Sensitivity analysis: A tool for tailoring environmentally friendly materials

Expe Sys ems Wi h Applica ions 208 (2022) 118039 A ailable online 8 July 2022 0957-4174/© 2022 The Au ho (s). Published by Else ie L d. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by- nc-nd/4.0/). Sensi i i y analysis: A ool o ailo ing en i onmen ally iendly ma e ials Da id Seidl a , * , I an Ruˇ ziak b , Zo a Koˇ s ialo ´ a Janˇ cíko ´ a a , Pa el Koˇ s ial a a Facul y o Elec ical Enginee ing and Compu e Science, VSB Technical Uni e si y o Os a a, Czech Republic b Facul y o Wood Sciences and Technology, Technical Uni e si y, Z olen, Slo akia ARTICLE INFO Keywo ds: Rubbe blends Plas icize s A i icial neu al ne wo ks Sensi i i y analysis P edic ions ABSTRACT In his a icle, we examine he use o sensi i i y analysis o he op imiza ion o selec ed physical p ope ies in ubbe compounds and de e mine objec i e c i e ia which allow o he educ ion o en i onmen al load du ing ubbe compound p oduc ion. The sensi i i y analysis shows how signi ican ly each inpu alue a ec s he ou pu alue, and he esponse g aphs exp ess he e ec o he selec ed pa ame e on he ou pu alue. The solu ions desc ibed in he a icle a e applicable o o he p oduc ion echnologies. We p esen a sensi i i y analysis based on he p edic ion o selec ed mechanical p ope ies o ubbe mix u es composed o S anda d Malaysian Rubbe (SMR). Two blends we e p e- pa ed by mixing SMR and oleic acid and di e en concen a ions o su ac an (2, 4, 6, 8, 10, 20, 30 w %). Tensile s eng h R m and moduli M100, M200, M300 we e measu ed and e alua ed. The sensi i i y analysis showed he signi icance o ce ain ing edien s which a ec he measu ed mechanical p ope ies. 1. In oduc ion Technological ea men s and he agen s o ille s added o ubbe ha e a s ong e ec on se e al p ope ies o ubbe mix u es. Ca bon black is an ac i e ille which is used o inc ease elec ical and he mal p ope ies due o i s s ong in e ac ion wi h na u al ubbe . Plas icize s play an impo an echnological ole in he ubbe indus y. Rubbe is a mul icomponen ma e ial which is e y sensi i e o many chemical species, and especially o echnological ea men . Fo mul icomponen sys em inpu s and ou pu s associa ed wi h he op imiza ion o chemis y o echnological p ocesses, i is con enien o use an a i icial neu al ne wo k (ANN). ANNs a e equen ly applied in ma e ial analysis, and in some cases also ha e success ully eplaced des uc i e es ing in ibe composi es (Fa hana e al., 2016). T ue s ess/s ain cu es we e success ully calcula ed o he au o- mobile indus y (Doh, Lee, & Lee, 2016), whe e he au ho s op imized in e connec ion weigh s ob ained wi h hidden laye s and ou pu laye s. A ma hema ical model o he ma e ial’s beha io was sugges ed h ough his eed- o wa d neu al ne wo k. The s udy Le, Y onne , and He (2015) applied a ini e elemen me hod and an ANN o desc ibe su ace s ess in ma e ials. A combina ion o an ANN and FEM was applied in a s udy o he dynamic p ope ies in glass lamina es wi h di e en ma e ial shaping (Seidl e al., 2011). Shape modes we e p esen ed o di e en ib a- ional exci a ions. ANNs we e used wi h success o op imize he composi ion o op ical glasses (Seidl e al., 2011) and ob ain equi ed op ical p ope ies. Feed o wa d neu al ne wo ks a e used o he classi ica ion o laye s o coal and shale coal acco ding o ash and mois u e con en (Janˇ cíko ´ a, Boˇ s´ ak, Zimný, Legoue a, Min´ a ik, Koˇ s ial, & Poulain, 2014). The wo k Ghosh, Cha e jee, and Shanke (2016) applied an ANN and neu o- uzzy in e e ence sys em which sepa a ed pe mea e lux and sal con en . A mul i esolu ion analysis o ime-dependen da a se s ob ained h ough Wa ele de- composi ion and e alua ed using an ANN was applied o he managemen o enewable ene gy sou ces in he s udy Salehi and Raza i (2016). A sound s a is ical model analysis was de eloped based on deep neu al ne wo ks. Using a new algo i hm on con en ional da a, he au- ho s a ained subs an ially be e esul s (Doucou e, Agbossou, & Ca - denas, 2016). A discolo a ion p ocess analysis was conduc ed wi h he applica ion o h ee in e media e laye s, a backp opaga ion lea ning algo i hm, and a