Appl. Sci. 2025, 15, x h ps://doi.o g/10.3390/xxxxx
A icle
Op imizing Bioleaching o P in ed Ci cui Boa d Coppe
Reco e y: An AI-D i en
RGB-Based App oach
Jo di Vi es Pons 1, Albe Come ma 2, Te esa Escobe 3, An onio D. Do ado 4 and Ma a I. Ta és-Pue as 1,*
1 Depa men o Mining, Indus ial and ICT Enginee ing, In o ma ics Enginee ing Sec ion, Uni e si a
Poli ècnica de Ca alunya—Ba celonaTech (UPC), 08242 Man esa, Ba celona, Spain;
2 Se a Hún e Fellow, Depa men o Mining, Indus ial and ICT Enginee ing, Elec onics Enginee ing
Sec ion, Uni e si a Poli ècnica de Ca alunya—Ba celonaTech (UPC), 08242 Man esa, Ba celona, Spain;
3 Depa men o Mining, Indus ial and ICT Enginee ing, Elec onics Enginee ing Sec ion, Uni e si a
Poli ècnica de Ca alunya—Ba celonaTech (UPC), 08242 Man esa, Ba celona, Spain;
4 Depa men o Mining, Indus ial and ICT Enginee ing, Chemical Enginee ing Sec ion, Uni e si a
Poli ècnica de Ca alunya—Ba celonaTech (UPC), 08242 Man esa, Ba celona, Spain;
* Co espondence: ma a.isabel. a e[email p o ec ed]
Fea u ed Applica ion: This AI-d i en op imiza ion esea ch ocuses on bioleaching o
Cu eco e y om e-was e, in eg a ing RGB senso da a and machine lea ning o enable
eal- ime moni o ing and p edic i e main enance. The IIoT-based app oach imp o es
eco e y a es, lowe s en i onmen al ha m, and p omo es sus ainable esou ce man-
agemen ac oss diffe en ope a ional scales in e-was e ecycling plan s.
Abs ac : Reco e ing coppe om end-o -li e elec onics, especially om p in ed ci cui
boa ds, p o ides significan economic benefi s, educes en i onmen al impac , and sup-
po s a ci cula economy. This case s udy p esen s a da a-d i en app oach o p edic ing
coppe eco e y in he elec olysis s age o a bioleaching p ocess by u ilizing RGB senso
eadings. We es ed nine eg ession models using RGB alues om expe imen al da a.
The g adien boos ing model, op imized ia esponse su ace me hodology (RSM), ou -
pe o med he o he s, wi h p edic ions ma ching 84% o obse ed pae ns. These esul s
demons a e s ong p edic i e capabili ies, wi h scope o u he accu acy enhancemen s.
We offe an open-sou ce, web-based digi al win designed specifically o moni o he bi-
oleaching plan , enabling eal- ime and his o ical da a analysis o suppo p edic i e
main enance. Ou esul s unde sco e he po en ial o op imize he en i e bioleaching p o-
cess, ma king a significan ad ancemen o la ge-scale coppe eco e y. This s udy is he
fi s o in es iga e p edic i e bioleaching con inuous p ocesses in a semi-indus ial e-
was e plan using RGB senso s, p esen ing a no el app oach in he field.
Keywo ds: a ificial in elligence; indus ial sys ems; machine lea ning; Indus ial IoT;
eal- ime sys ems; digi al win; coppe eco e y; bioleaching
1. In oduc ion
The Global E-was e Moni o 2020 analyzes global e-was e flows, emphasizing hei
ole in he ci cula economy h ough esou ce eco e y and ecycling o educe
Academic Edi o (s): Name
Recei ed: 29 No embe 2024
Re ised: 18 Decembe 2024
Accep ed: 25 Decembe 2024
Published: da e
Ci a ion: Vi es Pons, J.; Come ma,
A.; Escobe , T.; Do ado, A.D.;
Ta és-Pue as, M.I. Op imizing
Bioleaching o P in ed Ci cui
Boa d Coppe Reco e y: An
A ificial-In elligence-D i en
Red–G een–Blue-Based App oach.
Appl. Sci.
2024, 15, x.
hps://doi.o g/10.3390/xxxxx
Copy igh : © 2024 by he au ho s.
Submied o possible open access
publica ion unde he e ms and
condi ions o he C ea i e Commons
A ibu ion (CC BY) license
(hps://c ea i ecommons.o g/li-
censes/by/4.0/).
Appl. Sci. 2025, 15, x FOR PEER REVIEW 2 o 25
en i onmen al impac and inc ease ma e ial euse [1]. The BIOMETALLUM p ojec [2]
ocuses on de eloping bioleaching-based plan s o ex ac ing aluable me als om e-
was e, offe ing bo h ex ac ion and main enance se ices. Led by he BIOGAP g oup, his
p ojec de elops an al e na i e p ocess o eco e me als om e-was e ecycling using bio-
based echnology ha offe s en i onmen al ad an ages compa ed o con en ional me h-
ods. The p ocess significan ly educes ene gy use and emissions compa ed o py ome al-
lu gy, while eci cula ion minimizes he consump ion o eagen s and he p oduc ion o
liquid was e in compa ison o adi ional hyd ome allu gy [3]. Roo ed in ci cula econ-
omy p inciples [1], i suppo s a local business model ha educes esou ce dependency
and p omo es sus ainable ma e ial euse, add essing bo h en i onmen al and economic
challenges. Howe e , his a icle ocuses solely on he echnological aspec o he s udy, p e-
sen ing ini ial esul s om he applica ion o a digi al win based on he colo in o ma ion
o he solu ion wi hin he bioleaching p ocess. This echnological app oach aims o demon-
s a e he po en ial o he implemen ed me hodology, lea ing he ull e alua ion o eco-
nomic easibili y and b oade impac s wi hin he ci cula economy model o u u e wo k.
1.1. The Bioleaching Plan
Bioleaching le e ages mic oo ganisms o con e insoluble mine als in o soluble
o ms h ough hei me abolic p ocesses, enabling he eco e y o me als om low-g ade
o es whe e adi ional mining and efining a e no cos -effec i e [3,4]. A majo ad an age
o bioleaching is i s abili y o efficien ly p ocess bo h p ima y and seconda y esou ces
wi h a less en i onmen al impac han me hods like py ome allu gy [5].
Unlike con en ional chemical leaching, bioleaching minimizes ecological ha m by
employing na u ally occu ing mic oo ganisms, a oiding he use o oxic agen s such as
cyanide o sul u ic acid [3,4]. This makes he p ocess sa e and mo e sus ainable. Econom-
ically, bioleaching p o ides a cos -effec i e solu ion, allowing he ex ac ion o me als
om low-g ade o es ha would o he wise be imp ac ical. Addi ionally, bioleaching op-
e a es a ambien empe a u es and p essu es, educing ene gy consump ion compa ed
o con en ional chemical leaching, which ypically equi es high ene gy inpu s due o ex-
eme empe a u e and p essu e condi ions [6].
The BIOMETALLUM p ojec ope a es a semi-indus ial bioleaching plan a he
EPSEM-UPC onsi e acili y dedica ed o expe imen al esea ch o enhancing and op i-
mizing me al eco e y, wi h a ocus on coppe (Cu (II)) and i on (Fe (III)) ex ac ion om
elec onic was e like p in ed ci cui boa ds (PCBs) om compu e s and mobile phones.
Al hough he plan cu en ly ope a es wi h an effec i e moni o ing sys em, ou esea ch
is ad ancing owa ds he in eg a ion o p edic i e, moni o ing, and simula ion echnolo-
gies, including digi al wins and machine lea ning models. These enhancemen s a e ex-
pec ed o boos he p ecision and efficiency o me al eco e y. These echnologies could
significan ly educe he en i onmen al impac o he mining and chemical indus ies glob-
ally [7]. Figu e 1 p esen s a g aphical ep esen a ion o he bioleaching plan , ea u ing a
sc eensho aken om he Biome allum web-based digi al win, accessible ia [8]. The
plan is s uc u ed in o h ee p ima y s a ions, each assigned o dis inc ope a ional asks.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 3 o 25
Figu e 1. Diag am o he bioleaching plan , highligh ing i s key componen s: bio eac o , leaching
anks, and elec olysis uni .
