Design and Development of An Electrostatic Separator For Waste Segregation Lai Koon Chun Doctor of Philosophy (Engineering) Facu

29.11.
2019 DESIGN AND DEVELOPMENT OF AN ELECTROSTATIC SEPARATOR FOR WASTE SEGREGATION LAI KOON CHUN DOCTO…
http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf dosyasının HTML sürümüdür. Googlet aradığı belgeler n HTML

sürümler n otomat k olarak oluşturur.
İpucu: Bu sayfada arama ter m n z kolayca bulab lmek ç n Ctrl+F veya ⌘ -F (Mac) tuşlarına basın ve bulma çubuğunu kullanın.
Page 1
DESIGN AND DEVELOPMENT OF AN ELECTROSTATIC

SEPARATOR FOR WASTE SEGREGATION
LAI KOON CHUN
DOCTOR OF PHILOSOPHY (ENGINEERING)
webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 1/147

29.11.2019 DESIGN AND DEVELOPMENT OF AN ELECTROSTATIC SEPARATOR FOR WASTE SEGREGATION LAI KOON CHUN DOCTO…
FACULTY OF ENGINEERING AND GREEN

TECHNOLOGY
UNIVERSITI TUNKU ABDUL RAHMAN
APRIL 2015
Page 2

By
LAI KOON CHUN

A thes s subm tted to the Department of Electron c Eng neer ng,

Faculty of Eng neer ng and Green Technology,
Un vers t Tunku Abdul Rahman,
n part al fulf llment of the requ rements for the degree of
Doctor of Ph losophy (Eng neer ng)
Apr l 2015
Page 3
ABSTRACT

La Koon Chun
Household sol d waste s commonly d sposed of n landf lls, generat ng
hazardous by-products such as leachate and methane gas from ts organ c
contents. Thus, organ cs n landf lls wh ch cons st ma nly of food matter
should be reduced. In th s study, an electrostat c separator was des gned and
developed to segregate the organ c food waste from waste m xtures. Pr nc ples
of the electrostat c separat on were d scussed by referr ng to force models.
Bes des, the separat on process of food waste was character sed w th respect to
both system and no se factors. The hard-to-control no se factors, .e. s ze and
mo sture level of the feed ng part cles were then dent f ed by employ ng a

robust des gn based on Taguch 's method. Results revealed that the no se
factors are key factors that should be properly selected so as to avo d
undes rable sens t v ty and var at ons of the separat on process. The evaluat on
results conf rmed that the system factors, .e. rotat on speed, electr cal
potent al and electrodes nterval are the most s gn f cant factors for the
separat on process. A stat st cal analys s w th central compos te des gn was
conducted to analyse and model the performance of separat on. Ind v dual and
nteract ve effects of ndependent factors on separat on performance were
assessed n terms of the recovery eff c ency and pur ty of food waste matter.
Page 4
For the system cons dered, the opt mal operat onal cond t ons were deduced to
be 60 rpm rotat onal speed, 30 kV appl ed voltage and 54 mm electrodes
nterval on part cles w th 4.0 mm s ze and 20 % water content. Under these
cond t ons, food waste separat on eff c ency of 84.20% and pur ty of 93.00%
were exper mentally ach eved. Separat on eff c ency and pur ty of non-food
waste were respect vely 88.70% and 98.50% under the same operat onal
cond t on. These results f tted well w th the pred cted model. Results n th s
study concluded that the electrostat c separat on could be an effect ve pre-
treatment alternat ve n deal ng w th leachate and methane problems caused by
landf lled organ c wastes.

Page 5
ACKNOWLEDGEMENTS
Acknowledgements are due to my superv sors, Assoc ate Professor Dr. L m
Soo K ng and Dr. Teh Peh Ch ong for the r attent on and support throughout
th s study. I am part cularly grateful to Assoc. Prof. Dr. L m, w thout whose
adv ce th s study would be harder to understand and apprec ate. Thanks are
also due to Dr. Teh, whose encouragement ensured the complet on of the study.
Thanks are due to my employer, Un vers t Tunku Abdul Rahman for fund ng
the project. My s ncere thanks to the Dean of Faculty of Eng neer ng and
Green Technology, Un vers t Tunku Abdul Rahman for be ng support ve and
understand ng w th the d ff cult es n work ng and study ng at the same t me

