Determination of the Acidity of Waste Cooking Oils by Near Infrared Spectroscopy

p ocesses A icle De e mina ion o he Acidi y o Was e Cooking Oils by Nea In a ed Spec oscopy Juan F ancisco Ga cía Ma ín1,* , Ma ía del Ca men López Ba e a 1, Miguel To es Ga cía2, Qing-An Zhang 3and Paloma Ál a ez Ma eos 1 1Depa amen o de Ingenie ía Química, Facul ad de Química, Uni e sidad de Se illa, C/P o eso Ga cía González, 1, 41012 Se ille, Spain; [email p o ec ed] (M.d.C.L.B.); [email p o ec ed] (P.Á.M.) 2Depa amen o de Ingenie ía Ene gé ica. E.T.S. de Ingenie ía, Uni e sidad de Se illa, Camino de los Descub imien os, s/n, 41092 Se ille, Spain; [email p o ec ed] 3 School o Food Enginee ing and Nu i ion Sciences, Shaanxi No mal Uni e si y, Xi 0 an 710062, Shanxi, China; [email p o ec ed] *Co espondence: [email p o ec ed]; Tel.: +34-954-55-71-83 Recei ed: 29 Ap il 2019; Accep ed: 15 May 2019; Published: 21 May 2019   Abs ac : Was e cooking oils (WCO) ecycling companies usually ha e economic losses o buying WCO no sui able o biodiesel p oduc ion, e.g., WCO wi h high ee acidi y (FA). Fo his eason, he de e mina ion o FA o WCO by nea in a ed (NIR) spec oscopy was s udied in his wo k o assess i s po en ial o in si u applica ion. To do his, FA o 45 WCO was measu ed by he classical i a ion me hod, which anged be ween 0.15 and 3.77%. Then, he NIR spec a om 800 o 2200 nm o hese WCO we e acqui ed, and a pa ial leas squa es model was buil , ela ing he NIR spec a o FA alues. The accu acy o he model was qui e high, p o iding 2 o 0.970 and a a io o pe o mance o de ia ion (RPD) o 4.05. Subsequen ly, a model using an NIR ange simila o ha p o ided by po able NIR spec ome e s (950–1650 nm) was buil . The pe o mance was lowe ( 2 =0.905; RPD =2.66 ), bu e en so, wi h good accu acy, which demons a es he po en ial o NIR spec oscopy o he in si u de e mina ion o FA o WCO. Keywo ds: ee acidi y; NIRS; pa ial leas squa es; was e cooking oil 1. In oduc ion Fossil uel combus ion has nega i e e ec s on he en i onmen . The e o e, cu en esea ch is mainly ocused on he sea ch o economic, en i onmen ally iendly bio uels ha can eplace pe oleum uels. Diesel combus ion in engines leads o ai pollu ion by g eenhouse gas emissions, namely NO x , CO, and CO 2 , and o he des uc ion o he ozone laye , due o pho ochemical in e ac ions o he hyd oca bon, CO, and NO x emissions. As an al e na i e o pe oleum diesel uel, he cu en ly p oduced bio uel is biodiesel, which is composed o a y acid me hyl es e s (FAME). Biodiesel is ob ained by anses e i ica ion o ege able oils wi h sho -chain alcohols, mainly me hanol [1]. The main p oblems o biodiesel p oduc ion om ege able oils a e he high cos o he aw ma e ials, he h ea o ood secu i y [ 2 , 3 ], and he o e supply o glyce in as a byp oduc [ 4 ]. The cu en al e na i e o ege able oils is he use o was e cooking oils (WCO) as aw ma e ial o biodiesel p oduc ion. WCO a e much cheape han ege able oils om c ops o ees, a e no in con lic wi h ood secu i y, and a e a ailable as was e p oduc s [ 3 , 5 ]. The HORECA (ho els, es au an s, and ca e ing) sec o p oduces oughly 400,000 ons o WCO pe yea in Spain [3]. As a p o i able ma ke , many en e p ises dealing wi h he collec ion and ecycling o was e cooking oils o subsequen biodiesel p oduc ion ha e been se up [6]. P ocesses 2019,7, 304; doi:10.3390/p 