Acoustic emission signal characterisation of failure mechanisms in CFRP composites using dual-sensor approach and spectral clustering technique

Ci a ion: Šo e , M.; Šo e , P.; Pagáˇc, M.; Voloda skaja, A.; Babiuch, M.; G uˇn, F. Acous ic Emission Signal Cha ac e isa ion o Failu e Mechanisms in CFRP Composi es Using Dual-Senso App oach and Spec al Clus e ing Technique. Polyme s 2023,15, 47. h ps:// doi.o g/10.3390/polym15010047 Academic Edi o : Mau o Za elli Recei ed: 16 No embe 2022 Re ised: 16 No embe 2022 Accep ed: 15 Decembe 2022 Published: 22 Decembe 2022 Copy igh : © 2022 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/). polyme s A icle Acous ic Emission Signal Cha ac e isa ion o Failu e Mechanisms in CFRP Composi es Using Dual-Senso App oach and Spec al Clus e ing Technique Michal Šo e 1,*, Pa el Šo e 2, Ma ek Pagáˇc 3, Anas asia Voloda skaja 4, Ma ek Babiuch 2and Filip G uˇn 5 1 Depa men o Applied Mechanics, Facul y o Mechanical Enginee ing, VSB—Technical Uni e si y o Os a a, 17. lis opadu 2172/15, 708 00 Os a a, Czech Republic 2Depa men o Con ol Sys ems and Ins umen a ion, Facul y o Mechanical Enginee ing, VSB—Technical Uni e si y o Os a a, 17. lis opadu 2172/15, 708 00 Os a a, Czech Republic 3Depa men o Machining, Assembly and Enginee ing Technology, Facul y o Mechanical Enginee ing, VSB—Technical Uni e si y o Os a a, 17. lis opadu 2172/15, 708 00 Os a a, Czech Republic 4Depa men o RMSTC, Facul y o Ma e ials Science and Technology, VSB—Technical Uni e si y o Os a a, 17. lis opadu 2172/15, 708 00 Os a a, Czech Republic 5BREBECK Composi e s. .o., Volenská1718, 739 34 Šeno u Os a y, Czech Republic *Co espondence: [email p o ec ed]; Tel.: +420-731-664-248 Abs ac : The cha ac e isa ion o ailu e mechanisms in ca bon ib e- ein o ced polyme (CFRP) ma e ials using he acous ic emission (AE) echnique has been he opic o a numbe o publica ions. Howe e , i is o en challenging o ob ain comp ehensi e and eliable in o ma ion abou indi idual ailu e mechanisms. This si ua ion was he impe us o elabo a ing a comp ehensi e o e iew ha co e s all ailu e mechanisms wi hin he amewo k o CFRP ma e ials. Thus, we pe o med ensile and compac ension es s on specimens wi h a ious s acking sequences o induce speci ic ailu e modes and mechanisms. The AE ac i i y was moni o ed using wo di e en wideband AE senso s and u he analysed using a hyb id AE hi de ec ion p ocess. The da ase s ecei ed om bo h senso s we e sepa a ely subjec ed o clus e ing analysis using he spec al clus e ing echnique, which inco po a ed an unsupe ised k-means clus e ing algo i hm. The ailu e mechanism analysis also included a p oposed il e ing p ocess based on he powe dis ibu ion ac oss he conside ed equency ange, wi h which i was possible o dis inguish be ween he ib e pull-ou and ib e b eakage mechanisms. This unc ionali y was pa icula ly use ul in cases whe e i was e iden ha he abo e-men ioned damage mechanisms exhibi ed e y simila pa ame ic cha ac e is ics. The esul s o he clus e ing analysis we e compa ed o hose o he scanning elec on mic oscopy analysis, which con i med he conclusions o he AE da a analysis. Keywo ds: acous ic emission; CFRP; ailu e mechanism; spec al clus e ing 1. In oduc ion Composi e ma e ials in many o ms ha e become inc easingly popula in ecen decades, especially in he medical, au omo i e, and ae ospace indus ies. In pa icula , ca bon- ib e- ein o ced polyme composi es ha e been used in a wide ange o spo ing ac i i ies such as cycling (bicycle ames and accesso ies), yach ing (hull, mas , and o he cons uc ion elemen s), o mo o spo s (bonne , bike ame, and wheel ims). The main ad an ages o CFRP composi es a e ha hey ha e a e y high s eng h/weigh a io and ha hei p ope ies can be cus omised o a speci ic pu pose by changing he s acking sequence o he indi idual plies [ 1 ]. Howe e , hei main d awback is hei inabili y o ail in a duc ile manne [ 2 ]. Mo eo e , he inal ac u e occu s as a esul o g adual accumula ing damage, which is ba ely isible on he ou side su ace [ 3 ]. This is one o he main easons o adop ing a sui able me hod o he comp ehensi e analysis and cha ac e isa ion o ailu e mechanisms in CFRP composi es. Polyme s 2023,15, 47. h ps://doi.o g/10.3390/polym15010047 h ps://www.mdpi.com/jou nal/polyme s Polyme s 2023,15, 47 2 o 25 Cu en ly, nume ous non-des uc i e es ing (NDT) me hods, such as ul asonic es - ing [ 4 ], eddy cu en es ing [ 5 ], and X- ay omog aphy [ 6 ], ha e been applied o de ec and/o localise de ec s in composi es. Howe e , hese me hods ha e se e al limi a ions ega ding eal- ime e alua ion, di icul easibili y o he oluminal po ion o he ma e ial, which can be es ed in eal ime. Hence, he acous ic emission me hod [ 7 ] has been used in a la ge numbe o s udies [ 8 – 11 ] because i o e s a p ac ical and e ec i e solu ion compa ed o he me hods lis ed abo e. The AE me hod belongs o a g oup o NDT me hods ha a e ully suppo ed by in e na ional s anda ds o applica ions in echnical p ac ice. The me hod is highly e ec i e o es ing me allic p essu e equipmen [ 12 ], such as pipelines, boile s, seamless s eel essels, au ocla es, and composi e essels, whe e he modal AE echnique is o en used [ 13 ]. The AE me hod is based on he de ec ion o acous ic s ess wa es o igina ing om a ma e ial as a esul o sudden s ess edis ibu ion owing o he p esen mic os uc u al damage, which is igge ed by he applied load [ 12 ]. The cap u ed AE signals a e hen analysed, and hey may exhibi di e en deg ees o complexi y. These di e en deg ees o complexi y a e pa icula ly impo an in he case o composi e ma e ials, which ha e se e al ypes o ailu e mechanisms [ 14 ], such as ib e b eakage, ib e/ma ix debonding, delamina ion, o ma ix c acking. Nume ous s udies ha e been de o ed o he cha ac e isa ion and me hodology o ailu e mechanisms. The mos basic app oach is pa ame e -based analysis [ 15 ] o he de ec ed AE signals using ea u es such as ampli ude, ise ime, ene gy, du a ion, and equency pa ame e s in he o m o equency cen oids o peak equencies [ 16 ]. This app oach is less e ec i e o composi e s uc u es, mos ly because o he o e all complexi y o he ailu e mechanisms, which may occu simul aneously. As a esul , he de ec ed AE signals may be associa ed wi h mo e han one ailu e mechanism [ 2 , 17 ]. One o he majo d awbacks o pa ame e -based analysis is he conside able pa ame e sensi i i y o he geome y, senso ype, o dis ance be ween he senso and he AE sou ce ( he in luence o signal a enua ion) [2]. A mo e ecen app oach is signal-based analysis [ 16 ], whose de elopmen and subse- quen applica ion ha e been highly condi ioned by AE ins umen a ion. In his app oach, indi idually de ec ed AE wa e o ms a e eco ded and subjec ed o equency analysis, mos ly in he o m o he Fou ie ans o m [3], wa ele ans o m [18,19], wa ele packe ans o m [20], o Hilbe –Huang ans o m [21]. In he las wo decades, supe ised/unsupe ised clus e ing and pa e n ecogni ion algo i hms [ 22 ] based on mul i-pa ame e analysis [ 2 ] ha e been adop ed in a la ge num- be o s udies because o hei abili y o u ilise ea u es om bo h pa ame e -based and signal-based analyses. In he supe ised app oach, he inpu signals a e associa ed wi h a gi en class o clus e , which, o example, ep esen s a pa icula ailu e mechanism using he aining da ase . In he unsupe ised app oach, signals wi h simila cha ac e is ics a e g ouped wi hou using he aining da ase . The mos commonly used clus e ing me hod among he unsupe ised app oaches is k-means [ 23 – 25 ], ollowed by i s a ian s such as k-means++ clus e ing [ 26 ], uzzy c-means clus e ing [ 27 , 28 ], o gene ic k-means clus- e ing [ 29 ], which may be combined wi h o he app oaches such as sel -o ganising maps (SOMs) [30] o p incipal componen analysis (PCA) o imp o e hei e ec i eness [31]. Al hough he e is a ela i ely ex ensi e knowledge base o a ious me hodologies o iden i ying ailu e mechanisms in CFRP composi es, ce ain issues s ill a ise ega ding hei o e all cha ac e isa ion. Fo example, i he ailu e mechanisms a e classi ied based on he peak equency, some inconsis en esul s a e ob ained, al hough he same ype o AE senso (WD senso ; Physical Acous ics Co po a ion) is used [ 32 ] (see Table 1). This can be a ibu ed o a numbe o ac o s, such as he es ed ma e ial, specimen geome y, a ia ions in he senso esponse, o p ocessing me hodology. The g ea es a ia ions in e ms o peak equency occu in he case o ib e b eakage o ib e pull-ou , ollowed by delamina ion and ib e/ma ix debonding. Polyme s 2023,15, 47 3 o 25 Table 1. Peak equencies o AE signals (in kHz) o di e en ailu e mechanisms in CFRP composi es ( he esul s a e ob ained wi h a WD senso ). Re e ence Ma ix C acking Delamina ion Fib e/Ma ix Debonding Fib e B eakage Fib e Pull-Ou Peak F equency [kHz] [3] 80–120 120–170 170–200 ~250 - [33] 90–180 - 240–310 >300 180–240 [34] <50 50–150 200–300 400–500 500–600 [30] <150 150–300 - >400 - [35] 60–120 120–210 - 200–350 - The aim o his s udy is o in es iga e he cha ac e is ics o he ailu e mechanisms o CFRP composi e ma e ials. Two es se ies a e conduc ed in his s udy: he i s es se ies in ol es only one ailu e mechanism, which is ully cha ac e ised, and he subsequen es se ies in ol es se e al ypes o ailu e mechanisms. Two wideband senso s [ 22 ] a e used o moni o he AE ac i i y. Unlike mul i esonan / esonan senso s, wideband senso s p o ide a mo e accu a e cha ac e isa ion o AE signals in e ms o hei pa ame e s. The collec ed AE signals a e hen analysed using he spec al clus e ing echnique, which u ilises he k-means unsupe ised pa e n ecogni ion app