Al hough he ull indus ial p ocess is mo e in ica e, a simplified o e iew o i s
ope a ion is ou lined below:
Bio eac o : The bio eac o is a i al componen o he bioleaching sys em, esponsible
o main aining and egene a ing he biological agen s ha d i e he leaching p ocess. I
con inuously moni o s key pa ame e s, including edox po en ial, pH, empe a u e, and
dissol ed oxygen, all o which a e essen ial o op imizing mic obial ac i i y and ensu ing
efficien sys em pe o mance. The bio eac o ope a es con inuously, using an ai injec o
o main ain oxygen le els and o ensu e uni o m mixing o he bioleaching solu ion. The
empe a u e is egula ed a 31 °C, which is op imal o mic obial g ow h. pH is supe ised
wi hin he na ow ange o 1.6 o 1.8 using wo pumps: PB1 o acid addi ion and PB2 o
base addi ion. This egula ion is c ucial, as any de ia ion could deac i a e he mic obial
agen s, dis up ing he leaching p ocess. Real- ime moni o ing o dissol ed oxygen and
edox po en ial ensu es con inuous adjus men o ope a ional pa ame e s, while a maxi-
mum liquid le el senso p e en s o e flow. In he e en o anomalies, such as an abno -
mal pH inc ease o edox po en ial d op, he sys em au oma ically adjus s eagen addi-
ion o mixing in ensi y o es o e op imal condi ions.
Leaching Tanks: The leaching p ocess ans e s he biologically ac i e solu ion om
he bio eac o o he leaching anks, whe e i in e ac s wi h he elec onic was e ma e ial.
He e, mic obial agen s acili a e he b eakdown and dissolu ion o me al ions om he
solid ma ix. The anks a e equipped wi h senso s and ac ua o s o main ain necessa y
en i onmen al condi ions. pH, empe a u e, and edox le els a e moni o ed, while
pumps and al es egula e he flow o leaching agen s and emo al o dissol ed me als.
High- and low-le el senso s ensu e ank olumes emain wi hin sa e limi s o p e en
o e flow o unde flow. Abno mali ies igge au oma ed esponses, such as hal ing he
addi ion o new was e o ac i a ing eci cula ion pumps. The eci cula ion sys em, man-
aged by pump PL1, ensu es uni o m dis ibu ion o bioleaching agen s and p e en s solid
esidue sedimen a ion. Addi ionally, colo ime ic senso s measu e he solu ion’s ed,
g een, and blue componen s o es ima e me al ion concen a ion in eal ime, p o iding
an indi ec measu e o he leaching p ocess’s efficiency.
Elec olysis Uni : Commonly e e ed o as he coppe eco e y uni , his s age con-
s i u es he final s ep in he bioleaching p ocess. He e, dissol ed coppe ions a e sepa a ed
om he solu ion and deposi ed as solid coppe h ough he elec owinning echnique,
Appl. Sci. 2025, 15, x FOR PEER REVIEW 4 o 25
enabling efficien me al eco e y. In his p ocess, an elec ic cu en is applied o p ecipi-
a e coppe on o ca hodes. The sys em includes senso s o moni o ol age, cu en , and
solu ion composi ion. Ci cula ion pump PB4 ensu es con inuous solu ion flow be ween
he leaching anks and he eco e y uni , main aining a s eady supply o coppe ions o
deposi ion. P essu e and le el senso s egula e he inflow and ou flow o he solu ion o
p e en imbalances ha could dis up he elec owinning p ocess. Anomalies in ol age
o cu en igge au oma ic adjus men s o ope a ing pa ame e s, ensu ing efficien me al
eco e y and p o ec ing he equipmen om po en ial damage.
The bioleaching plan ope a es h ough a obus ha dwa e and so wa e sys em buil
a ound an indus ial A duino de ice in eg a ed in o a SCADA amewo k; u he de ails
a e a ailable in ou p e ious wo k [9]. Key componen s like he ESP32-PLC, Raspbe y
Pi, and Node-Red o m a flexible con ol s uc u e. The ESP32-PLC se es as he main
con olle , in e acing wi h senso s and ac ua o s, managing con ol loops, and adjus ing
pa ame e s such as pH o hal ing pumps du ing eme gencies. Da a collec ed by he
ESP32-PLC a e ansmied o a Raspbe y Pi unning Node-Red, which p ocesses and
isualizes he da a h ough a cloud-based Big Da a Sys em (BDS). The sys em uses he
MQTT p o ocol o eliable, low-la ency communica ion. Technicians can moni o and
con ol ope a ions in eal ime ia a ouchsc een panel onsi e o emo ely ia he Teleg am
app, wi h manual o e ides a ailable o in e en ion when necessa y.
A key ea u e o ou semi-indus ial bioleaching plan is he in eg a ion o an ad-
anced RGB digi al senso a he bio eac o s a ion. This non-in asi e RGB senso
measu es Fe (III) and Cu (II) concen a ions by de ec ing colo changes in he solu ion,
elimina ing he need o complex sampling and analysis. The applica ion o RGB senso s
in bioleaching is an unde explo ed a ea, offe ing po en ial o moni o ing mic obial ac i -
i y. RGB senso s ope a e wi hou di ec con ac wi h he co osi e and agg essi e liquids
in ol ed in chemical p ocesses, significan ly ex ending hei ope a ional li espan. They
a e no only cos -effec i e bu also offe a clea and eliable co ela ion wi h eac ion p o-
g ess, acili a ing apid da a acquisi ion. Unlike s anda d models ha ely on sample ex-
ac ion, ex ended analysis imes, and manual esul in e p e a ion, hese senso s enable
eal- ime moni o ing. Thei implemen a ion in ou bioleaching plan can op imize ope a-
ions by au oma ically hal ing he p ocess a he op imal h eshold, leading o subs an ial
sa ings in ime, ma e ials, and labo cos s. I s pe o mance has been alida ed by compa -
ing i s eadings wi h con en ional me hods, such as UV-VIS spec oscopy and a omic ab-
so p ion, demons a ing i s accu acy and eliabili y in he ha sh bioleaching en i onmen ,
as shown in ou p e ious wo k [10]. This senso can de ec sub le colo changes in he
solu ion (see Figu e 2), enabling eal- ime moni o ing and analysis o me al dissolu ion
and eco e y p ocesses.
Figu e 2. Colo e olu ion a diffe en s ages o he edox eac ion [10].
Mul iple RGB BH1749NUC senso s [11] a e s a egically placed a c i ical poin s in
he plan and isola ed om he chemically agg essi e pH ange o 1.6 o 1.8, whe e he
leaching p ocess occu s, using cus om 3D-p in ed enclosu es moun ed ex e nally o he
eci cula ion pipes (see Figu e 3). These anspa en enclosu es a e designed o fi a ound
he pipes, allowing p ecise measu emen s. To minimize ambien ligh in e e ence, a
Appl. Sci. 2025, 15, x FOR PEER REVIEW 5 o 25
pulsed LED ci cui (Figu e 3) p o ides con olled illumina ion, ensu ing accu a e colo
da a cap u e. The senso s demons a ed obus ness and eliabili y in a 24 h ambien ligh
es (day and nigh ), u he confi ming hei sui abili y o eal-wo ld bioleaching appli-
ca ions.
Figu e 3. In eg a ion o colo senso s wi h eci cula ion pipes.
The RGB senso , coupled wi h an ESP32-based p og ammable logic con olle (PLC),
p ocesses and ansmi s da a o analysis, enabling eal- ime moni o ing. This sys em en-
hances he efficiency and sus ainabili y o me al eco e y compa ed o adi ional me h-
ods and p o ides aluable insigh s o bio echnological esea ch in o me al eco e y om
elec onic was e and o he hyd ome allu gical p ocesses.