7050304 www.mdpi.com/jou nal/p ocesses P ocesses 2019,7, 304 2 o 7 F ying consis s o in oducing ood o an oil ba h a empe a u es be ween 160 and 200 ◦ C, du ing a ce ain pe iod in he p esence o ai . Because o his high empe a u e exposu e, he oil unde goes nume ous physical and chemical changes ha make he cha ac e is ics o his oil di e en . Among hese changes, he mos no o ious one is he inc easing o acidi y, mainly p o oked by he elease o ee a y acids om he pa ial hyd olysis o iglyce ides. WCO ecycling companies usually collec hese oils om clean poin s o special con aine s, o buy en i e ucks om smalle companies. A e il e ing in hei acili ies o emo e le o e ood, lou , e c., WCO a e le o decan o ob ain h ee phases: Clean oil, wa e , and sludge [ 6 ]. This oil is hen analysed. The main equi emen is ha he ee acidi y (FA) is lowe han 2.5%. I no , in o de o be sui able o biodiesel p oduc ion, his oil mus be subjec ed o a chemical p ocess (es e i ica ion) o educe FA. Mos WCO ecycling companies usually do no ha e ei he eac o s no quali ied s a o ca ied ou he es e i ica ion eac ion, so hese companies ha e o sell his oil o bigge companies ha ha e he equi ed equipmen (mainly biodiesel p oduce s) a a lowe p ice han ha o pu chase, o o pay o an en i onmen al manage o dispose i . This p oblem could be o e come i FA was measu ed in si u. The de e mina ion o oil FA is ca ied ou by acid–base i a ion, and canno be pe o med in si u because i needs eagen s, sample p epa a ion, and labo a o y glasswa e, and gene a es chemical was es. Nea -in a ed spec oscopy (NIRS) is a low-cos , sa e, and non-des uc i e echnique, which gene ally does no equi e sample p epa a ion o chemicals [ 7 , 8 ]. Nowadays, po able nea -in a ed spec ome e s can be pu chased o oughly 6000 € (e.g., Oceanop ics Flame-NIR Spec ome e ). The e o e, NIRS can be sui able o in si u analysis. The po en ial o NIRS o oli e and sun lowe oils’ ee acidi y de e mina ion has been demons a ed [ 7 – 9 ], which makes i easible ha WCO 0 s FA can also be measu ed by NIRS. The aim o his wo k is o assess he easibili y o de e mining he ee acidi y o was e cooking oils by nea -in a ed spec oscopy and e i y whe he his echnique can be used o i s in si u de e mina ion by WCO collec ion and ecycling companies. To do his, FA de e mina ion in WCO was i s assayed using he whole NIR spec um. Subsequen ly, i was assayed in a educed NIR in e al, simila o ha p o ided by po able NIR spec ome e s, hus p o ing he applicabili y o he echnique. 2. Ma e ials and Me hods 2.1. Was e Cooking Oils (WCO) Fi y WCO we e collec ed om he uni e si y can een o he Reina Me cedes Campus (Uni e si y o Se ille) and di e en p i a e households. These WCO we e oli e, sun lowe and pomace oils, and mix u es o hem. This ensu ed a wide a ie y o oil ypes and deg ada ion deg ees. Once ecei ed a he labo a o y, WCO we e il e ed o emo e solids such as le o e ood, lou , e c. 2.2. F ee Acidi y (FA) De e mina ion by Acid–Base Ti a ion The pe cen age o ee a y acids ha WCO con ain, exp essed as oleic acid pe cen age, was measu ed acco ding o he O icial Me hods o Analysis o he EC [ 10 , 11 ]. B ie ly, 20 g WCO we e placed in o 250-cm 3 wide-mou h E lenmeye lasks along wi h 50 cm 3 e hanol/e hyl e he solu ion (1:1 / ) and a ew d ops o phenolph halein, and hen neu alized wi h 0.1 N KOH, p e