oach. Subsequen ly, he clus e ing esul s a e associa ed wi h he indi idual ypes o es s pe o med, ollowed by a discussion ela ed o he ailu e mechanism cha ac e isa ion, which is addi ionally suppo ed by scanning elec on mic oscopy (SEM) esul s. 2. Expe imen s The expe imen s in ol ed wo es s: ensile and compac ension (CT) es s. The ensile es was conduc ed on ca bon ib e bundle (CFB) (Ha el Composi es CZ Company L d., Czech Republic) and c oss-ply specimens, whe eas he CT es was conduc ed on bulk esin (BRCT), unidi ec ional (UCT), and c oss-ply (CPCT) specimens, which a e shown in Figu e 1. No e ha each sample ype was ep esen ed by ou pieces. The BRCT specimens we e made o LH 385 epoxy esin wi h cu ing agen H512 (Ha el Composi es CZ Company L d., Czech Republic). The UCT and CPCT specimens we e made o high-pe o mance ca bon/epoxy p ep eg CM-P eg T-C-230/600 CP004 39 wi h a nominal ply hickness o 0.25 mm. The BRCT and UCT/CPCT pla es we e cu ed based on he schedule p o ided by he manu ac u e . De ailed in o ma ion abou he indi idual specimens is p esen ed in Tables 2and 3. Polyme s 2022, 13, x FOR PEER REVIEW 3 o 25 e ms o peak equency occu in he case o ib e b eakage o ib e pull-ou , ollowed by delamina ion and ib e/ma ix debonding. Table 1. Peak equencies o AE signals (in kHz) o di e en ailu e mechanisms in CFRP compo- si es ( he esul s a e ob ained wi h a WD senso ). Re e ence Ma ix C acking Delamina ion Fib e / Ma ix Debonding Fib e B eakage Fib e Pull-Ou Peak F equency [kHz] [3] 80–120 120–170 170–200 ~250 - [33] 90–180 - 240–310 >300 180–240 [34] <50 50–150 200–300 400–500 500–600 [30] <150 150–300 - >400 - [35] 60–120 120–210 - 200–350 - The aim o his s udy is o in es iga e he cha ac e is ics o he ailu e mechanisms o CFRP composi e ma e ials. Two es se ies a e conduc ed in his s udy: he i s es se ies in ol es only one ailu e mechanism, which is ully cha ac e ised, and he subsequen es se ies in ol es se e al ypes o ailu e mechanisms. Two wideband senso s [22] a e used o moni o he AE ac i i y. Unlike mul i esonan / esonan senso s, wideband senso s p o- ide a mo e accu a e cha ac e isa ion o AE signals in e ms o hei pa ame e s. The col- lec ed AE signals a e hen analysed using he spec al clus e ing echnique, which u ilises he k-means unsupe ised pa e n ecogni ion app oach. Subsequen ly, he clus e ing e- sul s a e associa ed wi h he indi idual ypes o es s pe o med, ollowed by a discussion ela ed o he ailu e mechanism cha ac e isa ion, which is addi ionally suppo ed by scanning elec on mic oscopy (SEM) esul s. 2. Expe imen s The expe imen s in ol ed wo es s: ensile and compac ension (CT) es s. The en- sile es was conduc ed on ca bon ib e bundle (CFB) (Ha el Composi es CZ Company L d., Czech Republic) and c oss-ply specimens, whe eas he CT es was conduc ed on bulk esin (BRCT), unidi ec ional (UCT), and c oss-ply (CPCT) specimens, which a e shown in Figu e 1. No e ha each sample ype was ep esen ed by ou pieces. The BRCT specimens we e made o LH 385 epoxy esin wi h cu ing agen H512 (Ha el Composi es CZ Company L d., Czech Republic). The UCT and CPCT specimens we e made o high- pe o mance ca bon/epoxy p ep eg CM-P eg T-C-230/600 CP004 39 wi h a nominal ply hickness o 0.25 mm. The BRCT and UCT/CPCT pla es we e cu ed based on he schedule p o ided by he manu ac u e . De ailed in o ma ion abou he indi idual specimens is p esen ed in Tables 2 and 3. (b) (a) (c) Figu e 1. Geome y o he es ed specimens [mm]: (a) BRCT, CPCT, and UCT samples; (b) CPT sample; (c) CPTS sample. Figu e 1. Geome y o he es ed specimens [mm]: ( a ) BRCT, CPCT, and UCT samples; ( b ) CPT sample; (c) CPTS sample. Polyme s 2023,15, 47 4 o 25 Table 2. Ma e ial p ope ies o he es specimens. Sample Iden i ica ion Ma e ial Pa ame e s Bulk esin (BRCT) LH 385 epoxy esin wi h cu ing agen H512 E = 3.1 GPa (ex), Rm= 58 MPa (ex) Unidi ec ional (UCT) C oss-ply (CPT, CPCT, CPTS) ca bon/epoxy p ep eg CM-P eg T-C-230/600 CP004 39 nominal ply hickness = 0.25 mm, esin con en = 39% E(0◦) = 135 GPa (mmd), Rm(0◦) = 1900 MPa (mmd) Ca bon ib e bundle (CFB) E = 240 GPa (mmd), Rm= 4 GPa (mmd) No es: E—Young’s modulus, Rm—Ul ima e ensile s eng h, ex—expe imen ally ob ained, mmd—manu ac u e ’s ma e ial da a. Table 3. Cha ac e isa ion o he es specimens. Type o Tes Name/Numbe o Specimens S acking Sequence Thickness [mm] Tensile es Ca bon ib e bundle (CFB)/4 - - Tensile es C oss-ply (CPT)/4 (90◦, 0◦)4S 2 Tensile es C oss-ply (CPTS)/4 (45◦,−45◦)4S 2 Compac ension Unidi ec ional (UCT)/4 (90◦)41 Compac ension C oss-ply (CPCT)/4 (90◦, 0◦)4S 2 Compac ension Bulk esin (BRCT)/4 - 5 Bo h es s we e conduc ed on he Tes ome ic M500-50CT uni e sal es ing machine in de o ma ion-con olled mode wi h an uppe g ip speed o 1 mm/min o he ensile es o he CFB specimens and 0.5 mm/min o he o he es s. Addi ionally, a load cell o 100 N was u ilised o he ensile es o he CFB specimens and he CT es o he UCT specimens, whe eas a 50 kN load cell was u ilised o he CT es o he BRCT and CPCT specimens, as well as he ensile es o he CPT and CPTS specimens. 2.1. Acous ic Emission Moni o ing The AE ac i i y was eco ded using wo wideband AE senso s, DWC 454 and Fujice a 1045S, which we e di ec ly connec ed o a Vallen AEP5H 40 dB p eampli ie . The al eady ampli ied signal is hen ans e ed o he AE uni , whe e basic signal p ocessing akes place, ollowed by a ans e o he AE da a o he PC o he en i e da a p ocessing p ocedu e (see Figu e 2). I mus be emphasised ha he maximum dis ance be ween he axis o he senso and he a ea in which he AE ac i i y occu ed was no g ea e han 25 mm. Thanks o his, i was possible o la gely neglec he dependence o he equency spec um on he dis ance om he AE sou ce. Bo h u ilised AE senso s a e designed as wideband senso s wi h a usable ange om 50 o 1200 kHz, while in he case o Fujice a, i is possible o coun on an e en wide band due o inc eased sensi i i y a equencies abo e 1 MHz. Polyme s 2022, 13, x FOR PEER REVIEW 4 o 25 Table 2. Ma e ial p ope ies o he es specimens. Sample Iden i ica ion Ma e ial Pa ame e s Bulk esin (BRCT) LH 385 epoxy esin wi h cu ing agen H512 E = 3.1 GPa (ex), Rm = 58 MPa (ex) Unidi ec ional (UCT) C oss-ply (CPT, CPCT, CPTS) ca bon/epoxy p ep eg CM-P eg T-C-230/600 CP004 39 nominal ply hickness = 0.25 mm, esin con en = 39% E(0°) = 135 GPa (mmd), Rm(0°) = 1900 MPa (mmd) Ca bon ib e bundle (CFB) E = 240 GPa (mmd), Rm = 4 GPa (mmd) No es: E—Young’s modulus, Rm—Ul ima e ensile s eng h, ex—expe imen ally ob ained, mmd— manu ac u e ’s ma e ial da a. Table 3. Cha ac e isa ion o he es specimens. Type o Tes Name/Numbe o Specimens S acking Sequence Thickness [mm] Tensile es Ca bon ib e bundle (CFB)/4 - - Tensile es Tensile es C oss-ply (CPT)/4 C oss-ply (CPTS)/4 (90o, 0o)4S (45o, −45o)4S 2 2 Compac ension Unidi ec ional (UCT)/4 (90o)4 1 Compac ension C oss-ply (CPCT)/4 (90o, 0o)4S 2 Compac ension Bulk esin (BRCT)/4 - 5 Bo h es s we e conduc ed on he Tes ome ic M500-50CT uni e sal es ing machine in de o ma ion-con olled mode wi h an uppe g ip speed o 1 mm/min o he ensile es o he CFB specimens and 0.5 mm/min o he o he es s. Addi ionally, a load cell o 100 N was u ilised o he ensile es o he CFB specimens and he CT es o he UCT speci- mens, whe eas a 50 kN load cell was u ilised o he CT es o he BRCT and CPCT spec- imens, as well as he ensile es o he CPT and CPTS specimens. 2.1. Acous ic Emission Moni o ing The AE ac i i y was eco ded using wo wideband AE senso s, DWC 454 and Fu- jice a 1045S, which we e di ec ly connec ed o a Vallen AEP5H 40 dB p eampli ie . The al eady ampli ied signal is hen ans e ed o he AE uni , whe e basic signal p ocessing akes place, ollowed by a ans e o he AE da a o he PC o he en i e da a p ocessing p ocedu e (see Figu e 2). I mus be emphasised ha he maximum dis ance be ween he axis o he senso and he a ea in which he AE ac i i y occu ed was no g ea e han 25 mm. Thanks o his, i was possible o la gely neglec he dependence o he equency spec um on he dis ance om he AE sou ce. Bo h u ilised AE senso s a e designed as wideband senso s wi h a usable ange om 50 o 1200 kHz, while in he case o Fujice a, i is possible o coun on an e en wide band due o inc eased sensi i i y a equencies abo e 1 MHz. Figu e 2. Schema ic diag am o he expe imen al se up. Figu e 2. Schema ic diag am o he expe imen al se up. Polyme s 2023,15, 47 5 o 25 Figu e 3displays he esul s o ace- o- ace ela i e calib a ion o bo h senso s using he Vallen Senso Tes e ® u ili y (Manu ac u e : Vallen Sys eme GmbH, Ge many). The cap u ed signals we e hen p ocessed by a Vallen AMSY-6 AE sys em uni comp ising an ASIP-2A signal p ocesso ca d wi h sampling equency and equency il e se o 10 MHz and 50–1100 kHz, espec i ely. Polyme s 2022, 13, x FOR PEER REVIEW 5 o 25 Figu e 3. Face− o− ace calib a ion esul s o DWC B454 and Fujice a 1045S AE senso s using he Vallen Senso Tes e ® u ili y. Figu e 3 displays he esul s o ace- o- ace ela i e calib a ion o bo h senso s using he Vallen Senso Tes e ® u ili y (Manu ac u e : Vallen Sys eme GmbH, Ge many). The cap u ed signals we e hen p ocessed by a Vallen AMSY-6 AE sys em uni comp ising an ASIP-2A signal p ocesso ca d wi h sampling equency and equency il e se o 10 MHz and 50–1100 kHz, espec i ely. Mos comme cially a ailable AE de ices use he ixed- h eshold p inciple o hi de- ec ion. The main disad an age o his p inciple is he endency o a i icially combine hi s ha eme ge in apid succession. The eason o his beha iou is ha he signal unde he gi en ci cums ances does no all below he h eshold; he e o e, he hi is no in e up ed. As a esul , such hi s