sigmoid ac i a ion unc ion implemen ed in Fo an. Th ee neu ons in he in e media e laye p o ided he bes esul s Hwang, Pa k, and Chang (2016). Technological ea men s and he agen s o ille s added o ubbe * Co esponding au ho . E-mail add esses: [email p o ec ed] (D. Seidl), [email p o ec ed] (I. Ruˇ ziak), [email p o ec ed] (Z. Koˇ s ialo ´ a Janˇ cíko ´ a), [email p o ec ed] (P. Koˇ s ial). Con en s lis s a ailable a ScienceDi ec Expe Sys ems Wi h Applica ions jou nal homepage: www.else ie .com/loca e/eswa h ps://doi.o g/10.1016/j.eswa.2022.118039 Recei ed 7 Feb ua y 2022; Recei ed in e ised o m 13 May 2022; Accep ed 30 June 2022 Expe Sys ems Wi h Applica ions 208 (2022) 118039 2 ha e a s ong e ec on se e al p ope ies o ubbe mix u es. The main addi i es and hei unc ions we e desc ibed in Lenzi e al. (2016). The pape Jonˇ s a e al. (2011) examines he e ec o nano-s y ene bu adiene ubbe in phenolic nanocomposi es. When applied in he amoun o only se e al pe cen , hese nanopa icles subs an ially in- c ease he no ched impac s eng h wi hou a signi ican change in lexu al pe o mance. The applica ion o ul asound du ing he ex usion p ocess o di e en ubbe ypes showed changes due o die p essu e (Liu, Ma, Zhan, & Wang, 2015). Inc easing he ul asound ampli ude dec eased he die p essu e o bo h na u al ubbe and mix u es o na u al and s y ene bu adiene ubbe . In he case o s y ene-bu adiene ubbe , he die p essu e inc eased. In (Choi & Isaye , 2015), he au ho s p esen ed liquid isop ene as a eplacemen o oil plas icize . In his case, liquid isop ene educed bo h Mooney and appa en iscosi y and supp essed plas icize mig a ion. The wo k (Ren, Zhao, Li, Zhang, & Zhang, 2015) p esen ed an imp o emen in he mechanical p ope ies o silica- illed ubbe caused by a ing opening be ween ubbe chains and Si-OH g oups. Ring opening p ocesses occu ed du ing bo h he ulcaniza ion and mixing p ocesses. Polye hylene-co- inyl ace a e, na u al ubbe and o ganoclay mix- u es we e s udied in (Xu, Jia, Luo, Jia, & Peng, 2015). The addi ion o o ganoclay supp essed he na u al ubbe amoun and aised Young’s modulus and yield s ess. Elonga ion a b eak and s ess a b eak we e dec eased. The pape (Raza i-Nou i & Ka ami, 2014) s udied he pe cola ion e ec and u he elec ic anspo phenomena caused by he p esence o g aphi e. Mansou , Hussein, and Moha am (2014) desc ibed he in luence o ille s on na u al ubbe p ope ies. The esul s o di e en ille s (including ca bon nano ubes) showed an inc ease in iscosi y c oss-link densi y, modulus and ha dness and a dec ease in cu e ime. The plas i- cize s in a ubbe blend inc eased hei p ocessing p ope ies. Ne e - heless, hey o en mig a ed om he ubbe ma ix and caused poo s abili y in he ubbe blend in he long e m. The ad e se en i on- men al impac o his p ocess was also subs an ial. Kopal e al. (2022) ha e p edic ed cu ing cha ac e is ics which play a signi ican ole in he ulcaniza ion p ocess o ubbe blends using ANN. Al hough some s udies, e.g. (Lubu a e al., 2021), (Sa a e al., 2019), (Vodka & Pog ebnyak e al., 2020) and (Lopes, Sil a, & Machado, 2021), ha e used ANN ne wo ks o he p edic ion o di e en p ocesses in ubbe echnology, hey ocused on he p edic ion o di e en pa- ame e s bu none o hem ake in o accoun he en i onmen al aspec o ubbe echnology. As he p oduc ion o ubbe compounds also has i s en i onmen al impac , i is necessa y o ocus on he op imiza ion o chemical compounds. The aim o his pape is o e alua e he in luence o exis ing chem- icals on he mechanical p ope ies o he mix u e in o de o educe en i onmen al e ec s in he p oduc ion p ocess o ubbe compounds, whe e he en i onmen al aspec is he main con ibu ion o his esea ch. In he esea ch, we deal wi h he e ec o plas icize and he amoun o su ac an on ce ain mechanical p ope ies such as s eng h R m and moduli M100, M200, M300. The analysis me hod we applied was based on ou p e ious expe imen al wo k (Ruˇ ziak e al., 2018). The a icle is s uc u ed as ollows: Sec ion 2 summa izes he heo y o a i icial neu al ne wo ks. Sec ion 3 desc ibes he expe imen al con- di ions. Sec ion 4 gi es a de ailed o e iew o esul s, and Sec ion 5 lis s he