Damage analysis of composite CFRP tubes using acoustic emission monitoring and pattern recognition approach

ma e ials A icle Damage Analysis o Composi e CFRP Tubes Using Acous ic Emission Moni o ing and Pa e n Recogni ion App oach Michal Šo e 1,*, Jakub Cienciala 1, Ma in Fusek 1, Pa el Pa líˇcek 1and Richa d Mo a ec 2   Ci a ion: Šo e , M.; Cienciala, J.; Fusek, M.; Pa líˇcek, P.; Mo a ec, R. Damage Analysis o Composi e CFRP Tubes Using Acous ic Emission Moni o ing and Pa e n Recogni ion App oach. Ma e ials 2021,14, 786. h ps://doi.o g/10.3390/ma14040786 Academic Edi o : Michele Bacciocchi Recei ed: 12 No embe 2020 Accep ed: 28 Janua y 2021 Published: 7 Feb ua y 2021 Publishe ’s No e: MDPI s ays neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a il- ia ions. Copy igh : © 2021 by he au ho s. Licensee MDPI, Basel, Swi ze land. This a icle is an open access a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion (CC BY) license (h ps:// c ea i ecommons.o g/licenses/by/ 4.0/). 1 Depa men o Applied Mechanics, Facul y o Mechanical Enginee ing, VŠB—Technical Uni e si y o Os a a, 17. lis opadu 2172/15, 708 00 Os a a, Czech Republic; [email p o ec ed] (J.C.); [email p o ec ed] (M.F.); [email p o ec ed] (P.P.) 2 Ha el Composi es CZ s. .o., S ésedlice 67, 783 54 Pˇ ásla ice, Czech Republic; mo a ec @ha el-composi es.com *Co espondence: [email p o ec ed]; Tel.: +420-731-664-248 Abs ac : The acous ic emission me hod has been adop ed o de ec ion o damage mechanisms in ca bon- ibe - ein o ced polyme composi e ubes du ing he h ee-poin bending es . The damage e olu ion p ocess o he indi idual samples has been moni o ed using he acous ic emission me hod, which is one o he non-des uc i e me hods. The ob ained da a we e hen subjec ed o a wo-s ep echnique, which combines he unsupe ised pa e n ecogni ion app oach u ilizing he sho - ime equency spec a wi h he bounda y cu e enabling he al eady clus e ed da a o be addi ionally il e ed. The bounda y cu e iden i ica ion has been ca ied ou on he basis o p elimina y ensile es s o he ca bon ibe shea s, whe e, by o e lapping he o ce e sus ime dependency by he acous ic emission ac i i y e sus ime dependency, i was possible o iden i y he bounda y which will sepa a e he signals o igina ing om he ibe b eak om unwan ed seconda y sou ces. The applica ion o he p esen ed wo-s ep me hod esul ed in he iden i ica ion o he ailu e mechanisms such as ma ix c acking, ibe b eak, decohesion, and debonding. Besides he compa ison o he esul s wi h al eady published esea ch pape s, he s udy p esen s he comp ehensi e pa ame ic acous ic emission signal analysis o he indi idual clus e s. Keywo ds: acous ic emission; CFRP composi e ube; unsupe ised lea ning app oach; ailu e mechanism 1. In oduc ion O e he pas decades, ca bon- ibe - ein o ced polyme (CFRP) composi es ha e shown a cons an inc ease in a a ie y o applica ions such as ca o ai c a componen s, spo s and medical equipmen [ 1 ], and ecen ly also addi i e manu ac u ing [ 2 – 4 ]. Thei main bene i lies p ima ily in he ela i ely high s eng h/weigh a io o he abili y o cus omize he ma e ial p ope ies o dedica ed pu poses by changing he s acking se- quence and ela ed ibe o ien a ion. A ela i e d awback o CRFP composi es is he lack o duc ile-like beha io and he co esponding absence o p e-wa ning phase be o e he s uc u al collapse [ 5 , 6 ] leading o he b i le ailu e. CRFP composi es a e also cha ac- e ized by he accumula ion o damage inside he s uc u e wi hou any e idence on he s uc u e su ace [ 6 ] hus leading o a ela i ely challenging damage assessmen . The e a e many non-des uc i e es ing app oaches, which can be applied on composi e s uc- u es, namely in a ed omog aphy [ 7 ], eddy cu en es ing [ 8 ], ul asonic es ing [ 9 ], and X- ay omog aphy [10]. One o he mos p omising app oaches, especially coupled wi h o he me hods [ 11 ] such as Scanning Elec on Mic oscopy (SEM), is he acous ic emission (AE) me hod, which is also used in a ious applica ions as a eal ime moni o ing ool [ 6 ]. The AE me hod exhibi s g ea sensi i i y including conside able eliabili y o ac i e c acks de ec ion [ 12 ], e en in he case o ini ia ion phase [ 13 ]. The AE echnique is e en capable o de ec ing he onse o plas ic de o ma ion [ 14 ], which has he cha ac e o whi e noise wi h low ene gy [ 15 ]. Fo gaining a mo e de ailed insigh in o he damage moni o ing p ocess wi hin Ma e ials 2021,14, 786. h ps://doi.o g/10.3390/ma14040786 h ps://www.mdpi.com/jou nal/ma e ials Ma e ials 2021,14, 786 2 o 16 he meaning o AE sou ce cha ac e iza ion, i is a o able o inco po a e an adequa e signal analysis ool. The supe ised/unsupe ised pa e n ecogni ion (UPR) app oach [ 16 ] has become a e y sui able and p omising app oach o ackle a wide a ie y o p oblems such as a igue es s [ 17 ], s uc u e heal h moni o ing [ 18 ], and condi ion assessmen o p essu e essels [ 19 ] and p essu e componen s in ope a ion [ 20 ]. Nume ous s udies [ 21 – 26 ] ha e been conduc ed in o de o assess cha ac e is ic ea u es o he AE ansien s o igi- na ing om a ious ailu e mechanism in he CFRP composi es such as ma ix c acking, delamina ion, ibe b eak, and debonding (see Table 1 o u he explana ion). Table 1. Basic cha ac e iza ion o damage mechanisms occu ing in ca bon- ibe - ein o ced polyme (CFRP) composi es. Damage Mechanism Cha ac