con ain a numbe o o he wise sepa a ed hi s, leading o signi ican inaccu acies in he cha ac e isa ion o AE signals in some cases ( ha is, specimen iola ion du ing he ensile es ). Al hough his beha iou can be educed o some ex en by using sui able pa ame e s o he hi sepa a ion p ocess o by inc easing he sensing h eshold, hese s eps do no lead o he comple e elimina ion o his phenomenon. In 2017, Sha e i e al. p oposed a oo mean squa e (RMS) AE hi de ec ion algo i hm [15] ha can o e come he p oblems associa ed wi h he h eshold-based echnique by using he in o ma ion s o ed in he RMS en elope o he AE signal. Subsequen ly, he join - alley peak- inding algo i hm is hen applied o de ec sepa a e e en s. The unc- ionali y o he en i e algo i hm is no desc ibed in his pape , as i has been explained in de ail by Sha e i e al. [15]. Howe e , we modi ied his algo i hm o u he imp o e i s pe o mance. The RMS-based AE hi de ec ion algo i hm uses h ee unc ionali ies o ensu e he de ec ion o possible AE hi s, namely wo h esholds: one o he RMS and he o he o he o iginal AE signal, and he a io be ween he p esen peak and alley wi hin he RMS en elope, which is hen used o alida e whe he he de ec ed AE hi is alid. Sha e i e al. [15] de i ed an exp ession o calcula ing a gi en peak/ alley a io depending on he RMS en elope and he o iginal ime signal h esholds. Howe e , we ound ha his a io depends on a numbe o ac o s, such as he size o he ime window o calcula ing he RMS alue, he ype o senso , na u e o he AE signal, and he geome y being es ed. Addi ionally, o a mo e e icien sepa a ion o he AE hi s, we implemen ed a wo- le el h eshold ( ixed and loa ing) o he RMS signal wa e o m. In p ac ice, his h esh- old segmen s po en ial hi s in he i s phase using a p ede e mined ixed RMS h eshold, and hen, depending on he RMS alue o he backg ound noise be o e he s a o he hi , he maximum ampli ude o he RMS wa e o m is se as a new RMS h eshold, allowing mo e e icien hi segmen a ion. The p ocedu e abo e can be applied o wa e ansien s, which a e ob ained by he con inuous eco ding o wa e o ms o a gi en measu ing channel o by using a ixed ampli ude h eshold by he p ocedu e below: Figu e 3. Face − o − ace calib a ion esul s o DWC B454 and Fujice a 1045S AE senso s using he Vallen Senso Tes e ®u ili y. Mos comme cially a ailable AE de ices use he ixed- h eshold p inciple o hi de ec ion. The main disad an age o his p inciple is he endency o a i icially combine hi s ha eme ge in apid succession. The eason o his beha iou is ha he signal unde he gi en ci cums ances does no all below he h eshold; he e o e, he hi is no in e up ed. As a esul , such hi s con ain a numbe o o he wise sepa a ed hi s, leading o signi ican inaccu acies in he cha ac e isa ion o AE signals in some cases ( ha is, specimen iola ion du ing he ensile es ). Al hough his beha iou can be educed o some ex en by using sui able pa ame e s o he hi sepa a ion p ocess o by inc easing he sensing h eshold, hese s eps do no lead o he comple e elimina ion o his phenomenon. In 2017, Sha e i e al. p oposed a oo mean squa e (RMS) AE hi de ec ion algo- i hm [15] ha can o e come he p oblems associa ed wi h he h eshold-based echnique by using he in o ma ion s o ed in he RMS en elope o he AE signal. Subsequen ly, he join - alley peak- inding algo i hm is hen applied o de ec sepa a e e en s. The unc- ionali y o he en i e algo i hm is no desc ibed in his pape , as i has been explained in de ail by Sha e i e al. [ 15 ]. Howe e , we modi ied his algo i hm o u he imp o e i s pe o mance. The RMS-based AE hi de ec ion algo i hm uses h ee unc ionali ies o ensu e he de ec ion o possible AE hi s, namely wo h esholds: one o he RMS and he o he o he o iginal AE signal, and he a io be ween he p esen peak and alley wi hin he RMS en elope, which is hen used o alida e whe he he de ec ed AE hi is alid. Sh a e i e al. [15] de i ed an exp ession o calcula ing a gi en peak/ alley a io depending on he RMS en elope and he o iginal ime signal h esholds. Howe e , we ound ha his a io depends on a numbe o ac o s, such as he size o he ime window o calcula ing he RMS alue, he ype o senso , na u e o he AE signal, and he geome y being es ed. Addi ionally, o a mo e e icien sepa a ion o he AE hi s, we implemen ed a wo- le el h eshold ( ixed and loa ing) o he RMS signal wa e o m. In p ac ice, his h eshold segmen s po en ial hi s in he i s phase using a p ede e mined ixed RMS h eshold, and hen, depending on he RMS alue o he backg ound noise be o e he s a o he hi , he maximum ampli ude o he RMS wa e o m is se as a new RMS h eshold, allowing mo e e icien hi segmen a ion. Polyme s 2023,15, 47 6 o 25 The p ocedu e abo e can be applied o wa e ansien s, which a e ob ained by he con inuous eco ding o wa e o ms o a gi en measu ing channel o by using a ixed ampli ude h eshold by he p ocedu e below: 1. The AE ac i i y was eco ded a a ixed h eshold alue o 28 dB AE , while he du a ion disc imina ion ime (DDT) and ea m ime (RAT) we e se o 100 µs; No e: Acco ding o Vallen documen a ion, DDT is he pe iod in which no h eshold c ossing mus occu o de e mine he end o a hi , whe eas RAT is he pe iod a e which he channel is eady o gene a e a new hi da ase ; 2 The eco ded wa e ansien s we e subsequen ly subjec ed o an addi ional hi seg- men a ion p ocess wi h he ollowing pa ame e s: wid h o he RMS window = 10 µ s, ixed alue o he RMS h eshold = 10 dB AE . The loa ing alue o he RMS h eshold is gi en by: ms loa ing = msmean +0.12[max( ms)− msmean](1) whe e ms mean is he mean alue o he RMS be o e he s a o he AE hi (mean alue om he i s 10 samples be o e he hi s a ), max( ms) is he maximum alue o he RMS eco ding o he gi en AE hi , and peak/ alley a io = 0.25 (maximum) o alid pai s exceeding he p incipal h eshold o 28 dBAE. The AE signal pa ame e s [ 36 ] o ea u es ac as inpu a iables o he subsequen clus e ing p ocess. These ea u es can be u he di ided in o di ec ly ex ac ed ea u es and hose ha a e calcula ed du ing pos -p ocessing. Table 4lis s all he ea u es used in his s udy and hei desc ip ions. Table 4. Cha ac e isa ion o AE ea u es used in his s udy. Fea u e Desc ip ion Ampli ude (E) La ges ol age peak o he gi en AE hi (in dB AE ) (dB el. o 1 µ V be o e he inpu o he p eampli ie ). Rise ime (E) Time in e al (in µ s) be ween he i s h eshold c ossing and he eached maximum ampli ude (in µs). Du a ion (E) Time in e al (in µ s) be ween he i s and las h eshold c ossings (in µ s). Ene gy (E) In eg al o he squa ed AE signal o e ime (E=RDU( )2d ,D e e s o he AE hi du a ion) (in aJ). p—Peak equency (E) F equency co esponding o he maximum magni ude in he equency spec um (in kHz). c—F equency cen oid (E) Cen e o mass o he equency spec um (in kHz). pw—Weigh ed peak equency (C) Squa e oo o he p oduc be ween he peak equency and equency cen oid [37], namely pw =p p c(in kHz). RA—RA alue (C) p I ÷p VI—Pa ial powe (C) Rise ime/peak ampli ude a io (in µs/dBAE) [38]. (RA is some imes called he ise angle). Non-dimensionalised a io be ween he powe in he equency in e al I÷VI and he powe o he en i e equency spec um, ha is, wi hin he 〈50, 1100〉[kHz] equency in e al. No e: I=h50,225), II =h225,400), III =h400,575), IV =h575, 750), V=h750,925), VI =h925,1100) [kHz]. No e: E—ex ac ed ea u e, C—calcula ed ea u e. 2.2. Failu e Mechanisms 2.2.1. Tensile Tes o Fib e Bundles and Compac Tension Tes o Bulk Resin The AE signals o igina ing om ib e b eakage we e in es iga ed by conduc ing ensile es s on he ib e bundles, which we e adequa ely adhe ed o he clamps using a cyanoac yla e adhesi e. To elimina e he possible occu ence o ib e pull-ou in he clamps, he pe iphe al pa s o he ib e bundle we e equipped wi h gua d senso s (DAKEL MIDI), as shown in Figu e 4a. Be o e he ac ual e alua ion o he AE da a om he ensile es esul s o he ca bon ib e bundles, i was necessa y o pe o m da a il e ing Polyme s 2023,15, 47 7 o 25 because he ailu e o ca bon ib es, which a e no sepa a ed by a ma ix, has comple ely di e en cha ac e is ics compa ed o con en ional composi es. Mo eo e , a ela i ely la ge pe cen age o he AE ac i i y comes om he mu ual ic ion o he ib es, which has an inc easing endency, especially when a la ge numbe o ib es ail a he same ime. The b oken ib es hen begin o slide o e hose ha s ill ca y a ce ain load, wi h he na u e o he de ec ed signal being e y simila o ha co esponding o he ib e ailu e. Hence, i was necessa y o ind a sui able me hodology o sepa a ing he use ul signal (in ou case, ib e ailu e) om he seconda y AE. Fo he gi en pu poses, an e alua ion algo i hm in which he il e ing o AE hi s coupled wi h ib e b eakage is based on he de ec ion o a dec ease in o ce o e a ime cou se was de eloped wi h MATLAB so wa e. The in e al o in e es is de ined by he de ec ion o a nega i e g adien and he momen o e u ning he o ce o i s o iginal alues, which includes he ex ension o bo h bo de s by he o ce sampling pe iod (100 ms). Only hi s alling wi hin hese in e als a e subsequen ly analysed. The only sho coming o he gi en app oach is ha i is no possible o il e ou he seconda y AE emission i i occu s simul aneously wi h a ib e b eak, which is associa ed wi h in e al epo ing o ce d ops. 1 (a) (b) Figu e 4. In si u pho og aphs o he ( a ) ensile es o he ca bon ib e bundle, ( b ) BRCT specimen, including he p esen uns able mac oc ack. Figu e 5shows he ela ionship be ween he peak equency and o ce e sus displace- men o he ep esen a i e specimen o he ca bon ib e bundle a e he il e ing p ocess o he esul s om bo h AE senso s ( he blue highligh ed segmen s indica e he de ec ed o ce dips). The