles, making hem g een and sus ainable, suppo ing
de-manu ac u ing ope a ions and op imiza ion, and c ea ing a mo e sus-
P ep in submi ed o Addi i e Manu ac u ing Le e s Oc obe 22, 2024
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
ainable p oduc design o build a sus ainable and esponsi e p oduc ion
en i onmen [1, 2]. These ac o s pose con inuous challenges o p oduc de-
elope s and manu ac u ing companies in s aying compe i i e in he global
ma ke , demanding p oduc and p oduc ion line design op imisa ion [3]. In
his scena io, addi i e manu ac u ing (AM) is a well-known echnique in he
indus ial sec o . I is a cus omized pa ab ica ion by adding laye s wi h
indus ial ma e ials [4]. I is ime-e icien by educing sc ap and elimina ing
supply chain managemen and negligible ooling equi emen s. Fu he mo e,
AM signi ican ly educes ene gy o uel wi h less ca bon p in and g een-
house gases wi h a iendly use in e ac ion be ween he obo ha will do
he ask and he compu e -aided design (CAD) o diminish su ace e o s [3].
Unde AM s anda ds, wi e a c addi i e manu ac u ing (WAAM) u ilizes an
elec ic a c as a hea sou ce o deposi me al wi e laye by laye , enabling
he p oduc ion o la ge and complex me al componen s wi h educed ma e-
ial was e and lead ime [5]. In mul i-laye wall geome ies, he cold me al
ans e (CMT) p ocess migh be he sui able ab ica ion con igu a ion since
i p o ides he bes deposi ion con ol, leading o ine weld beads and be -
e su ace quali y [4]. Howe e , hese bene i s come wi h challenges. The
lowe hea inpu could esul in insu icien usion be ween laye s, leading o
po en ial de ec s such as lack o pene a ion o bonding issues [6]. Addi ion-
ally, he educed hea can exace ba e esidual s esses and mic os uc u al
a ia ions, pa icula ly in hicke componen s, a ec ing he inal p oduc ’s
mechanical p ope ies and dimensional accu acy [7]. Elec ically, he CMT
p ocess ies o main ain s able ol age and cu en pulses o es ablish he
a c be ween he me al wi e and he mel pool. Howe e , he success o his
indus ial p ocess depends on in e nal ac o s, such as he wi e speed, he
laye -by-laye wai ing ime and he mel ed ma e ial [8]. Also, ex e nal pa-
ame e s, such as mel pool empe a u e, humidi y and empe a u e, could
injec noise in o he con ol sys em and deg ade he quali y o he me al piece
buil [9].
In a WAAM p ocess wi h a CMT con igu a ion, he con ol sys em usu-
ally uns in o he welding machine, and he a icula ed obo execu es he
designed ajec o ies; bo h a e in con inuous communica ion whe e an ope -
a o moni o s he p ocess [4]. Howe e , o he ope a o , he con ol sys em
is a black box whe e i does i s bes o only see he in e nal pa ame e s,
and is unawa e o he ex e nal condi ions [10]. The e o e, ex e nal moni o -
ing sys ems could help o p o ide he con ol sys em wi h ex a in o ma ion
abou he p ocess o make be e con ol decisions. Fu he mo e, a new con-
3
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
ol sys em buil on he op o he obo ’s ajec o ies could de e mine i he
laye is done co ec ly o wi h de ec s o ec ea e new ajec o ies o pa ch
he mel ing issues [7]. In his scena io, senso s could be added o he WAAM
sys em o collec da a om he CMT con igu a ion (elec ic signals), and a
came a could de ec bias in he laye s [6]. To ma ch bo h signals and co e-
la e hei da a, he came a could be ocused on he con ac - ip-wo k-dis ance
(CTWD), which is he dis ance be ween he con ac ip o he welding o ch
(whe e he wi e exi s) and he su ace o he wo kpiece being welded o ad-
di i ely manu ac u ed [11]. This c i ical pa ame e in luences he welding
cu en , a c s abili y, and o e all hea inpu . A sho e CTWD can inc ease
he welding cu en and hea inpu , while a longe CTWD can educe hem,
po en ially leading o issues such as poo pene a ion o uns able a c condi-
ions.