iously s anda dized wi h benzoic acid. The i a ion ends when a eddish-b own colo change is obse ed. De e mina ions we e pe o med in duplica e. The pe cen age o acidi y o he oil was calcula ed acco ding o he ollowing equa ion: FA (%)=V×0.1 N ×0.282 m×100 whe e V is he spen olume o KOH in mL, 0.1 N s ands o he KOH no mali y, 0.282 is he equi alen weigh o oleic acid in mequi , and m is he mass o WCO sample in g ams. The FA con en s anged be ween 0.15% and 3.77%, being he mean alue and s anda d de ia ion, 0.94% and 0.79%, espec i ely. P ocesses 2019,7, 304 3 o 7 2.3. Spec a Acquisi ion Be o e spec a acquisi ion, WCO samples we e kep a 32 ◦ C o 30 min in a wa e ba h because NIR adia ion e lec ed and abso bed by a sample depends on i s empe a u e and his empe a u e ensu es ha all oil compounds a e comple ely dissol ed. A Vis/NIR Labspec P o model LSP 350-2500P (Analy ical Spec al De ices Inc., Boulde , CO, USA) spec opho ome e wi h h ee de ec o s was used o spec al acquisi ion, as desc ibed in [ 7 ]. The spec opho ome e is equipped wi h in e nal shu e s and au oma ic o se co ec ion, he scanning speed ime being 100 ms. Samples we e in oduced in a 10-mm qua z cu e e, which was placed in a cu e e accesso y joined by ib e op ic connec o s o he spec opho ome e ligh sou ce on one side, and o he spec opho ome e de ec o on he opposi e side. This op ical pa h leng h was selec ed because i showed highe abso p ion in ensi y han 1 mm, 2 mm, and 5 mm pa h-leng h qua z cu e es when acqui ing oli e oil NIR spec a [ 7 ]. NIR spec a om 800 o 2200 nm we e hen acqui ed in ansmi ance mode and eco ded using he Indico P o so wa e (Analy ical Spec al De ices Inc. Boulde , CO, USA), each spec al a iable ma ching o a 1-nm in e al. The spec um o each sample was acqui ed in duplica e. 2.4. Calib a ion P ocedu e and Model E alua ion Re lec ance da a we e i s ans o med o abso bance and hen maximum no malised. Pa ial leas squa es (PLS) models coupled o ull-c oss in e nal alida ion we e buil wi h The Unsc amble so wa e (CAMO So wa e AS, Oslo, No way) o he ull NIR spec um (800–2200 nm) and o he wa eleng h in e al speci ied o he Ocean Op ics Flame-NIR spec ome e (950–1650 nm). The s anda d e o o calib a ion (SEC) and he mul iple co ela ion coe icien o calib a ion ( 2c ) we e used o assess models 0 i ness. The p edic ion pe o mance o he models was e alua ed based on he s anda d e o o p edic ion (SEP), which co esponds o he s anda d e o o ull-c oss alida ion, he mul iple co ela ion coe icien o ull-c oss alida ion ( 2c ), and he a io o pe o mance o de ia ion (RPD). The RPD was de ined as he a io be ween he s anda d de ia ion om he FA e e ence da a and SEP. Among hem, he mos impo an pa ame e s o assess he pe o mance o a PLS model a e 2c (calib a ion) and RPD ( alida ion): The highe hese pa ame e s a e, he highe he accu acy o he model. To be speci ic, models wi h 2c≥ 0.90 a e conside ed o ha e excellen p ecision, while models wi h 2c =0.70–0.89 a e ega ded as good p ecision models [ 12 ]. As o RPD, his pa ame e mus be highe han 3 o a PLS model o be conside ed o excellen p ecision [ 13 ], al hough ano he au ho has poin ed ou ha p edic i e models sui able o ou ine analysis should ha e RPD alues be ween 2 and 10 [14]. 