Characterization and Differentiation between Olive Varieties through Electrical Impedance Spectroscopy, Neural Networks and IoT

senso s A icle Cha ac e iza ion and Di e en ia ion be ween Oli e Va ie ies h ough Elec ical Impedance Spec oscopy, Neu al Ne wo ks and IoT JoséMiguel Madueño Luna 1,* , An onio Madueño Luna 2 and Ra ael E. Hidalgo Fe nández 3 1G aphics Enginee ing Depa men , Uni e si y o Se ille, 41013 Se ille, Spain 2Ae ospace Enginee ing and Fluid Mechanical Depa men , Uni e si y o Se ille, 41013 Se ille, Spain; [email p o ec ed] 3G aphics Enginee ing and Geoma ics Depa men , Uni e si y o Có doba, 14014 Có doba, Spain; [email p o ec ed] *Co espondence: [email p o ec ed] Recei ed: 20 Sep embe 2020; Accep ed: 16 Oc obe 2020; Published: 20 Oc obe 2020   Abs ac : Elec ical impedance has shown i sel o be use ul in measu ing he p ope ies and cha ac e is ics o ag i- ood p oduc s: ui quali y, mois u e con en , he ge mina ion capaci y in seeds o he os - esis ance o ui . In he case o oli es, i has been used o de e mine a con en and op imal ha es ime. In his pape , a sys em based on he Sys em on Chip (SoC) AD5933 unning a 1024-poin disc e e Fou ie ans o m (DFT) o e u n he impedance alue as a magni ude and phase and which, wo king oge he wi h wo ADG706 analog mul iplexe s and an ex e nal p og ammable clock based on a syn hesized DDS in a FPGA XC3S250E-4VQG100C, allows o he impedance measu emen in ag i- ood p oduc s wi h a equency sweep om 1 Hz o 100 kHz. This pape demons a es how elec ical impedance is a ec ed by he empe a u e bo h in eshly picked oli es and in hose p ocessed in b ine and p o ides a way o cha ac e ize cul i a s by making use o only he elec ical impedance, neu al ne wo ks (NN) and he In e ne o Things (IoT), allowing in o ma ion o be collec ed om he oli e samples analyzed bo h on a ms and in ac o ies. Keywo ds: elec ical impedance; SoC AD5933; a i icial neu al ne wo ks (ANNs); in e ne o hings (IoT); empe a u e 1. In oduc ion 1.1. The E ec o Tempe a u e Tempe a u e a ec s he ex u e o ui and ege ables, bo h aw [ 1 ] and cooked [ 2 ], and i also a ec s hei elec ical pa ame e s such as he elec ical impedance o hei pulp. A wide a ie y o p ocedu es and equipmen [ 3 , 4 ] exis o ob ain a ious pa ame e s associa ed wi h he pulp o hese ui s and ege ables (pene a ion esis ance, iscosi y, e c.) In he case o oli es, he e a e no s udies explici ly analyzing he in luence o empe a u e on he ui , especially in i s elec ical pa ame e s, al hough i s e ec is known in ce ain mechanized oli e p ocessing p ocedu es. This is he case, o example, in he indus ial oli e pi ing p ocess [ 5 ], which is done by au oma ic machines whe e clamps a e used o ap he oli e, while a punch needle goes h ough he clamps and, consequen ly, he oli es, causing hem o be de-s oned. This ype o machine eaches a conside able pi ing speed, which ob iously means a minimum cos pe uni o ui de-s oned. Howe e , he e y clamps ha hold he ui in place and he de-s oning ool cause b eakage in a conside able numbe o oli es du ing no mal ope a ion, which can be o he o de o 14%, especially Senso s 2020,20, 5932; doi:10.3390/s20205932 www.mdpi.com/jou nal/senso s Senso s 2020,20, 5932 2 o 20 in he “Go dal Se illana” oli e a ie y. While he p oduc i i y o hese pi ing machines makes hem mo e cos -e ec i e despi e he pe cen age o oli e b eakage compa ed o p e ious de-s oning sys ems, hese machines p esen a se ious p oblem, as he pe cen age o b oken oli es is qui e conside able. While mechanical solu ions educing he pe cen age o oli e b eakage in such de-s oning machines do no exis , i has been p o en ha cooling he oli es p io o pi ing minimizes he p oblem o he poin ha , wi h adequa e e ige a ion, he pe cen age o b oken oli es du ing he pi ing p ocess is o he o de o 2%, o he “Go dal Se illana” oli e a ie y p e iously ci ed. Fu he mo e, he cu en consume end o lowe sal and acidi y [ 6 ] means ha a signi ican p opo ion o he oli es de e io a e du ing he ime hey spend in e men e s. A e a ew mon hs a an a e age empe a u e o 25 ◦ C, he p obabili ies o ha ing a signi ican loss o p oduc in he pi ing p ocess a e e y high, due o a loss in i mness in he oli es, especially he la ge ones, which a e he mos in demand on he ma ke . By cooling o empe a u es close o 7 ◦ C (in some cases i eaches 0◦C , depending on he cooling sys em and he cos s he company wan s o ake on in he p ocess), i is possible o inc ease he i mness o he oli es while main aining all o hei cha ac e is ics, he eby educing he loss o p oduc due o he oli es ha ing a mo e igid ex u e. In addi ion, his cu s down on he hou s he machine is wo king, meaning a dec ease in machine jams and hou s when he punch needles a e s opped. One way o cha ac e izing he s a e o he oli es p io o de-s oning would be by doing an elec ical impedance measu emen on a ep esen a i e sample. As will be demons a ed in his s udy, he elec ical impedance