Consequen ly, his wo k aims o de elop a hyb id moni o ing sys em ha
co ela es he elec ical signal o he welding machine wi h he o ch be-
ha iou (CTWD). The e o e, a wo-channel oscilloscope is selec ed o ga he
elec ical da a (i.e., ol age and cu en pulses) and a welding came a o de-
ec and measu e he CTWD [12]. Then, an unsupe ised machine-lea ning
echnique is applied o ecognize he cu en pulse ha is classi ied as an
ou lie o no a e wa ds. On he o he hand, he welding came a ook
ames om he WAAM p ocess, whe e an objec de ec ion model ecog-
nizes he CTWD and con e s i in o a wi e measu emen o co ela e wi h
he elec ical signal and double-check he ou lie [13]. As a main esul , he
unsupe ised algo i hm could de ec ou lie s, and hey we e con i med by
he CTWD de ec ed by YoloV9 nano, demons a ing a s ong ela ion o
u he con ol sys em deploymen . The es o he manusc ip is s uc u ed
as ollows: Sec ion II shows he backg ound and he li e a u e. The me hod-
ology is p esen ed in sec ion III. The Resul s a e shown in sec ion IV, and
hei discussion is gi en in sec ion V. Finally, conclusions and u u e wo ks
a e desc ibed in sec ion VI.
2. Backg ound and li e a u e e iew
Gi en his scena io, se e al wo ks ha e been p esen ed da a acquisi ion
sys ems o imp o e he WAAM p ocess. Consequen ly, Xiong e al [14].,
p esen ed an ea ly wo k abou adding a passi e ision-sensing sys em o de-
ec he bead heigh and wid h h ough he deposi ion o hin walls. Then,
Li e al. demons a ed a he mal beha iou analysis based on ini e ele-
4
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
men s analysis o de ine a sui able laye unde pass ime. Con e sely, Yan
e al [15]. designed one o he i s app oaches o using neu al ne wo ks, es-
pecially in he addi i e manu ac u ing p ocess. Following his end, wo ks
such as [16, 9, 7, 12, 17] p esen ed moni o ing sys ems o collec da a om
he WAAM p ocess and ex ac knowledge by using ML me hods and im-
p o e he sys em pe o mances. Fu he mo e, [6, 18] also added a lexible
con ol sys em ha enhances i s pe o mance by senso in o ma ion. Nowa-
days, wo ks such as Holsche e al [19]., p esen ed a no el CTWD de ec ion
in a WAAM p ocess wi h a closed-loop laye heigh con ol. Fo i s pa ,
No elino e al [11]. show how he CTWD in luences he wall geome y o an-
alyze he in luence o he pa ame e s in bead and mul i-laye wall geome ies
ab ica ed by he CMT p ocess o selec he con igu a ions ha esul in he
bes deposi ion con ol. Finally, F anke e al [20]., p esen ed a ision-based
p ocess moni o ing sys em, and image segmen a ion o he welding wi e is
used o moni o he wo king dis ance as well as he ho izon al posi ion o
he wi e du ing welding, and classic image p ocessing echniques a e applied
o cap u e spa e o ma ion. Howe e , e en when hose wo ks al eady de-
ec and apply he CTWD in di e en con ol s a egies, co ela ing wi h
he elec ic signals ha ep esen he welding machine con igu a ion is s ill
an open challenge. This in o ma ion could p o ide insigh s in o new con ol
echniques and a new senso con igu a ion o imp o e he WAAM p ocess.
3. Ma e ials and Me hods
This sec ion shows he es bench se up wi h all ha dwa e speci ica ions o
each componen . The ollowing sec ion de ails he welding machine y se up.
Nex , he cu en pulse de ec ion is p esen ed wi h i s s eps, s a ing om
he aw da a o de ec ou lie s. Las ly, he CWTD de ec ion is desc ibed,
especially he image ans o ma ion in o a dis ance measu emen .
3.1. Tes bench se up
The WAAM cell comp ises an ABB obo and a welding machine o make
weld beads. As he moni o ing sys em, one oscilloscope is placed nea he
welding machine o collec elec ical pa ame e s, such as he ol age and
he cu en . The e o e, a ol me e and ampe ome ic clamp a e needed.
Fu he mo e, a welding came a is placed in on o he welding o ch o
cap u e ames o he CTWD. The Edge se e synch onises he Robo ’s
ajec o ies wi h hei axis o a ions, he welding pa ame e s con igu ed, he
5
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
ol age and cu en om he oscilloscope, and akes ames om he welding
came a. Then, aw da a is s o ed in .cs iles and ames in di e en olde s.
Also, he moni o ing sys em me ges all hese .cs iles wi h he ames’ names
in o a single HDF5 ile o u he analysis. All ha dwa e componen s a e
desc ibed as ollows:
•ABB obo : I has a 6-axis mo emen whe e he welding o ch is
placed. The ABB obo publishes i s ajec o ies and welding machine
da a om a TCP po as an a ay.