throughout all these years.
I also would l ke to ded cate th s thes s to my parents and colleagues. Many
thanks are due to Professor Andrew Raga Henry R g t for h s encouragement
towards the end of the project, Dr. KH Yeap for the proofread ng, Mr. Peter
Cha and Mr. M chael Lee for the r ass stances and adv ces n fabr cat ng and
sett ng up the separator, Mr. J yuan Lee w th the h gh voltage power source,
and fr ends who have helped me n other ways espec ally Wymen and
Chr st ne. Last but not least, spec al thanks are due to my w fe and my son,
who have been very co-operat ve and also my source of nsp rat on.
Page 6
APPROVAL SHEET
Th s d ssertat on/thes s ent tled “DESIGN AND DEVELOPMENT OF AN
ELECTROSTATIC SEPARATOR FOR WASTE SEGREGATION” was
prepared by LAI KOON CHUN and subm tted as part al fulf llment of the
requ rements for the degree of Doctor of Ph losophy (Eng neer ng) at
Un vers t Tunku Abdul Rahman.
Approved by:
___________________________
(ASSOC. PROF. DR. LIM SOO KING)
Date:
Assoc ate Professor/Superv sor

Department of Electr cal and Electron c Eng neer ng
LKC Faculty of Eng neer ng and Sc ence
Un vers t Tunku Abdul Rahman
___________________________
(DR. TEH PEH CHIONG)
Date:
Ass stant Professor/Co-superv sor
Department of Electron c Eng neer ng
Faculty of Eng neer ng and Green Technology
Un vers t Tunku Abdul Rahman
Page 7
FACULTY OF ENGINEERING AND GREEN TECHNOLOGY
UNIVERSITI TUNKU ABDUL RAHMAN
Date: __________________
SUBMISSION OF FINAL YEAR PROJECT /DISSERTATION/THESIS
It s hereby cert f ed that LAI KOON CHUN (ID No: 10AED04870) has
completed th s f nal year project/ d ssertat on/ thes s* ent tled “Des gn and
Development of an Electrostat c Separator for Waste Segregat on” under the
superv s on of Assoc. Prof. Dr. L m Soo K ng (Superv sor) from the Department
of Electr cal and Electron c Eng neer ng, LKC Faculty of Eng neer ng and
Sc ence , and Ass st. Prof. Dr. Teh Peh Ch ong (Co-Superv sor) from the
Department of Electron c Eng neer ng, Faculty of Eng neer ng and Green
Technology.
I understand that Un vers ty w ll upload softcopy of my f nal year project /
d ssertat on/ thes s* n pdf format nto UTAR Inst tut onal Repos tory, wh ch may
be made access ble to UTAR commun ty and publ c.
Yours truly,
____________________
(LAI KOON CHUN)
*Delete wh chever not appl cable
Page 8
DECLARATION
I hereby declare that the d ssertat on s based on my or g nal work except for
quotat ons and c tat ons wh ch have been duly acknowledged. I also declare
that t has not been prev ously or concurrently subm tted for any other degree
at UTAR or other nst tut ons.

Name LAI KOON CHUN N
Date ____________________________
Page 9
TABLE OF CONTENTS
Page
ABSTRACT
ACKNOWLEDGEMENT v
APPROVAL SHEET v
SUBMISSION SHEET v

DECLARATION v
TABLE OF CONTENT v
LIST OF FIGURES x
LIST OF TABLES x
LIST OF NOMENCLATURES xv
LIST OF ABBREVIATIONS xv
CHAPTER
1.0 INTRODUCTION 1
1.1 Research Background 1
1.2 Problem Statements 4
1.3 Object ves of the Thes s 6
1.4 Outl ne of the Thes s 6
2.0 LITERATURE REVIEW 8