o he oli e pulp is no only a ec ed by he oli e a ie y o he ype o indus ial p ocessing (g een “Se illana S yle” o black “Cali o nian S yle” oli es) [ 7 ], bu also by he empe a u e. Fo his eason, cha ac e izing his elec ical pa ame e e sus he empe a u e (especially be ween 7 ◦ C and 0 ◦ C) o a gi en ype o indus ial p ocessing would show whe he o no oli es a e ap be o e pi ing. 1.2. Elec ical Impedance: Measu emen The de e mina ion o elec ical p ope ies is used in a wide ange o disciplines and indus ies [ 8 ]. In he ag i- ood sec o , he use o elec ical conduc i i y is applied o he de e mina ion o di e se cha ac e is ics in ag i- ood p oduc s [ 9 , 10 ] such as os sensi i i y, eezing ole ance, mois u e con en , and seed ge mina ion [ 11 ]. The use o elec ical impedance spec oscopy is a echnique ha can p o ide e y good esul s in he ma u i y s age o ui and ege ables. In his pape , we a e going o ocus on a Sys em on Chip (SoC) capable o a complex impedance measu emen : he SoC AD5933 [12]. Impedance is a pa ame e o g ea impo ance o he cha ac e iza ion o ci cui s and elec onic componen s [ 13 ], as well as he ma e ials used in hei p oduc ion. Impedance (Z) is gene ally de ined as he o al opposi ion a de ice o ci cui o e s o he low o an al e na ing cu en (AC) o a speci ic equency, and i is ep esen ed as a complex numbe wi h a g aphic ep esen a ion on a complex plane. An impedance ec o consis s o he eal pa ( esis ance, R) and he imagina y pa ( eac ance, X). Impedance can be exp essed by using he ec angula coo dina es in he o m o R +j · X, o in he pola o m as a magni ude and phase angle: |Z|∠Ø [14]. The ins umen s mos commonly used o measu e impedance a e: he LCR me e o LCR b idge and he impedance analyze . The i s p o ides a simple and exac impedance measu emen o a speci ic equency alue. Howe e , o componen s o he han pu e induc o s (L), capaci o s © , o esis o s (R), i is inadequa e o de e mine alue. In hese cases, an impedance analyze is used o measu e and g aphically ep esen he complex impedance o he de ice being es ed o e a ange o equencies [ 15 ]. As hey a e high cos de ices [ 16 , 17 ], he eason why designs using LCR me e s (which a e cheape ) exis , is o ob ain a sys em impedance analyze combining i wi h i ual ins umen a ion [12,18]. The e a e a ious con igu a ions o he design o impedance measu ing b idges such as Sche ing [ 19 ] and he Maxwell b idge [ 20 ]. The di icul y wi h hese is ha hey need o ha e he balance condi ion. Mo eo e , hey a e gene ally used o pu e induc i e o capaci i e impedance Senso s 2020,20, 5932 3 o 20 measu emen s. In o de o ob ain he complex impedance, elec onic me hods a e used, such as he ec o al me hod and he me hod using wo quad a u e sinusoidal wa es [ 21 , 22 ].Ano he way o measu e he complex impedance is he h ee- ol age me hod [ 23 ]; howe e , his equi es ol ages o be aised o he squa e, which makes he measu emen e o s g ea e , in addi ion o he ac ha i equi es e y p ecise ins umen s o un he measu emen s. An Impedance/Gain-Phase analyze [ 24 ] is a measu ing ins umen o g ea alue o he s udy and design o elec onic ci cui s. This powe ul de ice is capable o ob aining diag ams sepa a ely, bo h in magni ude and in phase o any ne wo k o elec onic ci cui which has an inpu and an ou pu . Wi h his in o ma ion, i is possible, o example, o ob ain he ans e unc ion o a speci ic ci cui , al hough i s implemen a ion is no known. An Impedance/Gain-Phase analyze should be capable o ob aining a eliable ampli ude and phase diag ams co esponding o a speci ic ci cui . In bo h cases, hese a iables will be ep esen ed acco ding o equency. These ypes o analyze s mus be capable o gene a ing a sinusoidal signal and di ec ly applying i o he inpu o he ne wo k o be measu ed. Thus, a equency sweep is done on he ne wo k wi h a speci ic c i e ion ega ding he ini ial, in e media e, inal equency alues, numbe o poin s, linea o loga i hmic sweep, e c., which is no mally selec ed by he use . To gene a e he magni ude diag am, i is necessa y o ind he quo ien be ween he ampli udes o he ne wo k ou pu and inpu o each o he alues o be measu ed in he equency sweep. Fu he mo e, in o de o do he phase diag am, i is necessa y o ob ain he phase di e ence be ween he ne wo k ou pu and inpu signals again o each o he equency alues. The e o e, i is ob ious ha in o de o build an impedance/gain-phase analyze , a sine wa e gene a o is necessa y, a he e y leas , o gene a e a ci cui equency sweep capable o measu ing ampli ude and o ano he ci cui capable o measu ing he phase di e ence be ween wo signals. 