conclusions a ising om he esea ch. 2. Theo y A i icial neu al ne wo ks a e used o p edic ma e ial p ope ies when an analy ical ma hema ical app oxima ion canno be ound. Using his e y obus ma hema ical ool, ma e ial p ope ies can be p edic ed. A i icial neu al ne wo ks use di e en opologies which, in mos cases, con ain h ee o ou laye s. Th ee ypes o neu ons occu in hese ne wo ks: •inpu neu ons •hidden neu ons •ou pu neu ons. Inpu neu ons con ain in o ma ion abou he pa ame e s which change in ou da ashee , such as ma e ial composi ion, he mal ea - men , and o he s. Ou pu neu ons co espond o he ma e ial p ope ies we wan o p edic . Some o he inpu neu ons a e used o aining and o he s a e used o gene alizing. The ad an age o a neu al ne wo k is ha i can bo h lea n and gene alize. The main disad an age o a neu al ne wo k is ha i needs mo e alues o one o mo e pa ame e s which change in each inpu da ase han equi ed by s anda d i ing p ocedu es such as he leas squa es and o he me hods. The use o an ANN is es ed wi h he s a is ical pa ame e s REL_RMSE and R 2 , which a e de ined by Eqs. (1) and (2). RMSE = ∑n i=1(yi−oi)2 n.(n−1) √(1) R2=(E(y).E(o) E(y.o))2 (2) whe e: y i −measu ed alues o i −p edic ed alues. n −numbe o inpu alues. E (y) −s a is ical mean alue o inpu s. E (o) −s a is ical mean alue o p edic ed alues. E (y.o) −mean alue o mul iplica ion o p edic ed and inpu alues. 3. Expe imen al condi ions 3.1. Ma e ials Plas icize s play an impo an ole in ubbe echnology and ha e an e ec on he blend mixing. We he e o e p esen he e ec o he amoun o plas icize on selec ed mechanical p ope ies such as s eng h R m, modulus M100, M200 and M300. The expe imen al alues we e measu ed on he appa a us Z250 All oundLine (250 kN). The ensile s eng h R m and moduli M100, M200, M300 measu emen s we e pe - o med on his de ice by ensile es . The chemical composi ion o s anda d ubbe mix u e in PHR (pa s pe hund ed o ubbe ) is shown in Table 1. The ubbe was SMR (S anda d Malaysian Rubbe ). Sul enax is used in ubbe indus y in p ocessing o na u al and syn he ic ubbe in ubbe compounds as a as accele a o o ulcani- za ion wi h delayed ac ion. I p o ides good physical and mechanical p ope ies, high c osslinking e iciency and good modulus. N339 CB is a ype o ca bon black (CB) and Gumodex is a plas icize . The gene al p ocedu e o he ubbe ulcaniza es p epa a ion is as ollows and is p o ec ed by he manu ac u e . In he wo-s age mixing, which we used in ou case, he mix u e is p epa ed in he i s phase a a Table 1 Chemical composi ion o s anda d ubbe mix u e. Ing edien PHR con en SMR 100 Sulphu 2 ZnO 5 S ea ine 2 Sul enax 2 N339 CB 50 Gumodex 10 D. Seidl e al. Expe Sys ems Wi h Applica ions 208 (2022) 118039 3 empe a u e o abou 150◦C. The p essu e exe ed by he kneading wedge is abou 3 MPa and he mixing ime is 4 o 4.5 min. These pa- ame e s depend on he ype o mix u e. The hickness o he bel is de e mined by he gap be ween he olle s. To p e en he mix u e om s icking, i is imme sed in a we ing ba h con aining a sepa a ion solu ion. The empe a u e o he e luen mix u e is a maximum o 36◦C. In he second s age o mixing he mix u e ( inal mix u e), he mix u e is weighed again in he o m o a chopped honeycomb. I is i s p o- cessed in a mixe , and when he equi ed plas ici y is eached, he ulcanizing agen and o he subs ances a e added. The empe a u e o he mixing chambe is lowe han in he p e ious mixing (abou 110◦C) and he mixing akes a e y sho ime. The e e ence samples we e mixed wi h oleic acid plas icize wi h a con en o 1 PHR and 3 PHR, oge he wi h weigh pe cen age o su - ac an (0, 2, 4, 6, 8, 10, 20, 30 w %). The su ac an was sodium 1- Table 2 Numbe o hidden uni s. Type o ne wo k Minimum hidden inpu s Maximum hidden inpu s Radial basis unc ion RBF 1 8 Th ee-laye MLP, laye 2 1 16 Fou -laye MLP, laye 2 1 16 Fou -laye MLP, laye 3 1 16 Table 3 Bes neu al ne wo ks. P ope y De ails o ne wo k R m [MPa] MLP 2:2–16-6:6 M100 [MPa] MLP 2:2–16-6:6 M200 [MPa] MLP 2:2–16-16–6:6 M300 [MPa] MLP 2:2–16-16–6:6 Fig. 1. MLP2:2-16-6:6 is bes ne wo k o he p edic ion o ensile s eng h R m and modulus M100. Fig. 2. MLP2:2-16-16-6:6 id he bes