Da a om he RGB senso and o he plan componen s a e ansmied o a cloud-
based Big Da a Sys em (BDS) ia Web Socke s, acili a ing bo h da a isualiza ion and
command execu ion. In e ac i e dashboa ds display key me ics such as pH, empe a u e,
and me al eco e y a es, while his o ical da a analysis helps iden i y ends, anomalies,
and oppo uni ies o p edic i e main enance o p ocess op imiza ion.
Secu i y is a undamen al aspec o he SCADA sys em. The plan ’s so wa e secu ely
s o es clien keys o se e au hen ica ion, which a e alida ed by a oo ce ifica ion au-
ho i y (CA). This ensu es ha only au hen ica ed de ices can access he se e , sa e-
gua ding agains unau ho ized access and da a b eaches. Enc yp ed communica ion
channels and obus au hen ica ion p o ocols a e employed o main ain he in eg i y and
confiden iali y o he ansmied da a [9].
1.2. AI and Machine Lea ning Algo i hms o Bioleaching P ocess Op imiza ion
A ificial in elligence and machine lea ning offe significan po en ial o op imize bi-
oleaching p ocesses and ex end hei use in hyd ome allu gy. The KEEN p ojec [12] ex-
emplifies he applica ion o AI in enhancing efficiency, educing de elopmen ime, and
lowe ing ope a ional cos s in enginee ing and p oduc ion. AI is op imizing chemical p o-
cesses, inc easing ene gy efficiency, and imp o ing me al ecycling, as seen in coppe
leaching using neu al ne wo ks [13,14]. AI is also applied in p edic i e main enance, op-
imizing ope a ions, and imp o ing sa e y in bioleaching and chemical plan s [15].
S anda d models like Linea Reg ession can be applied o bioleaching bu a e limi ed
in cap u ing i s nonlinea ela ionships, such as in e ac ions be ween mic obial g ow h,
pa icle size, and empe a u e. Polynomial eg ession offe s an in e media e solu ion, bal-
ancing complexi y, and pe o mance, hough high-deg ee polynomials isk o e fiing
noisy da a. Bo h app oaches a e compu a ionally efficien , wi h complexi y p ima ily de-
penden on da ase size and he numbe o a iables [16].
Bioleaching is a complex, nonlinea p ocess influenced by a iables like mic obial
g ow h, pH, empe a u e, pa icle size, and me al ion concen a ion. Decision ees effec-
i ely model hese in e ac ions by c ea ing in e p e able ules bu a e p one o o e fiing,
Appl. Sci. 2025, 15, x FOR PEER REVIEW 6 o 25
especially wi h deep ees. This can be add essed using p uning o ensemble me hods like
Random Fo es s o imp o e gene aliza ion and p e en o e fiing [17].
In bioleaching p ocesses, whe e da a complexi y and nonlinea i y a e common, an-
dom o es eg ession (RFR) and g adien boos ing machines (GBMs) offe subs an ial ad-
an ages o e adi ional machine lea ning models. Thei capaci y o p ocess complex,
high-dimensional da ase s and gene a e accu a e p edic ions makes hem in aluable o
enhancing bioleaching efficiency and esou ce managemen . RFR and GBM ou pe o m
simple models like linea eg ession o decision ees because hey can effec i ely accoun
o he nonlinea and in e dependen na u e o bioleaching ac o s. Fo ins ance, linea
models a e o en oo simplis ic, ailing o cap u e complex ela ionships be ween a iables
such as pa icle size, mic obial g ow h, and empe a u e, which a e key de e minan s o
me al eco e y a es [17].
AdaBoos effec i ely cap u es nonlinea ela ionships in bioleaching p ocesses and
can ou pe o m simple models like decision ees o suppo ec o machines in e ms o
obus ness and accu acy. Howe e , XGBoos may pe o m be e in handling missing o
high-dimensional da a [16]. Fo smalle -scale p oblems, simple models such as decision
ees o suppo ec o machines (SVR) can be used [18], wi h GBM and andom o es eg ession
o e ing s ong pe o mance o pa ame e op imiza ion and eco e y a e p edic ions [16].
A ificial neu al ne wo ks (ANN), including mul ilaye pe cep on (MLP) [19], a e e -
ec i e o modeling complex bioleaching p ocesses wi h in e ac ing a iables. While MLP
ou pe o ms linea and polynomial eg ession, me hods like Random Fo es and XGBoos
may offe simila pe o mance wi h lowe compu a ional cos . Resea ch highligh s he
op imiza ion o me al eco e y in bioleaching, especially in elec onic was e ecycling.
S udies using machine lea ning, such as ANN and SVM models, ha e achie ed o e 99%
me al eco e y efficiency in bio-Fen on p ocesses [20].
F om a compu a ional pe spec i e, he complexi y o machine lea ning algo i hms
a ies significan ly. G adien boos ing machines (GBMs) and andom o es eg ession
(RFR) s and ou due o hei ensemble na u e: GBM’s i e a i e p ocess can inc ease com-
pu a ional load, while RFR scales wi h he numbe o ees [21]. Howe e , op imized im-
plemen a ions like XGBoos imp o e efficiency, making hese models bo h compu a ion-
ally easible and effec i e o la ge da ase s.
Fu he s udies applied machine lea ning o u ban mining, op imizing coppe eco -
e y om PCBs, wi h g adien boos ing machines (GBMd) yielding he bes esul s o pa am-
e e combina ion op imiza ion [20,22]. Addi ionally, andom o es eg ession demon-
s a ed he highes accu acy in p edic ing eco e y a es in bioleaching p ocesses, high-
ligh ing he impo ance o ac o s like esou ce ype, pa icle size, and mic oo ganism ype
in me al eco e y p edic ions [20]. RFR and GBM excel a gene alizing ac oss unseen da a,
educing o e fiing, and ensu ing eliable p edic i e models o bioleaching unde di-
e se condi ions.
The s udy in [16] p esen s he e alua ion o 40 eg ession-based machine lea ning
(ML) algo i hms o bioleaching p ocess op imiza ion, iden i ying andom o es eg ession
as he mos effec i e, achie ing a p edic ion accu acy o 77%. The da ase , comp ising 871
samples ex ac ed om 206 jou nal a icles, included nine independen a iables—such
as mic oo ganism ype, empe a u e, and pa icle size—while he eco e y a e se ed as
he a ge a iable. Da a we e ga he ed using sea ch keywo ds like “bioleaching” and
“mic obial leaching” om da abases including Google Schola , Scopus, and Web o Sci-
ence. Building on his, he s udy in [16] applied andom o es eg ession and epo ed
p omising esul s.
In ou esea ch, we analyzed he key machine lea ning algo i hms applied o bi-
oleaching da a and achie ed a p edic ion accu acy o 84% using a g adien boos ing machine
(GBM), unde sco ing i s po en ial o op imizing bioleaching p ocesses. Key pa ame e s
Appl. Sci. 2025, 15, x FOR PEER REVIEW 7 o 25
like pH, coppe concen a ion, and inno a i e RGB-based me ics we e inco po a ed o
imp o e p edic i e pe o mance. These esul s highligh he subs an ial po en ial o ad-
anced machine lea ning echniques o enhance he efficiency and effec i eness o bi-
oleaching.