e iza ion Fibe b eak Disin eg a ion o single and/o mul iple ca bon ibe s Delamina ion Sepa a ion o wo adjacen plies (In e ace ailu e) Debonding In eg i y ailu e be ween ibe and ma ix (In e ace ailu e) Ma ix c acking Nuclea ion and u he p opaga ing o (mic o)c acks in he ma ix Al hough Chou [ 21 ] poin s o a disc epancy conce ning, in pa icula , he signal ampli ude, du a ion as well as equency spec a o he indi idual damage mechanisms, i was possible o compile a gene al o e iew, which is gi en in he ollowing able (Table 2). Table 2. Summa y o he acous ic emission (AE) signal cha ac e is ics o gi en damage mechanisms in CFRP composi es Damage Mechanism AE Signal Cha ac e is ics * Fibe b eak A: 50–100 dBAE, D: 100–10,000 µs, = 300–700 kHz Ma ix mic o c acks A: 30–40 dBAE, D: <1000 µs, = 100–250 kHz Ma ix mic o c acks (p opaga ion) A: 40–80 dBAE, D: 1000–10,000 µs, = 100–250 kHz Delamina ion A: >70 dBAE, D: 1000–10,000 µs, = 250–300 kHz Debonding A: <60 dBAE, ∼ =300 kHz * A—ampli ude, D—du a ion. In he las decade, se e al/nume ous s udies u ilizing ad anced echniques o classi- ica ion o ailu e modes, such as he use o s a is ical analysis o wa ele coe icien s [ 27 ] o in a ed he mog aphy (IT) [ 28 ], ha e been conduc ed. Ano he in e es ing app oach can be ound in he wo k published by Munoz e al. [ 29 ], who iden i ied and u he cha ac e - ized he damage mechanisms in he unidi ec ional CFRP composi es subjec ed o axis and o -axis s a ic ensile es s using he acous ic emission me hod and in a ed he mog aphy. Fu he u iliza ion o unsupe ised pa e n ecogni ion echnique oge he wi h he IT me hod esul ed in he iden i ica ion o he ailu e mechanisms such as ma ix c acking, ibe b eakage, and in e ace ailu e, o which he cha ac e iza ion in e ms o he signal ampli ude o ene gy has been pe o med. In 2011, Gu kin e al. published an ex ensi e esea ch [ 30 ], in which he AE signal da a om a ious es con igu a ions we e analyzed by h ee di e en pa e n ecogni ion app oaches. The analysis esul ed in cha ac e is ic equency spec a o ma ix c acking, delamina ion, debonding, ibe pull-ou , and ibe ailu e. I has o be no ed ha he gi en indings in e ms o he equency spec a a e o some ex en simila o he esul s summa ized in Table 2and he e o e con i ms he ac ual accu acy o he s udy [30]. The main objec i e o his s udy is o in es iga e and comp ehensi ely desc ibe he AE signal cha ac e is ics o he damage mechanisms in h ee di e en ypes o CFRP composi e ubes using a wo-s ep me hod combining he unsupe ised pa e n ecogni ion app oach wi h he u iliza ion o he bounda y cu e. The cons uc ion o he bounda y cu e has been conduc ed on he da a om he p elimina y ca bon ibe shea ensile es s. The al eady iden i ied bounda y cu e has hen been used o u he e inemen o he da a ac oss indi idual clus e s. Using he p esen ed app oach, i was possible o iden i y a Ma e ials 2021,14, 786 3 o 16 o al o ou damage mechanisms p esen ed in Table 1wi h subsequen compa ison o he ob ained esul s wi h he al eady published esea ch pape s. The pa o he s udy is also he comp ehensi e AE wa e o m analysis o he ep esen a i e signals belonging o he indi idual clus e s. 2. Expe imen al P ocedu e 2.1. Tes Sample Cha ac e iza ion The expe imen s we e ca ied ou on h ee ypes o CFRP ubes wi h a di e en numbe o laye s, hei o ien a ion, and wo en ibe densi y o he used ma e ial (see Figu e 1), whe e each ype o CFRP ube has been ep esen ed by h ee es samples. The samples labeled “A” we e manu ac u ed using ou laye s o unidi ec ional ca bon wo en ab ic wi h densi y o 200 g/m 2 and one laye o a amid/ca bon wo en ab ic (0–90 ◦ ) wi h densi y o 175 g/m 2 and a e age wall hickness o 1.45 mm. The p oduc ion o samples labeled “B” included he use o wo laye s o unidi ec ional ca bon wo en ab ic wi h densi y 300 g/m 2 and one laye o ca bon wo en ab ic (0–90 ◦ ) wi h densi y 280 g/m 2 wi h a e age wall hickness o 0.9 mm, while samples labeled “C” we e manu ac u ed using solely ou laye s o unidi ec ional ca bon wo en ab ic wi h densi y o 300 g/m 2 wi h a e age wall hickness o 1.42 mm. Table 3summa izes he speci ica ion o he es ed CFRP ubes. Ma e ials 2021, 14, x FOR PEER REVIEW 4 o 16 Figu e 1. CFRP composi e ubes unde es ; A—A se ies sample, B—B se ies sample, C— C se ies sample. 2.2. Th ee-Poin Bending Tes The h ee conside ed ypes o CFRP composi e ubes unde es a e being used o paddle p oduc ion; he e o e, he h ee-poin bending es has been selec ed in o de o simula e as much as possible he eal na u e o he loading p ocess du ing he use o he gi en spo s equipmen . The expe imen s we e ca ied ou on he uni e sal Tes ome ic M500-50CT es ing machine (The Tes ome ic Company L d., Rochdale, UK) wi h dedica ed weldmen , which enables i s geome y o be modi ied wi h i s mo ing pa s o a wide a ie y o such expe imen s. The dis ance be ween he suppo s was equal o 1040 mm wi h he o ce ac ing poin in a dis ance o 440 mm om AE senso #1 (see Figu e 2). The es ed CFRP ube wi h a ached AE senso s was addi ionally placed in a plas ic pipe o p e en damage o he AE senso s and o he equipmen due o sudden s uc u al in eg i y iola ion. The suppo s we e co e ed by hin el o allow ee mo emen o he ube du ing i s bending. The es has been de o ma ion-con olled wi h he uppe an il speed equal o 10 mm/min. (a) (b) Figu e 2. (a) In-Si u pho og aph o he es ig