ne wo k o moduli M200 and M300. Table 4 Slope be ween he p edic ed and measu ed cha ac e is ics, pe cen age de iance, co ela ion coe icien R and mean squa ed e o RMSE. P ope y Slope Pe cen age di e ence (%) R RMSE (MPa) R m [MPa] 1,048 +4,8 0,982 0,082 M100 [MPa] 0,983 −1,7 0,985 0,028 M200 [MPa] 0,999 −0,1 0,978 0,068 M300 [MPa] 0,938 −6,2 0,981 0,095 Table 5 Sensi i i y coe icien s o he inpu pa ame e s o all he s udied mechanical cha ac e is ics. P ope y/Sensi i i y coe icien Su ac an w % Oleic acid amoun R m [MPa] 2.491 1.163 M100 [MPa] 2.632 1.159 M200 [MPa] 2.931 1.203 M300 [MPa] 2.514 1.073 Fig. 3. Es ima ed R m alues e sus he eal measu ed alues in he subg oup o 16 alues. Fig. 4. Es ima ed M100 alues e sus he eal measu ed alues in he subg oup o 16 alues. D. Seidl e al. Expe Sys ems Wi h Applica ions 208 (2022) 118039 4 nonyl-2-(2-nonylphenoxy benzene sul a e). These amphiphilic sub- s ances educe he su ace ension o he luids and he in e acial en- sion be ween wo liquids. The use o oleic acid plas icize as a subs i u e o gumodex has a lowe ecological impac . 3.2. Sensi i i y analysis The sensi i i y analysis me hod applied in he s udy allows o a s a is ical analysis o he e ec o echnological p ocesses on he me- chanical p ope ies o he ubbe blend (in his case). The esponse g aphs, which a e addi ional esul s om he ANN p ocedu es, indica e he alues o he s udied p ope ies in combina ions o inpu pa ame e s which we e no measu ed. Neu al ne wo ks we e c ea ed in STATISTICA – Neu al Ne wo ks so wa e. Neu al ne wo k so wa e enables he execu ion o a sensi i i y analysis and he c ea ion o esponse g aphs. The sensi i i y analysis e eals how signi ican ly each inpu alue a ec s he ou pu alue, and he esponse g aphs exp ess he e ec o he selec ed pa ame e on he ou pu alue. A neu al ne wo k conduc s a sensi i i y analysis by ea ing each inpu a iable in u n as i i we e “una ailable”. Each model de ines a missing alue subs i u ion p ocedu e o allow p edic ions in he absence o alues o one o mo e inpu s. To de ine he sensi i i y o a pa icula a iable, we i s un he ne wo k on a se o es cases and accumula e he ne wo k e o . We hen un he ne wo k again using he same cases, bu his ime eplace he obse ed alues wi h he alue es ima ed by he missing alue p ocedu e, and again he ne wo k e o accumula es. Gi en ha we e ec i ely emo e some in o ma ion which he ne wo k p esumably uses (i.e., one o i s inpu a iables), we would easonably expec some de e io a ion in e o o occu . The basic mea- su e o sensi i i y is he a io o he e o wi h he missing alue sub- s i u ion and he o iginal e o . The mo e sensi i e he ne wo k is o a pa icula inpu , he g ea e he de e io a ion we can expec and he e o e he g ea e he a io. I he a io is one o less, hen making he a iable “una ailable” ei he has no e ec on he pe o mance o he ne wo k o enhances i s pe o mance. This means ha he s udied p ope y is no a unc ion o he “missing” inpu pa ame e . Fig. 5. Es ima ed M200 alues e sus he eal measu ed alues in he subg oup o 16 alues. Fig. 6. Es ima ed M300 alues e sus he eal measu ed alues in he subg oup o 16 alues. Fig. 7. P edic ion o ensile s eng h R m e sus su ac an amoun . D. Seidl e al. Expe Sys ems Wi h Applica ions 208 (2022) 118039 5 Fo he neu al ne wo k inpu pa ame e s, we used he su ac an amoun as a con inuous neu on inpu and he oleic acid amoun as a ca ego ical neu on inpu . We selec ed oleic acid as a ca ego ical inpu because he s udied ma e ials only con ain wo di e en alues o his inpu pa ame e , which is a e y low le el o lea ning and gene aliza- ion. The measu ed alues o he mechanical p ope ies we e used as con inuous neu on ou pu s. Since he concen a ions o he su ac an equaled 0, 2, 4, 6, 8, 10, 20, and 30 w % and he wo concen a ions o oleic acid we e 1 PHR and 3 PHR, he neu al ne wo ks lea ned wi h 16 inpu neu ons. The o al analyzed neu al ne wo ks o each a iable and each s a e consis ed o en neu al ne wo ks, om which we used he bes i e neu al ne wo ks. The selec ion c i e ion o e ain a ne wo k was he lowes e o . The ypes o ne wo k which we applied we e linea , PNN o GRNN, Radial Basis unc ion, h ee laye and ou -laye pe cep on. The minimum and maximum alues o he hidden inpu s o each Fig. 8. P edic ion o modulus M100 e sus su ac an amoun . Fig. 9. P edic ion o modulus M200 e sus su ac an amoun . D. Seidl e al. Expe Sys ems Wi h Applica ions 208 (2022) 118039 6 ne wo k a e shown in Table 2. 