1.3. P e ious Examples o he Use o Elec ical Impedance in Oli e P oduc ion and O he F ui s Elec ic conduc i i y has been used (only i s magni ude) o cha ac e ize di e en oli e a ie ies (Olea eu opa L.) [ 25 ], speci ically, ou di e en cul i a s: “Picual”, “Manzanilla de Se illa”, “Hojiblanca” and “Go dal Se illana” wi h he objec i e o es ablishing a ui ma u i y index and, hus o de e mining he op imal ime o ha es ing based on pa ame e s such as oil yield and quali y, demons a ing ha he conduc i i y inc eased wi h ui ma u i y and ha each a ie y had an a e age cha ac e is ic elec ic conduc i i y alue in he las s ages o ma u a ion. P e ious s udies ha e ocused in pa icula on de e mining ma u i y indexes and quali y pa ame e s [ 7 , 26 , 27 ], o on ways o slowing he ma u a ion pe iod, abo e all in climac e ic ui s, wi h he aim o leng hening he pe iod be ween ha es ing and comme cial consump ion [28]. 1.4. Neu al Ne wo ks (NN) o Adjus men and So ing A i icial neu al ne wo ks a e compu a ional models inspi ed by he beha io obse ed in hei biological coun e pa s [ 29 ]. Thei use is widesp ead in an abundance o ields. Focusing on he able oli e, we can see hei use in [30] o in [31]. In his s udy, we a e going o use wo ypes o neu al ne wo k. On he one hand, a neu al ne wo k o adjus men [ 32 ] o e alua e he e ec i eness o his echnique in gene a ing a alid impedance beha io mode o oli e pulp and, on he o he hand, a neu al ne wo k o so ing [ 33 ] o dis inguish be ween oli e a ie ies a di e en empe a u es. 1.5. The In e ne o Things (IoT) The use o he IoT is widely ex ended nowadays and he e a e nume ous cases ( o example [ 34 – 38 ]) in p ecision ag icul u e, i iga ion, empe a u e con ol, moni o ing o he ag icul u al p oduc ion p ocess, in au oma ed oli e chain p ocesses, o in he ope a ion o he pi ing, slicing, and s u ing machines hemsel es (DRR) [ 30 , 31 ]. This pape shows he de elopmen o equipmen based on he Senso s 2020,20, 5932 4 o 20 SoC AD5933 [ 12 ] which, oge he wi h a neu al ne wo k and an IoT sys em, allows o he analysis o he s a e o he oli es on he a m o in he ac o y be o e pi ing. The gene al objec i e se ou by his pape is o de elop an elec ical impedance measu emen sys em adap ed o able oli es, making use o he SoC AD5933. In o de o do his, wo asks a e going o be done: (1) Impedance modeling h ough neu al ne wo ks o wo a ie ies o oli es (“Go dal Se illana” and “Hojiblanca”), cu ed in caus ic soda and e men ed in b ine (an indus ial p ocess known as “Es ilo Se illano”). (2) Classi ica ion ia neu al ne wo ks o each oli e a ie y a h ee empe a u es (25 ◦ C, 7 ◦ C, and 0◦C). To his end, hese sys ems will be de eloped: • A speci ic de ice wi h he SoC AD5933 [ 12 ], ha includes an I 2 C communica ion in e ace [ 39 ], an ex e nal DDS gene a o based on a FPGA XC3S250E-4VQG100C [ 40 ] o conduc a comple e sweep om 1 Hz o 100 kHz and wo ADG706 analog mul iplexe s [ 41 ] o se he impedance ange o be measu ed. This de ice is con olled by a 32-bi ARM CORTEX M3 AT91SAM 3 × 8 E mic ocon olle wo king a 84 MHz [42]. •The sys em con ol so wa e using Ma lab p og aming language [43]. •IoT communica ion is based on [44] o gene a e a da abase wi h he ial esul s. 2. Ma e ials and Me hods 2.1. Oli e Va ie ies and Indus ial P ocess Used The oli e ui (Olea eu opaea L.) [ 25 ] is an o oid d upe whose size oscilla es om 0.6 o 2 cm in diame e and om 1 o 4 cm in leng h. This size depends on he a ie y (see Figu e 1 in [ 25 ]), he ege a i e s a e o he ee, he en i onmen al condi ions, and he cul i a ion echniques [ 45 ]. We used samples o 100 un ipe oli es om oli e o cha ds (200 ees ha −1 , Se ille, Spain), unde i iga ion and non-limi ing nu ien condi ions wi h mechanical ha es ing, and samples o 100 oli es in b ine om a ac o y in Se ille. In his wo k we used wo a ie ies: “Hojiblanca” and “Go dal Se illana” eshly picked in class 0, 1, and 2 [ 46 ] and p ocessed in b ine he “Es ilo Se illano” way. This ype o oli e, seasoned his way, is p ocessed in ou s ages [47–49]: •Lye ea men in NaOH 2–4% (p/ ) o 6–12 h •Rinsed in wa e (12–15 h) •Fe men ed in b ine (10–12% (p/ ) o 60–300 days) •Pi ed and s u ed o sliced. 2.2. The SoC AD5933 o Measu e Impedance Acco ding o he SoC AD5933 da ashee [ 12 ], “ he AD5933 is a high p ecision impedance con e e combining an in eg a ed equency gene a o wi h a 12-bi and 1 MSPS analogue- o-digi al con e e (ADC)”. The impedance o he de ice unde es (DUT) is sampled by he in e nal ADC and is p ocessed wi h a disc e e Fou ie ans o m (DFT)” [ 50 ]. The ou pu exci a ion ol age and he measu emen equency a e o ally p og ammable. Communica ion is done ia an I2C in e ace. Re iewing he a ailable bibliog aphy con i ms ha he SoC AD5933 has signi ican biological applica ions, such ha i has been used o moni o he g ow h o cell cul u es [ 51 , 52 ] in measu emen s in isola ed cells [ 53 ], de ec ion o blood clo ing [ 54 ], biosenso applica ions [ 55 ], and bio-impedance measu emen s [ 56 – 60 ]. Likewise, i is used in he moni o ing o “ echnical objec s”, o example, o co osion analysis in s eel s uc u es [ 61 , 62 ]. In o de o ob ain comp ehensi e in o ma ion abou he elec ical p ope ies o a measu ed objec and o use a sui able impedance spec um analysis me hod (equi alen ci cui modelling), impedance