2.1 Introduct on 8
2.2 Food Waste Overv ew and Treatment 8
2.3 Fundamental of Electr cal D scharges 11
2.3.1 Dark d scharges 12
2.3.1.1 Corona d scharge 14
2.3.1.2 Spark d scharge 16
2.3.2 Glow d scharges 16
2.3.3 Arc d scharge 18
2.4 Electrostat c Separator 19
2.4.1 Typ cal separat on techn ques 19
2.4.2 Appl cat ons and des gn cons derat on 24
2.5 Taguch ’s Method 25
2.5.1 Array des gn 27
2.5.2 S gnal-to-no se rat o 28
2.6 Response Surface Methodology 29
2.6.1 Operat onal des gn 30
2.6.2 Analys s of var ance 30
2.7 Force Model 33
2.7.1 Food waste 35
2.7.2 Non-food waste 38
2.8 Summary 40
Page 10
3.0 MATERIALS AND METHODS 44

3.1 Introduct on 44
3.2 Waste Granule Preparat on 45
3.3 Electrostat c Separator Des gn and Setup 47
3.3.1 Test r g 48
3.3.2 The separator des gn 50
3.4 Analyt cal Procedures 53
3.4.1 Eff c ency and pur ty determ nat on for 54
OVAT evaluat ons
3.5 Robust Des gn w th Taguch ’s Method 55
3.6 Separat on Process Opt m sat on us ng Response 59
Surface Methodology
3.7 Summary 60
4.0 RESULTS AND DISCUSSION 62

4.1 Character sat on of Recovery Eff c ency 62
4.1.1 Effect of appl ed voltage 62
4.1.2 Effect of roller rotat on speed 66
4.1.3 Effect of angular pos t on of electrodes 68
4.1.4 Effect of m xture compos t on 72
4.1.5 Summary 74
4.2 Robust Des gn 75
4.2.1 Exper mental results 75
4.2.2 SNR analys s 78
4.2.3 Summary 80
4.3 Opt m sat on and Modell ng 81
4.3.1 Operat onal process des gn analys s and 85
opt m sat on
4.3.2 Surface plot analys s 95
4.3.3 Model opt m sat on and val dat on 101
4.3.4 Summary 102
5.0 CONCLUSION AND FURTHER WORK 105

5.1 Introduct on 105
5.2 Conclus on 105
5.3 Further Work 107
LIST OF REFERENCES 108

LIST OF PUBLICATIONS 118
Page 11
LIST OF FIGURES

F gures Page
1.1 Food waste n d fferent reg ons of the world 1
1.2 Sol d waste compos t on (wt %) n Malays a 3
2.1 The voltage-current character st c between parallel 12

plate electrodes n a low pressure env ronment
2.2 Type of corona d scharges (a) pass ve corona (b) 15

act ve corona
2.3 Typ cal electrostat c separat on techn ques (a) 20

tr boelectr c (b) nduct on (c) corona charg ng
2.4 Cyclone electrostat c separator 22
2.5 Induct on type electrostat c separator 23
2.6 Corona charg ng type electrostat c separator 23
2.7 General model of control system 26
2.8 Forces act on part cles (magn tude not accord ng 33

to scale)
2.9 Forces exerted on food part cles n (a) feed ng, (b) 36
on z ng and (c) detach ng stages
2.10 Forces exerted on non-food part cles n (a) 38

feed ng, (b) on z ng and (c) detach ng stages
3.1 Flowchart of the proposed method 45
3.2 D fferent conduct v t es of part cles (a) food, (b) 47

plast c and (c) glass
3.3 D agram of electrostat c separator 48
3.4 Photograph of the separator 51
3.5 Des gn schemat c of the separator 51
3.6 Inner and outer arrays by Taguch des gn 58
4.1 Effect of appl ed voltage on food waste recovery 63

and pur ty (rotat on speed = 70 rpm; feed content
Page 12

FW:NF = 40:60)
4.2 Effect of appl ed voltage on mass of recovered 64

food waste and m ddl ng (rotat on speed = 70 rpm;
feed content FW:NF = 40:60)
4.3 Effect of appl ed voltage on pur ty and recovered 65

mass of non-food waste (rotat on speed = 70 rpm;
4.4 Effect of roller rotat on speed on FW separat on 66

eff c ency (feed content FW:NF = 40:60)
4.5 Effect of roller rotat on speed on m ddl ng (feed 67

content FW:NF = 40:60)
4.6 Effect of electrodes gap on separat on eff c ency 69

(rotat on speed = 70 rpm; feed content FW:NF =
40:60)
4.7 Effect of electrodes gap on m ddl ng (rotat on 70

speed = 70 rpm; feed content FW:NF = 40:60)
4.8 Mass and pur ty w th d fferent electrodes gap 71