•F onius TP400i: I is a high-pe o mance welding powe sou ce de-
signed o demanding indus ial applica ions. The TPS 400i suppo s
MIG/MAG, Pulse, and TIG welding p ocesses. I has a use - iendly
in e ace and adap i e con ols ha op imize welding pa ame e s in
eal- ime, enhancing weld quali y and educing spa e .
•Oscilloscope Pico2206b: This oscilloscope is a compac , high-pe o mance
oscilloscope o a ious elec onic es ing and diagnos ic applica ions.
This de ice o e s an 8-bi esolu ion and a bandwid h o 50 MHz, mak-
ing i sui able o cap u ing de ailed wa e o ms in eal- ime. I ea u es
wo inpu channels and a 1 GS/s sampling a e, ensu ing accu a e sig-
nal acquisi ion and analysis.
•Ca i a Came a C400: The Ca i a C400 is a high-pe o mance
lase illumina ion sys em designed o challenging imaging applica ions
in indus ial en i onmen s. I deli e s a monoch oma ic lase ligh ,
making i ideal o high-speed imaging and isualizing as p ocesses
such as welding, combus ion, and luid dynamics.
•Edge se e : The Dell P ecision 5860, equipped wi h an NVIDIA
T1000 g aphics ca d, is a obus wo ks a ion ailo ed o da a mon-
i o ing sys ems and eal- ime con ol. I has a powe ul In el Xeon
W-Se ies wi h 8 co es and 32 GB o RAM: 2 x 16 GB, DDR5. The
NVIDIA T1000 GPU, wi h 4 GB o GDDR6 memo y, enhances isual
pe o mance and suppo s mul iple displays ia ou Mini DisplayPo
connec ions.
Figu e 5 summa izes he es bench se up wi h all he ha dwa e connec-
ions.
6
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
Figu e 1: Tes bench se up desc ip ion o he hyb id anomaly de ec ion in a WAAM cell
3.2. Welding Pa ame e Con igu a ions
F onius TP400I is con igu ed as ollows:
•Wi e-speed: 3.5 m/min
•Cu en : 131 Amps - eal: 113
•Vol age: 13 V - eal: 15.2
•Laye high: 1.8 mm
•laye wid h: 4.4 mm
•Con igu a ion: CMT
•Shielding Gas: A gon, 20 L/min
•CTWD: 1.5 cm
Con e sely, he designed es in ol es welding single-meld beads o con-
igu e he came a and synch onize senso s. Then, a 16-laye wall wi h a
wai ing ime be ween laye s o 2 seconds is welded o alida e ha he ou -
lie de ec ion sys em is wo king p ope ly.
3.3. Cu en pulse de ec ion
•Raw da a: The da a was ga he ed om he oscilloscope, which has
a sample a e 10KHz o bo h he ol age and cu en signals. Since,
in he CMT con igu a ion, he ol age signal emains s able/cons an ;
he cu en signal is he a ge o de ec ing ou lie s.
7
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
•Spli down in laye s: A e elec ical da a was collec ed, i was
b oken down in o laye s o p ocess each sample indi idually since he
numbe o samples was as . Also, he da ase con ains dead zones
whe e he o ch was no wo king, and he oscilloscope sa es emp y
alues ha migh a ec he ollowing s eps. Figu e 2 (a) shows an
example o he 16 laye s-wall.
•Take he i s pulse ou : The i s elec ical pulse is indeed anoma-
lous since he wi e c ea es he i s a c wi h he subs a e pla e o mel
bead. The e o e, i can make a unique clus e o con use he clus e ing
algo i hm in he ollowing s ep. Figu e 2 (b) shows he i s cu en
pulse compa ed wi h he es .
•Clus e ing pulses: When each laye is indi idually s o ed and he
i s pulse is emo ed, he DBSCAN algo i hm is applied. This non-
pa ame ic clus e ing algo i hm g oups poin s ha a e closely packed
and ma ks as ou lie s poin s ha lie alone in low-densi y egions. Con-
sequen ly, 4 clus e s we e made; clus e 1 g oups he alley poin s,
middle poin (clus e 2), peak poin s (clus e 3), and noise (poin s be-
ween clus e 1 and 2 ). Figu e 2 (c) shows he clus e s applied o he
cu en signal.
•De ec peaks and alleys: Wi h clus e 3, he maximum poin is
de ec ed on each cu en pulse and he alley poin s could be emo ed
o de e mine ou lie s only on he cu en pulse wi hou noise.
•Selec pulses: Wi h he maximum-cu en -pulse-poin , a mo ing le -
o- igh algo i hm is de eloped o ga he only he pulse wi hou alleys.