mus be measu ed a a wide ange o equencies [63]. Senso s 2020,20, 5932 5 o 20 In Table 1, he echnical da a o a ious impedance me e s based on he SoC AD5933 a e shown. Only h ee o he de ices desc ibed allow impedance we e o be measu ed in mo e han h ee o de s o magni ude o equency. The impedance ange measu ed is ypically om 10 Ω o mo e han 10 MΩ . Howe e , in many cases, i does no gi e he exac ange. The majo i y o he impedance me e s men ioned equi e addi ional analog on -ends o p o ide an adequa e in e ace be ween he SoC AD5933 and he de ice unde es (DUT) [64]. Table 1. Lis o elec ical impedance me e s based on he SoC AD5933 ( aken om [64]). Au ho Pu pose F equency Range Impedance Range Maximum E o C. J. Chen e al. [ 51 ] Moni o ing Cell Cul u es Se o 10 Hz No speci ied No speci ied T. Schwa zenbe ge e al. [52] Moni o ing Cell Cul u es 100 Hz–100 kHz No speci ied 2%–magni ude, 2%–a gumen M. H. Wang e al. [53] (uses an AD5934) Measu ing Isola ed Cells 0.1 Hz–100 kHz 100 Ω–10 MΩA ound 10% o cell measu emen J. B oede s e al. [55] Biosenso Applica ion 10 Hz–100 kHz 10 Ω–5 MΩNo speci ied P. Bogónez-F anco e al. [57] Bioimpedance Moni o 100 Hz–200 kHz 10 Ω–1 kΩ2.5%–magni ude, 4.5%–a gumen J. Fe ei a e al. [ 58 ] Bioimpedance Elec odes In Clo hing 5 kHz–450 kHz No speci ied 0.7%– esis ance, 17%– eac ance C. Ma go e al. [59] “Embedded” applica ions o bioimpedance 1 kHz–100 kHz No speci ied. No da a 2.5%–magni ude, 1.3%–a gumen A. Melwin y K. Rajaseka an [60] Body composi ion measu emen s Se o 50 kHz No speci ied 2% (no speci ied) J. Hoja y G. Len ka [61,62] Objec moni o ing echnique 0.01 Hz–100 kHz 10 Ω–10 GΩ1.6%–magni ude, 0.6%–a gumen O he au ho s [ 65 – 67 ] p oposed modi ica ions in he o iginal opology p o ided by he manu ac u e wi h he use o a mul iplex sys em o adjus he ange o he impedance measu emen , al hough his does no include changes in he sou ce clock inpu in o he SoC AD5933. In [ 68 ] he ope a ion o he SoC AD5933, he ha dwa e and so wa e de eloped, and he IoT sys em a e desc ibed. 2.3. Neu al Ne wo ks o Modeling and So ing Two ypes o neu al ne wo ks we e used om he Ma lab lib a ies: • A neu al ne wo k o adjus men , “ i ne ” [ 32 ], allows a desc ip ion o he e olu ion o he complex impedance in he pulp o 2 a ie ies o oli es o be ob ained. I is a conside able imp o emen o e he Hayden model [69]. • A so ing ne wo k, “pa e nne ” [ 33 ], o dis inguish be ween 6 cases (2 a ie ies and 3 empe a u es) in un ipe and ano he 6 cases (2 a ie ies and 3 empe a u es) in oli es p ocessed he “Es ilo Se illano” way. 3. Resul s Th ee e i ica ion es s o he de eloped ha dwa e ha e been ca ied ou (Sec ions 3.1–3.3) as well as a s udy on he e ec o a ie y on elec ical impedance in un ipe oli es (Sec ion 3.4) and on oli es in b ine (Sec ion 3.5). A model wi h neu al ne wo ks on he e olu ion o impedance in un ipe and b ined oli es o he wo a ie ies s udied in Sec ion 3.6. and a classi ie based on neu al ne wo ks capable o dis inguishing be ween 6 di e en cases in un ipe oli es was ained (Sec ion 3.7.1), in b ined oli es (Sec ion 3.7.2), and inally (Sec ion 3.8) he IoT sys em used is shown. Senso s 2020,20, 5932 6 o 20 3.1. Impedance Me e Ve i ica ion Tes ing: DUT Made Up o a Pu e Resis ance o 10 kΩ A es has been done using a pu e esis ance o 10 K Ω , he esul s a e in Figu e 1. The objec o he es is o make su e ha he equipmen is wo king co ec ly. As can be seen, he answe is he same in bo h he magni udes as in he phase h oughou he sweep om 1 Hz o 100 kHz. Figu e 1. Pola diag am: Comple e sweep om 1 Hz o 100 kHz wi h ZIN =10 K Ω (Magni ude 10 k Ω , Phase 0◦). 3.2. Impedance Me e Ve i ica ion Tes ing: DUT Comp ised o a SERIAL RLC Ci cui We conduc ed a es on he SERIAL RLC ci cui wi h an induc ance o L =56 mH, C =1500 pF and R =5.1 k Ω , wi h Zcal =10 k, ZBF =10 k, Range 1: 2 Vpp, PGA = × 1, Mul iplie = × 1 om 1 Hz o 100 kHz (see Tables 1 and 2 in [68]). Fo he es , h ee elemen s we e connec ed in se ies o build a ypical RLC ci cui . The heo e ical esonance equency is: o= 1 2·π·√L·C = 1 2·π·p56 ×10−3×1500 ×10−12 =17365.22 Hz (1) A low equency p edomina es he capaci i e pa wi h nega i e phases ha end o − 90 ◦ . In he p oximi y o he esonance zone, he phase ended o ze o and om ha momen he induc i e pa began o p edomina e, he phase in ha case posi i e ends o 90 ◦ . The heo e ical and expe imen al esul s a e shown in Figu e 2. The small disc epancies be ween heo e ical and expe imen al da a a e due o he ole ance o he componen s used in he eal es . Senso s 2020,20, 5932 7 o 20 Figu e 2. Pola diag am o he es wi h SERIAL RLC ci cui be ween 1 Hz and 100 KHz. The blue alues a e measu ed and he ed ones a e heo e ical. 3.3. Impedance Me e Ve i ica ion Tes ing: DUT Comp ised o a PARALLEL RLC Ci cui The es on he PARALLEL RLC ci cui wi h an induc ance o L =56 mH, C =1500 pF and R=5.1 kΩ , wi h Zcal =10 k, ZBF =10 k, Range 1: 2 Vpp, PGA = × 1, Mul iplie = × 1 om 