including he specimen equipped wi h AE senso s; (b) Schema ic ep esen a ion o he h ee-poin bending es se up. 2.3. Acous ic Emission Moni o ing The acous ic emission ac i i y has been moni o ed using he Vallen AMSY-6 AE sys em (Vallen Sys eme GmbH, Icking, Ge many) wi h wo u ilized measu ing channels (ASIP-2A dual channel signal p ocesso ca d), equipped wi h he AEP5H 34 dB p eampli ie s and he b oad-band Vallen VS-900 AE senso s. The senso s we e a ached on o he ube wi h he use o oil-based plas icine. The sampling equency o he AE da a Figu e 1. CFRP composi e ubes unde es ; A—A se ies sample, B—B se ies sample, C—C se ies sample. Table 3. Speci ica ion o he es ed CFRP ubes P ope y A Se ies B Se ies C Se ies Wall h. (mm)/Diame e (mm) 1.45/32 0.9/32 1.42/32 Fab ica ion 4 laye s o 200 g/m2unidi . ca bon ab ic1 laye o 175 g/m2a amid/ca bon ab ic (0◦–90◦) 2 laye s o 300 g/m2unidi . ca bon ab ic1 laye o 280 g/m2ca bon ab ic (0◦–90◦) 4 laye s o 300 g/m2unidi . ca bon ab ic Ma e ials 2021,14, 786 4 o 16 2.2. Th ee-Poin Bending Tes The h ee conside ed ypes o CFRP composi e ubes unde es a e being used o paddle p oduc ion; he e o e, he h ee-poin bending es has been selec ed in o de o simula e as much as possible he eal na u e o he loading p ocess du ing he use o he gi en spo s equipmen . The expe imen s we e ca ied ou on he uni e sal Tes ome ic M500-50CT es ing machine (The Tes ome ic Company L d., Rochdale, UK) wi h dedica ed weldmen , which enables i s geome y o be modi ied wi h i s mo ing pa s o a wide a ie y o such expe imen s. The dis ance be ween he suppo s was equal o 1040 mm wi h he o ce ac ing poin in a dis ance o 440 mm om AE senso #1 (see Figu e 2). The es ed CFRP ube wi h a ached AE senso s was addi ionally placed in a plas ic pipe o p e en damage o he AE senso s and o he equipmen due o sudden s uc u al in eg i y iola ion. The suppo s we e co e ed by hin el o allow ee mo emen o he ube du ing i s bending. The es has been de o ma ion-con olled wi h he uppe an il speed equal o 10 mm/min. Ma e ials 2021, 14, x FOR PEER REVIEW 4 o 16 Figu e 1. CFRP composi e ubes unde es ; A—A se ies sample, B—B se ies sample, C— C se ies sample. 2.2. Th ee-Poin Bending Tes The h ee conside ed ypes o CFRP composi e ubes unde es a e being used o paddle p oduc ion; he e o e, he h ee-poin bending es has been selec ed in o de o simula e as much as possible he eal na u e o he loading p ocess du ing he use o he gi en spo s equipmen . The expe imen s we e ca ied ou on he uni e sal Tes ome ic M500-50CT es ing machine (The Tes ome ic Company L d., Rochdale, UK) wi h dedica ed weldmen , which enables i s geome y o be modi ied wi h i s mo ing pa s o a wide a ie y o such expe imen s. The dis ance be ween he suppo s was equal o 1040 mm wi h he o ce ac ing poin in a dis ance o 440 mm om AE senso #1 (see Figu e 2). The es ed CFRP ube wi h a ached AE senso s was addi ionally placed in a plas ic pipe o p e en damage o he AE senso s and o he equipmen due o sudden s uc u al in eg i y iola ion. The suppo s we e co e ed by hin el o allow ee mo emen o he ube du ing i s bending. The es has been de o ma ion-con olled wi h he uppe an il speed equal o 10 mm/min. (a) (b) Figu e 2. (a) In-Si u pho og aph o he es ig including he specimen equipped wi h AE senso s; (b) Schema ic ep esen a ion o he h ee-poin bending es se up. 2.3. Acous ic Emission Moni o ing The acous ic emission ac i i y has been moni o ed using he Vallen AMSY-6 AE sys em (Vallen Sys eme GmbH, Icking, Ge many) wi h wo u ilized measu ing channels (ASIP-2A dual channel signal p ocesso ca d), equipped wi h he AEP5H 34 dB p eampli ie s and he b oad-band Vallen VS-900 AE senso s. The senso s we e a ached on o he ube wi h he use o oil-based plas icine. The sampling equency o he AE da a Figu e 2. ( a ) In-Si u pho og aph o he es ig including he specimen equipped wi h AE senso s; ( b ) Schema ic ep esen a- ion o he h ee-poin bending es se up. 2.3. Acous ic Emission Moni o ing The acous ic emission ac i i y has been moni o ed using he Vallen AMSY-6 AE sys em (Vallen Sys eme GmbH, Icking, Ge many) wi h wo u ilized measu ing channels (ASIP-2A dual channel signal p ocesso ca d), equipped wi h he AEP5H 34 dB p eampli ie s and he b oad-band Vallen VS-900 AE senso s. The senso s we e a ached on o he ube wi h he use o oil-based plas icine. The sampling equency o he AE da a was se o 10 MHz while he ansien da a (wa e ansien s) we e sampled wi h 20 MHz in he equency ange be ween 50 and 1100 kHz. The de ec ion h eshold has been se o 32 dB owing o a ela i ely g ea e dis ance o bo h senso s om he a ea in which he b each will mos likely occu and ela ed highe a enua ion o he AE signal in composi es. Only localized AE e en s, which all wi hin he h 250,650 i (mm) (see Figu e 2b) in e al o he x coo dina e will be included o u he da a p ocessing. The il e ed da a was ollowingly analyzed wi h he Vallen VisualClass so wa e package (Vallen Sys eme GmbH, Icking, Ge many), which uses he pa e n ecogni ion me hod [ 15 ] o associa e simila wa e o m ypes in o sepa a e g oups. Due o he na u e o he ask, an unsupe ised lea ning app oach was chosen. The p ocedu e s a s wi h loading he selec ed da abase o AE ansien s in o Vallen VisualClass so wa e, whe e he numbe o ime windows including hei span and he s a ing poin o he segmen a ion analysis in