I means only using he poin s be ween he las alley poin o he le
o he cu en pulse and he i s cu en poin o he igh alley poin s.
•De ec ou lie s: When all pulses we e aken solely, hey we e me ged
in o single columns o he same size. I one pulse does no con ain he
same sample poin s, a mo ing a e age algo i hm is applied o ill emp y
spaces. Then, he Isola ion Fo es algo i hm is applied o he da ase
o ecognize ou lie s. Then, ou lie cu en pulses e u n o hei ini ial
o m a ached o he alleys o ge he exac leng h hey we e ini ially.
In addi ion, pulses ha ha e been assigned as no mal ones a e se o
a ze o in his a ay o ma ch he o iginal samples. Figu e 2 (e) shows
he ou lie s a he same ime o he o iginal samples.
8
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
(a) Cu en pulses o he 16 laye s-wall (b) Fi s cu en pulse compa ed wi h he
es
(c) Clus e ing elec ic pulses (d) Ou lie samples
Figu e 2: Cu en pulses in a WAAM p ocess
3.4. CWTD de ec ion
•Raw ames: The Ca i a came a ook 1440x1080 ames a 60
ames pe second (FPS) om he WAAM p ocess. The ideo was
s o ed in a .wa ile; hen, he ideo was spli in o ames in .png o -
ma . Figu e 3 (a) shows a ame example.
•Anno a e ames: The da ase was anno a ed wi h bounding boxes
o highligh he CTWD loca ion. This s ep is c ucial o ain objec
de ec ion models. In he end, 150 images we e anno a ed.
•P e-p ocess da a: Since he anno a ed image numbe could be a ew
samples o ain ML models, p e-p ocessing da a is he ollowing s ep;
anno a ed images could be o a ed, sa u a ed, esized, lipped, and
escaled. As a esul , he da ase inc eased in numbe and go ex a
in o ma ion o de ec he CTWD. 300 s e ched images in o 604x640
is he da ase whe e 70% is o he aining se , 10% o he alida ion
and 20% o he es .
•T ain objec de ec ion model: The objec de ec ion model selec ed
was YOLO 9 since i p esen s a high accu acy wi h p e- ained weigh s.
9
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
[4] A. Waqas, Q. Xiansheng, X. Jiang ao, W. Hongbo, M. Muzamil, A. Ma-
jeed, Di ec ional ensile p ope ies o s eel s uc u e manu ac u ed by
obo ic assis ed GMAW addi i e manu ac u ing, in: 2019 16 h In e -
na ional Bhu ban Con e ence on Applied Sciences and Technology (IB-
CAST), pp. 239–242, ISSN: 2151-1411. doi:10.1109/IBCAST.2019.
8667262.
URL h ps://ieeexplo e.ieee.o g/documen /8667262
[5] Y. Li, H. Mu, J. Polden, H. Li, L. Wang, C. Xia, Z. Pan, To-
wa ds in elligen moni o ing sys em in wi e a c addi i e manu ac u -
ing: a su ace anomaly de ec o on a small da ase 120 (7) 5225–5242.
doi:10.1007/s00170-022-09076-5.
URL h ps://doi.o g/10.1007/s00170-022-09076-5
[6] H. Mu, J. Polden, Y. Li, F. He, C. Xia, Z. Pan, Laye -by-laye model-
based adap i e con ol o wi e a c addi i e manu ac u ing o hin-wall
s uc u es 33 (4) 1165–1180. doi:10.1007/s10845-022-01920-5.
URL h ps://doi.o g/10.1007/s10845-022-01920-5
[7] A. ˇ
Sˇce inec, D. Klobˇca , D. B aˇcun, In-p ocess pa h eplanning
and online laye heigh con ol h ough deposi ion a c cu en
o gas me al a c based addi i e manu ac u ing 64 1169–1179.
doi:10.1016/j.jmap o.2021.02.038.
URL h ps://www.sciencedi ec .com/science/a icle/pii/
S1526612521001249
[8] J. Xiong, G. Zhang, Adap i e con ol o deposi ed heigh in
GMAW-based laye addi i e manu ac u ing 214 (4) 962–968.
doi:10.1016/j.jma p o ec.2013.11.014.
URL h ps://www.sciencedi ec .com/science/a icle/pii/
S0924013613003415
[9] N. Kozame nik, D. B aˇcun, D. Klobˇca , WAAM sys em wi h in e -
pass empe a u e con ol and o ced cooling o nea -ne -shape p in -
ing o small me al componen s 110 (7) 1955–1968. doi:10.1007/
s00170-020-05958-8.