1.3. Digi al Twins in he Bioleaching Indus y
Digi al wins a e i ual eplicas o physical sys ems ha enable eal- ime simula ion,
moni o ing, and op imiza ion. They can be classified by hei unc ionali y: Moni o ing
digi al wins collec and isualize eal- ime da a o ack sys em s a us. Analy ical digi al
wins use machine lea ning o iden i y pae ns and p edic issues. P edic i e digi al wins
analyze bo h his o ical and eal- ime da a o o ecas u u e beha io s, while p esc ip i e
digi al wins sugges ac ions based on op imiza ion models. The mos ad anced con ol
digi al wins in e ac di ec ly wi h he physical sys em o adjus ope a ional pa ame e s
in eal ime. In he chemical indus y, digi al wins enhance p ocess efficiency, educe
cos s, and minimize en i onmen al impac by enabling eal- ime modeling and simula-
ion o indus ial p ocesses [14]. In [7], he po en ial o digi al wins in biomanu ac u ing
is explo ed, demons a ing how his echnology can imp o e p oduc ion moni o ing and
con ol while in eg a ing wi h AI o op imize p oduc ion and minimize was e. Digi al
wins also acili a e machine lea ning model aining, p edic ing sys em beha io and im-
p o ing esou ce managemen [22,23].
The syne gy o p ocess in ensifica ion and digi al wins offe s significan benefi s,
such as educing equipmen size, ene gy consump ion, and aw ma e ials, while inc eas-
ing p oduc i i y, sa e y, and sus ainabili y. This combina ion is ans o ming he p ocess
indus y, pa icula ly in he ene gy ansi ion. In bo h he mining and chemical indus ies,
digi al wins ha e p o en effec i e in op imizing p ocesses, including mine al beneficia-
ion and educing ene gy cos s in sepa a ion p ocesses [24,25].
By in eg a ing AI-d i en p ocess in ensifica ion wi h digi al wins, he s age is se o
ad anced me hodologies in bioleaching. The nex sec ion ou lines he me hodology used
in ou s udy o apply an IIoT and machine lea ning amewo k o op imize bioleaching.
2. Ma e ials and Me hods
The s a e-o - he-a e iew o machine lea ning applied o bioleaching, p esen ed in
Sec ion 1, iden ifies andom o es eg ession as one o he leading algo i hms o es ima -
ing eco e y a es using a iables such as ini ial pH and a omic numbe (e.g., [6]). Ou
app oach in oduces a no el app oach, u ilizing RGB da a om sma senso s o p edic
coppe and i on eco e y. RGB senso s, as ou lined in Sec ion 1, p o ide eal- ime moni-
o ing by acking eac ion p og ess wi hou di ec con ac wi h co osi e liquids, enhanc-
ing du abili y and cos -efficiency.
In ou esea ch, we selec ed nine machine lea ning algo i hms o analyze hei pe -
o mance in he bioleaching p ocess o he elec olysis uni . These algo i hms, anging
om s anda d models o hose iden ified as op-pe o ming in p io heo e ical s udies,
a e widely alida ed in he academic li e a u e. Thei p o en efficacy in bo h p ac ical
applica ions and heo e ical in es iga ions ensu es a obus ounda ion o ep oducible
and compa able esul s wi hin he scien ific communi y. In ou analysis, he g adien
boos ing machine su passes he andom o es algo i hm, showing g ea e p edic i e ac-
cu acy and efficiency. The compa ibili y o hese algo i hms wi h a ious da a olumes
and ypes is ano he eason o hei selec ion. Me hods like SVM and g adien boos ing
a e pa icula ly effec i e in con ex s wi h high dimensionali y o nume ous explana o y
a iables, whe eas ee-based and ensemble algo i hms a e mo e ole an o missing al-
ues o asymme ic dis ibu ions.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 8 o 25
2.1. Expe imen al Sys em Se up
The bioleaching p ocess was di ided in o ou key ope a ional s ages o main ain
consis ency and con ol o e expe imen al condi ions:
Ma e ial P epa a ion: Sh edded coppe cables we e chosen as he s a ing ma e ial
and p ocessed in a biologically ac i e solu ion wi hin a bio eac o . The solu ion was en-
iched wi h Fe (III) ions p oduced by i on-oxidizing bac e ia, es ablishing a con olled en-
i onmen o coppe leaching.
Bioleaching P ocess: Ope a ional pa ame e s in he bio eac o we e main ained
wi hin op imal anges o mic obial ac i i y, specifically a 31 °C and a pH ange o 1.6 o
1.8. These condi ions acili a ed efficien coppe ion dissolu ion om he solid ma ix in o
he solu ion, enhancing mic obial effec i eness.
Senso Ne wo k Moni o ing: An a ay o senso s con inuously measu ed c i ical pa-
ame e s such as pH, empe a u e, and me al ion concen a ion du ing he bioleaching
p ocess. RGB senso s acked colo changes in he solu ion, which we e linked o Fe (III)
and Cu (II) concen a ions, offe ing a non-in asi e me hod o moni o me al dissolu ion,
as explained in Sec ion 1.
Da a T ansmission and Logging: Senso da a we e ansmied in eal- ime o a cen-
al se e ia he MQTT p o ocol, ensu ing low-la ency and eliable da a ans e . This
se up enabled con inuous da a acquisi ion o e mul iple bioleaching cycles (72–96 h) and
in e al-based sample analyses, gene a ing a comp ehensi e da ase o model aining
ha cap u ed bo h s anda d and edge-case ope a ional condi ions.
2.2. Expe imen al Da a Acquisi ion
A comp ehensi e ange o da ase s was collec ed om eal- ime senso da a o de-
elop and e alua e nine machine lea ning models o p edic ing coppe eco e y. These
da ase s we e sys ema ically ga he ed o ensu e di e si y and ep esen a i eness o he
ope a ional condi ions. Emphasis was placed on da a quali y and consis ency o acili a e
accu a e model aining, c oss- alida ion, and pe o mance e alua ion.
The model de elopmen pipeline comp ised se e al key s eps:
Da a P ep ocessing: The da ase unde wen p ep ocessing o handle missing alues
using median impu a ion. Fea u e no maliza ion was pe o med using min–max scaling
o ensu e all ea u es had compa able anges, p e en ing bias due o scale diffe ences
ac oss a iables.
Fea u e Enginee ing: Addi ional ea u es we e de i ed o imp o e model pe o -
mance, including colo in ensi y a ios and in e ac ion e ms ex ac ed om he RGB sen-
so da a. These enginee ed ea u es aimed o cap u e mo e complex pae ns in he da a.
Model T aining: The da ase was spli in o aining (70%) and es ing (30%) subse s.
A a ie y o eg ession models we e e alua ed, including linea eg ession, decision ee
eg ession, andom o es eg ession, polynomial eg ession, mul ilaye pe cep on (MLP),
suppo ec o eg ession (SVR), g adien boos ing eg ession, XGBoos , and AdaBoos .
Model E alua ion: Model pe o mance was assessed using se e al me ics, includ-
ing oo mean squa ed e o (RMSE), mean absolu e e o (MAE), R2 (coefficien o de e -
mina ion), and adjus ed R2. These me ics p o ided insigh s in o he accu acy and obus -
ness o each model.
K- old C oss-Valida ion: To ensu e eliable pe o mance and mi iga e o e fiing, K-
old c oss- alida ion was applied. This echnique, which pa i ions he da ase in o k sub-
se s, helps alida e he model and es ima e i s pe o mance. In each i e a ion, one subse
is used as alida ion da a while he emaining subse s a e used o aining. The mean pe -
o mance ac oss all i e a ions p o ides an assessmen o he model’s accu acy. Common k-
alues a e 5, 10, o 20. Fo ou s udy, he aining- o- es ing se a io was 70% o 30%.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 9 o 25
To c ea e each da ase , 25 bioleaching expe imen s we e conduc ed o es ablish a co -
ela ion be ween he colo s o Fe (II)/Fe (III) and hei concen a ions. Each expe imen
u ilized a p o o ype plan column filled wi h 16 g o sh edded cables and 10 g o a 3D-
p in ed fille , combined wi h 1.3 L o Fe (III) solu ion p oduced by mic oo ganisms.
Th oughou he expe imen s, mul iple samples we e collec ed o moni o Fe (II)/Fe (III)
le els and o co ela e ion concen a ions wi h he solu ion colo de ec ed by he plan ’s
sma colo senso and an onsi e pH senso . The solu ion samples we e hen analyzed o
Fe (III) and Cu (II) concen a ions, ollowing he me hodology ou lined in [10]. Table 1
p esen s he measu ed concen a ions o Fe (III) and Cu (II) ions ob ained du ing he elec-
olysis s age o he expe imen .