he ime domain mus be speci ied (see Figu e 3). Ma e ials 2021,14, 786 5 o 16 Ma e ials 2021, 14, x FOR PEER REVIEW 5 o 16 was se o 10 MHz while he ansien da a (wa e ansien s) we e sampled wi h 20 MHz in he equency ange be ween 50 and 1100 kHz. The de ec ion h eshold has been se o 32 dB owing o a ela i ely g ea e dis ance o bo h senso s om he a ea in which he b each will mos likely occu and ela ed highe a enua ion o he AE signal in composi es. Only localized AE e en s, which all wi hin he 〈250,650〉 (mm) (see Figu e 2b) in e al o he x coo dina e will be included o u he da a p ocessing. The il e ed da a was ollowingly analyzed wi h he Vallen VisualClass so wa e package (Vallen Sys eme GmbH, Icking, Ge many), which uses he pa e n ecogni ion me hod [15] o associa e simila wa e o m ypes in o sepa a e g oups. Due o he na u e o he ask, an unsupe ised lea ning app oach was chosen. The p ocedu e s a s wi h loading he selec ed da abase o AE ansien s in o Vallen VisualClass so wa e, whe e he numbe o ime windows including hei span and he s a ing poin o he segmen a ion analysis in he ime domain mus be speci ied (see Figu e 3). Figu e 3. Se ing up he Hamming windowed ime segmen s wi h co esponding esul s in he equency domain. The cu en analysis uses he ollowing se ings o he AE ansien s wi h ela ion o he VisualClass so wa e package: Numbe o ime segmen s: 5; Size o single ime segmen in e ms o poin s: 4096; Rel. igge o se : −256 poin s; min/max equency limi : 50/800 kHz. The so wa e hen pe o ms he assembly o mul idimensional ea u e ec o , he size o which depends on he chosen numbe o ime segmen s including hei size. The pa e n ecogni ion analysis will esul in he basic ea u e space iden i ica ion, which is hen linea ly ans o med o maximizing in e -class dis ance and minimizing he in a- class ex ension, a he same ime (see Figu e 4). Figu e 4. T ans o med ea u es p ojec ion— esul s o ou selec ed clus e s. Figu e 3. Se ing up he Hamming windowed ime segmen s wi h co esponding esul s in he equency domain. The cu en analysis uses he ollowing se ings o he AE ansien s wi h ela ion o he VisualClass so wa e package: Numbe o ime segmen s: 5; Size o single ime segmen in e ms o poin s: 4096; Rel. igge o se : − 256 poin s; min/max equency limi : 50/800 kHz. The so wa e hen pe o ms he assembly o mul idimensional ea u e ec o , he size o which depends on he chosen numbe o ime segmen s including hei size. The pa e n ecogni ion analysis will esul in he basic ea u e space iden i ica ion, which is hen linea ly ans o med o maximizing in e -class dis ance and minimizing he in a-class ex ension, a he same ime (see Figu e 4). Ma e ials 2021, 14, x FOR PEER REVIEW 5 o 16 was se o 10 MHz while he ansien da a (wa e ansien s) we e sampled wi h 20 MHz in he equency ange be ween 50 and 1100 kHz. The de ec ion h eshold has been se o 32 dB owing o a ela i ely g ea e dis ance o bo h senso s om he a ea in which he b each will mos likely occu and ela ed highe a enua ion o he AE signal in composi es. Only localized AE e en s, which all wi hin he 〈250,650〉 (mm) (see Figu e 2b) in e al o he x coo dina e will be included o u he da a p ocessing. The il e ed da a was ollowingly analyzed wi h he Vallen VisualClass so wa e package (Vallen Sys eme GmbH, Icking, Ge many), which uses he pa e n ecogni ion me hod [15] o associa e simila wa e o m ypes in o sepa a e g oups. Due o he na u e o he ask, an unsupe ised lea ning app oach was chosen. The p ocedu e s a s wi h loading he selec ed da abase o AE ansien s in o Vallen VisualClass so wa e, whe e he numbe o ime windows including hei span and he s a ing poin o he segmen a ion analysis in he ime domain mus be speci ied (see Figu e 3). Figu e 3. Se ing up he Hamming windowed ime segmen s wi h co esponding esul s in he equency domain. The cu en analysis uses he ollowing se ings o he AE ansien s wi h ela ion o he VisualClass so wa e package: Numbe o ime segmen s: 5; Size o single ime segmen in e ms o poin s: 4096; Rel. igge o se : −256 poin s; min/max equency limi : 50/800 kHz. The so wa e hen pe o ms he assembly o mul idimensional ea u e ec o , he size o which depends on he chosen numbe o ime segmen s including hei size. The pa e n ecogni ion analysis will esul in he basic ea u e space iden i ica ion, which is hen linea ly ans o med o maximizing in e -class dis ance and minimizing he in a- class ex ension, a he same ime (see Figu e 4). Figu e 4. T ans o med ea u es p ojec ion— esul s o ou selec ed clus e s. Figu e 4. T ans o med ea u es p ojec ion— esul s o ou selec ed clus e s. Ma e ials 2021,14, 786 6 o 16 The esul s a e hen ans e ed in o VisualAE so wa e o u he pos p ocessing. Fou clus e s we e chosen o subsequen analysis, since he addi ional inc ease o he numbe o clus e s did no lead o be e di e en ia ion o indi idual ansien s. One sample om each se ies has been subjec ed o he a enua ion measu emen o he AE signal using Hsu-Nielsen sou ce [ 31 ] (pencil lead diame e : 0.35 mm; ha dness: 2H) in o de o p ope ly e alua e he eal AE signal ampli ude in subsequen da a analysis. The esul s o he a enua ion measu emen s a e shown in Table 4. The p opaga ion eloci y has been de e mined expe imen ally using Hsu-Nielsen sou ce wi h alue a ying be ween 3200 and 3300 m/s ac oss A/B/C se ies samples. Table 4. A enua ion measu emen on A/B/C se ies o CFRP