3. Resul s 3.1. Fea u es o he NIR Spec a o Was e Cooking Oils Abso bance in he NIR egion is linea wi h he concen a ion o o ganic compounds. The NIR spec um o a sample consis s o i s and second o e ones (800–1800 nm) and combina ion bands (1800–2700 nm) o undamen al, la gely hyd ogenic ib a ions ha occu in he MIR egion. The acqui ed WCO abso bance spec a (Figu e 1) we e p ac ically iden ical o hose p e iously ob ained o oli e oils (Figu e 2), so hei main ea u es a e desc ibed elsewhe e [ 7 ]. B ie ly, om le o igh , a b oad abso bance band is obse ed a 1210 nm, which is ela ed o C–H second o e ones and CH=CH– s e ching ib a ions. Nex , a wide abso p ion band, due o he wa e i s o e one, is obse ed in he 1350–1450 nm egion. The in ense abso p ion ound a 1720 nm is ela ed o he i s o e one o he C–H ib a ion o se e al chemical g oups (=CH–, –CH 3 , –CH 2 –) and is cha ac e is ic o iglyce ides and a y acids o ege able oils [ 7 ]. Ano he b oad wa e combina ion band is loca ed a 1880–2100 nm. The wo wa e bands (1350–1450 nm and 1880–2100 nm) show mul iple o e lapping bands. Finally, he abso p ion band o he C–H ib a ion o cis-unsa u a ion occu s a 2140 nm. P ocesses 2019,7, 304 4 o 7 Figu e 1. Nea -in a ed spec a o he was e cooking oils (WCO) used in his esea ch. Figu e 2. Visible/nea -in a ed spec a o oli e oils [7]. 3.2. NIR Pa ial Leas Squa es Model o he 800–2200 nm Spec um In ligh o he simila i y be ween he NIR spec a o ege able and was e cooking oils, i is easonable o hink ha ee acidi y o WCO can be ob ained om hei NIR spec a, as i has been demons a ed o oli e oil [ 7 , 8 ]. Indeed, he PLS calib a ion model o FA de e mina ion, buil using he 800–2200 NIR abso bance spec a o he 45 WCO samples, achie ed high 2c and RPD (0.970 and 4.05, espec i ely), which accoun s o he excellen p ecision o he model aking in o accoun he a o emen ioned c i e ia [ 12 – 14 ]. Figu es 3and 4illus a e he i o he da a o he p oposed model, along wi h he model s a is ics. The numbe o op imal p incipal componen s o build his PLS model was nine, explaining 95.1% o he a ia ion in he FA da a. Models wi h la ge o al explained a iance (close o 100%) explain mos o he a ia ion in he da a. Ideally, in o de o ha e simple models, he esidual a iance has o go down o ze o wi h as ew p incipal componen s as possible. I his we e no he case, i would mean ha he e migh be a la ge amoun o noise in he da a. 3.3. NIR Pa ial Leas Squa es Model o he 950–1650 nm Spec um Once i was p o ed he NIRS was sui able o de e mine FA in WCO, he second objec i e o his wo k was o y o quan i y FA in WCO using a na owe NIR ange, simila o he ypical ange p o ided by po able NIR spec ome e s. In his case, he selec ed ange was 950–1650 nm. The PLS model was buil using nine p incipal componen s, which explained 90.5% o he a ia ion in he FA da a. Acco ding o he c i e ia es ablished [ 12 – 14 ], his PSL model can be ega ded as a good p ecision model. As shown in Figu e 5, he s a is ics o he calib a ion model we e sa is ac o y, achie ing an accep able 2c (0.90). Howe e , SEP inc eased o 0.30% (Figu e 6), hus dec easing he RPD alue o 2.66. Compa ed o he model ob ained o he 800–2200 nm NIR ange, he p ecision o he model ob ained o 950–1650 nm is ma kedly lowe (Table 1). P ocesses 2019,7, 304 5 o 7 Figu e 3. Pe o mance o he 800–2200 pa ial leas squa es (PLS) model o ee acidi y (FA) de e mina ion. Figu e 4. Valida ion o he 800–2200 PLS model o