4. Resul s We p esen de ails o he imp o ed eal ime ne wo ks. In he inal s ep, we compu ed he p edic ions o all i e bes neu al ne wo ks and com- pa ed hem o he measu ed alues in he eaching subse . As he bes neu al ne wo k, we used a neu al ne wo k in which he co ela ion be ween he measu ed and ANN ap- p oxima ed ne wo k we e closes o 1 by sus aining he oo mean squa e e o be ween he measu ed and app oxima ed alues closes o ze o. The bes neu al ne wo ks o each a iable a e lis ed in Table 3. whe e MLP e e s o he mul ilaye pe cep on ne wo k. Figs. 1 and 2 p esen he bes ne wo k o moduli 100, 200 and 300. Table 4 p esen s he slope alues be ween he p edic ed alues o he s udied mechanical cha ac e is ics and he eal measu ed alues. The able also p esen s de iances o his slope om he alue o 1 (i cha - ac e izes he de iance) and he co ela ion coe icien , oge he wi h he mean squa ed e o RMS. Based on he esul s p esen ed in Table 4, we can d aw he ollowing conclusions: The maximal size o he e o in de e mining he mechanical cha - ac e is ics by neu on ne wo ks in a 16-membe base is 6.2% ( o he M300 cha ac e is ic), which is below he measu ing e o limi o he s a ed mechanical cha ac e is ics o ubbe compounds. Since he e o in de e mining mechanical cha ac e is ics o neu al ne wo ks is smalle han he measu ing e o , we can s a e ha ANN a e able o p edic ensile s eng h and s eng h o he M100, M200, M300 samples be o e ageing. The s a ed conclusion is also suppo ed by high alues o he co e- la ion coe icien – a leas 0.978. The maximal mean squa ed e o be ween he p edic ed and measu ed alues o he mechanical cha ac e is ics is 0.095 MPa, which is a he measu abili y limi o he gi en cha ac e is ics. Mo e use ul pa ame e o assessing he di e ences be ween he p edic ed and measu ed alues o mechanical cha ac e is ics is he ela i e mean squa ed e o , de ined as a quo ien o he mean squa ed e o and he minimal alue o he gi en. This pa ame e amoun s o 0.005; 0.008; 0.009 and 0.009 o he ensile s eng h and he M100, M200 and M300 moduli, which co e- spond o he pe cen age de ia ion o 0.5% o he ensile s eng h, 0.8% o he M100 modulus and 0.9% o he M200 and M300 moduli. •These alues ep esen he maximal de iance be ween he p edic ed and measu ed cha -ac e is ics, wi h he s ipula ion ha hey a e a leas 10 imes smalle han he unce ain ies in de e mining he s udied mechanical cha ac e is ics. •The heo e ical e o in de e mining he mechanical p ope ies o ubbe compounds based on a leas 10 es specimens is app oxi- ma ely equal o 5%. The expe imen ally de ec ed e o s in he de e mina ion o Rm, M100, M200 and M300 o all measu ed sample ypes and all p ope ies do no exceed he e o alue o 5%. Based on diag ams 3 o 6, we can d aw he ollowing conclusions: All he s udied ma e ial p ope ies inc eased along wi h a ise in he amoun o oleic acid and su ac an . •Tensile s eng h inc eased app ox. 29% o 1 PHR and 22% o 3 PHR. •Modulus M100 inc eased app ox. 53% o 1 PHR and 46% o 3PHR. •Modulus M200 inc eases app ox. 51% o 1 PHR and 44% o 3PHR. •Modulus M300 inc eases app ox. 53% o 1 PHR and 45% o 3PHR. Fig. 10. P edic ion o modulus M300 e sus su ac an amoun . Table 6 To al sha e o he sha e o he su ac an and he olume o oleic acid o he s udied mechanical cha ac e is ics. P ope y/Sensi i i y coe icien Su ac an w % Oleic acid amoun R m [MPa] 68.2 % 31.8 % M100 [MPa] 69.4 % 30.6 % M200 [MPa] 70.9 % 29.1 % M300 [MPa] 70.1 % 29.9 % D. Seidl e al. Expe Sys ems Wi h Applica ions 208 (2022) 118039 7 •All hese inc eases a e highly abo e he p ope y measu emen un- ce ain y; he e o e, he su ac an weigh pe cen age has a signi i- can ole in he ensile s eng h and moduli. •The di e ences be ween 1 PHR and 3 PHR in inc ease we e 7%, 7%, 7%, and 8%, which a e all below he measu emen e o s o hese ypes o ubbe blends, and we can he e o e conclude ha he e ec o su ac an