Table 1. Example o measu ed concen a ions o Fe (III) and Cu (II) ions om an expe imen .
Reac ion Time (K ), min
[Fe(III)], mg/L [Cu(II)], mg/L
0 6,134.6 0.0
5 5,632.2 ~0.0
10 5,144.2 310.66
15 4,915.9 514.71
25 4,064.9 1,063.42
35 3,557.7 1,405.33
45 3,564.9 1,622.24
75 2,247.6 2,348.35
105 1,882.2 2,577.21
195 1,521.6 3,081.80
225 1,185.1 3,081.8
342 1,341.3 3,250.00
A e da a collec ion, he da a we e o maed in o a s uc u e compa ible wi h ma-
chine lea ning algo i hms. The p ep ocessing s eps included cleaning he da a, add essing
missing alues (less han 5%) using median impu a ion, and no malizing he a iables o
a 0–1 scale h ough min–max scaling.
Da a isualiza ion and co ela ion analyses e ealed key ela ionships be ween RGB
alues, pH, and coppe concen a ion (Cu) du ing he bioleaching p ocess. The co ela ion
ma ix in Figu e 3 shows a nega i e co ela ion be ween he ed componen (C1R1) and
Cu concen a ion (−0.44), indica ing ha highe C1R1 alues co espond o lowe Cu con-
cen a ions. Con e sely, mode a e posi i e co ela ions we e obse ed be ween he g een
componen (C1G1) and pH (pH1) wi h Cu concen a ion (0.44 and 0.36, espec i ely), sug-
ges ing ha inc eases in hese pa ame e s a e associa ed wi h highe Cu le els.
Figu e 5 p esen s he scae plo ma ix, which isually confi ms he co ela ions
shown in Figu e 4. I illus a es ha Cu concen a ion ypically dec eases wi h highe
C1R1 alues, while posi i e ends a e obse ed be ween C1G1, pH1, and Cu concen a-
ion, consis en wi h he nume ical co ela ions. Figu e 6 p o ides scae plo s o each
p edic o agains Cu concen a ion, u he highligh ing he influence o each a iable on
Cu eco e y, in alignmen wi h he ends iden ified in he ma ix analysis.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 16 o 25
Figu e 15. P edic ed s. ac ual coppe eco e y a es using mul ilaye pe cep on.
Figu e 16. P edic ed s. ac ual coppe eco e y a es using polynomial eg ession.
Figu e 17. P edic ed s. ac ual coppe eco e y a es using andom o es eg ession.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 17 o 25
Figu e 18. P edic ed s. ac ual coppe eco e y a es using suppo ec o machine.
Figu e 19. P edic ed s. ac ual coppe eco e y a es using XGBoos .
The compu a ional complexi y o he algo i hms used in ou esea ch (see Table 2)
o bioleaching p ocesses is influenced by ac o s such as da ase size, model a chi ec u e,
and he efficiency o suppo ing lib a ies. F amewo ks like enso Flow-2.14.0, PyTo ch
2.x, and Sciki lea n 1.6.0 enhance pe o mance h ough GPU/TPU accele a ion, pa allel
p ocessing, and op imized compu a ion, allowing efficien handling o la ge da ase s.
Simila ly, XGBoos 2.1.3 employs ad anced g adien boos ing echniques, including da a
p uning, ea u e selec ion, and dis ibu ed p ocessing, o minimize compu a ional o e -
head. Le e aging hese lib a ies in ou implemen a ions [26] ensu es scalabili y, ep o-
ducibili y, and flexibili y o expe imen ing wi h diffe en configu a ions. These capabili-
ies make he algo i hms pa icula ly well-sui ed o bioleaching, whe e high speed and
scalabili y a e essen ial.
2.4. The Web-Based Digi al Twin Sys em
Ou p e ious esea ch on plan moni o ing and con ol so wa e a chi ec u e [8] has
been enhanced by in eg a ing machine lea ning algo i hms and implemen ing a web-
based digi al win sys em. The en i e in as uc u e has been adap ed om he ini ial p o-
o ype o a la ge-scale plan se up.
The new sys em, accessible a [8,26], in eg a es HTTPS enc yp ion and ad anced use
au hen ica ion p o ocols o sa egua d sensi i e da a and ensu e ha access is es ic ed o
au ho ized pe sonnel only. Access con ol is managed h ough ole-based access con ol
(RBAC), which assigns pe missions based on use oles, ensu ing ha only au ho ized
indi iduals can modi y c i ical sys em pa ame e s o access sensi i e da a. Addi ionally,
cookie-based au hen ica ion enhances secu i y o he web in e ace, wi h enc yp ed
Appl. Sci. 2025, 15, x FOR PEER REVIEW 18 o 25
HTTP cookies p o ec ing he confiden iali y and in eg i y o session da a h oughou use
in e ac ions.
Fo imp o ed sys em pe o mance and secu i y, he se e -side a chi ec u e uses
Gunico n in combina ion wi h Nginx, whe e Nginx ac s as a e e se p oxy and load bal-
ance . This se up no only adds an ex a laye o secu i y bu also op imizes sys em effi-
ciency. Fu he mo e, SSL/TLS p o ocols enc yp all da a communica ions be ween he
Raspbe y Pi and ex e nal de ices, sa egua ding agains unau ho ized in e cep ion and
enhancing o e all secu i y.
The Flask-based web in e ace includes use - iendly, eal- ime dashboa ds wi h in-
e ac i e cha s and his o ical da a, enabling end analysis o key pa ame e s. This unc-
ionali y suppo s in o med decision making and dynamic adjus men s, p omo ing p o-
ac i e p ocess managemen o main ain op imal bioleaching condi ions. Figu es 20–24 a e
sc eensho s aken om he Biome allum web-based digi al win, accessible ia [7]. Figu e
20 illus a es he Dashboa d Page, showing senso eadings and ac ua o s a uses o he
bioleaching plan desc ibed in Sec ion 1 and depic ed in Figu e 1.
Addi ionally, he web applica ion ea u es a cus omizable da a ex ac ion in e ace,
whe e ope a o s can speci y ime anges, sampling in e als, and choose senso s o ac u-
a o s om a ious p ocess s ages. The Ge Da a page allows o immedia e da a isuali-
za ion h ough gene a ed g aphs o CSV downloads o u he analysis. CSV files a e
o maed wi h imes amps, senso /ac ua o names, and eco ded alues, acili a ing a -
ge ed analysis and in o med decision making o enhance ope a ional efficiency. Figu e 21
shows g aphical ep esen a ions o senso eadings ele an o he biological s age o he
p ocess, wi h adjus able ime pe iod seings. Simila ly, Figu e 22 illus a es senso ead-
ings o he leaching s age, also allowing o ime pe iod adjus men s.
The sys em upg ade in oduces p edic i e models o o ecas beha io based on his-
o ical da a. While s ill in he ea ly s ages, his unc ionali y pa es he way o in eg a ing
ad anced machine lea ning algo i hms o op imize me al eco e y.
Cu en ly, he sys em uses he me hods de ailed in Sec ion 2 (AdaBoos , decision ee,
g adien boos ing machine, linea eg ession, mul ilaye pe cep on, polynomial eg es-
sion, andom o es , suppo ec o machine, and XGBoos ) o analyze his o ical da a and
p edic coppe eco e y efficiency unde a ious condi ions. This p edic i e capabili y
enables echnicians o make p oac i e adjus men s, p e en ing c i ical h esholds om
being su passed. By inco po a ing hese models, he sys em can op imize ope a ional pa-
ame e s, imp o ing me al eco e y efficiency and he sus ainabili y o he bioleaching
p ocess. Fo ins ance, he sys em can iden i y pae ns in coppe dissolu ion a es, ecom-
mending op imal eagen addi ion imes o adjus ing leaching solu ion flow a es.