URL h ps://doi.o g/10.1007/s00170-020-05958-8
[10] M. Ka muhilan, A. k. sood, In elligen p ocess model o
bead geome y p edic ion in WAAM 5 (11) 24005–24013.
16
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
doi:10.1016/j.ma p .2018.10.193.
URL h ps://www.sciencedi ec .com/science/a icle/pii/
S221478531832488X
[11] A. No elino, G. Ca alho, M. Zibe o , In luence o waam-
cm deposi ion pa ame e s on wall geome y, Ad ances in
Indus ial and Manu ac u ing Enginee ing 5 (2022) 100105.
doi:h ps://doi.o g/10.1016/j.aime.2022.100105.
URL h ps://www.sciencedi ec .com/science/a icle/pii/
S2666912922000320
[12] P. K. Nalajam, R. Va ada ajan, A hyb id deep lea ning model o
laye -wise mel pool empe a u e o ecas ing in wi e-a c addi i e man-
u ac u ing p ocess 9 100652–100664, con e ence Name: IEEE Access.
doi:10.1109/ACCESS.2021.3097177.
URL h ps://ieeexplo e.ieee.o g/abs ac /documen /9483890
[13] C. Xia, Z. Pan, Y. Li, J. Chen, H. Li, Vision-based mel pool moni o ing
o wi e-a c addi i e manu ac u ing using deep lea ning me hod 120 (1)
551–562. doi:10.1007/s00170-022-08811-2.
URL h ps://doi.o g/10.1007/s00170-022-08811-2
[14] J. Xiong, G. Zhang, Online measu emen o bead geome y in GMAW-
based addi i e manu ac u ing using passi e ision 24 (11) 115103, pub-
lishe : IOP Publishing. doi:10.1088/0957-0233/24/11/115103.
URL h ps://dx.doi.o g/10.1088/0957-0233/24/11/115103
[15] L. Yang, H. Sohn, Z. Ma, I. Jeon, P. Liu, J. C. Cheng, Real-
ime laye heigh es ima ion du ing mul i-laye di ec ed ene gy
deposi ion using domain adap i e neu al ne wo ks 148 103882.
doi:10.1016/j.compind.2023.103882.
URL h ps://linkinghub.else ie .com/ e ie e/pii/
S0166361523000325
[16] C. Halisch, T. Radel, D. Ty alla, T. See eld, Measu ing he mel pool
size in a wi e a c addi i e manu ac u ing p ocess using a high dynamic
ange wo-colo ed py ome ic came a 64 (8) 1349–1356. doi:10.1007/
s40194-020-00892-5.
URL h ps://doi.o g/10.1007/s40194-020-00892-5
17
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
[17] C. Lee, G. Seo, D. B. Kim, M. Kim, J.-H. Shin, De elopmen o de ec
de ec ion AI model o wi e + a c addi i e manu ac u ing using high
dynamic ange images 11 (16) 7541, numbe : 16 Publishe : Mul idisci-
plina y Digi al Publishing Ins i u e. doi:10.3390/app11167541.
URL h ps://www.mdpi.com/2076-3417/11/16/7541
[18] S. Adel, E. B. Njaas ad, M. H. A bo, F. Lindse h, Laye heigh es i-
ma ion using lase line scanne s o closed-loop di ec ed ene gy depo-
si ion, in: 2023 11 h In e na ional Con e ence on Con ol, Mecha on-
ics and Au oma ion (ICCMA), pp. 308–313, ISSN: 2837-5149. doi:
10.1109/ICCMA59762.2023.10374694.
URL h ps://ieeexplo e.ieee.o g/documen /10374694
[19] T. M. H. J. H¨olsche , Lenna Vincen Hassel, De ec ion o he con ac
ube o wo king dis ance in wi e and a c addi i e manu ac u ing, The
In e na ional Jou nal o Ad anced Manu ac u ing Technology (2022).
doi:10.1007/s00170-022-08805-0.
URL h ps://doi.o g/10.1007/s00170-022-08805-0
[20] H. F. . R. R. F anke, J., Vision based p ocess moni o ing in wi e a c
addi i e manu ac u ing (waam), Jou nal o In elligen Manu ac u ing
(2022). doi:h ps://doi.o g/10.1007/s10845-023-02287-x.
URL h ps://link-sp inge -com-443.web pn.synu.edu.cn/
a icle/10.1007/s10845-023-02287-x#ci eas
18
This p ep in esea ch pape has no been pee e iewed. Elec onic copy a ailable a : h ps://ss n.com/abs ac =5007393
P ep in no pee e iewed