(rotat on speed = 70 rpm; feed content FW:NF =
40:60, appl ed voltage = 25 kV)
4.9 Effect of m x ng rat os on separat on eff c ency 73

and pur ty (rotat on speed = 70 rpm; appl ed
voltage = 25 kV)
4.10 Effect of d fferent m x ng rat os (rotat on speed = 74

70 rpm; appl ed voltage = 25 kV)
4.11 Effect of factors on SNR for max mal food waste 79

recovery
4.12 Effect of factors on SNR for m n mal m ddl ng 80

product
4.13 Pareto charts for (a) FW separat on eff c ency, (b) 83

m ddl ng and (c) NF separat on eff c ency
4.14 Pred cted values versus actual values (a) FW 94

separat on eff c ency (b) NF separat on eff c ency
4.15 Pred cted values versus actual values (a) FW 95

separat on pur ty (b) NF separat on pur ty

Page 13
4.16 Surface plots for comb ned effects of two 99

ndependent factors on FW separat on eff c ency.
(a) Potent al level and rotat on speed (electrodes
gap = 65 mm), (b) Potent al level and electrodes
gap (rotat on speed = 75 rpm) and (c) Rotat on
speed and electrodes gap (potent al level = 25 kV)
4.17 Surface plots for comb ned effects of two 100

ndependent factors on NF separat on eff c ency
gap = 65 mm); (b) Potent al level and electrodes
gap (rotat on speed = 75 rpm); (c) Rotat on speed
and electrodes gap (potent al level = 25 kV)
4.18 F gure 4.18: Surface plots for comb ned effects of 101
two ndependent factors on FW separat on pur ty
4.19 F gure 4.19: Surface plots for comb ned effects of 102
two ndependent factors on NF separat on pur ty

Page 14
LIST OF TABLES
Table Page
2.1 Tr boelectr c ser es 21
2.2 L-9 orthogonal array des gn 27
2.3 Summary of rev ew on development of waste 42

separator
2.4 Summary of rev ew on des gn cons derat on 43
3.1 Typ cal propert es of test samples 46
3.2 System parameters and the r range 52
3.3 Factors and the r levels 57
3.4 L-9 orthogonal array and the factors 58
4.1 Corona electrode angle and correspond ng 69

electrodes gap
4.2 M xture w th d fferent m x ng rat os 72
4.3 Exper mental response and the correspond ng SNR 76
4.4 Percentage mpact of d fferent factors on food 77

waste and m ddl ng
4.5 Screen ng w th PB factor al des gn 82
4.6 ANOVA table for FW separat on eff c ency n PB 84

des gn
4.7 ANOVA table for m ddl ng product n PB des gn 84
4.8 ANOVA table for NF separat on eff c ency n PB 84

des gn
4.9 Exper mental levels of ndependent process factors 86
4.10 CCD for var ous exper mental cond t ons 87
4.11 ANOVA results for quadrat c model of S1 89
4.12 ANOVA results for quadrat c model of S2 90

Page 15
4.13 ANOVA results for quadrat c model of P1 91
4.14 ANOVA results for quadrat c model of P2 92
4.15 Compar son results of observat on and pred ct on 101

xv
Page 16
LIST OF NOMENCLATURES
Notat on Descr pt on Un t
2
A surface area mm
B Pareto percentage %
CAD A r drag coeff c ent -
d Data d spers on -
d1 Corona electrode d stance from roller mm
d2 Electrostat c electrode d stance from roller mm
D Part cle s ze mm
-1
E Electr c f eld strength Vm
Fad A r drag force N
Fct Centr fugal force N
Fe Electrostat c force N
Fg Grav ty force N
F Image force N
FW0 Mass of n t al food waste n feeder g
GE Electrodes gap mm
k Number of factor -
K Constant -
m Mass g
mFW Mass n food waste tank g

m’FW Mass of food waste n food waste tank g
mNF Mass n non-food tank g
m’NF Mass of non-food n non-food tank g
xv
Page 17
n Number -
-1
N Rotat on speed mn
NF0 Mass of n t al non-food n feeder g
PFW Separat on pur ty of food waste %
PNF Separat on pur ty of non-food %
Q Part cle charge C
R Rad us of roller mm
SFW Separat on eff c ency of food waste %
SNF Separat on eff c ency of non-food %
t Th ckness mm
U Suppl ed voltage V
v Number of level -
-1
vr Relat ve veloc ty ms
WC Water content %
y Response -
Greek
α1 Corona electrode angle deg
α2 Electrostat c electrode angle deg
β0 Constant coeff c ent -
Coeff c ent of l near -