composi e ubes Sample Se ies Nea Field A enua ion (dB/m) Fa ield A enua ion (dB/m) A 90 33.3 B 66.6 33.2 C 222.2 36.2 The p elimina y measu emen s also included a se ies o six ensile es s o ca bon ibe shea s (see Figu e 5) in o de o cons uc he abo e-men ioned bounda y cu e, which will be u he used o he de ec ion o he ca bon ibe b eaks ac oss he iden i ied clus e s. The de o ma ion-con olled es s (uppe an il speed equal o 0.5 mm/min) we e ca ied ou on he uni e sal Tes ome ic M500-50CT es ing machine equipped wi h 100 N load cell. The ela i ely small scale o he load cell enabled us o de ec in ime he e en s co esponding o he ailu e o ce ain numbe o ibe s hanks o he egis e ed o ce d op. The AE ac i i y has been moni o ed using he Vallen AMSY-6 AE sys em wi h h ee u ilized measu ing channels (ASIP-2S dual channel signal p ocesso ca d), equipped wi h he AEP5H 34 dB p eampli ie s and he b oad-band DAKEL MIDI AE senso s, whe e he op and bo om senso ac ed as gua d elemen s, while he middle senso has been used o he da a acquisi ion. The ibe shea s we e glued on hei ends hus p o iding a clamping suppo o he a achmen in o he jaws. Ma e ials 2021, 14, x FOR PEER REVIEW 6 o 16 The esul s a e hen ans e ed in o VisualAE so wa e o u he pos p ocessing. Fou clus e s we e chosen o subsequen analysis, since he addi ional inc ease o he numbe o clus e s did no lead o be e di e en ia ion o indi idual ansien s. One sample om each se ies has been subjec ed o he a enua ion measu emen o he AE signal using Hsu-Nielsen sou ce [31] (pencil lead diame e : 0.35 mm; ha dness: 2H) in o de o p ope ly e alua e he eal AE signal ampli ude in subsequen da a analysis. The esul s o he a enua ion measu emen s a e shown in Table 4. The p opaga ion eloci y has been de e mined expe imen ally using Hsu-Nielsen sou ce wi h alue a ying be ween 3200 and 3300 m/s ac oss A/B/C se ies samples. Table 4. A enua ion measu emen on A/B/C se ies o CFRP composi e ubes Sample Se ies Nea Field A enua ion (dB/m) Fa ield A enua ion (dB/m) A 90 33.3 B 66.6 33.2 C 222.2 36.2 The p elimina y measu emen s also included a se ies o six ensile es s o ca bon ibe shea s (see Figu e 5) in o de o cons uc he abo e-men ioned bounda y cu e, which will be u he used o he de ec ion o he ca bon ibe b eaks ac oss he iden i ied clus e s. The de o ma ion-con olled es s (uppe an il speed equal o 0.5 mm/min) we e ca ied ou on he uni e sal Tes ome ic M500-50CT es ing machine equipped wi h 100 N load cell. The ela i ely small scale o he load cell enabled us o de ec in ime he e en s co esponding o he ailu e o ce ain numbe o ibe s hanks o he egis e ed o ce d op. The AE ac i i y has been moni o ed using he Vallen AMSY-6 AE sys em wi h h ee u ilized measu ing channels (ASIP-2S dual channel signal p ocesso ca d), equipped wi h he AEP5H 34 dB p eampli ie s and he b oad-band DAKEL MIDI AE senso s, whe e he op and bo om senso ac ed as gua d elemen s, while he middle senso has been used o he da a acquisi ion. The ibe shea s we e glued on hei ends hus p o iding a clamping suppo o he a achmen in o he jaws. Figu e 5. Expe imen al se up o ensile es o he ca bon ibe s. The basic equency analysis o he AE signal is, besides he exploi ed pa e n ecogni ion app oach, ela i ely e icien and powe ul ool o il e ing he AE signal, which can be hen a ilia ed o di e en ailu e mechanisms [32]. Chou in his wo k [21] s a es ha ibe b eakages in he case o ca bon ibe /glass ibe composi e sys ems p oduce ex ensional wa e signals wi h equencies be ween 350–700 kHz, while ma ix c acks gene a e lexu al wa e modes wi h equencies up o 350 kHz. I has o be no ed ha he gi en inding has been e i ied on he p elimina y ensile es s o he ca bon ibe s, whe e he AE signals o igina ing om ibe ailu e exhibi a highe powe ac ion in he 300(350)–600 kHz equency in e al. Based on his conside a ion, he e will be Figu e 5. Expe imen al se up o ensile es o he ca bon ibe s. Ma e ials 2021,14, 786 7 o 16 The basic equency analysis o he AE signal is, besides he exploi ed pa e n ecogni- ion app oach, ela i ely e icien and powe ul ool o il e ing he AE signal, which can be hen a ilia ed o di e en ailu e mechanisms [ 32 ]. Chou in his wo k [ 21 ] s a es ha ibe b eakages in he case o ca bon ibe /glass ibe composi e sys ems p oduce ex ensional wa e signals wi h equencies be ween 350–700 kHz, while ma ix c acks gene a e lexu al wa e modes wi h equencies up o 350 kHz. I has o be no ed ha he gi en inding has been e i ied on he p elimina y ensile es s o he ca bon ibe s, whe e he AE signals o igina ing om ibe ailu e exhibi a highe powe ac ion in he 300(350)–600 kHz equency in e al. Based on his conside a ion, he e will be de ined a a iable deno ed as p ac o , which will ela e he powe ac ion o he AE signal in a ce ain equency band o he AE signal powe in he en i e conside ed equency ange: P =P(350–800)kHz P(50–800)kHz 100(%)(1) whe e P (350–800)kHz ep esen s he powe ac ion o he AE signal in he 350–800 kHz equency ange and P (50–800)kHz is he powe o he AE signal in he en i e conside ed equency ange, i.e., 50–800 kHz. No e ha he powe o he AE signal in he gi en equency in e al has been calcula ed using Pa se al’s heo em. The bounda y cu e hen s a es he p ac o and he AE signal ampli ude in ela ion wi h subsequen in en ion o il e he AE signal o igina ing om he ibe ailu e om he o he ailu e