F om he p edic ion page in Figu e 23, plan echnicians can inpu a ious pa ame-
e s in o he machine lea ning models o gene a e coppe eco e y p edic ions.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 19 o 25
Figu e 20. O e iew o he dashboa d page, displaying senso eadings and ac ua o s a uses o
he en i e plan .
Figu e 21. Senso eadings ele an o he biological s age page.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 20 o 25
Figu e 22. Senso eadings ele an o he leaching s age page.
Figu e 23. P edic ions on coppe eco e y om machine lea ning models.
Figu e 24. Colo on coppe eco e y om machine lea ning models.
3. Resul s
Appl. Sci. 2025, 15, x FOR PEER REVIEW 21 o 25
The newly de eloped web-based digi al win in eg a es senso da a in o a unified
dashboa d, accessible on bo h desk op and mobile de ices. The in ui i e in e ace enables
echnicians o moni o key pa ame e s, such as Cu (II) and Fe (III) concen a ions, pH le -
els, and empe a u e, in eal ime. In e ac i e cha s and his o ical da a isualiza ions a-
cili a e end analysis and anomaly de ec ion, enabling p omp adjus men s o main ain
op imal condi ions.
Remo e accessibili y enhances ope a ional flexibili y, enabling echnicians o moni o
and con ol he sys em offsi e. This is especially aluable o bioleaching p ocesses ha
equi e cons an o e sigh , as i educes he need o onsi e p esence, lowe s labo cos s,
and ensu es con inuous ope a ions.
Da a expo unc ionali y suppo s p ocess op imiza ion by enabling he download
o his o ical da ase s o u he analysis. This acili a es ad anced analy ics and machine
lea ning, con ibu ing o con inuous imp o emen s and efinemen s in sys em pe o -
mance. Analyzing his o ical da a iden ifies ends ha in o m ope a ional adjus men s.
To ensu e secu i y and da a in eg i y, he sys em employs mode n p o ocols, includ-
ing enc yp ed communica ion and ole-based access con ol. This ensu es ha only au-
ho ized pe sonnel can access o modi y plan da a. Addi ionally, he sys em logs all use
ac ions, c ea ing an audi ail ha p omo es anspa ency and accoun abili y.
Ex ensi e es ing was conduc ed du ing he de elopmen o he so wa e a chi ec-
u e o ensu e unc ionali y, compa ibili y, and e iciency. The sys em unde wen igo ous
compa ibili y es ing ac oss a ious ope a ing sys ems, including Windows 10, And oid 12,
and Ubun u 24.04 LTS, and on di e en web b owse s such as Ch ome, Fi e ox, and Edge.
These es s con i med he web applica ion’s seamless ope a ion ac oss di e se pla o ms.
To op imize pe o mance, he sys em in eg a es a ge en -based concu ency man-
agemen configu a ion wi hin Flask, enabling efficien handling o mul iple simul aneous
connec ions, which is c ucial o eal- ime moni o ing in indus ial en i onmen s. Addi-
ionally, a caching mechanism was implemen ed o educe da a loading imes, enhancing
page speed and p o iding a smoo he use expe ience du ing emo e moni o ing.
P elimina y es ing demons a ed he sys em’s abili y o o ecas p ocess beha io s,
de ec anomalies, and educe manual in e en ion, ep esen ing significan ad ancemen s
in au oma ion and ope a ional efficiency. By analyzing eal- ime senso da a, including
RGB colo ime ic eadings and pH alues, he sys em accu a ely es ima es coppe eco -
e y a es, enhancing esou ce efficiency, lowe ing cos s, and p omo ing sus ainabili y by
minimizing non-p oduc i e un ime, p e en ing solu ion o e exposu e, and op imizing
chemical and ene gy use. These capabili ies, combined wi h subs an ial imp o emen s in
accessibili y, usabili y, and scalabili y, posi ion he a chi ec u e o deploymen in la ge -
scale ope a ions.
The in eg a ion o a web-based in e ace, p edic i e modeling, and eal- ime da a ac-
cess es ablishes a obus ounda ion o a mo e au onomous, efficien , and da a-d i en
bioleaching plan . This scalable pla o m suppo s u u e enhancemen s, such as addi-
ional p edic i e models o ad anced senso s, making i a powe ul ool o op imizing
bioleaching p ocesses and ad ancing sus ainable me al eco e y.
4. Discussion
This s udy p esen s a no el app oach o me al eco e y om e-was e using bio-
based echnology. The p ocess offe s significan en i onmen al ad an ages o e con en-
ional me hods, including educed ene gy consump ion, lowe emissions compa ed o
py ome allu gy, and minimized eagen use and liquid was e p oduc ion compa ed o
adi ional hyd ome allu gy [27]. These benefi s highligh he po en ial o his app oach
o con ibu e o a ci cula economy and sus ainable e-was e managemen .
Appl. Sci. 2025, 15, x FOR PEER REVIEW 22 o 25
While he cu en s age o de elopmen emains a he labo a o y and semi-indus ial
scale, he ocus o his s udy is no on e alua ing i s economic easibili y bu a he on
demons a ing i s echnical iabili y and en i onmen al benefi s. Economic easibili y will
be add essed in u u e s udies as he p ocess ad ances owa d ull-scale indus ial imple-
men a ion.
The implemen a ion challenges and complexi ies encoun e ed du ing his s udy p o-
ide aluable insigh s in o he echnological hu dles ha mus be add essed du ing scale-
up. Howe e , he p ima y ocus o his wo k emains on he po en ial applica ions and
ad an ages o he p oposed echnology. By add essing hese key aspec s, his app oach
lays he g oundwo k o a sus ainable and economically iable al e na i e o con en ional
ecycling p ocesses.
The challenges in ensu ing eliable senso da a, including senso d i , en i onmen al
in e e ence, and calib a ion inconsis encies, ha e been add essed. A dedica ed sys em
le e aging RGB digi al senso s has been de eloped o p o ide con inuous, eal- ime moni-
o ing o he bioleaching p ocess, minimizing he impac o hese ac o s. The non-in asi e
design o he RGB senso sys em elimina es di ec con ac wi h co osi e solu ions, he eby
ex ending senso li espan and main aining consis en measu emen accu acy o e ime.
Addi ionally, he need o da a alida ion o ensu e consis ency and eliabili y has
been ecognized. To ensu e high-quali y da a, he RGB senso s ha e been alida ed
agains es ablished me hods, such as UV-VIS spec oscopy and a omic abso p ion, con-
fi ming hei eliabili y in bioleaching en i onmen s. In eg a ing hese ad anced senso s
wi h machine lea ning algo i hms is expec ed o enhance bo h he accu acy and efficiency
o he bioleaching p ocess, effec i ely add essing challenges associa ed wi h senso cali-
b a ion and da a a iabili y.
Scalabili y and sys em load managemen offe oppo uni ies o op imiza ion. The
a chi ec u e effec i ely suppo s emo e moni o ing and eal- ime da a isualiza ion,
hough pe o mance may be impac ed unde high da a olumes and simul aneous use
access. This p o ides a aluable insigh o u u e imp o emen s, as he use o ligh weigh
componen s like SQLi e and Flask can be efined o eplaced o bee sui la ge-scale in-
dus ial applica ions.
The di e se ange o algo i hms es ed, om in e p e able linea models o ad anced
ensemble me hods and neu al ne wo ks, enabled a comp ehensi e e alua ion o p edic-
i e pe o mance in he bioleaching con ex . This s a egic app oach acili a ed he iden i-
fica ion o op imal me hods o imp o ing coppe eco e y p ocesses while e ealing he
s eng hs and limi a ions o each model. The g adien boos ing machine model pe o med
well o e all, bu s uggled o handle ex eme alues ou side he no mal ope a ing ange,
sugges ing a need o imp o ed ou lie de ec ion and model efinemen .