β,
β Coeff c ent of quadrat c -
βj Coeff c ent of nteract on equat ons -
-1
ε D electr c constant Fm
-1
σ Electr cal conduct v ty Sm
-3
ρ Dens ty kgm
xv
Page 18
-3
ρc Surface charge dens ty kgm
ρr Res st v ty Ωm
-1
ω Angular veloc ty rads

xv
Page 19
LIST OF ABBREVIATIONS
Notat on Descr pt on
ANOVA Analys s of var ance
CCD Central compos te des gn
DC D rect current
ESD Electrostat c d scharge
FAO Food and Agr culture Organ zat on
FSC Food supply cha n
FW Food waste
NF Non-food waste
NIMBY Not In My Back Yard
NSP Nat onal Strateg c Plan
OA Orthogonal array
OVAT One Var able At T me
PB Plackett-Burman

PET Polyethylene terephthalate
PS Polystyrene
PTFE Polytetrafluoroethylene
PVC Polyv nyl chlor de
RH Relat ve hum d ty
RSM Response surface methodology
SNR S gnal-to-no se rat o
SS Sum of squares
xv
Page 20
CHAPTER ONE
INTRODUCTION
1.1 Research Background
Waste of food appears as a global d lemma n many countr es. A study
from Food and Agr culture Organ zat on of the Un ted Nat ons (FAO) reveals
that 1.3 b ll on tonnes of food are wasted every year (FAOSTAT, 2012). Both
the ndustr al sed world and develop ng countr es are suffered from th s global
threat (Gustavsson, 2010). Food loss and waste n d fferent reg ons of the
world s shown n F gure 1.1. Sol d waste generat on ncreases due to rural-
urban m grat on, ncome per-cap ta ncrement and h gh demand of qual ty l fe
from the c t zens (Manaf, Samah and Zukk , 2009; Per athamby, Ham d and

Kh dz r, 2009). Rap d urban sat on and ndustr al sat on n Malays a make th s
country on a par as developed countr es, wh ch lead to ncrement of waste
generat on (Chua, Sah d and Leong, 2011).
F gure 1.1: Food waste n d fferent reg ons of the world (Gustavsson, 2010)
Page 21
The sol d waste generated per day n Malays a has reached 17000
tonnes accord ng to the Nat onal Sol d Waste Management Department
(2013). It s est mated that the da ly wastes w ll ncrease to 30000 tonnes n
2020. The Nat onal Strateg c Plan (NSP) for Sol d Waste Management n
Malays a has ntroduced pol cy on waste management to pr or t se waste
reduct on through processes of reduc ng, reus ng and recycl ng s nce 2001.
However, the pol cy does not lead to a pos t ve result due to low awareness of
c t zens (Meen-Chee and Narayanan, 2006). Most food waste s d sposed at
the d sposal s te due to the lack of food waste recovery fac l t es, poor waste
management n th s country and NIMBY (Not In My Back Yard) syndrome
(Saeed, Hassan and Mujeebu, 2009; Badg e et al., 2011). To date, source
segregat on of food waste s not commonly pract sed n Malays a (Samsud n
and Don, 2013).