1 Hz o 100 kHz (see Tables 1 and 2 in [68]). Fo his es , h ee elemen s ha e been connec ed in pa allel o build a ypical RLC ci cui in pa allel. The heo e ical equency will be he same as in he p e ious es . The heo e ical and expe imen al esul s a e shown in Figu e 3. Once again, he di e ences a e due o he ole ance o he componen s used. Figu e 3. Pola diag am o he es wi h he PARALLEL RLC ci cui be ween 1 Hz and 100 kHz. The blue alues a e measu ed and he ed ones a e heo e ical. Senso s 2020,20, 5932 8 o 20 3.4. Tes on Un ipe Oli es: The E ec o Oli e Va ie y on Elec ic Impedance This s udy has deal wi h oli e samples om he “Go dal Se illana” and “Hojiblanca” a ie ies, shown in Figu e 4, om wo oli e ees si ua ed side by side and he e o e subjec o he same en i onmen al and wa e ing condi ions. These samples we e ha es ed on he same day a he same ime (16/06/2019). Figu e 4. ( a ) “Go dal Se illana” and ( b ) “Hojiblanca” oli e a ie ies du ing he elec ical impedance measu emen . In Figu e 5, he e olu ion o he impedance p o ile ob ained (an a e age o 100 es s) appea s p esen ed as i s componen s X-R o un ipe oli es o bo h a ie ies a h ee empe a u es: 0 ◦ C (blue), 7◦C (yellow), and 25 ◦C ( ed). As can be seen, hey p esen a cha ac e is ic p o ile di e en ia ed bo h by empe a u e and maximum alues, he “Hojiblanca” being he one wi h highe alues in bo h componen s X, R, and he eby he module Z, bo h a 0 ◦ C and 7 ◦ C, while a 25 ◦ C i s p o ile is simila o he “Go dal Se illana”. P e ious es s ha e demons a ed ha hese impedance p o iles do no adjus well o models o he ype desc ibed [ 70 ] (see Figu e 3b o said ci a ion) wi h minimum a X close o ze o. Con e sely, in un ipe oli es a a low equency, he e a e high alues in bo h R and X ( ed ci cles on he g aphs). Likewise, wo ela i e minimums can be obse ed in he eac ance X alue (g een diamonds) a 7 ◦ C and 25 ◦ C. The end on he g aph o 0 ◦ C sugges s ha a equencies o e 100 kHz i would also exis in his case. Senso s 2020,20, 5932 9 o 20 Figu e 5. E olu ion o he impedance in ( a ) un ipe “Go dal Se illana” and ( b ) “Hojiblanca” a ie ies a 3 empe a u es: 0 ◦C, 7 ◦C, and 25 ◦C (a e age o 100 es s). The ed ci cles co esponds o =1 Hz. One o he p oblems ha some imes appea s is dis inguishing be ween he un ipe a ie ies, due o he ac ha hei ou wa d appea ance and size cause doub . This es demons a ed how hese a ie ies (un ipe) a ec he elec ical impedance he same as he es o he pa ame e s (especially empe a u e). This sys em would allow he sample a ie y o oli e o be iden i ied om he elec ical impedance wi hou using o he cha ac e ized pa ame e s. In Figu e 6, he impedance spec um o bo h a ie ies is shown oge he (in pola diag am). Each one has a cha ac e is ic p o ile a each empe a u e and he maximum di e ence is ound a low equencies, which could se e o dis inguish be ween un ipe oli e a ie ies as is also indica ed in [ 25 ]. Figu e 6. F equency sweep o measu e he elec ical impedance o he “Go dal Se illana” and “Hojiblanca” a ie ies when un ipe (a e age alues om 100 es s). 3.5. Tes on P ocessed Oli es: The e ec he Oli e Va ie y Has on Elec ical Impedance in Oli es P ocessed he “Es ilo Se illano” Way Jus as wi h un ipe oli es, i is some imes di icul o dis inguish be ween p ocessed a ie ies because o hei shape, size, and special colo (due o he chemical p ocesses hey unde go) and hey Senso s 2020,20, 5932 16 o 20 On he o he hand, o each hese esul s, he design o a ha dwa e has been chosen ha has allowed us o ob ain he maximum bene i s o he SoC AD5933: • A ci cui implemen ing an SoC AD5933 has been de eloped wi h all he pe iphe al elemen s necessa y o i o un. This p o o ype includes a pai o ADG706 analog mul iplexe s in o de o con e he ange in he impedance module o be measu ed. • In o de o achie e he maximum esolu ion in he DFT, a DDS based on an FPGA has been used o gene a e a clock signal o be p og ammed a will acco ding o he limi s o he equency sweep o be ca ied ou du ing he impedance measu emen . • P og amming he main applica ion has been done in Ma lab. In o de o con ol all o he elemen s, an ARM CORTEX M3 (AT91SAM3X8E) mic ocon olle has been used wi h an A duino DUE, implemen ing all o he i mwa e necessa y o con ol he ha dwa e. Las ly, he Picoblaze ou ine con ol embedded in he FPGA o he DDS has been implemen ed in ASM. • The esis ance es s and he RLC se ies/pa allel ci cui s ha e shown ha he sys em wo ks p ope ly. • The e is a unc ional limi a ion o he chip whe e he in e nal DSP speed is p opo ional o he clock speed applied ex e nally. In hese ci cums ances, o he low equency measu emen ( om 1 Hz o 30 Hz), a 25 kHz clock has been used which, compa ed o he equency used (16 MHz), makes he measu emen p ocess 640 imes slowe a low equencies. • The ci cui buil is expe imen al and, in o de o use i di ec ly on he a ms whe e he c op is loca ed, a e sion capable o wi hs anding hose wo king condi ions ough