esul s om he DWC 454 AE senso showed ha , pa icula ly in he ini ial phase, AE signals wi h a peak equency abo e 500 kHz exis ed (see Figu e 5a). A e eaching he maximum o ce, he ailu e o he ib es began o appea on a sligh ly wide peak equency scale. This was mainly due o he supe posi ion o signals o igina ing om bo h he ib e b eakage and he abo e-men ioned ic ion be ween he ib es in he bundle, which occu s along he en i e leng h o he bundle. A simila scena io was obse ed in he esul s ob ained om he Fujice a 1045S AE senso , whe e he uppe limi o he peak equency in e al shi ed o 1000 kHz, whe eas he lowe in e al s a ed a app oxima ely 300 kHz (see Figu e 5b). Fo he subsequen clus e ing p ocess, we conside ed only he AE hi s om bo h senso s wi h a peak equency highe han 400 kHz, which seemed like a easonable comp omise o bo h da ase s. Rega ding he CT es s o he bulk esin specimens, he only mechanisms obse ed we e ma ix mic oc acking and mac oc acking, which ha e he cha ac e is ic o b i le ac u e. Because he ini ia ion p ocess o mic oc acks is p ac ically unde ec able, he only mechanism co e ed by he AE ac i i y is he uns able p opaga ion o mac oc acks associa ed wi h an o e all collapse o he s uc u al in eg i y (see Figu e 4b). Polyme s 2023,15, 47 8 o 25 Polyme s 2022, 13, x FOR PEER REVIEW 8 o 25 o igina ing om bo h he ib e b eakage and he abo e-men ioned ic ion be ween he ib es in he bundle, which occu s along he en i e leng h o he bundle. A simila scena io was obse ed in he esul s ob ained om he Fujice a 1045S AE senso , whe e he uppe limi o he peak equency in e al shi ed o 1000 kHz, whe eas he lowe in e al s a ed a app oxima ely 300 kHz (see Figu e 5b). Fo he subsequen clus e ing p ocess, we conside ed only he AE hi s om bo h senso s wi h a peak e- quency highe han 400 kHz, which seemed like a easonable comp omise o bo h da- ase s. (a) (b) Figu e 5. Peak equency and o ce e sus displacemen esul s o he ib e bundle ensile es . (a) DWC 454 AE senso ; (b) Fujice a 1045S senso . Rega ding he CT es s o he bulk esin specimens, he only mechanisms obse ed we e ma ix mic oc acking and mac oc acking, which ha e he cha ac e is ic o b i le ac u e. Because he ini ia ion p ocess o mic oc acks is p ac ically unde ec able, he only mechanism co e ed by he AE ac i i y is he uns able p opaga ion o mac oc acks associ- a ed wi h an o e all collapse o he s uc u al in eg i y (see Figu e 4b). 2.2.2. Compac Tension Tes o Unidi ec ional Specimens The CT es s o he UCT specimens e ealed he p esence o se e al ailu e mecha- nisms, namely ma ix mic oc acking, ib e/ma ix debonding, and ib e ailu e, as shown in Figu e 6b. A e he ini ial phase o mic oc ack ini ia ion and coalescence, he p opaga- ion o he mac oc ack began, which was cha ac e ised by a sudden d op in o ce. Mean- while, he ib e-b idging phenomenon [39] (Figu e 6a) was obse ed in he CT es s o he UCT specimens, indica ing ha his phenomenon signi ican ly con ibu es o ib e/ma ix debonding and ib e ailu e, which ag ees wi h he esul s o Gu kin e al. [34]. Figu e 5. Peak equency and o ce e sus displacemen esul s o he ib e bundle ensile es . (a) DWC 454 AE senso ; (b) Fujice a 1045S senso . 2.2.2. Compac Tension Tes o Unidi ec ional Specimens The CT es s o he UCT specimens e ealed he p esence o se e al ailu e mecha- nisms, namely ma ix mic oc acking, ib e/ma ix debonding, and ib e ailu e, as shown in Figu e 6b. A e he ini ial phase o mic oc ack ini ia ion and coalescence, he p opaga ion o he mac oc ack began, which was cha ac e ised by a sudden d op in o ce. Mean- while, he ib e-b idging phenomenon [39] (Figu e 6a) was obse ed in he CT es s o he UCT specimens, indica ing ha his phenomenon signi ican ly con ibu es o ib e/ma ix debonding and ib e ailu e, which ag ees wi h he esul s o Gu kin e al. [34]. Polyme s 2022, 13, x FOR PEER REVIEW 9 o 25 (a) (b) Figu e 6. In si u pho og aph o he CT es o UCT specimens. (a) De ail o he ib e-b idging p ocess; (b) SEM image o he ac u e su ace o a UCT ep esen a i e specimen. 2.2.3. Compac Tension and Tensile Tes o 0–90° C oss-Ply Specimens The 0–90° c oss-ply specimens we e subjec ed o ensile and CT es s, and hei load- ing cha ac e is ics we e compa ed. The CT es was comple ed o e a longe pe iod, du - ing which he inal s age o he s uc u e iola ion was moni o ed. Meanwhile, he ensile es was cha ac e ised by an ab up collapse, du ing which a ela i ely la ge amoun o in o ma ion was los owing o he sudden elease o ene gy and he supe imposi ion o he indi idual AE signals. The SEM analysis o ep esen a i e samples om bo h ypes o es s e ealed he occu ence o delamina ion, ib e/ma ix debonding, ma ix c acking, ib e ailu e, and ib e pull-ou . The CPCT samples showed he p esence o egions wi h ensile and com- p essi e loading cha ac e is ics (Figu e 7a). Comp essi e loading was mainly esponsible o he delamina ion p ocess a he back pa o he ep esen a i e sample (Figu e 7b,c), including ib e/ma ix debonding and ma ix mic oc acking (Figu e 7e), ollowed by ib e ailu e (Figu e 7 ). In con as , he egion wi h ensile loading cha ac e is ics had no icea- ble aces o ib e pull-ou and ib e b eakage (Figu e 7d), ollowed by egions showing ib e/ma ix debonding and delamina ion. Simila obse a ions we e ound o he ensile es o he CPT samples, and he same cha ac e is ics we e obse ed o he abo e-men- ioned ailu e mechanisms. (a) (b) Figu e 6. In si u pho og aph o he CT es o UCT specimens. ( a ) De ail o he ib e-b idging p ocess; (b) SEM image o he ac u e su ace o a UCT ep esen a i e specimen. 2.2.3. Compac Tension and Tensile Tes o 0–90◦C oss-Ply Specimens The 0–90 ◦ c oss-ply specimens we e subjec ed o ensile and CT es s, and hei loading cha ac e is ics we e compa ed. The CT es was comple ed o e a longe pe iod, du ing Polyme s 2023,15, 47 9 o 25 which he inal s age o he s uc u e iola ion was moni o ed. Meanwhile, he ensile es was cha ac e ised by an ab up collapse, du ing which a ela i ely la ge amoun o in o ma ion was los owing o he sudden elease o ene gy and he supe imposi ion o he indi idual AE signals. The SEM analysis o ep esen a i e samples om bo h ypes o es s e ealed he occu ence o delamina ion, ib e/ma ix debonding, ma ix c acking, ib e ailu e, and ib e pull-ou . The CPCT samples showed he p esence o egions wi h ensile and comp essi e loading cha ac e is ics (Figu e 7a). Comp essi e loading was mainly esponsible o he delamina ion p ocess a he back pa o he ep esen a i e sample (Figu e 7b,c), including ib e/ma ix debonding and ma ix mic oc acking (Figu e 7e), ollowed by ib e ailu e (Figu e 7 ). In con as , he egion wi h ensile loading cha ac e is ics had no iceable aces o ib e pull-ou and ib e b eakage (Figu e 7d), ollowed by egions showing ib e/ma ix debonding and delamina ion. Simila obse a ions we e ound o he ensile es o he CPT samples, and he same cha ac e is ics we e obse ed o he abo e-men ioned ailu e mechanisms. Polyme s 2022, 13, x FOR PEER REVIEW 9 o 25 (a) (b) Figu e 6. In si u pho og aph o he CT es o UCT specimens. (a) De ail o he ib e-b idging p ocess; (b) SEM image o he ac u e su ace o a UCT ep esen a i e specimen. 2.2.3. Compac Tension and Tensile Tes o 0–90° C oss-Ply Specimens The 0–90° c oss-ply specimens we e subjec ed o ensile and CT es s, and hei load- ing cha ac e is ics we e compa ed. The CT es was comple ed o e a longe pe iod, du - ing which he inal s age o he s uc u e iola ion was moni o ed. Meanwhile, he ensile es was cha ac e ised by an ab up collapse, du ing which a ela i ely la ge amoun o in o ma ion was los owing o he sudden elease o ene gy and he supe imposi ion o he indi idual AE signals. The SEM analysis o ep esen a i e samples om bo h ypes o es s e ealed he occu ence o delamina ion, ib e/ma ix debonding, ma ix c acking, ib e ailu e, and ib e pull-ou . The CPCT samples showed he p esence o egions wi h ensile and com- p essi e loading cha ac e is ics (Figu e 7a). Comp essi e loading was mainly esponsible o he delamina ion p ocess a he back pa o he ep esen a i e sample (Figu e 7b,c), including ib e/ma ix debonding and ma ix mic oc acking (Figu e 7e), ollowed by ib e ailu e (Figu e 7 ). In con as , he egion wi h ensile loading cha ac e is ics had no icea- ble aces o ib e pull-ou and ib e b eakage (Figu e 7d), ollowed by egions showing ib e/ma ix debonding and delamina ion. Simila obse a ions we e ound o he ensile es o he CPT samples, and he same cha ac e is ics we e obse ed o he abo e-men- ioned ailu e mechanisms. (a) (b) Polyme s 2022, 13, x FOR PEER REVIEW 10 o 25 (c) (d) (e) ( ) Figu e 7. SEM images o a CPCT ep esen a i e specimen: (a) gene al iew, (b) s uc u e collapse due o comp essi e load, (c) ib e pull-ou , (d) delamina ion and ib e/ma ix debonding, (e) i- b e/ma ix debonding and ma ix mic oc acking, ( ) ib e ailu e. 2.2.4. Tensile Tes o ± 45° C oss-ply Specimens The SEM analysis o he CPTS specimens e ealed he p esence o delamina ion (Fig- u e 8a), ma ix mic oc acking (Figu e 8b), and ib e/ma ix debonding. (a) (b) Figu e 8. SEM images o a CPTS ep esen a i e specimen: (a) delamina ed plies, (b) ma ix mi- c oc acking and ib e/ma ix debonding. Figu e 7. Con . Polyme s 2023,15, 47 16 o 25 has o be no ed ha clus e one is ep esen ed in he g ea es abundance—see Figu e 17a,b. Clus e wo hen ep esen s ib e/ma ix debonding, which was also con i med by he SEM analysis o his ype o sample (Figu e 6b). Polyme s 2022, 13, x FOR PEER REVIEW 16 o 25 ha e lowe ene gy alues (see Figu es 15b,c and 16b,c). Based on he cumula i e AE hi s o indi idual clus e s (Figu e 17a,b), i is qui e e iden ha he clus e ou ac i i y, espe- cially i s low- equency pa , occu s immedia ely a e he onse o mac oc ack p opaga- ion. Ano he clea ly assignable sou ce is ib e ailu e, ep esen ed by clus e one (in bo h da ase s). I has o be no ed ha clus e one is ep esen ed in he g ea es abundance—see Figu e 17a,b. Clus e wo hen ep esen s ib e/ma ix debonding, which was also con- i med by he SEM analysis o his ype o sample (Figu e 6b). (a) (b) (c) Figu e 15. Clus e ing esul s o a UCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 16. Clus e ing esul s o a UCT ep esen a i e sample using he spec al clus e ing algo- i hm (DWC da ase ). (a) (b) Figu e 17. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — UCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Figu e 15. Clus e ing esul s o a UCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Polyme s 2022, 13, x FOR PEER REVIEW 16 o 25 ha e lowe ene gy alues (see Figu es 15b,c and 16b,c). Based on he cumula i e AE hi s o indi idual clus e s (Figu e 17a,b), i is qui e e iden ha he clus e ou ac i i y, espe- cially i s low- equency pa , occu s immedia ely a e he onse o mac oc ack p opaga- ion. Ano he clea ly assignable sou ce is ib e ailu e, ep esen ed by clus e one (in bo h da ase s). I has o be no ed ha clus e one is ep esen ed in he g ea es abundance—see Figu e 17a,b. Clus e wo hen ep esen s ib e/ma ix debonding, which was also con- i med by he SEM analysis o his ype o sample (Figu e 6b). (a) (b) (c) Figu e 15. Clus e ing esul s o a UCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 16. Clus e ing esul s o a UCT ep esen a i e sample using he spec al clus e ing algo- i hm (DWC da ase ). (a) (b) Figu e 17. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — UCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Figu e 16. Clus e ing esul s o a UCT ep esen a i e sample using he spec al clus e ing algo i hm (DWC da ase ). Polyme s 2022, 13, x FOR PEER REVIEW 16 o 25 ha e lowe ene gy alues (see Figu es 15b,c and 16b,c). Based on he cumula i e AE hi s o indi idual clus e s (Figu e 17a,b), i is qui e e iden ha he clus e ou ac i i y, espe- cially i s low- equency pa , occu s immedia ely a e he onse o mac oc ack p opaga- ion. Ano he clea ly assignable sou ce is ib e ailu e, ep esen ed by clus e one (in bo h da ase s). I has o be no ed ha clus e one is ep esen ed in he g ea es abundance—see Figu e 17a,b. Clus e wo hen ep esen s ib e/ma ix debonding, which was also con- i med by he SEM analysis o his ype o sample (Figu e 6b). (a) (b) (c) Figu e 15. Clus e ing esul s o a UCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 16. Clus e ing esul s o a UCT ep esen a i e sample using he spec al clus e ing algo- i hm (DWC da ase ). (a) (b) Figu e 17. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — UCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Figu e 17. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —UCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Polyme s 2023,15, 47 17 o 25 No e ha he Fujice a 1045S da ase shows a pa ial in e mingling o clus e s one and wo, i.e., ib e ailu e and ib e/ma ix debonding, which is mos likely due o he la e equency spec um compa ed o he DWC 454 AE senso in he gi en ange—see Figu e 3. The esul s o he compac ension es o CPCT samples as well as he ensile es o CPT and CPTS samples (Figu es 18–26) poin o he p esence o i e ailu e mechanisms, namely ib e b eakage (clus e one), ib e/ma ix debonding (clus e wo), ib e pull-ou (clus e h ee), ma ix c acking and delamina ion (clus e ou ), which a e di ec ly ela ed o he es ed s acking sequence con igu a ion. The CT es o he CPCT specimens e ealed a somewha di e en na u e o he ailu e e olu ion p ocess compa ed o he UCT specimens. While he AE ac i i y o he UCT specimens was igge ed by he onse o mac oc ack p opaga ion, he CPCT specimens exhibi ed conside able AE ac i i y be o e he collapse o he s uc u e. In addi ion, bo h da ase s showed he p esence o ib e/ma ix debonding in he ea ly phase, ollowed by ma ix c acking, and ib e pull-ou , especially in he case o he Fujice a 1045S da ase (see Figu es 18,19 and 24). Polyme s 2022, 13, x FOR PEER REVIEW 17 o 25 No e ha he Fujice a 1045S da ase shows a pa ial in e mingling o clus e s one and wo, i.e., ib e ailu e and ib e/ma ix debonding, which is mos likely due o he la e e- quency spec um compa ed o he DWC 454 AE senso in he gi en ange—see Figu e 3. The esul s o he compac ension es o CPCT samples as well as he ensile es o CPT and CPTS samples (Figu es 18÷26) poin o he p esence o i e ailu e mechanisms, namely ib e b eakage (clus e one), ib e/ma ix debonding (clus e wo), ib e pull-ou (clus e h ee), ma ix c acking and delamina ion (clus e ou ), which a e di ec ly ela ed o he es ed s acking sequence con igu a ion. The CT es o he CPCT specimens e ealed a somewha di e en na u e o he ailu e e olu ion p ocess compa ed o he UCT speci- mens. While he AE ac i i y o he UCT specimens was igge ed by he onse o mac- oc ack p opaga ion, he CPCT specimens exhibi ed conside able AE ac i i y be o e he collapse o he s uc u e. In addi ion, bo h da ase s showed he p esence o ib e/ma ix debonding in he ea ly phase, ollowed by ma ix c acking, and ib e pull-ou , especially in he case o he Fujice a 1045S da ase (see Figu es 18, 19, and 24). Fib e b eakage (clus e one) occu ed when a mo e signi ican s uc u al collapse was p esen (see Figu es 18b and 19b). A ela i ely simila desc ip ion o he esul s can be adop ed in he case o he ensile es o he CPT samples (Figu es 20, 21, and 25), whe e he ib e/ma ix debonding mechanism and ma ix c acking we e e en mo e p edominan han in he case o CPCT samples. The esul s o he CPTS samples o a ce ain ex en eplica e hose o he CPT samples, bu wi h he di e ence ha he p edominan de ec ed mechanism is ma ix c acking (see Figu es 22, 23, and 26). (a) (b) (c) Figu e 18. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 18. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Polyme s 2022, 13, x FOR PEER REVIEW 17 o 25 No e ha he Fujice a 1045S da ase shows a pa ial in e mingling o clus e s one and wo, i.e., ib e ailu e and ib e/ma ix debonding, which is mos likely due o he la e e- quency spec um compa ed o he DWC 454 AE senso in he gi en ange—see Figu e 3. The esul s o he compac ension es o CPCT samples as well as he ensile es o CPT and CPTS samples (Figu es 18÷26) poin o he p esence o i e ailu e mechanisms, namely ib e b eakage (clus e one), ib e/ma ix debonding (clus e wo), ib e pull-ou (clus e h ee), ma ix c acking and delamina ion (clus e ou ), which a e di ec ly ela ed o he es ed s acking sequence con igu a ion. The CT es o he CPCT specimens e ealed a somewha di e en na u e o he ailu e e olu ion p ocess compa ed o he UCT speci- mens. While he AE ac i i y o he UCT specimens was igge ed by he onse o mac- oc ack p opaga ion, he CPCT specimens exhibi ed conside able AE ac i i y be o e he collapse o he s uc u e. In addi ion, bo h da ase s showed he p esence o ib e/ma ix debonding in he ea ly phase, ollowed by ma ix c acking, and ib e pull-ou , especially in he case o he Fujice a 1045S da ase (see Figu es 18, 19, and 24). Fib e b eakage (clus e one) occu ed when a mo e signi ican s uc u al collapse was p esen (see Figu es 18b and 19b). A ela i ely simila desc ip ion o he esul s can be adop ed in he case o he ensile es o he CPT samples (Figu es 20, 21, and 25), whe e he ib e/ma ix debonding mechanism and ma ix c acking we e e en mo e p edominan han in he case o CPCT samples. The esul s o he CPTS samples o a ce ain ex en eplica e hose o he CPT samples, bu wi h he di e ence ha he p edominan de ec ed mechanism is ma ix c acking (see Figu es 22, 23, and 26). (a) (b) (c) Figu e 18. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 19. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). Polyme s 2023,15, 47 18 o 25 Polyme s 2022, 13, x FOR PEER REVIEW 18 o 25 Figu e 19. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) (c) Figu e 20. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 21. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) (c) Figu e 22. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Figu e 20. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Polyme s 2022, 13, x FOR PEER REVIEW 18 o 25 Figu e 19. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) (c) Figu e 20. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 21. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) (c) Figu e 22. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Figu e 21. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). Polyme s 2022, 13, x FOR PEER REVIEW 18 o 25 Figu e 19. Clus e ing esul s o a CPCT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) (c) Figu e 20. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). (a) (b) (c) Figu e 21. Clus e ing esul s o a CPT ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) (c) Figu e 22. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Figu e 22. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm and seconda y il e ing using a pa ial powe a iable (Fujice a 1045S da ase ). Polyme s 2023,15, 47 19 o 25 Polyme s 2022, 13, x FOR PEER REVIEW 19 o 25 (a) (b) (c) Figu e 23. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) Figu e 24. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — CPCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . (a) (b) Figu e 25. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —CPT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Figu e 23. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). Polyme s 2022, 13, x FOR PEER REVIEW 19 o 25 (a) (b) (c) Figu e 23. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) Figu e 24. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — CPCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . (a) (b) Figu e 25. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —CPT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Figu e 24. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —CPCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Polyme s 2022, 13, x FOR PEER REVIEW 19 o 25 (a) (b) (c) Figu e 23. Clus e ing esul s o a CPTS ep esen a i e sample using a spec al clus e ing algo i hm (DWC 454 da ase ). (a) (b) Figu e 24. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — CPCT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . (a) (b) Figu e 25. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —CPT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Figu e 25. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —CPT samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Polyme s 2023,15, 47 20 o 25 Polyme s 2022, 13, x FOR PEER REVIEW 20 o 25 (a) (b) Figu e 26. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — CPTS samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . b. Cha ac e isa ion o ailu e mechanisms 4.1. Delamina ion and Ma ix C acking To he bes o ou knowledge, delamina ion is a syne gy o AE signals, which a e cha ac e ised by a wide spec um o ene gy (up o 107 aJ) and ampli ude alues (up o 94 dBAE and occasionally mo e). Rega ding he peak equency, se e al au ho s ha e cha ac- e ised his phenomenon as ha ing a low peak equency, o en below 100 kHz [3,30,34,35]. This conclusion is pa ially con i med by he esul s o he BRCT samples om bo h da ase s. Howe e , clus e ing esul s om bo h da ase s, which a e ela ed o UCT, CPCT, CPT, and CPTS samples, clea ly demons a e ha he ma ix c acking/de- lamina ion phenomenon spans om 50 o app oxima ely 200 kHz (see Figu e 27). Figu e 27. Rep esen a i e AE signal o ma ix c acking (DWC 454 da ase ). 4.2. Fib e/Ma ix Debonding The epo ed ene gy and ampli ude alues o he ib e/ma ix debonding mecha- nism a e somewha lowe han hose o delamina ion/ma ix c acking, ha is, <106 aJ and <85 dBAE, espec i ely, wi hin he maximum limi ing alues, while mos AE hi s belonging o ib e/ma ix debonding ha e ampli ude and ene gy alues below 70 dBAE and 105 aJ, espec i ely. The published esul s om bo h da ase s also con i m ha bo h he ampli- ude and ene gy alues a e closely ela ed o he loading cha ac e is ics and s acking se- quence o he plies. A simila conclusion applies in he case o he peak equency, in which he equency in e al is ela ed o he loading cha ac e is ics and s acking se- quence o he plies. The 200-400 kHz peak equency ange we ound was he wides com- pa ed o hose o o he s udies (see Figu e 28) [3,33,34]. Figu e 26. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen —CPTS samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . Fib e b eakage (clus e one) occu ed when a mo e signi ican s uc u al collapse was p esen (see Figu es 18b and 19b). A ela i ely simila desc ip ion o he esul s can be adop ed in he case o he ensile es o he CPT samples (Figu es 20,21 and 25), whe e he ib e/ma ix debonding mechanism and ma ix c acking we e e en mo e p edominan han in he case o CPCT samples. The esul s o he CPTS samples o a ce ain ex en eplica e hose o he CPT samples, bu wi h he di e ence ha he p edominan de ec ed mechanism is ma ix c acking (see Figu es 22,23 and 26). 4.1. Delamina ion and Ma ix C acking To he bes o ou knowledge, delamina ion is a syne gy o AE signals, which a e cha - ac e ised by a wide spec um o ene gy (up o 10 7 aJ) and ampli ude alues (up o 94 dB AE and occasionally mo e). Rega ding he peak equency, se e al au ho s ha e cha ac e ised his phenomenon as ha ing a low peak equency, o en below 10 0 kHz [3,30,34,35] . This conclusion is pa ially con i med by he esul s o he BRCT samples om bo h da ase s. Howe e , clus e ing esul s om bo h da ase s, which a e ela ed o UCT, CPCT, CPT, and CPTS samples, clea ly demons a e ha he ma ix c acking/delamina ion phenomenon spans om 50 o app oxima ely 200 kHz (see Figu e 27). Polyme s 2022, 13, x FOR PEER REVIEW 20 o 25 (a) (b) Figu e 26. Cumula i e sum o hi s o indi idual clus e s, including o ce e sus displacemen — CPTS samples: (a) Fujice a 1045S da ase ; (b) DWC 454 da ase . b. Cha ac e isa ion o ailu e mechanisms 4.1. Delamina ion and Ma ix C acking To he bes o ou knowledge, delamina ion is a syne gy o AE signals, which a e cha ac e ised by a wide spec um o ene gy (up o 107 aJ) and ampli ude alues (up o 94 dBAE and occasionally mo e). Rega ding he peak equency, se e al au ho s ha e cha ac- e ised his phenomenon as ha ing a low peak equency, o en below 100 kHz [3,30,34,35]. This conclusion is pa ially con i med by he esul s o he BRCT samples om bo h da ase s. Howe e , clus e ing esul s om bo h da ase s, which a e ela ed o UCT, CPCT, CPT, and CPTS samples, clea ly demons a e ha he ma ix c acking/de- lamina ion phenomenon spans om 50 o app oxima ely 200 kHz (see Figu e 27). Figu e 27. Rep esen a i e AE signal o ma ix c acking (DWC 454 da ase ). 4.2. Fib e/Ma ix Debonding The epo ed ene gy and ampli ude alues o he ib e/ma ix debonding mecha- nism a e somewha lowe han hose o delamina ion/ma ix c acking, ha is, <106 aJ and <85 dBAE, espec i ely, wi hin he maximum limi ing alues, while mos AE hi s belonging o ib e/ma ix debonding ha e ampli ude and ene gy alues below 70 dBAE and 105 aJ, espec i ely. The published esul s om bo h da ase s also con i m ha bo h he ampli- ude and ene gy alues a e closely ela ed o he loading cha ac e is ics and s acking se- quence o he plies. A simila conclusion applies in he case o he peak equency, in which he equency in e al is ela ed o he loading cha ac e is ics and s acking se- quence o he plies. The 200-400 kHz peak equency ange we ound was he wides com- pa ed o hose o o he s udies (see Figu e 28) [3,33,34]. Figu e 27. Rep esen a i e AE signal o ma ix c acking (DWC 454 da ase ). 4.2. Fib e/Ma ix Debonding The epo ed ene gy and ampli ude alues o he ib e/ma ix debonding mechanism a e somewha lowe han hose o delamina ion/ma ix c acking, ha is, <10 6 aJ and <85 Polyme s 2023,15, 47 21 o 25 dB AE , espec i ely, wi hin he maximum limi ing alues, while mos AE hi s belonging o ib e/ma ix debonding ha e ampli ude and ene gy alues below 70 dB AE and 10 5 aJ, espec i ely. The published esul s om bo h da ase s also con i m ha bo h he ampli ude and ene gy alues a e closely ela ed o he loading cha ac e is ics and s acking sequence o he plies. A simila conclusion applies in he case o he peak equency, in which he equency in e al is ela ed o he loading cha ac e is ics and s acking sequence o he plies. The 200-400 kHz peak equency ange we ound was he wides compa ed o hose o o he s udies (see Figu e 28) [3,33,34]. Polyme s 2022, 13, x FOR PEER REVIEW 21 o 25 Figu e 28. Rep esen a i e AE signal o ib e/ma ix debonding (DWC 454 da ase ). 4.3. Fib e Failu e AE hi s associa ed wi h he ib e ailu e mechanism we e cha ac e ised by ene gies lowe han 104 aJ and ampli udes up o 80 dBAE. As in o he cases, bo h a iables end o be somewha lowe in he case o he Fujice a 1045S da ase . Howe e , a majo di e ence was ound in he case o he peak equency, whe e he DWC 454 da ase e ealed a 400– 600 kHz equency ange (see Figu e 29), whe eas he Fujice a 1045S da ase ex ended he uppe bound o 1000 kHz. This obse a ion was con i med by he subsequen il e ing o clus e h ee in he Fujice a 1045S da ase . By compa ing he achie ed esul s wi h hose in Table 1, we can s a e ha he lowe limi is in acco dance wi h he gi en alues. Figu e 29. Rep esen a i e AE signal o ib e ailu e (DWC 454 da ase ). 4.4. Fib e Pull-Ou Along wi h ib e b eakage, ib e pull-ou equi ed he addi ional il e ing o clus e h ee da a belonging o he Fujice a 1045S da ase . This mechanism can be easily iden i ied in he case o he DWC 454 da ase , which is pa o clus e one, excep o he esul s in he case o FCB samples. In addi ion, conside ing he il e ed da a om he Fujice a 1045S da ase , he AE hi s belonging o ib e pull-ou we e mainly cha ac e ised by peak e- quencies highe han 700 kHz (see Figu e 30), ampli udes below 60 dBAE, and ene gies no exceeding 103 aJ. The u ilisa ion o Fujice a 1045S in simila measu emen s may he e o e in ol e he es ima ion o he pa ial powe o he wa e ansien o dis inguish ib e b eak- age and ib e pull-ou . The achie ed esul s in e ms o peak equency a e undoub edly highe han hose gi en in he s udies lis ed in Table 1. Figu e 28. Rep esen a i e AE signal o ib e/ma ix debonding (DWC 454 da ase ). 4.3. Fib e Failu e AE hi s associa ed wi h he ib e ailu e mechanism we e cha ac e ised by ene gies lowe han 10 4 aJ and ampli udes up o 80 dB AE . As in o he cases, bo h a iables end o be somewha lowe in he case o he Fujice a 1045S da ase . Howe e , a majo di e ence was ound in he case o he peak equency, whe e he DWC 454 da ase e ealed a 400–600 kHz equency ange (see Figu e 29), whe eas he Fujice a 1045S da ase ex ended he uppe bound o 1000 kHz. This obse a ion was con i med by he subsequen il e ing o clus e h ee in he Fujice a 1045S da ase . By compa ing he achie ed esul s wi h hose in Table 1, we can s a e ha he lowe limi is in acco dance wi h he gi en alues. Polyme s 2022, 13, x FOR PEER REVIEW 21 o 25 Figu e 28. Rep esen a i e AE signal o ib e/ma ix debonding (DWC 454 da ase ). 