weigh pe cen age is app oxima ely he same o 1 PHR and 3 PHR mix u es, which is also suppo ed by a e y high co e- la ion coe icien . •be ween he p edic ions and measu emen s o bo h 1 PHR and 3 PHR mix u es. •The inc ease in R m was p obably caused by a highe deg ee o s uc u e c oss- linking Sk obak e al. (2016). In he ollowing esul s, we show p edic ions o he s udied p ope - ies. Because o same inc ease in he s udied mechanical p ope ies o 1 PHR and 3 PHR oleic acid mix u es, we show he p edic ion o he a e age alues o p ope y e sus su ac an amoun . Fo each neu al model esponse, g aphs we e c ea ed. The esponse g aphs exp ess he in luence o he chosen inpu pa ame e on he p e- dic ed ou pu pa ame e – see Figs. 7-10. These g aphs exp ess he in- luence o he su ac an amoun on he p edic ed mechanical p ope ies. Fig. 7 shows he p edic ion o ensile s eng h, while Figs. 8- 10 show he p edic ion esul s o modulus M100, M200 and M300 espec i ely. Based on Figs. 7-10, we can d aw he ollowing conclusions: •The bigges change in he s udied mechanical cha ac e is ics wi h a change in he sha e o he su ac an can be obse ed wi hin he weigh pe cen age ange o 0 o 10%; he e ec o he su ac an is educed o e 10%. •An inc ease in he alues o he s udied mechanical cha ac e is ics o up o 10% ep e- sen s 12%; 27%; 26% and 22%, while he inc ease o up o 30% in he weigh pe cen age o he su ac an only co e- sponds o inc eases o 25.5%; 49.5%; 47.5% and 49%. •An inc ease in he sha e o he su ac an by 1/3 o he o e all in- c ease leads o a 48% inc ease o he o al inc ease o he ensile s eng h R m . •An inc ease in he sha e o he su ac an by 1/3 o he o e all in- c ease leads o a 55% inc ease in he o al inc ease o he M100 modulus. •An inc ease in he sha e o he su ac an by 1/3 o he o e all in- c ease leads o a 55% inc ease in he o al inc ease o he M200 modulus. •An inc ease in he sha e o he su ac an by 1/3 o he o e all in- c ease leads o a 45% inc ease in he o al inc ease o he M300 modulus. •On he o he hand, an inc ease in he su ac an by 2/3 o he o e all inc ease leads o an inc ease in he ensile s eng h and o he M100, M200 and M300 moduli by 82.5%; 85%; 87.5% and 91%. •Such an inc ease o a 20% sha e o he su ac an is su icien o imp o ing he s udied mechanical cha ac e is ics and, a he same ime, educing he sha e o he su ac an has a posi i e impac on he en i onmen al bu den o he inal mix u e o e ime since he inal mix u e as well as su ac an age due o he impac o hea and ime. Below, we discuss he sensi i i y analysis ob ained om he p edic- ion using he ANN ne wo ks. Table 5 shows he sensi i i y coe icien s o he inpu pa ame e s o all he s udied mechanical cha ac e is ics. Based on Table 5, we can s a e ha : •The inc ease in he e o by omi ing he sha e o he oleic acid (in he way he sensi i i y coe icien s o neu al ne wo ks a e calcula ed) amoun s o 16%, 16%, 20% and 7% o R m , M100, M200 and M300, which is close o he measu ing e o , i.e., he change in he me- chanical cha ac e is ics wi h a change in he sha e o he su ac an can be desc ibed by a single unc ion o bo h concen a ions o he oleic acid, which can be also deduced om Figs. 3-6. •On he o he hand, he same ope a ion o he sha e o he su ac an leads o, a leas , a 149% inc ease in he mechanical cha ac e is ics, which means ha he su ac an has a signi ican ly g ea e e ec on he mechanical cha ac e is ics. In Table 6 we calcula ed he o e all sensi i i y coe icien s o bo h inpu pa ame e s and all s udied mechanical cha ac e is ics. I is clea based on Table 6 ha he sha e o he su ac an has app oxima ely a 70% e ec on he s udied mechanical cha ac e is ics. Oleic acid has a nega i e en i onmen al impac especially du ing he ageing p ocess. Based on he esul s p esen ed in he able, we can hus ecommend a educ ion in he sha e o he so ene wi h a simul aneous inc ease in he sha e o he su ac an o be e en i onmen al mix u es wi h he same mechanical cha ac e is ics. 5. Conclusion Based on he esul s o he p edic ion o he mechanical cha ac e - is ics o ubbe com- pounds wi h a ious olume sha es o oleic acid and a ious weigh sha es o he su ac an , using a s a is ical se o 16 samples, we