o be p oduced. Likewise, he applica ion should be an app, o example on a cell phone o a able , whe e i could connec o he compu e ia Blue oo h. • One possible op ion is o implemen all he ou ines h ough a mic ocon olle embedded in he FPGA, using a p og amable mic ocon olle di ec ly in C ( o example, Mic oblaze) in his case. •Finally, among he u u e a eas o wo k, we a e conside ing: • Inc easing he equency ange o 25 MHz (likely on a sys em which allows i o go o e 100 kHz ) wi h he objec i e o seeing i , a e ha poin , hey a e s anda d applica ion models like in Hayden’s. •S udying o he oli e a ie ies o comme cial ele ance like “Manzanilla” o “Cace eña” oli es. •S udying o he indus ial ea men s like he oxidized black oli e (Cali o nia s yle). • Running an analysis which co ela es he b eakage pe cen age in DRR machines di ec ly wi h he measu ed impedance alue o di e en a ie ies, p ocesses, and empe a u es. Fu u e s udies will include he applica ion o his me hodology in o he ui s such as oma o o che y. Au ho Con ibu ions: Concep ualiza ion, A.M.L.; Da a cu a ion, J.M.M.L.; In es iga ion, A.M.L. and R.E.H.F.; Supe ision, J.M.M.L., A.M.L. and R.E.H.F.; Valida ion, J.M.M.L., A.M.L. and R.E.H.F. All au ho s ha e ead and ag eed o he published e sion o he manusc ip . Funding: This esea ch ecei ed no ex e nal unding. Con lic s o In e es : The au ho s decla e no con lic o in e es . Re e ences 1. Bou ne, M.C. E ec o Tempe a u e on Fi mness o Raw F ui s and Vege ables. J. Food Sci. 1982 ,47, 440–444. [C ossRe ] 2. Bou ne, M.C.; Coms ock, S.H. E ec o Tempe a u e on Fi mness o The mally P ocessed F ui s and Vege ables. J. Food Sci. 1986,51, 531–533. [C ossRe ] 3. Bou ne, M.C. Food Tex u e and Viscosi y: Concep and Measu emen ; Academic P ess: London, UK, 2002; ISBN 9780080491332. Senso s 2020,20, 5932 17 o 20 4. Kılıçkan, A.; Güne , M. Physical p ope ies and mechanical beha io o oli e ui s (Olea eu opaea L.) unde comp ession loading. J. Food Eng. 2008,87, 222–228. [C ossRe ] 5. Gu ie ez Rubio, J. Maquina En iado a de Acei unas; ES 1 003 639 U. 1988. A ailable online: h p: //www.oepm.es/pd /ES/0000/000/01/00/36/ES-1003639_U.pd (accessed on 15 Sep embe 2020). 6. Lei a, D.; Tapia, F. Elabo aci ó n de Acei unas Con Bajo Con enido de Sodio. In P oducci ó n de Acei unas con Bajo Con enido de Sodio (“Ligh ”); Ins i u o de In es igaciones Ag opecua ias, Cen o Regional de In es igaci ó n In ihuasi: La Se ena, Chile, 2015; A ailable online: h p://biblio eca.inia.cl/medios/biblio eca/bole ines/ NR40474.pd (accessed on 13 Sep embe 2020). 7. Gómez, A.H.S.; Ga cía, P.; Na a o, L.R. Elabo a ion o able oli es. G asas Acei es 2006,57, 86–94. 8. Hla á ˇco á , Z. Low equency elec ic p ope ies u iliza ion in ag icul u e and ood ea men . Res. Ag ic. Eng. 2003,49, 125–136. [C ossRe ] 9. Mi chell, F.R.G.; Alwis, A.A.P.D. Elec ical conduc i i y me e o ood samples. J. Phys. E 1989 ,22, 554–556. [C ossRe ] 10. Nelson, S.O. Dielec ic p ope ies o ag icul u al p oduc s-measu emen s and applica ions. IEEE T ans. Elec . Insul. 1991,26, 845–869. [C ossRe ] 11. Repo, T.; Paine, D.H.; Taylo , A.G. Elec ical impedance spec oscopy in ela ion o seed iabili y and mois u e con en in snap bean (Phaseolus ulga is L.). Seed Sci. Res. 2002,12, 17–29. [C ossRe ] 12. AD5933; Analog De ices: No wood, MA, USA, 2005. A ailable online: h p://www.analog.com/media/en/ echnical-documen a ion/da a-shee s/AD5933.pd (accessed on 15 Sep embe 2020). 13. Rod í guez G ó mez, R.; C uz Hu ado, J. Sis ema de medici ó n y an á lisis de impedancia. Ing. Elec ó nica Au omá ica Comun. 2015,36, 56–66. 14. Okada, K.; Sekino, T. Agilen Impedance Measu emen Handbook. In A Guide o Measu emen Technology and Techniques; Agilen Technologies: San a Cla a, CA, USA, 2009. 15. Tegam. The LCR Me e as an Impedance Analyze . A ailable online: h p://www. egam.com/wp-con en / uploads/2015/10/AN303.pd (accessed on 15 Sep embe 2020). 16. Keysigh 4395A Ne wo k/Spec um/Impedance Analyze . A ailable online: h p://www.keysigh .com/en/ pd-1000000864%3Aepsg%3Ap o-pn-4395A/ne wo k-%0Aspec um-impedance-analyze ?cc=ES&lc=eng (accessed on 15 Sep embe 2020). 17. Keysigh 4194A Impedance/Gain-Phase Analyze . A ailable online: h ps://www.keysigh .com/en/ pd-1000003398%3Aepsg%3Ap o-pn-4194A/impedance-gain-phase-analyze ?cc=ES&lc=eng (accessed on 5 Sep embe 2020). 18. Juping, G.; Long, J.; Shenbei, Q.; Xinjian, W.; Zhike, X. Resea ching on he au oma ic impedance measu emen sys em. In P oceedings o he Eigh h In e na ional Con e ence on Elec ical Machines and Sys ems, Nanjing, China, 27–29 Sep embe 2005. 19. Ingenie ia Eléc ica F a edsa. A ailable online: h p://ingenie iaelec ica a edsa.blogspo .com.es/2014/11/ puen e-sche ing.h ml (accessed on 15 Sep embe 2020). 20. Wikipedia. Maxwell B idge. A ailable online: h ps://en.wikipedia.o g/wiki/Maxwell_b idge (accessed on 18 Sep embe 2020). 