4.3. Fib e Failu e AE hi s associa ed wi h he ib e ailu e mechanism we e cha ac e ised by ene gies lowe han 104 aJ and ampli udes up o 80 dBAE. As in o he cases, bo h a iables end o be somewha lowe in he case o he Fujice a 1045S da ase . Howe e , a majo di e ence was ound in he case o he peak equency, whe e he DWC 454 da ase e ealed a 400– 600 kHz equency ange (see Figu e 29), whe eas he Fujice a 1045S da ase ex ended he uppe bound o 1000 kHz. This obse a ion was con i med by he subsequen il e ing o clus e h ee in he Fujice a 1045S da ase . By compa ing he achie ed esul s wi h hose in Table 1, we can s a e ha he lowe limi is in acco dance wi h he gi en alues. Figu e 29. Rep esen a i e AE signal o ib e ailu e (DWC 454 da ase ). 4.4. Fib e Pull-Ou Along wi h ib e b eakage, ib e pull-ou equi ed he addi ional il e ing o clus e h ee da a belonging o he Fujice a 1045S da ase . This mechanism can be easily iden i ied in he case o he DWC 454 da ase , which is pa o clus e one, excep o he esul s in he case o FCB samples. In addi ion, conside ing he il e ed da a om he Fujice a 1045S da ase , he AE hi s belonging o ib e pull-ou we e mainly cha ac e ised by peak e- quencies highe han 700 kHz (see Figu e 30), ampli udes below 60 dBAE, and ene gies no exceeding 103 aJ. The u ilisa ion o Fujice a 1045S in simila measu emen s may he e o e in ol e he es ima ion o he pa ial powe o he wa e ansien o dis inguish ib e b eak- age and ib e pull-ou . The achie ed esul s in e ms o peak equency a e undoub edly highe han hose gi en in he s udies lis ed in Table 1. Figu e 29. Rep esen a i e AE signal o ib e ailu e (DWC 454 da ase ). Polyme s 2023,15, 47 22 o 25 4.4. Fib e Pull-Ou Along wi h ib e b eakage, ib e pull-ou equi ed he addi ional il e ing o clus e h ee da a belonging o he Fujice a 1045S da ase . This mechanism can be easily iden i ied in he case o he DWC 454 da ase , which is pa o clus e one, excep o he esul s in he case o FCB samples. In addi ion, conside ing he il e ed da a om he Fujice a 1045S da ase , he AE hi s belonging o ib e pull-ou we e mainly cha ac e ised by peak equencies highe han 700 kHz (see Figu e 30), ampli udes below 60 dB AE , and ene - gies no exceeding 10 3 aJ. The u ilisa ion o Fujice a 1045S in simila measu emen s may he e o e in ol e he es ima ion o he pa ial powe o he wa e ansien o dis inguish ib e b eakage and ib e pull-ou . The achie ed esul s in e ms o peak equency a e undoub edly highe han hose gi en in he s udies lis ed in Table 1. Polyme s 2022, 13, x FOR PEER REVIEW 22 o 25 Figu e 30. Rep esen a i e AE signal o ib e pull-ou (DWC 454 da ase ). Table 5 summa ises he cha ac e is ics o he epo ed ailu e mechanisms, which e- sul ed om clus e ing analysis o he Fujice a 1045S and DWC 454 da ase s. Table 5. Cha ac e isa ion o epo ed ailu e mechanisms. Failu e Mechanism A [dBAE] E [aJ] Peak F equency [kHz] Delamina ion/Ma ix c acking 40÷94 (occas. 100) <107 50÷200 Fib e/Ma ix debonding 40÷70 (occas. 85) <105 (occas. 106) 200÷400 Fib e ailu e <80 <104 400÷600(1000) Fib e pull-ou <60 <103 >700 5. Conclusions The iden i ica ion o indi idual ailu e mechanisms, including hei de ailed cha ac- e isa ion, has been he subjec o a numbe o publica ions. Howe e , many au ho s ha e only deal wi h speci ic mechanisms om he abo e scale, which p o ided an incen i e o u he esea ch in his a ea by ou eam. To ob ain a mo e gene al cha ac e is ic o he indi idual ailu e modes, we used wo di e en wideband senso s. This solu ion has he ad an age o mu ually e i ying he esul s om bo h da ase s. The expe imen al sec ion included he ealisa ion o ensile o compac ension es s on bulk esin and ca bon ib e bundle specimens, o which only one damage mechanism could be expec ed, ollowed by ensile/compac ension es s on uni/bidi ec ional ply spec- imens, whe e mul iple ailu e mechanisms we e assumed o occu . The measu ed da ase s we e hen indi idually subjec ed o clus e analysis, which in ol ed a spec al clus e ing app oach u ilising an unsupe ised k-means clus e ing al- go i hm. The ea u e space included besides he con en ional AE signal pa ame e s such as ampli ude, ene gy, ise ime, o equency cha ac e is ics, non-dimensionalised AE sig- nal powe in p eselec ed equency in e als. The clus e ed da a om bo h da ase s we e hen compa ed o each o he and o he SEM analysis esul s. This combina ion p o ed o be e y e ec i e, especially in he case o cha ac e ising ma ix c acking/delamina ion, o ib e/ma ix debonding ailu e mecha- nisms. Fo a mo e de ailed cha ac e isa ion o ib e ailu e o ib e pull-ou mechanisms in he case o he Fujice a 1045S da ase , a me hodology based on seconda y il e ing using he non-dimensionalised AE signal powe in p eselec ed equency in e als was p o- posed. Finally, he p esen ed me hod o e s a obus me hodology o he classi ica ion o AE signals in e ms o hei pa ame e s, which bene i s om he use o wo di e en wide- band senso s. The indispu able ad an age o he a o emen ioned me hod is also i s ull applicabili y in indus ial p ac ice, which, howe e , en ails he necessi y o use wice he numbe o senso s compa ed o con en ional echnology. Figu e 30. Rep esen a i e AE signal o ib e pull-ou (DWC 454 da ase ). Table 5summa ises he cha ac e is ics o he epo ed ailu e mechanisms, which esul ed om clus e ing analysis o he Fujice a 1045S and DWC 454 da ase s. Table 5. Cha ac e isa ion o epo ed ailu e mechanisms. Failu e Mechanism A [dBAE] E [aJ] Peak F equency [kHz] Delamina ion/Ma ix c acking 40÷94 (occas. 100) <10750÷200 Fib e/Ma ix debonding 40÷70 (occas. 85) <105(occas. 106)200÷400 Fib e ailu e <80 <104400÷600(1000) Fib e pull-ou <60 <103>700 5. Conclusions The iden i ica ion o indi idual ailu e mechanisms, including hei de ailed cha ac e - isa ion, has been he subjec o a numbe o publica ions. Howe e , many au ho s ha e only deal wi h speci ic mechanisms om he abo e scale, which p o ided an incen i e o u he esea ch in his a ea by ou eam. To ob ain a mo e gene al cha ac e is ic o he indi idual ailu e modes, we used wo di e en wideband senso s. This solu ion has he ad an age o mu ually e i ying he esul s om bo h da ase s. The expe imen al sec ion included he ealisa ion o ensile o compac ension es s on bulk esin and ca bon ib e bundle specimens, o which only one damage mechanism could be expec ed, ollowed by ensile/compac ension es s on uni/bidi ec ional ply specimens, whe e mul iple ailu e mechanisms we e assumed o occu . The measu ed da ase s we e hen indi idually subjec ed o clus e analysis, which in ol ed a spec al clus e ing app oach u ilising an unsupe ised k-means clus e ing algo i hm. The ea u e space included besides he con en ional AE signal pa ame e s such Polyme s 2023,15, 47 23 o 25 as ampli ude, ene gy, ise ime, o equency cha ac e is ics, non-dimensionalised AE signal powe in p eselec ed equency in e als. The clus e ed da a om bo h da ase s we e hen compa ed o each o he and o he SEM analysis esul s. This combina ion p o ed o be e y e ec i e, especially in he case o cha ac e ising ma ix c acking/delamina ion, o ib e/ma ix debonding ailu e mecha- nisms. Fo a mo e de ailed cha ac e isa ion o ib e ailu e o ib e pull-ou mechanisms in he case o he Fujice a 1045S da ase , a me hodology based on seconda y il e ing using he non-dimensionalised AE signal powe in p eselec ed equency in e als was p oposed. Finally, he p esen ed me hod o e s a obus me hodology o he classi ica ion o AE signals in e ms o hei pa ame e s, which bene i s om he use o wo di e en wideband senso s. The indispu able ad an age o he a o emen ioned me hod is also i s ull applicabili y in indus ial p ac ice, which, howe e , en ails he necessi y o use wice he numbe o senso s compa ed o con en ional echnology. Au ho Con ibu ions: M.Š. and P.Š.; me hodology, so wa e, da a cu a ion, w i ing, M.P.; w i ing —o iginal d a p epa a ion, unding acquisi ion, A.V.; in es iga ion, M.B.; isualiza ion, F.G.; unding acquisi ion. All au ho s ha e ead and ag eed o he published e sion o he manusc ip . Funding: This wo k was suppo ed by he p ojec Inno a i e and Addi i e Manu ac u ing Tech- nology —New Technological Solu ions o he 3D P in ing o Me als and Composi e Ma e ials (CZ.02.1.01/0.0/0.0/17_049/0008407) and by he Eu opean Regional De elopmen Fund in he Re- sea ch Cen e o Ad anced Mecha onic Sys ems p ojec , CZ.02.1.01/0.0/0.0/16_019/0000867 wi hin he Ope a ional P og amme Resea ch, De elopmen , and Educa ion and he p ojec SP2022/60 Ap- plied Resea ch in he A ea o Machines and P ocess Con ol suppo ed by he Minis y o Educa ion, You h, and Spo s. Con lic s o In e es : The au ho s decla e no con lic o in e es . Re e ences 1. Šo e , M.; Cienciala, J.; Fusek, M.; Pa líˇcek, P.; Mo a ec, R. Damage analysis o composi e CFRP ubes using acous ic emission moni o ing and pa e n ecogni ion app oach. Ma e ials 2021,14, 786. [C ossRe ] [PubMed] 2. Xu, J.; Wang, W.; Han, Q.; Liu, X. Damage pa e n ecogni ion and damage e olu ion analysis o unidi ec ional CFRP endons unde ensile loading using acous ic emission echnology. Compos. S uc . 2020,238, 1–10. [C ossRe ] 3. A umugam, V.; Sidha h, A.A.P.; San ulli, C. Failu e modes cha ac e iza ion o impac ed ca bon ib e ein o ced plas ics lamina es unde comp ession loading using acous ic emission. J. Compos. Ma e . 2014,48, 1–12. [C ossRe ] 4. Dong, J.; Kim, B.; Locque , A.; McKeon, P.; Decle cq, N.; Ci in, D.S. Nondes uc i e e alua ion o o ced delamina ion in glass ibe ein o ced composi es by e ahe z and ul asonic wa es. Compos. Pa B Eng. 2015,79, 667–675. [C ossRe ] 5. Heue , H.; Schulze, M.; Pooch, M.; Gäble , S.; Nocke, A.; Ba dl, G. Re iew on quali y assu ance along he CFRP alue chain— Nondes uc i e es ing o ab ics, p e o ms and CFRP by HF adio wa e echniques. Compos. Pa B Eng. 2015 ,77, 494–501. [C ossRe ] 6. Ga cea, S.C.; Wang, Y.; Wi he s, P.J. X- ay compu ed omog aphy o polyme composi es. Compos. Sci. Technol. 