can d aw he ollowing conclusions: •ANN ne wo ks a e sui able ools o he p edic ion o ensile s eng h and modulus o ubbe mix u es con aining oleic acid, wi h a co - ela ion o a leas 0.98 and REL_RMSE maximum 0.8 %. •The dependence o mechanical p ope ies e sus su ac an amoun can be desc ibed by a single unc ion o 1 PHR and 3 PHR mix u es. •ANN ne wo ks can also p edic he mechanical p ope ies o non- measu ed concen a- ions o su ac an . •F om he p edic ion o mechanical p ope ies, we can conclude ha he le el o su - ac an amoun a 20 w % inc eases he mechanical p ope ies o a leas 80 % o he le el o inc ease o 30 w % o su ac an , and he e o e wi h he aid o ANNs, we can lowe he en i onmen al impac o he mix u es, especially as he ma e ial ages. •F om he sensi i i y analysis, we can conclude ha unca ing he changes in he oleic acid amoun leads only o a maximum 20 % inc ease in e o , he e o e we can conclude ha he oleic acid amoun in he egion o 1 PHR o 3 PHR does no ha e a signi ican e ec on he s udied mechanical p ope ies. •T unca ing he changes o su ac an amoun , howe e , leads o an inc ease in e o o a leas 150 %, he e o e we can conclude ha he su ac an amoun s be ween 3 w % and 30 w % ha e a e y la ge e ec on he mechanical p ope ies. •No malizing he sensi i i y coe icien s leads o a 70 % e ec on mechanical p ope ies by he su ac an amoun and only a 30 % e ec by he oleic acid. •The lesse e ec o oleic acid is because oleic acid is mainly used in ubbe mix u es o imp o e he p ocessing o mix u es which a e no c ea ed o any speci ic mechanical p ope ies. •Su ac an s, howe e , bond wi h ille s, and he e o e he amoun o su ac an has a signi ican e ec on mechanical p ope ies. •The submi ed s udy demons a es he possibili ies o how o op i- mize he ma e ial composi ion o p oduc s o ma e ial ailo ing wi h an accen on he en i onmen al bu den. Decla a ion o Compe ing In e es The au ho s decla e ha hey ha e no known compe ing inancial in e es s o pe sonal ela ionships ha could ha e appea ed o in luence he wo k epo ed in his pape . D. Seidl e al. Expe Sys ems Wi h Applica ions 208 (2022) 118039 8 Acknowledgemen s The esea ch was suppo ed by G an o SGS No. SP2022/81, VSB – Technical Uni e si y o Os a a, Czech Republic. Re e ences Choi, J., & Isaye , A. (2015). Na u al ubbe /s y ene bu adiene ubbe blends p epa ed by ul asonically aided ex usion. Jou nal o Elas ome s &Amp; Plas ics, 47(2), 170–193. h ps://doi.o g/10.1177/0095244313507803 Doh, J., Lee, S., & Lee, J. (2016). Back-p opaga ion neu al ne wo k-based app oxima e analysis o ue s ess-s ain beha io s o high-s eng h me allic ma e ial. Jou nal o Mechanical Science and Technology, 30(3), 1233–1241. h ps://doi.o g/10.1007/ s12206-016-0227-1 Doucou e, B., Agbossou, K., & Ca denas, A. (2016). Time se ies p edic ion using a i icial wa ele neu al ne wo k and mul i- esolu ion analysis: Applica ion o wind speed da a. Renewable Ene gy, 92, 202–211. h ps://doi.o g/10.1016/j. enene.2016.02.003 Fa hana, N., Abdul Majid, M., Paul aj, M., Ahmadhilmi, E., Fakhzan, M., & Gibson, A. (2016). A no el ib a ion based non-des uc i e es ing o p edic ing glass ib e/ ma ix olume ac ion in composi es using a neu al ne wo k model. Composi e S uc u es, 144, 96–107. h ps://doi.o g/10.1016/j.comps uc .2016.02.066 Ghosh, S., Cha e jee, R., & Shanke , P. (2016). Es ima ion o ash, mois u e con en and de ec ion o coal li ho acies om well logs using eg ession and a i icial neu al ne wo k modelling. Fuel, 177, 279–287. h ps://doi.o g/10.1016/j. uel.2016.03.001 Hwang, I., Pa k, H., & Chang, J. (2016). Ensemble o deep neu al ne wo ks using acous ic en i onmen classi ica ion o s a is ical model-based oice ac i i y de ec ion. Compu e Speech &Amp. Language, 38, 1–12. h ps://doi.o g/10.1016/j. csl.2015.11.003 Janˇ cíko ´ a, Z., Boˇ s´ ak, O., Zimný, O., Legoue a, M., Min´ a ik, S., Koˇ s ial, P., & Poulain, M. (2014). The Neu al Ne wo k Analysis o Op ical Glasses T ansmi ance. In In e na ional Ca pa ian Con ol Con e ence (pp. 196–200). Velke Ka lo ice. Jonˇ s a, Z., Koˇ s ial, P., Ruˇ ziak, I., Jonˇ s a, P., Ju ˇ cio ´ a, J., Janˇ cíko ´ a, Z., e al. (2011). Ad anced Fille s