mechanisms o he seconda y AE signals, which igu e as noise (in e ac ion o he indi idual ibe s be ween each o he ( ubbing) and/o AE ac i i y a ising om he sample a achmen poin s). The iden i ica ion p ocedu e o he bounda y cu e is based on he o e lapping he o ce e sus ime dependency by he acous ic emission ac i i y e sus ime dependency, whe e he de ec ed o ce d ops caused by he ailu e o he indi idual/mul iple ibe s can be di ec ly ma ched o he eme ged AE signals. The ollowing igu es display he dependency be ween he o ce and ampli ude o indi idual AE hi s on ime (Figu e 6) and he ela ion be ween he p ac o and he AE signal ampli ude o he en i e se ies (Figu e 7). No e ha he AE signal ampli ude is being e e ed o dB AE uni , hus exp essing he ol age ampli ude o non-ampli ied signal as a gain ela ed o 1 µ V. The blue line in Figu e 7 ep esen s he bounda y line sepa a ing AE signals belonging o he ibe b eak om signals o igina ing om he in e ac ion be ween he indi idual ibe s and he o he in e e ing AE sou ces. A his poin , i has o be no ed ha he high- equency ange appea ing in nume a o in Equa ion (1) has been o u he applica ion o he CFRP ubes ex ended o 300–800 kHz ange due o a equency a enua ion a ising om he geome ic dimensions o he CFRP ubes and he mu ual dis ance be ween he AE sou ce and he AE senso s. Ma e ials 2021, 14, x FOR PEER REVIEW 7 o 16 de ined a a iable deno ed as p ac o , which will ela e he powe ac ion o he AE signal in a ce ain equency band o he AE signal powe in he en i e conside ed equency ange: p =P(350–800)kHz P(50–800)kHz 100 (%), (1) whe e P(350–800)kHz ep esen s he powe ac ion o he AE signal in he 350–800 kHz equency ange and P(50–800)kHz is he powe o he AE signal in he en i e conside ed equency ange, i.e., 50–800 kHz. No e ha he powe o he AE signal in he gi en equency in e al has been calcula ed using Pa se al’s heo em. The bounda y cu e hen s a es he p ac o and he AE signal ampli ude in ela ion wi h subsequen in en ion o il e he AE signal o igina ing om he ibe ailu e om he o he ailu e mechanisms o he seconda y AE signals, which igu e as noise (in e ac ion o he indi idual ibe s be ween each o he ( ubbing) and/o AE ac i i y a ising om he sample a achmen poin s). The iden i ica ion p ocedu e o he bounda y cu e is based on he o e lapping he o ce e sus ime dependency by he acous ic emission ac i i y e sus ime dependency, whe e he de ec ed o ce d ops caused by he ailu e o he indi idual/mul iple ibe s can be di ec ly ma ched o he eme ged AE signals. The ollowing igu es display he dependency be ween he o ce and ampli ude o indi idual AE hi s on ime (Figu e 6) and he ela ion be ween he p ac o and he AE signal ampli ude o he en i e se ies (Figu e 7). No e ha he AE signal ampli ude is being e e ed o dBAE uni , hus exp essing he ol age ampli ude o non-ampli ied signal as a gain ela ed o 1 µV. The blue line in Figu e 7 ep esen s he bounda y line sepa a ing AE signals belonging o he ibe b eak om signals o igina ing om he in e ac ion be ween he indi idual ibe s and he o he in e e ing AE sou ces. A his poin , i has o be no ed ha he high- equency ange appea ing in nume a o in Equa ion (1) has been o u he applica ion o he CFRP ubes ex ended o 300–800 kHz ange due o a equency a enua ion a ising om he geome ic dimensions o he CFRP ubes and he mu ual dis ance be ween he AE sou ce and he AE senso s. Figu e 6. Fo ce and AE ac i i y as he unc ion o ime o he selec ed sample. Figu e 6. Fo ce and AE ac i i y as he unc ion o ime o he selec ed sample. Ma e ials 2021,14, 786 8 o 16 Ma e ials 2021, 14, x FOR PEER REVIEW 8 o 16 Figu e 7. Dependency be ween p ac o and AE signal ampli ude o he en i e se ies. 3. Resul s and Discussion 3.1. Mechanical P ope ies and Basic AE Signal Analysis Figu e 8 shows ela ion be ween o ce and displacemen o he an il o indi idual A/B/C se ies p oduc ion samples. As expec ed, he highes s i ness is being eached by he C se ies samples, ollowed by A and B se ies samples. All h ee manu ac u ing modi ica ions show wi hin hei g oup e y simila end in e ms o he o ce- displacemen cou se excep he A se ies samples, namely A2 sample, which exhibi s ma ginally lowe s i ness, mos likely due o he ab ica ion p ocess, which is no in he o m o he au oma ed p oduc ion. The abo e-men ioned consis ency in e ms o he sample s i ness is, howe e , no alid o he maximum o ce ac oss indi idual se ies, whe e di e ences om 10 o 28 pe cen ela ed o he maximum achie ed o ce in each p oduc ion se ies can be obse ed. Again, he eason o such esul s a ia ion can be ound in he p oduc ion o m i sel . A somewha simila end can be egis e ed in he case o he numbe o loca ed AE e en s ac oss he 0–Fmax ange o he indi idual samples (see Figu e 9), whe e a ela i ely la ge a ia ion has been egis e ed. Howe e , e en despi e his ac , A/B se ies epo conside ably highe le el o he loca ed AE e en s, which is mos likely caused due o he p esence o (0°–90°) ab ic igu ing as a op laye . No e ha he gi en s a emen is cu en ly a hypo hesis, which needs o be e i ied in he u u e. Figu e 8. Fo ce as he unc ion o displacemen o indi idual A/B/C se ies p oduc ion samples Figu e 7. Dependency be ween p ac o and AE signal ampli ude o he en i e se ies. 3. Resul s and Discussion 3.1. Mechanical P ope ies and Basic AE Signal Analysis Figu e 8shows ela ion be ween o ce and