2018 ,156, 305–319. [C ossRe ] 7. Panasiuk, K.; Dudzik, K. De e mining he s ages o de o ma ion and des uc ion o composi e ma e ials in a s a ic ensile es by acous ic emission. Ma e ials 2022,15, 313. [C ossRe ] 8. Fe dinánd, M.; Vá dai, R.; Móczó, J.; Pukánszky, B. De o ma ion and ailu e mechanism o pa icula e illed and sho ibe ein o ced he moplas ics: De ec ion and analysis by acous ic emission es ing. Polyme s 2021,13, 3931. [C ossRe ] 9. Gholizadeh, A.; Mansou i, H.; Nikbakh , A.; Sagha i, H.; Fo ouhi, M. Applying acous ic emission echnique o de ec ing a ious damages occu ed in PCL nanomodi ied composi e lamina es. Polyme s 2021,13, 3680. [C ossRe ] 10. Pacheco-Salaza , O.F.; Wakayama, S.; Can-He e a, L.A.; Dzul-Ce an es, M.A.A.; Ríos-Sobe anis, C.R.; Ce an es-Uc, J.M. Damage e olu ion and ac u e e en s sequence analysis o co e-shell nanopa icle modi ied bone cemen s by acous ic emission echnique. Polyme s 2020,12, 208. [C ossRe ] 11. Guo, Y.; Zhu, S.; Yuxia, C.; Liu, D.; Li, D. Acous ic emission-based s udy o cha ac e ize he c ack ini ia ion poin o wood ibe /HDPE composi es. Polyme s 2019,11, 701. [C ossRe ] [PubMed] 12. Šo e , M.; Kuˇce a, P.; Mazanco á, E.; K ejˇcí, L. Acous ic emission and ac og aphic analysis o seamless s eel p essu e cylinde s wi h a i icial laws unde hyd os a ic bu s es ing. J. Nondes uc . E al. 2019,38, 1–13. [C ossRe ] 13. Yaacoubi, S.; Dahmene, F.; Bouzenad, A.; el Moun assi , M.; Aouini, M. Modal acous ic emission o composi e s uc u es heal h moni o ing: Issues o sa e compu ing ime and algo i hmic implemen a ion. Compos. S uc . 2018,183, 338–346. [C ossRe ] 14. Godin, N.; Hugue , S.; Gae ne , R. In eg a ion o he Kohonen’s sel -o ganising map and k-means algo i hm o he segmen a ion o he AE da a collec ed du ing ensile es s on c oss-ply composi es. NDTE In . 2005,38, 299–309. [C ossRe ] Polyme s 2023,15, 47 24 o 25 15. Sha e i, M.; Ghaib, M.; S eco a, D.; Thomson, D. On acous ic emission o damage de ec ion and ailu e p edic ion in ibe ein o ced polyme ods using pa e n ecogni ion analysis. Sma Ma e . S uc . 2017,26, 1–17. [C ossRe ] 16. G osse, C.U.; Oh su, M. Acous ic Emission Tes ing; Sp inge : Be lin/Heidelbe g, Ge many, 2008; ISBN 3540699724. 17. Fo ouhi, M.; Naja abadi, M.A. Acous ic emission-based s udy o cha ac e ize he ini ia ion o delamina ion in composi e ma e ials. J. The moplas . Compos. Ma e . 2014,29, 1–19. [C ossRe ] 18. Sause, M.G.R.; Mülle , T.; Ho oschenko , A.; Ho n, S. Quan i ica ion o ailu e mechanisms in mode-I loading o ibe ein o ced plas ics u ilizing acous ic emission analysis. Compos. Sci. Technol. 2012,72, 167–174. [C ossRe ] 19. Dongsheng, L.; Qian, H.; Jinping, O. Fa igue damage e olu ion and moni o ing o ca bon ibe ein o ced polyme b idge cable by acous ic emission echnique. In . J. Dis ib. Sens. Ne w. 2012,8, 1–7. 20. Mohammadi, R.; Naja abadi, M.A.; Saeedi a , M.; Youse i, J.; Minak, G. Co ela ion o acous ic emission wi h ini e elemen p edic ed damages in open-hole ensile lamina ed composi es. Compos. Pa B Eng. 2017,108, 427–435. [C ossRe ] 21. No den, H.E.; Zheng, S.; S e en, R.L.; Manli, C.W.; Hsing, H.S.; Quanan, Z.; Yen, N.-C.; Tung, C.C.; Liu, H.H. The empi ical mode decomposi ion and he hilbe spec um o nonlinea and non-s a iona y ime se ies analysis. P oc. R. Soc. Lond. Se . A 1998 ,454, 903–995. 22. Sause, M.G.R. In-Si u Moni o ing o Fibe -Rein o ced Composi es: Theo y, Basic Concep s, Me hods, and Applica ions; Sp inge : Be lin/Heidelbe g, Ge many, 2018; ISBN 9783319309538. 23. Roundi, W.; el Mahi, A.; el Gha ad, A.; Rebie e, J.-L. Acous ic emission moni o ing o damage p og ession in Glass/Epoxy composi es du ing s a ic and a igue ensile es s. Appl. Acous . 2018,132, 124–134. [C ossRe ] 24. Ameu , M.B.; el Mahi, A.; Rebie e, J.-L.; Gimenez, I.; Beyaoui, M.; Abdennadhe , M.; Hadda , M. In es iga ion and iden i ica ion o damage mechanisms o unidi ec ional ca bon/ lax hyb id composi es using acous ic emission. Eng. F ac . Mech. 2019 ,216, 1–18. 25. Tab izi, I.E.; Ke al, A.; Zanjani, J.S.M.; Akalin, C.; Yildiz, M. Expe imen al and nume ical in es iga ion on ac u e beha io o glass/ca bon ibe hyb id composi es using acous ic emission me hod and e ined zigzag heo y. Compos. S uc . 2019 ,223, 1–21. [C ossRe ] 26. Li, L.; Swol s, Y.; S aumi , I.; Yan, X.; Lomo , S.V. Clus e analysis o acous ic emission signals o 2D and 3D wo en ca bon ibe /epoxy composi es. J. Compos. Ma e . 2015,50, 1921–1935. [C ossRe ] 27. Saidane, E.H.; Scida, D.; Assa a , M.; Ayad, R. Damage mechanisms assessmen o hyb id lax-glass ib e composi es using acous ic emission. Compos. S uc . 2017,174, 1–11. [C ossRe ] 28. Ma ec, A.; Thomas, J.H.; el Gue jouma, R. Damage cha ac e iza ion o polyme -based composi e ma e ials: Mul i a iable analysis and wa ele ans o m o clus e ing acous ic emission da a. Mech. Sys . Signal P ocess. 2008,22, 1441–1464. [C ossRe ] 29. Pashm o oush, F.; Khamedi, R.; Fo ouhi, M.; Hajikhani, M.; Ahmadi, M. Damage classi ica ion o sandwich composi es using acous ic emission echnique and k-means gene ic algo i hm. J. Nondes uc . E al. 2014,33, 481–492. [C ossRe ] 30. Saeedi a , M.; Naja abadi, M.A.; Za ouchas, D.; Toudeshky, H.H.; Jalal and, M. Clus e ing o in e lamina and in alamina damages in lamina ed composi es unde inden a ion loading using acous ic emission. Compos. Pa B Eng. 2018 ,144, 206–219. [C ossRe ] 31. Oskouei, A.R.; Heida y, H.; Ahmadi, M.; Fa ajpu , M. Unsupe ised acous ic emission da a clus e ing o he analysis o damage mechanisms in glass/polyes e composi es. Ma e . Des. 2012,37, 416–422. [C ossRe ] 32. Saeedi a , M.; Za ouchas, D. Damage cha ac e iza ion o lamina ed composi es using acous ic emission: A e iew. Compos. Pa B 2020,195, 1–21. [C ossRe ] 33. De G oo , P.J.; Wijnen, P.A.M.; Janssen, R.B.F. Real- ime equency de e mina ion o acous ic emission o di e en ac u e mechanisms in ca bon/epoxy composi es. Compos. Sci. Technol. 1995,55, 405–412. [C ossRe ] 34. 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, 1393–1406. [C ossRe ] 35. Boomina han, R.; A umugam, V.; San ulli, C.; Sidha h, A.A.P.; Sanka , R.A.; S idha , B.T.N. Acous ic emission cha ac e iza ion o he empe a u e e ec on alling weigh impac damage in ca bon/epoxy lamina es. Compos. Pa B Eng. 2014 ,56, 591–598. [C ossRe ] 36. Schull, P.J. Nondes uc i e E alua ion: Theo y, Techniques, and Applica ions; M. Dekke : New Yo k, NY, USA, 2002. 37. Sause, M.G.R.; G ibo , A.; Unwin, A.R.; Ho n, S.; Schull, P.J. Pa e n ecogni ion app oach o iden i y na u al clus e s o acous ic emission signals. Pa e n Recogni . Le . 2012,33, 17–23. [C ossRe ] 38. Liu, X.; Liu, Z.; Li, X.; Cui, J. Acous ic emission RA- alue and g ani e ac u e modes unde dynamic and s a ic loads. In Ad ances in Acous ic Emission Technology; Sp inge P oceedings in Physics; Sp inge : Cham, Swi ze land, 2019; Volume 218, WCAE 2017. 39. Khan, R. Fibe b idging in composi e lamina es: A li e a u e e iew. Compos. S uc . 2019,229, 1–15. [C ossRe ] 40. Ba ile, C.; Casa ola, C.; Pappale e a, G.; Kannan, V.P. Laplacian sco e and K-means da a clus e ing o damage cha ac e iza ion o adhesi ely bonded CFRP composi es by means o acous ic emission echnique. Appl. Acous . 2022,185, 1–11. [C ossRe ] 41. Ichenihi, A.; Li, W.; Gao, Y.; Rao, Y. Fea u e selec ion and clus e ing o damage o pseudo-duc ile unidi ec ional ca bon/glass hyb id composi e using acous ic emission. Appl. Acous . 2021,182, 1–10. [C ossRe ] 42. He, X.; Cai, D.; Niyogi, P. Laplacian sco e o ea u e selec ion. Ad . Neu al In . P ocess. Sys . 2005,18, 507–514. 43. Li, L.; Lomo , S.V.; Yan, X.; Ca elli, V. Clus e analysis o acous ic emission signals o 2D and 3D wo en glass/epoxy composi es. Compos. S uc . 2014,116, 286–299. [C ossRe ] Polyme s 2023,15, 47 25 o 25 44. Alia, A.; Fan ozzi, G.; Godin, N.; Osmani, H.; Reynaud, P. Mechanical beha iou o ju e ib e- ein o ced polyes e composi e: Cha ac e iza ion o damage mechanisms using acous ic emission and mic os uc u al obse a ions. J. Compos. Ma e . 2019 ,53, 3377–3394. [C ossRe ] 45. A ailable online: h ps://uk.ma hwo ks.com/help/s a s/ sulaplacian.h ml (accessed on 3 Feb ua y 2022). 46. Ahmed, M.; Se aj, R.; Islam, S.M.S. The k-means Algo i hm: A Comp ehensi e su ey and pe o mance e alua ion. Elec onics 2020,9, 1295. [C ossRe ] 47. Von Luxbu g, U. A Tu o ial on spec al clus e ing. S a . Compu . 2007,17, 395–416. [C ossRe ] 48. Ng, A.Y.; Jo dan, M.; Weiss, Y. On spec al clus e ing: Analysis and an algo i hm. In P oceedings o he Ad ances in Neu al In o ma ion P ocessing Sys ems 14; MIT P ess: Camb idge, MA, USA, 2001; pp. 849–856. 49. A ailable online: h ps://www.ma hwo ks.com/help/s a s/spec alclus e .h ml (accessed on 3 Feb ua y 2022). 50. Chaimon ee, S.; A kinson, K.; Coenen, F. Bes clus e ing con igu a ion me ics: Towa ds mul iagen based clus e ing. In Ad anced Da a Mining and Applica ions; Sp inge : Be lin/Heidelbe g, Ge many, 2010; pp. 48–59. 51. Kau man, L.; Rousseeuw, P.J. Finding G oups in Da a: An In oduc ion o Clus e Analysis; Wiley-In e science: Hoboken, NJ, USA, 1990. Disclaime /Publishe ’s No e: The s a emen s, opinions and da a con ained in all publica ions a e solely hose o he indi idual au ho (s) and con ibu o (s) and no o MDPI and/o he edi o (s). MDPI and/o he edi o (s) disclaim esponsibili y o any inju y o people o p ope y esul ing om any ideas, me hods, ins uc ions o p oduc s e e ed o in he con en .