Enhancing The mal and Mechanical P ope ies o Rubbe Blends. Jou nal o Nano Resea ch, 13, 27–32. h ps://doi.o g/10.4028/www.scien i ic.ne / jnano .13.27 Kopal, I., Labaj, I., V ˇ sko ´ a, J., Ha niˇ c´ a o ´ a, M., Valíˇ cek, J., Ond uˇ so ´ a, D., e al. (2022). A Gene alized Reg ession Neu al Ne wo k Model o P edic ing he Cu ing Cha ac e is ics o Ca bon Black-Filled Rubbe Blends. Polyme s, 14(4), 653. h ps:// doi.o g/10.3390/polym14040653 Le, B., Y onne , J., & He, Q. (2015). Compu a ional homogeniza ion o nonlinea elas ic ma e ials using neu al ne wo ks. In e na ional Jou nal o Nume ical Me hods in Enginee ing, 104(12), 1061–1084. h ps://doi.o g/10.1002/nme.4953 Lenzi, G., E angelis a, R., Dua e, E., Colpini, L., Fo na i, A., Menechini Ne o, R., e al. (2016). Pho oca aly ic deg ada ion o ex ile eac i e dye using a i icial neu al ne wo k modeling app oach. Desalina ion and Wa e T ea men , 57(30), 14132–14144. h ps://doi.o g/10.1080/19443994.2015.1064035 Liu, W., Ma, J., Zhan, M., & Wang, K. (2015). The oughening e ec and mechanism o s y ene-bu adiene ubbe nanopa icles o no olac esin. Jou nal o Applied Polyme Science, n/a-n/a.. h ps://doi.o g/10.1002/app.41533 Lopes, H., Sil a, S., & Machado, J. (2021). Applica ion o a i icial neu al ne wo ks o p edic mechanical beha iou o co k- ubbe composi es. Neu al Compu ing and Applica ions, 33(20), 14069–14078. h ps://doi.o g/10.1007/s00521-021-06048-w Lubu a, J., Kojic, P., Pa lice ic, J., Ikonic, B., Omo jan, R., & Be a, O. (2021). P edic ion o ubbe ulcaniza ion using an a i icial neu al ne wo k. Chemical Indus y, 75(5), 277–283. h ps://doi.o g/10.2298/hemind210511026l Mansou , S., Hussein, M., & Moha am, A. (2014). The moelec ic Powe P ope ies o G aphi e-Loaded Ni ile Rubbe /Poly( inyl chlo ide) Blends Abo e he Pe cola ion Th eshold. Ad ances in Polyme Technology, 33(S1), n/a-n/a. h ps://doi.o g/ 10.1002/ad .21439 Raza i-Nou i, M., & Ka ami, M. (2014). E ec o ubbe con en on mo phology and he mal and heological beha io s o ac yloni ile-bu adiene ubbe /poly(e hylene- co- inyl ace a e)/o ganoclay nanocomposi es. Polyme , 55(26), 6940–6947. h ps:// doi.o g/10.1016/j.polyme .2014.10.050 Ren, Y., Zhao, S., Li, Q., Zhang, X., & Zhang, L. (2015). In luence o liquid isop ene on heological beha io and mechanical p ope ies o polyisop ene ubbe . Jou nal o Applied Polyme Science, 132(8), n/a-n/a. h ps://doi.o g/10.1002/app.41485 Ruˇ ziak, I., Koˇ s ial, P., Janˇ cíko ´ a, Z., Gaj ansk´ a, M., Kopal, I., K is ak, L., e al. (2018). A i icial neu al ne wo ks p edic ion o ubbe mechanical p ope ies in aged and nonaged s a e. Ad anced S uc u ed Ma e ials, 72, 27–35. h ps://doi.o g/10.1007/ 978-3-319-59590-0_3 Sa a , M., Lin, B., Chen, W., Langaue , D., Racla ska, H., P´ e issans, A., e al. (2019). E ec s o imp egna ed po assium on biomass o e ac ion. Ene gy P ocedia, 158, 55–60. h ps://doi.o g/10.1016/j.egyp o.2019.01.035 Salehi, F., & Raza i, S. (2016). Modeling o was e b ine nano il a ion p ocess using a i icial neu al ne wo k and adap i e neu o- uzzy in e ence sys em. Desalina ion and Wa e T ea men , 57(31), 14369–14378. h ps://doi.o g/10.1080/ 19443994.2015.1063087 Seidl, D., Koˇ s ial, P., Janˇ cíko ´ a, Z., Ruˇ ziak, I., Rusn´ ako ´ a, S., & Fa kaˇ so ´ a, M. (2011). Modal analysis – Measu emen s e sus FEM and a i icial neu al ne wo ks simula ion. Digi al In o ma ion P ocessing and Communica ions, 170–175. Sk obak, A., Janoˇ s ík, V., S anek, M., Manas, D., O sik, M., Senke ik, V., e al. (2016). Mechanical P ope ies o Injec ion Molded and Comp ession Molded Samples om Na u e-Bu adiene Rubbe . MATEC Web O Con e ences, 76, 02023. h ps://doi.o g/ 10.1051/ma eccon /20167602023 Vodka, O., Pog ebnyak, S. (2020). App oxima ion o S ess-S ain Cu e o Rubbe -Like Ma e ial Using An A i icial Neu al Ne wo k. In 2020 IEEE KHPI Week on Ad anced Technology (KHPI WEEK) (pp. 221-226). Kha ki , UKRAINE. Xu, T., Jia, Z., Luo, Y., Jia, D., & Peng, Z. (2015). In e acial in e ac ion be ween he epoxidized na u al ubbe and silica in na u al ubbe /silica composi es. Applied Su ace Science, 328, 306–313. h ps://doi.o g/10.1016/j.apsusc.2014.12.029 D. Seidl e al.