displacemen o he an il o indi idual A/B/C se ies p oduc ion samples. As expec ed, he highes s i ness is being eached by he C se ies samples, ollowed by A and B se ies samples. All h ee manu ac u ing mod- i ica ions show wi hin hei g oup e y simila end in e ms o he o ce-displacemen cou se excep he A se ies samples, namely A2 sample, which exhibi s ma ginally lowe s i ness, mos likely due o he ab ica ion p ocess, which is no in he o m o he au o- ma ed p oduc ion. The abo e-men ioned consis ency in e ms o he sample s i ness is, howe e , no alid o he maximum o ce ac oss indi idual se ies, whe e di e ences om 10 o 28 pe cen ela ed o he maximum achie ed o ce in each p oduc ion se ies can be obse ed. Again, he eason o such esul s a ia ion can be ound in he p oduc ion o m i sel . A somewha simila end can be egis e ed in he case o he numbe o loca ed AE e en s ac oss he 0–F max ange o he indi idual samples (see Figu e 9), whe e a ela i ely la ge a ia ion has been egis e ed. Howe e , e en despi e his ac , A/B se ies epo conside ably highe le el o he loca ed AE e en s, which is mos likely caused due o he p esence o (0 ◦ –90 ◦ ) ab ic igu ing as a op laye . No e ha he gi en s a emen is cu en ly a hypo hesis, which needs o be e i ied in he u u e. Ma e ials 2021, 14, x FOR PEER REVIEW 8 o 16 Figu e 7. Dependency be ween p ac o and AE signal ampli ude o he en i e se ies. 3. Resul s and Discussion 3.1. Mechanical P ope ies and Basic AE Signal Analysis Figu e 8 shows ela ion be ween o ce and displacemen o he an il o indi idual A/B/C se ies p oduc ion samples. As expec ed, he highes s i ness is being eached by he C se ies samples, ollowed by A and B se ies samples. All h ee manu ac u ing modi ica ions show wi hin hei g oup e y simila end in e ms o he o ce- displacemen cou se excep he A se ies samples, namely A2 sample, which exhibi s ma ginally lowe s i ness, mos likely due o he ab ica ion p ocess, which is no in he o m o he au oma ed p oduc ion. The abo e-men ioned consis ency in e ms o he sample s i ness is, howe e , no alid o he maximum o ce ac oss indi idual se ies, whe e di e ences om 10 o 28 pe cen ela ed o he maximum achie ed o ce in each p oduc ion se ies can be obse ed. Again, he eason o such esul s a ia ion can be ound in he p oduc ion o m i sel . A somewha simila end can be egis e ed in he case o he numbe o loca ed AE e en s ac oss he 0–Fmax ange o he indi idual samples (see Figu e 9), whe e a ela i ely la ge a ia ion has been egis e ed. Howe e , e en despi e his ac , A/B se ies epo conside ably highe le el o he loca ed AE e en s, which is mos likely caused due o he p esence o (0°–90°) ab ic igu ing as a op laye . No e ha he gi en s a emen is cu en ly a hypo hesis, which needs o be e i ied in he u u e. Figu e 8. Fo ce as he unc ion o displacemen o indi idual A/B/C se ies p oduc ion samples Figu e 8. Fo ce as he unc ion o displacemen o indi idual A/B/C se ies p oduc ion samples Ma e ials 2021,14, 786 9 o 16 Ma e ials 2021, 14, x FOR PEER REVIEW 9 o 16 Figu e 9. Cumula i e numbe o loca ed e en s as he unc ion o o ce o indi idual A/B/C se ies samples The ene gy o accumula ed AE e en s e sus o ce (Figu e 10) is ano he impo an dependency, which can b ing us close o he o e all s uc u e beha io . The maximum alue o he eleased AE ene gy is o all samples be ween 5 × 108 and 109 aJ. The di e ence, howe e , lies in he cha ac e how he ene gy is being eleased du ing he loading p ocess. The A and B se ies specimen exhibi almos g adual AE ene gy elease, wi h he di e ence in he inal loading s age. While he A se ies specimen end o g adually con inue wi h he cumula ion o he AE e en s and g adual elease o he AE ene gy, he B se ies samples end o suddenly lose in eg i y wi hou any signi ican wa ning phase. A comple ely di e en beha io can be ound in he case o he C se ies samples, which ha e conside ably la ge ene gy pe e en a io wi h a lack o any wa ning phase be o e he in eg i y lose. Figu e 10. Ene gy as he unc ion o o ce o indi idual A/B/C se ies samples 3.2. AE Signal Analysis Using Pa e n Recogni ion App oach The u ilized unsupe ised pa e n ecogni ion analysis esul ed in iden i ica ion o ou clus e s o AE signals wi h he ollowing ea u es. The signals a ilia ed o he i s clus e a e cha ac e ized by high ampli ude, in mos cases exceeding 90 dBAE wi h ene gy alue usually abo e 1 × 106 aJ and equencies in he span om 50 kHz o 150 kHz (see Figu e 11a), whe eas his clus e also pa ially con ains signals wi h equency con en abo e 300 kHz. The second clus e is cha ac e ized by he ampli udes mos ly below 65 dBAE wi h AE ene gy in he o de o hund eds o he ens o housands o aJ and he equency in he 50–450 kHz ange (see Figu e 11b). The hi d clus e is ep esen ed by Figu e 9. Cumula i e numbe o loca ed e en s as he unc ion o o ce o indi idual A/B/C se ies samples The ene gy o accumula ed AE e en s e sus o ce (Figu e 10) is ano he impo an dependency, which can b ing us close o he o e all s uc u e beha io . The maximum alue o he eleased AE ene gy is o all samples be ween 5 × 10 8 and 10 9 aJ. The di e ence, howe e , lies in he cha ac e how he ene gy is being eleased du ing he loading p ocess. The A and B se ies specimen exhibi almos g adual AE ene gy elease, wi h he di e ence in he inal loading s age. While he A se ies specimen end o g adually con inue wi h he cumula