Las ly, he sys em s ill elies on manual in e en ion o p edic i e con ols. While he
models p o ide ecommenda ions o p ocess adjus men s, ope a o s mus s ill make deci-
sions ega ding hal ing ope a ions o adjus ing pa ame e s. This manual dependency can
cause delays, especially i ope a o s a e no immedia ely a ailable, highligh ing he need o
a ully au onomous con ol sys em o enhance esponsi eness and ope a ional e iciency.
These challenges unde sco e a eas o imp o emen , pa icula ly in da a quali y,
scalabili y, au oma ion, and secu i y, which will guide u u e de elopmen effo s. Fu u e
de elopmen s will ocus on he ollowing aspec s:
1. Enhance model accu acy: In eg a e addi ional da a sou ces, including mic obial pop-
ula ion densi y and solu ion flow a e, o imp o e p edic i e accu acy.
2. Boos p edic ion p ecision and e iciency: Apply ime se ies models (TDSM), such as
ecu en neu al ne wo ks (RNNs) and long sho - e m memo y (LSTM) ne wo ks,
which a e well-sui ed o cap u ing empo al pa e ns in e ol ing biological p ocesses.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 23 o 25
3. C ea e a 3D digi al win o he physical sys em using game engines like Uni y, ena-
bling ope a o s and esea che s o:
(a) Visualize p ocesses in eal ime: Access li e da a in a i ual en i onmen o
quick insigh s.
(b) Simula e scena ios: Expe imen wi h pa ame e changes o simula e aul s
sa ely, wi hou affec ing he eal plan .
(c) Enhance in e ac ion: Use in e ac i e con ols o explo e and manipula e p o-
cesses, boos ing use engagemen and comp ehension.
4. Ad anced model op imiza ion: Valida e ad anced machine lea ning echniques, in-
cluding deep lea ning models, o accu a ely cap u e complex in e ac ions ac oss all
s ages o he bioleaching p ocess. In eg a e hese esul s in o he digi al win o acil-
i a e eal- ime analy ics and imp o e in o med decision making.
5. Conclusions
This s udy p esen s a bio-based echnology o me al eco e y om e-was e ha sig-
nifican ly educes ene gy consump ion, emissions, and eagen and liquid was e com-
pa ed o adi ional me hods. By combining machine lea ning wi h he Indus ial In e ne
o Things (IIoT), he esea ch enhances bioleaching e iciency by means o indus ial p ocess
con ol. A key inno a ion is he de elopmen o a web-based supe ision sys em imple-
men ed as a digi al win, enabling eal- ime moni o ing and con ol o he bioleaching plan .
The sys em uses colo da a and senso inpu s o c ea e a i ual model o he plan ,
imp o ing decision making and p ocess op imiza ion. I o e comes he limi a ions o p e-
ious solu ions de eloped by ou esea ch eam based on And oid o hi d-pa y apps,
making plan managemen mo e e icien and collabo a i e. The s udy also highligh s he
use o non-in asi e RGB senso s in he elec olysis uni , which a oid di ec con ac wi h
co osi e solu ions. These senso s o e ad an ages such as a longe li espan, p ecise in si u
measu emen s, and he elimina ion o sample ex ac ion and analysis ime consump ion.
A p edic i e model buil wi h g adien boos ing achie ed 84% accu acy, showcasing
he po en ial o machine lea ning o imp o e bioleaching p ocesses. The in eg a ion o he
digi al win no only p o es he echnical iabili y o his app oach bu also p o ides no-
able en i onmen al benefi s. The design adhe es o human-cen e ed design p inciples,
ensu ing ha he echnology is in ui i e, use - iendly, and accessible o ope a o s,
he eby p omo ing collabo a ion and imp o ing he use expe ience. Addi ionally, he ap-
p oach suppo s Indus y 5.0 by os e ing collabo a ion be ween humans and machines, en-
hancing decision-making and p ocess con ol while mee ing sus ainabili y goals.
While economic easibili y and scalabili y will be explo ed in u u e wo k, he cu en
findings lay a solid ounda ion o u he de elopmen . This esea ch is he fi s o apply
con inuous p edic i e echniques using RGB senso s and digi al win echnology in a
semi-indus ial e-was e plan , ma king a c ucial s ep in op imizing and scaling bioleach-
ing. Ul ima ely, he in eg a ion o ad anced senso s, machine lea ning, and digi al win
echnology offe s he po en ial o enhance he efficiency, sus ainabili y, and scalabili y o
bioleaching, wi h applica ions in bo h semi-indus ial and ull indus ial seings.
Au ho Con ibu ions: Concep ualiza ion, M.I.T.-P. and J.V.P.; me hodology, M.I.T.-P., A.C., T.E.
and J.V.P.; so wa e, J.V.P.; alida ion, M.I.T.-P. and J.V.P.; o mal analysis, M.I.T.-P. and J.V.P.; in-
es iga ion, M.I.T.-P. and J.V.P.; esou ces, M.I.T.-P., A.C. and J.V.P.; da a cu a ion, M.I.T.-P., A.C.
and J.V.P.; w i ing—o iginal d a p epa a ion, M.I.T.-P. and J.V.P.; w i ing— e iew and edi ing,
M.I.T.-P., A.C., T.E., J.V.P. and A.D.D.; isualiza ion, M.I.T.-P. and J.V.P.; supe ision, M.I.T.-P.; A.C.
and A.D.D.; p ojec adminis a ion, A.D.D.; unding acquisi ion, A.D.D. All au ho s ha e ead and
ag eed o he published e sion o he manusc ip .
Appl. Sci. 2025, 15, x FOR PEER REVIEW 24 o 25
Funding: This esea ch was unded by he Spanish Minis e io de Inno ación unde G an PID2020-
117520RA-I00 o MCIN/AEI/10.13039/501100011033.
Ins i u ional Re iew Boa d S a emen : No applicable.
In o med Consen S a emen : No applicable.
Da a A ailabili y S a emen : The da a p esen ed in his s udy a e openly a ailable in Biome allum
Gi Lab eposi o y a [26].
Acknowledgmen s: Au ho s acknowledge he Spanish Go e nmen , h ough p ojec PID2020-
117520RA-I00, o he financial suppo p o ided o conduc his esea ch. The au ho s also
acknowledge Joan Bello o his con ibu ion and his pa icipa ion in he p ojec .
Conflic s o In e es : The au ho s decla e no conflic s o in e es .
Re e ences
1. Fo i, V.; Balde, C.P.; Kueh , R.; Bel, G. The Global E-Was e Moni o 2020: Quan i ies, Flows and he Ci cula Economy Po en ial;
Uni ed Na ions Uni e si y (UNU)/UNITAR, SCYCLE P og amme: Bonn, Ge many; ITU: Gene a, Swie land; ISWA: Roe -
dam, The Ne he lands, 2020.
2. Do ado, A.D.; Ma inez, A.; Escobe , A.; Escobe , T.; Sole, M.; Lao, C.; Bascomp e, M.; SanMiquel, L.; Ta és, M.I; Benzal, E.; e al.
De elopmen o a sma au oma ed bio-based p ocess o he eco e y o aluable me als om end-o -li e phones (BIOMETAL).
R&D P ojec 2021–2024. UPC. Uni e si a Poli ècnica de Ca alunya. A ailable online: hps://biome allum.upc.edu/es (accessed
on 24 No embe 2024).
3. Rawlings, D.E. Hea y me al mining using mic obes. Annu. Re . Mic obiol. 2002, 56, 65–91. hps://doi.o g/10.1146/annu e .mi-
c o.56.012302.161052.
4. Benzal, E.; Cano, A.; Solé, M.; Gamisans, X.; Do ado, A.D. Coppe Reco e y om PCBs by Acidi hiobacillus e ooxidans: Toxici y
o Bioleached Me als on Biological Ac i i y. Was e Biomass Valo iza ion 2020, 11, 5483–5492. hps://doi.o g/10.1007/s12649-020-
01036-y.