21. Ib ahim, K.M.; Abdul-Ka im, M.A.H. Digi al Impedance Measu emen by Gene a ing Two Wa es. IEEE T ans. Ins um. Meas. 1985,IM-34, 2–5. [C ossRe ] 22. Taha, S.M.R. Digi al measu emen o he pola and ec angula o ms o impedances. IEEE T ans. Ins um. Meas. 1989,38, 59–63. [C ossRe ] 23. S ebe , G.R. A Low Cos RF Impedance Analyze . Nu s and Vol s. Feb ua y 2008, pp. 38–41. A ailable online: h ps://www.nu s ol s.com/magazine/a icle/a_low_cos _ _impedance_analyze (accessed on 15 Sep embe 2020). 24. Cas ell ó , J.; Esp í , J.; Ga c í a, R.; Es e e, V. Analizado de Impedancia/Ganancia-Fase pa a PC. Re . Española Elec ónica 2001,565, 70–75. 25. Jus icia, M.; Madueño, A.; Ruiz-Canales, A.; Molina, J.M.; L ó pez, M.; Madueño, J.M.; G anados, J.A. Low- equency cha ac e isa ion o mesoca p elec ical conduc i i y in di e en a ie ies o oli es (Olea eu opaea L.). Compu . Elec on. Ag ic. 2017,142, 338–347. [C ossRe ] Senso s 2020,20, 5932 18 o 20 26. Wea e , G.M.; Jackson, H.O. Elec ic impedance, an objec i e index o ma u i y in peach. Can. J. Plan Sci. 1966,46, 323–326. [C ossRe ] 27. Ezeike, G.O.I. A esis i e p obe mois u e senso o opical oo c ops and ege ables. J. Ag ic. Eng. Res. 1987,37, 15–26. [C ossRe ] 28. Mon oya Li ola, M. Es udio de la Conduc i idad El é c ica Como Í ndice de Madu ez en F u os Clima é icos y su E oluci ó n Du an e la Conse aci ó n F igo í ica en A mos e a No mal y Modi icada. Ph.D. Thesis, UNED, Mad id, Spain, 1992. A ailable online: h ps://dialne .uni ioja.es/se le / esis?codigo=40951 (accessed on 15 Sep embe 2020). 29. Van Ge en, M.; Boh e, S. Edi o ial: A i icial Neu al Ne wo ks as Models o Neu al In o ma ion P ocessing. F on . Compu . Neu osci. 2017,11. [C ossRe ] 30. De J ó da L á za o, M.; Madueño Luna, A.; Lucas Pascual, A.; Ma í nez, J.M.M.; Canales, A.R.; Madueño Luna, J.M. ; Sego ia, M.J.; S á nchez, M.B. Deep lea ning in oli e pi ing machines by compu e ision. Compu . Elec on. Ag ic. 2020,171, 105304. [C ossRe ] 31. Lucas Pascual, A.; Madueño Luna, A.; de J ó da L á za o, M.; Molina Ma í nez, J.M.; Ruiz Canales, A.; Madueño Luna, J.M. ; Jus icia Sego ia, M. Analysis o he Func ionali y o he Feed Chain in Oli e Pi ing, Slicing and S u ing Machines by IoT, Compu e Vision and Neu al Ne wo k Diagnosis. Senso s 2020 ,20, 1541. [C ossRe ] 32. The Ma hWo ks Inc. Func ion Fi ing Neu al Ne wo k (Fi ne ). A ailable online: h ps://es.ma hwo ks.com/ help/deeplea ning/ e / i ne .h ml (accessed on 15 Sep embe 2020). 33. The Ma hWo ks Inc. Pa e n Recogni ion Ne wo k (Pa e nne ). A ailable online: h ps://es. ma hwo ks.com/help/deeplea ning/ e /pa e nne .h ml;jsessionid=370562d44 3c46b93a717 92677 (accessed on 18 Sep embe 2020). 34. Lougheed, E.C.; Mille , S.R.; Ripley, B.D.; Cline, R.A. Elec ical impedance o daminozide- and calcium- ea ed McIn osh apples. Expe ien ia 1981,37, 835–837. [C ossRe ] 35. Jackson, P.J.; Ha ke , F.R. Apple B uise De ec ion by Elec ical Impedance Measu emen . Ho Science 2000 , 35, 104–107. [C ossRe ] 36. S ou , D.G.; Hall, J.W.; McLaughlin, N.B. In i o plan impedance measu emen s and cha ac e iza ion o memb ane elec ical p ope ies: The in luence o cold acclima ion. C yobiology 1987 ,24, 148–162. [C ossRe ] 37. S ou , D.G. E ec o Cold Acclima ion on Bulk Tissue Elec ical Impedance. Plan Physiol. 1988 ,86, 283–287. [C ossRe ] 38. Baucho , A.D.; Ha ke , F.R.; A nold, W.M. The use o elec ical impedance spec oscopy o assess he physiological condi ion o kiwi ui . Pos ha es Biol. Technol. 2000,18, 9–18. [C ossRe ] 39. NXP Semiconduc o s. I 2 C Bus. A ailable online: h p://www.in e acebus.com/Design_Connec o _I2C.h ml (accessed on 14 Sep embe 2020). 40. Xilinx Inc. Spa an-3E Fpga. A ailable online: h ps://www.digikey.es/es/da ashee s/xilinxinc/xilinx-inc- ds312 (accessed on 15 Sep embe 2020). 41. ADG706 Analog Mul iplexe ; Analog De ices: No wood, MA, USA, 2005. A ailable online: h ps://www.analog.com/media/en/ echnical-documen a ion/da a-shee s/ADG706_707.pd (accessed on 15 Sep embe 2020). 42. Mic ochip ARM Co ex-M3. A ailable online: h ps://www.mic ochip.com/wwwp oduc s/en/ATSAM3X8E (accessed on 16 Sep embe 2020). 43. The Ma hWo ks Inc. Ma lab So wa e. A ailable online: h ps://es.ma hwo ks.com/p oduc s/ma lab.h ml (accessed on 11 Sep embe 2020). 44. D opbox. A ailable online: h ps://www.d opbox.com/(accessed on 19 Sep embe 2020). 45. Rappopo , H. Bo á nica y Mo olog í a. In El Cul i o del Oli o; Mundi-P ensa: Mad id, Spain, 2008; pp. 35–60. 46. Fe ei a, J. Explo aciones Oli a e as Colabo ado as; Numbe 5; Minis e io de Ag icul u a: Mad id, Spain, 1979. 47. Ga ido, A.; Ga c í a, P.; B enes, M. Oli e e men a ions. In Bio echnology: A Mul i olume Comp ehesi e T ea ise; Reed, H.J., Nagodawi ana, T.W., Eds.; Wiley-VCH: Weinheim, Ge man, 1995; pp. 593–625. 