ion o he AE e en s and g adual elease o he AE ene gy, he B se ies samples end o suddenly lose in eg i y wi hou any signi ican wa ning phase. A comple ely di e en beha io can be ound in he case o he C se ies samples, which ha e conside ably la ge ene gy pe e en a io wi h a lack o any wa ning phase be o e he in eg i y lose. Ma e ials 2021, 14, x FOR PEER REVIEW 9 o 16 Figu e 9. Cumula i e numbe o loca ed e en s as he unc ion o o ce o indi idual A/B/C se ies samples The ene gy o accumula ed AE e en s e sus o ce (Figu e 10) is ano he impo an dependency, which can b ing us close o he o e all s uc u e beha io . The maximum alue o he eleased AE ene gy is o all samples be ween 5 × 108 and 109 aJ. The di e ence, howe e , lies in he cha ac e how he ene gy is being eleased du ing he loading p ocess. The A and B se ies specimen exhibi almos g adual AE ene gy elease, wi h he di e ence in he inal loading s age. While he A se ies specimen end o g adually con inue wi h he cumula ion o he AE e en s and g adual elease o he AE ene gy, he B se ies samples end o suddenly lose in eg i y wi hou any signi ican wa ning phase. A comple ely di e en beha io can be ound in he case o he C se ies samples, which ha e conside ably la ge ene gy pe e en a io wi h a lack o any wa ning phase be o e he in eg i y lose. Figu e 10. Ene gy as he unc ion o o ce o indi idual A/B/C se ies samples 3.2. AE Signal Analysis Using Pa e n Recogni ion App oach The u ilized unsupe ised pa e n ecogni ion analysis esul ed in iden i ica ion o ou clus e s o AE signals wi h he ollowing ea u es. The signals a ilia ed o he i s clus e a e cha ac e ized by high ampli ude, in mos cases exceeding 90 dBAE wi h ene gy alue usually abo e 1 × 106 aJ and equencies in he span om 50 kHz o 150 kHz (see Figu e 11a), whe eas his clus e also pa ially con ains signals wi h equency con en abo e 300 kHz. The second clus e is cha ac e ized by he ampli udes mos ly below 65 dBAE wi h AE ene gy in he o de o hund eds o he ens o housands o aJ and he equency in he 50–450 kHz ange (see Figu e 11b). The hi d clus e is ep esen ed by Figu e 10. Ene gy as he unc ion o o ce o indi idual A/B/C se ies samples 3.2. AE Signal Analysis Using Pa e n Recogni ion App oach The u ilized unsupe ised pa e n ecogni ion analysis esul ed in iden i ica ion o ou clus e s o AE signals wi h he ollowing ea u es. The signals a ilia ed o he i s clus e a e cha ac e ized by high ampli ude, in mos cases exceeding 90 dB AE wi h ene gy alue usually abo e 1 × 10 6 aJ and equencies in he span om 50 kHz o 150 kHz (see Figu e 11a), whe eas his clus e also pa ially con ains signals wi h equency con en abo e 300 kHz. The second clus e is cha ac e ized by he ampli udes mos ly below Ma e ials 2021,14, 786 16 o 16 21. Chou, H.-Y. Damage Analysis o Composi e P essu e Vessels Using Acous ic Emission Moni o ing. Ph.D. Thesis, School o Ae ospace, Mechanical & Manu ac u ing Enginee ing College o Science Enginee ing and Heal h, RMIT Uni e si y, Melbou ne, Aus alia, 2011. 22. Ono, K.; Kawamo o, K. Digi al signal analysis o acous ic emission om ca bon ibe /epoxy composi es. J. Acous . Emiss. 1990 ,9, 109–116. 23. Ono, K. Acous ic emission beha io o lawed unidi ec ional ca bon ibe -epoxy composi es. J. Rein . Plas . Compos. 1988 ,7, 90–105. [C ossRe ] 24. Bohse, J. Acous ic emission cha ac e is ics o mic o- ailu e p ocesses in polyme blends and composi es. Compos. Sci. Technol. 2000,60, 1213–1226. [C ossRe ] 25. Komai, K.; Minoshima, K.; Shibu ani, T. In es iga ions o he ac u e mechanism o ca bon/epoxy composi es by AE signal analyses. JSME In . J. Se . 1 Solid Mech. S eng h Ma e . 1991,34, 381–388. [C ossRe ] 26. Godin, N.; Hugue , S.; Gae ne , R.; Salmon, L. Clus e ing o acous ic emission signals collec ed du ing ensile es s on unidi ec- ional glass/polyes e composi e using supe ised and unsupe ised classi ie s. NDT&E In . 2004,37, 253–264. 27. Bacca , D.; Sö ke , D. Iden i ica ion and classi ica ion o ailu e modes in lamina ed composi es by using a mul i a iable s a is ical analysis o wa ele coe icien s. Mech. Sys . Signal P ocess. 2017,96, 77–87. [C ossRe ] 28. P akash, R.V.; Maha ana, M. Damage de ec ion using in a ed he mog aphz in a ca bon- lax ibe hyb id composi e. P ocedia S uc . In eg . 2017,7, 283–290. [C ossRe ] 29. Munoz, V.; Valés, B.; Pe in, M.; Pas o , M.L.; Welemane, H.; Can a el, A. Damage de ec ion in CFRP by coupling acous ic emission and in a ed he mog aphy. Compos. Pa B 2016,85, 68–75. [C ossRe ] 30. Gu kin, R.; G een, C.J.; Vang a anachai, S.; Pinho, S.T.; Robinson, P.; Cu is, P.T. On acous ic emission o ailu e in es iga ion in CFRP: Pa e n ecogni ion and peak equency analyses. Mech. Sys . Signal P ocess. 2011,25, 1393–1407. [C ossRe ] 31. Sause, M.G.R. In es iga ion o pencil lead b eaks as acous ic emission sou ces. J. Acous . Emiss. 2011,29, 184–196. 32. K ie sch, T.; Bohse, J. Selec ion o acous ic emissions and classi ica ion o damage mechanisms in ibe composi e ma e ials. J. Acous . Emiss. 1998,16, 233–242. 33. Nam, K.-W.; Ahn, S.-H.; Moon, C.-K. F ac u e beha io o ca bon ibe ein o ced plas ics de e mined by he ime- equency analysis me hod. J. Appl. Polym. Sci. 2003,88, 1659–1664. [C ossRe ] 34. Si on, O.; Tsuda, H. Acous ic emission in ca bon ib e- ein o ced plas ic ma e ials. Ann. Chim. Sci. Ma e . 2000 ,25, 533–537. [C ossRe ]