FA de e mina ion. Table 1. S a is ics o he PLS models ob ained o he wo assayed nea -in a ed spec oscopy (NIRS) in e als. NIR In e al n 2c 2c SEC SEP RPD 800–2200 1401 0.970 0.914 0.113 0.195 4.05 950–1650 701 0.905 0.800 0.201 0.297 2.66 n: Spec al a iable numbe ; 2c : Co ela ion coe icien o calib a ion; SEC: S anda d e o o calib a ion; SEP: S anda d e o o p edic ion; RPD: Ra io o pe o mance o de ia ion. Figu e 5. Pe o mance o he 950–1650 PLS model o FA de e mina ion. This could be due o he missing in o ma ion ela ed o he i s o e one o he C–H ib a ion o se e al chemical g oups (=CH–, –CH 3 , –CH 2 –), cha ac e is ic o a y acids o ege able oils, which was ound o be 1720 nm. Tha aside, WCO con ain no only he ege able oil componen s, bu also deg ada ion p oduc s, due o he mal, oxida i e, and hyd oly ic eac ions ha occu du ing ying and ood es s ( lou , e c.). These componen s di e among samples and can dec ease he PLS model pe o mance. Thei nega i e e ec can be balanced, inc easing he numbe o con ibu ing spec al P ocesses 2019,7, 304 6 o 7 a iables o he model, which accoun s o he highe accu acy o he PLS model buil o he NIR ange 800–2200 nm. The p ecision o bo h PLS models could be imp o ed, elimina ing spec al a iables wi hou in o ma ion ela ed o he measu ed pa ame e , such as noise and backg ound, o emo ing ou lie s in he calib a ion and alida ion se s [ 7 ]. Howe e , he accu acy o he 950–1650 nm PLS model is enough o he p oposed pu pose: The in si u de e mina ion o FA o WCO. In ac , he achie ed SEP is no high when compa ed o o he au ho s 0 esul s. In his sense, he SEP using he whole NIR spec um ( om 750 o 2500 nm) was 0.35% o i gin oli e oils [ 8 ], while in his wo k, he SEP o was e cooking oils was 0.30%. Fu he mo e, he use o echniques o unin o ma i e spec al a iable emo ing equi es knowledge in p og ams such as MATLAB (The Ma hWo ks, Inc., Na ick, MA, USA), and he pu chase o hese p og ams and powe ul compu e s, which will ob iously dec ease he easibili y and p ac ical applica ion o NIRS o in si u FA de e mina ion. Figu e 6. Valida ion o he 950–1650 PLS model o FA de e mina ion. 4. Conclusions In he absence o es s a companies’ acili ies wi h di e en po able NIR spec ome e s, he po en ial o nea -in a ed spec oscopy o in si u was e cooking oils 0 ee acidi y de e mina ion has been demons a ed. High accu acy was achie ed when using he whole NIR ange p o ided by a bench op NIR spec ome e ( 2 =0.970; RPD =4.05). Al hough no pe ec , he accu acy o he pa ial leas squa es model, buil using a wa eleng h ange simila o ha inco po a ed in po able NIR spec ome e s (950–1650 nm), was good ( 2 =0.905; RPD =2.66). Fu he mo e, he s anda d e o o p edic ion achie ed wi h his na ow NIR in e al (0.30%) was in he ange o ha ob ained by o he au ho s o ege able oils using he ull NIR spec um, which accoun s o he abili y o NIR spec oscopy o de e mine ee acidi y o was e cooking oils. Au ho Con ibu ions: Concep ualiza ion, J.F.G.-M.; me hodology, J.F.G.-M. and P. Á .-M.; o mal analysis, J.F.G.-M. and Q.-A.Z.; in es iga ion, M.d.C.L.-B.; da a cu a ion, M.d.C.L.-B. and P. Á .-M.; so wa e, M.d.C.L.-B. and J.F.G.-M.; w i ing—o iginal d a p epa a ion, J.F.G.-M.; w i ing— e iew and edi ing, J.F.G.-M.; supe ision, J.F.G.-M and P.Á.-M.; p ojec adminis a ion, P.Á.-M.; unding acquisi ion, P.Á.-M. and M.T.