48. Es ada, J.M. La Acei una de Mesa: Nociones Sob e sus Ca ac e í s icas, Elabo aci ó n y Cualidades; Fundacion Pa a El Fomen o y P omocion de la Acei una de Mesa: Se illa, Spain, 2011; A ailable online: h p: //www.besana.es/si es/de aul / iles/lib oacei unamaque a080411.pd (accessed on 15 Sep embe 2020). Senso s 2020,20, 5932 19 o 20 49. Ga ido, A.; Ga c í a, P.; L ó pez, A.; A oyo, F.N. P ocessing Technology in Oli e Oil and Table Oli e; In e na ional Oli e Council: Mad id, Spain, 2006; A ailable online: h ps://pd s.seman icschola .o g/ 5999/ 039244e30eda8538c20a2 2b47092c4 55e.pd (accessed on 17 Sep embe 2020). 50. S ockham, T.G. High-speed con olu ion and co ela ion. In P oceedings o he Sp ing Join Compu e Con e ence on XX—AFIPS ’66 (Sp ing), Bos on, MA, USA, 26–28 Ap il 1966; ACM P ess: New Yo k, NY, USA, 1966; p. 229. 51. Chen, C.J.; Liu, J.T.; Chang, S.J.; Lee, M.W.; Tsai, J.Z. De elopmen o a po able impedance de ec ion sys em o moni o ing he g ow h o mouse L929 cells. J. Taiwan Ins . Chem. Eng. 2012,43, 678–684. [C ossRe ] 52. Schwa zenbe ge , T.; Wol , P.; B ischwein, M.; Kleinhans, R.; Demmel, F.; Lechne , A.; Becke , B.; Wol , B. Impedance senso echnology o cell-based assays in he amewo k o a high-con en sc eening sys em. Physiol. Meas. 2011,32, 977–993. [C ossRe ] 53. Wang, M.H.; Kao, M.F.; Jang, L.S. Single HeLa and MCF-7 cell measu emen using minimized impedance spec oscopy and mic o luidic de ice. Re . Sci. Ins um. 2011,82, 064302. [C ossRe ] 54. Helen Be ney, H.; O’Rio dan, J.J. Impedance Measu emen Moni o s Blood Coagula ion. Analog Dialogue 2008,42, 2–4. 55. B oede s, J.; Ducha eau, S.; an G ins en, B.; Vanaken, W.; Pee e s, M.; Cleij, T.; Thoelen, R.; Wagne , P.; de Ceuninck, W. Minia u ised eigh -channel impedance spec oscopy uni as senso pla o m o biosenso applica ions. Phys. S a us Solidi A 2011,208, 1357–1363. [C ossRe ] 56. Seoane, F.; Fe ei a, J.; Sanch é z, J.J.; B ag ó s, R. An analog on -end enables elec ical impedance spec oscopy sys em on-chip o biomedical applica ions. Physiol. Meas. 2008,29, S267–S278. [C ossRe ] 57. Bog ó nez-F anco, P.; Bay é s-Gen í s, A.; Rosell, J.; B ag ó s, R. Pe o mance o an implan able impedance spec oscopy moni o using ZigBee. J. Phys. Con . Se . 2010,224, 012163. [C ossRe ] 58. Fe ei a, J.; Seoane, F.; Lindec an z, K. AD5933-based elec ical bioimpedance spec ome e . Towa ds ex ile-enabled applica ions. In P oceedings o he Annual In e na ional Con e ence o he IEEE Enginee ing in Medicine and Biology Socie y, Bos on, MA, USA, 30 Augus –3 Sep embe 2011; IEEE: Pisca away, NY, USA, 2011; pp. 3282–3285. 59. Ma go, C.; Ka ib, J.; Nadi, M.; Rouane, A. A ou -elec ode low equency impedance spec oscopy measu emen sys em using he AD5933 measu emen chip. Physiol. Meas. 2013,34, 391–405. [C ossRe ] 60. Melwin, A.; Rajaseka an, K. Implemen a ion o Bioimpedance Ins umen Ki in ARM7. In . J. Ad . Res. Compu . Sci. So w. Eng. 2013,5, 1271–1273. 61. Hoja, J.; Len ka, G. In e ace ci cui o impedance senso s using wo specialized single-chip mic osys ems. Sens. Ac ua o s A Phys. 2010,163, 191–197. [C ossRe ] 62. Hoja, J.; Len ka, G. A Family o New Gene a ion Minia u ized Impedance Analyze s o Technical Objec Diagnos ics. Me ol. Meas. Sys . 2013,20, 43–52. [C ossRe ] 63. G imnes, S.; Ma insen, O.G. Bioimpedance and Bioelec ici y Basics, 3 d ed.; Else ie Academic P ess: Ams e dam, The Ne he lands, 2014; ISBN 9780124115330. 64. Chabowski, K.; Piasecki, T.; Dzie ka, A.; Ni sch, K. Simple Wide F equency Range Impedance Me e Based on AD5933 In eg a ed Ci cui . Me ol. Meas. Sys . 2015,22, 13–24. [C ossRe ] 65. Simic, M. Realiza ion o Complex Impedance Measu emen Sys em Based on he In eg a ed Ci cui AD5933. In P oceedings o he 21s Telecommunica ions Fo um Tel o (TELFOR), Belg ade, Se bia, 26–28 No embe 2012; IEEE: Pisca away, NY, USA, 2013; pp. 573–576. 66. Simi´c, M. Complex Impedance Measu emen Sys em o he F equency Range om 5 kHz o 100 kHz. Key Eng. Ma e . 2015,644, 133–136. [C ossRe ] 67. Simic, M. Realiza ion o digi al LCR me e . In P oceedings o he In e na ional Con e ence and Exposi ion on Elec ical and Powe Enginee ing (EPE), Iasi, Romania, 16–18 Oc obe 2014; IEEE: Pisca away, NY, USA, 2014; pp. 769–773. 68. Madueño, J.M. Pape s Appendix. A ailable online: h ps://www.d opbox.com/sh/wg1hmyxdg 558nc/ AACPWe1XJi_ YcJoFzVih 3Ya?dl=0(accessed on 15 Sep embe 2020). 69. Hayden, R.I.; Moyse, C.A.; Calde , F.W.; C aw o d, D.P.; Fensom, D.S. Elec ical Impedance S udies on Po a o and Al al a Tissue. J. Exp. Bo . 1969,20, 177–200. [C ossRe ] Senso s 2020,20, 5932 20 o 20 70. Wu, L.; Ogawa, Y.; Tagawa, A. Elec ical impedance spec oscopy analysis o eggplan pulp and e ec s o d ying and eezing- hawing ea men s on i s impedance cha ac e is ics. J. Food Eng. 2008 ,87, 274–280. [C ossRe ] 71. The Ma hWo ks Inc. Ne wo k Pe o mance. A ailable online: h ps://es.ma hwo ks.com/help/deeplea ning/ e /pe o m.h ml (accessed on 14 Sep embe 2020). Publishe ’s No e: MDPI s ays neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a ilia ions. © 2020 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/).