5. Rohwe de , T.; Geh ke, T.; Kinzle , K.; Sand, W. Bioleaching e iew pa A: P og ess in bioleaching: Fundamen als and mecha-
nisms o bac e ial me al sul ide oxida ion. Appl. Mic obiol. Bio echnol. 2003, 63, 239–248. h ps://doi.o g/10.1007/s00253-003-1448-7.
6. Moka ian, P.; Bakhshayeshi, I.; Taghikhah, F.; Bo oumand, Y.; E ani, E.; Razmjou, A. The ad anced design o bioleaching p o-
cess o me al eco e y: A machine lea ning app oach. Sep. Pu i . Technol. 2022, 291, 120919. hps://doi.o g/10.1016/j.sep-
pu .2022.120919.
7. Huynh, T.A.; Zonde an, E. P ocess in ensifica ion and digi al win—The po en ial o he ene gy ansi ion in p ocess indus-
ies. Phys. Sci. Re . 2023, 8, 4859–4877. hps://doi.o g/10.1515/ps -2022-0058.
8. Biome allum, Uni e si a Poli ècnica de Ca alunya. In e nal Si e, Fo iClien VPN. A ailable online: hps://biome al-
lum2.epsem.upc.edu/ egis e (accessed on 4 Oc obe 2024).
9. Ta és-Pue as, M.I.; B osa, L.; Come ma, A.; Rossell, J.M.; Do ado, A.D. A chi ec ing an open-sou ce IIoT amewo k o eal-
ime con ol and moni o ing in he bioleaching indus y. Appl. Sci. 2024, 14, 10350. hps://doi.o g/10.3390/app14010350.
10. Mo ell, J.; Escobe , A.; Do ado, A.D.; Escobe , T. Design o a RGB-A duino de ice o moni o ing coppe eco e y om PCBs.
P ocesses 2023, 11, 1319. hps://doi.o g/10.3390/p 11051319.
11. BH1749NUC Digi al 16-bi Se ial Ou pu Type Colo Senso IC Technical Specifica ions. A ailable online: hps://www. a -
nell.com/da ashee s/2630939.pd (accessed on 17 Decembe 2024).
12. KEEN—A ificial In elligence Pla o m o he P ocess Indus y. KEEN P ojec . A ailable online: hps://keen-pla-
o m.de/keen/en/ (accessed on 2 Augus 2024).
13. O daz-Oli e , M.; Jiménez-Muñoz, E.; Gu ié ez-Mo eno, E.; Gu ié ez-Segu a, E.; Rangel-Mendez, J.R.; Ga cía-González, A.;
Iqbal, H.M.N. Applica ion o a ificial neu al ne wo ks o eco e y o Cu om elec onic was e by dynamic acid leaching: A
sus ainable app oach. Was e Biomass Valo iza ion 2024, 15, 7057–7076. hps://doi.o g/10.1007/s12649-024-02644-8.
14. Ga galo, C.; He as, S.; Jones, M.; Udugama, I.A.; Mansou i, S.; K ühne, U.; Ge naey, K. Towa ds he de elopmen o digi al
wins o he bio-manu ac u ing indus y. Ad . Biochem. Eng./Bio echnol. 2020, 176, 1–34. hps://doi.o g/10.1007/10_2020_142.
15. Nase i, T.; Beiki, V.; Mousa i, S.M.; Fa naud, S. A Comp ehensi e Re iew o Bioleaching Op imiza ion by S a is ical Ap-
p oaches: Recycling Mechanisms, Fac o s Affec ing, Challenges, and Sus ainabili y. RSC Ad . 2023, 13, 23570–23589.
hps://doi.o g/10.1039/d3 a03498d.
Appl. Sci. 2025, 15, x FOR PEER REVIEW 25 o 25
16. Moka ian, P.; Bakhshayeshi, I.; Taghikhah, F.; Bo oumand, Y.; E ani, E.; Razmjou, A. The Ad anced Design o Bioleaching
P ocess o Me al Reco e y: A Machine Lea ning App oach. Sep. Pu i . Technol. 2022, 291, 120919. hps://doi.o g/10.1016/j.sep-
pu .2022.120919.
17. Zhang, Z.; Zhang, X.; Zhang, D.; Zhang, X.; Qiu, F.; Li, W.; Liu, Z.; Shu, J.; Tang, C. Applica ion o Machine Lea ning in a Mine al
Leaching P ocess—Taking Py olusi e Leaching as an Example. ACS Omega 2022, 7, 48130–48138.
hps://doi.o g/10.1021/acsomega.2c06129.
18. T i edi, A.; Hai , S. Fungal bioleaching o me als om WPCBs o mobile phones employing mixed Aspe gillus spp.: Op imiza-
ion and p edic i e modelling by RSM and AI models. J. En i on. Manag. 2024, 349, 119565. hps://doi.o g/10.1016/j.jen-
man.2023.119565.
19. Leona d, K.C.; Hasan, F.; Sneddon, H.F.; You, F. Can a ificial in elligence and machine lea ning be used o accele a e sus ain-
able chemis y and enginee ing? ACS Sus ain. Chem. Eng. 2021, 9, 6126–6129. hps://doi.o g/10.1021/acssuschemeng.1c02741.
20. Oli ei a, L.M.C.; Dias, R.; Rebello, C.M.; Ma ins, M.A.F.; Rod igues, A.E.; Ribei o, A.M.; Noguei a, I.B.R. A ificial in elligence
and cybe -physical sys ems: A e iew and pe spec i es o he u u e in he chemical indus y. AI 2021, 2, 429–443.
hps://doi.o g/10.3390/ai2030027.
21. Dawa e, S.; Chandel, S.; Rai, B. A machine lea ning amewo k o u ban mining: A case s udy on eco e y o coppe om
p in ed ci cui boa ds. Mine . Eng. 2022, 180, 107479. hps://doi.o g/10.1016/j.mineng.2022.107479.
22. Golo ianko, M.; Te ziyan, V.; B any skyi, V.; Malyk, D. Indus y 4.0 s. Indus y 5.0: Co-exis ence, ansi ion, o a hyb id. P o-
cedia Compu . Sci. 2023, 217, 102–113. hps://doi.o g/10.1016/j.p ocs.2022.12.206.
23. T i edi, A.; Hai , S. Me al bioleaching om p in ed ci cui boa ds by bio-Fen on p ocess: Op imiza ion and p edic ion by e-
sponse su ace me hodology and a ificial in elligence models. J. En i on. Manag. 2023, 326 P B, 116797.
24. Yusupbeko , N.; Adilo , F.; I anyan, A.; Abdu asulo , F. Applica ion o digi al win echnologies in mining indus ial b anch.
E3S Web Con . 2023, 417, 05016. hps://doi.o g/10.1051/e3scon /202341705016.
25. Schweid mann, A.M.; Esche, E.; Fische , A.; Klo , M.; Repke, J.-U.; Sage , S.; Mi sos, A. Machine lea ning in chemical enginee -
ing: A pe spec i e. Chem. Ing. Tech. 2021, 93, 2029–2039. hps://doi.o g/10.1002/ci e.202100083.
26. Biome allum Gi Lab Reposi o y. A ailable online hps://gi em.epsem.upc.edu/biome allum/biome allum (accessed on 4 Oc o-
be 2024).
27. Benzal, E.; Solé, M.; Lao, C.; Gamisans, X.; Do ado, A.D. Elemen al Coppe Reco e y om e-Was es Media ed wi h a Two-S ep
Bioleaching P ocess. Was e Biomass Valo iza ion 2020, 11, 5457–5465. hps://doi.o g/10.1007/s12649-020-01040-2.
Disclaime /Publishe ’s No e: The s a emen s, opinions and da a con ained in all publica ions a e solely hose o he indi idual au-
ho (s) and con ibu o (s) and no o MDPI and/o he edi o (s). MDPI and/o he edi o (s) disclaim esponsibili y o any inju y o
people o p ope y esul ing om any ideas, me hods, ins uc ions o p oduc s e e ed o in he con en .