-G. Funding: This esea ch was unded by Eu opean Union unde g an LIFE 13-Biose ille ENV/1113. Con lic s o In e es : The au ho s decla e no con lic o in e es . Re e ences 1. Pe eda Ma í n, J.; Ba iga Ma eos, F.; Á l a ez Ma eos, P. Ap o echamien o de las oleinas esiduales p oceden es del p oceso de e inado de los acei es ege ales comes ibles, pa a la ab icaci ó n de biodiesel. G asas y Acei es 2003,54, 130–137. 2. S á nchez, S.; Cue as, M.; Ga c í a-Ma í n, J.F. Bioe hanol p oduc ion: Co n /s lignocellulose biomass om oli e oil indus y and he po en ial ole in ensu ing ood secu i y. J. Fundam. Renew. Ene gy Appl. 2015 ,5, 18. 3. Ga c í a-Ma í n, J.F.; Ba ios, C.C.; Al é s- Á l a ez, F.J.; Dominguez-S á ez, A.; Al a ez-Ma eos, P. Biodiesel p oduc ion om was e cooking oil in an oscilla o y low eac o . Pe o mance as a uel on a TDI diesel engine. Renew. Ene gy 2018,125, 546–556. [C ossRe ] P ocesses 2019,7, 304 7 o 7 4. Ga c í a-Ma í n, J.F.; Al é s- Á l a ez, F.J.; To es-Ga c í a, M.; Feng, C.-H.; Á l a ez-Ma eos, P. P oduc ion o Oxygena ed Fuel Addi i es om Residual Glyce ine Using Bioca alys s Ob ained om Hea y-Me al- Con amina ed Ja opha cu cas L. Roo s. Ene gies 2019,12, 740. [C ossRe ] 5. Ga c í a-Ma í n, J.F.; Al é s- Á l a ez, F.J.; del Ca men L ó pez-Ba e a, M.; Ma í n-Dom í nguez, I.; Á l a ez- Ma eos, P. Ce ane numbe p edic ion o was e cooking oil-de i ed biodiesel p io o anses e i ica ion eac ion using nea in a ed spec oscopy. Fuel 2019,240, 10–15. [C ossRe ] 6. Á l a ez-Ma eos, P.; Ga c í a-Ma í n, J.F.; Gue e o-Vacas, F.J.; Na anjo-Calde ó n, C.; Ba ios, C.C.; P é ez-Camino, M.C. Valo iza ion o a high-acidi y esidual oil gene a ed in he was e cooking oils ecycling indus ies. G asas y Acei es 2019,40, e335. 7. Ga c í a Ma í n, J.F. Op ical pa h leng h and wa eleng h selec ion using Vis/NIR spec oscopy o oli e oil’s ee acidi y de e mina ion. In . J. Food Sci. Technol. 2015,50, 1461–1467. [C ossRe ] 8. Ma quez, A.J.; D í az, A.M.; Regue a, M.I.P. Using op ical NIR senso o on-line i gin oli e oils cha ac e iza ion. Sens. Ac ua o s B Chem. 2005,107, 64–68. [C ossRe ] 9. Laga de e, L.; Lecha , H.; Lacos e, F. D é e mina ion de l’acidi é e de l’indice de pe oxyde dans les huiles d’oli e ie ges e dans les huiles a in é es pa spec om é ie p oche in a ouge à ans o m é e de Fou ie . OCL Ol Co ps G as Lipides 2004,11, 70–75. [C ossRe ] 10. EU S anda d Me hods. Commission Regula ion (EEC) No 2568/91 on he cha ac e is ics o oli e oil and oli e- esidue oil and on he ele an me hods o analysis. O . J. Eu . Communi ies 1991,24, 1–15. 11. Eu opean Union. Commission implemen ing egula ion (EU) No 1348/2013 amending Regula ion (EEC) No 2568/91. O . J. Eu . Union 2013,2013, 31–67. 12. Shenk, J.S.; Wes e haus, M. Calib a ion he ISI way. In Nea In a ed Spec osc Fu u Wa es; Da ies, A.M.C., Williams, P.C., Eds.; NIR Publica ions: Chiches e , UK, 1996; pp. 198–202. 13. Shenk, J.S.; Wo kman, J.J.J.; Wes e haus, M.O. Applica ion o NIR Spec oscopy o Ag icul u al P oduc s. In Handbook o Nea -In a ed Analysis; Bu ns, D.A., Ciu czak, E.W., Eds.; Dekke Inc.: New Yo k, NY, USA, 2001; pp. 383–431. 14. Fea n, T. Assessing calib a ions: SEP, RPD, RER and R2. NIR News 2002,13, 12–13. [C ossRe ] © 2019 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 p://c ea i ecommons.o g/licenses/by/4.0/).