In general, sol d waste s d sposed as landf lls. Landf ll ng s the most
general d sposal method as compared to other approaches such as nc nerat on
and compost ng. Th s s h ghly attr buted to ts s mplest and cheapest d sposal
procedures (Renou et al., 2008; Magdalena and Dana, 2014). Approx mately
95% of collected mun c pal wastes are landf lled n Malays a (Bash r et al.,
2010). Although th s could be the most pract cal waste treatment solut on,
landf ll ng does not seem to be the most rat onal approach to manage waste.
Desp te the nert sol ds, landf ll ng of food waste generates two ma n k nds of
by-products, namely gaseous em ss on and flu d c leachate wh ch can cause
h gh contam nat on (Chr stensen et al., 2001; Des der et al., 2003; Jaffr n et
al., 2003). Landf ll leachate s generally def ned as a complex l qu d
Page 22
conta n ng large amounts of organ c (ma nly food) and norgan c matter
(Ch an and De Walle, 1976). Landf ll leachate becomes a hazardous source to
groundwater aqu fer and surround ng water sources for ts h gh concentrat on
of pollutants (Kjeldsen et al. 2002). Ra nfall s the pr mary cause of leachate
generat on, followed by the b olog cal decompos t on act v t es tak ng place n
the landf ll. Bes des, landf ll gas such as methane and carbon d ox de due to
the decompos t on of b odegradable organ c matter s a great source of
pollut on. It would pollute the a r and cause publ c nu sance such as global
warm ng and cl mate change (Tagar s et al., 2003).
Landf lls n Malays a are generally crowded and t s mpract cal to

f nd new locat ons (Kath rvale et al., 2004). Therefore, proper food waste
management s cruc al n conserv ng a clean env ronment. It s not an
uncommon pract ce to sort and reuse the waste mater als. As shown n F gure
1.2, a h gh amount of organ c mater al, part cularly food waste (FW) (~45%)
can be found n the mun c pal sol d waste n Malays a, followed by plast c
(~24%), paper (~7%), metal (~6%), glass (~3%) and others (~15%).
F gure 1.2: Sol d waste compos t on (wt %) n Malays a (Nat onal Sol d Waste
Management Department, 2013)
Page 23
At present, stud es of the recovery of plast c, glass and metal from
sol d waste have been w dely carr ed out. However, to date, there s a lack of
progress made for the FW recovery (L n et al., 2013). The FW res dues are n
general turned nto landf lls or f rst generat on recycl ng pract ces such as
compost ng and an mal feed (Kofoworola, 2007). Lately, researchers have
placed h gh emphas s n food waste. Th s s because the h gh organ c content
n FW has the potent al of be ng turned nto h ghly added value end products,
such as ethanol or a source of b ofuel (Van Wyk, 2001; Le Man, Behera and
Park, 2010; Moukamnerd, Kawahara and Katakur, 2013). B oethanol s one of

the most prom s ng alternat ve energy sources to d m n sh the dependence on
foss l fuel (McM lan, 1997). It can be produced by the fermentat on of sugar-
r ch crops (e.g. sugar cane) and food wastes.
1.2 Problem Statements
Landf lls consume large land area and t may cause undes red
pollut ons. Proper captur ng and process ng of landf lls w ll turn the b ogases
em tted from landf lls nto renewable energy (Holm-N elsen, Al Sead and
Oleskow cz-Pop el, 2009). The recovery of b ogas from landf lls can be
prof table (Whalen, Reeburgh and Sandbeck, 1990). Inc nerat on appears to be
the lowest cost method, but the d rect heavy metal em ss ons generated from
the nc nerat on process contr bute s gn f cantly to human tox c ty and
env ronmental burdens (Tammemag and Tammemag , 1999; Xu, Chen and
Hong, 2014). Hence, food waste contents from mun c pal sol d waste are to be
segregated n an env ronmental fr endly way for the sake of ndependent
Page 24
b ogas product on and landf ll reduct on. Source segregat on s cruc al for
enabl ng the food waste to be reused and thus protect ng the env ronment.
Thus, the ma n a m of th s study s to nvest gate the feas b l ty of an
electrostat c separator n segregat ng non-food part cles from the recoverable
food waste.
Electrostat c separator s capable of separat ng part cles based on the
conduct v t es of the const tuent components. It s w dely used to sort out

part cles w th h gh conduct v t es from those w th relat vely low
conduct v t es. A number of stud es have shown the capab l ty of the separator
n treat ng the electron c waste (Mohabuth and M les, 2005; Yamane et al.,
2011). Nevertheless, to the best of our knowledge, there s st ll a lack of
research of electrostat c separator on the recovery of food waste be ng
documented.
Food waste segregat on w th electrostat c separat on process reduces
the water and a r pollut ons and m n m ses the land usage for landf ll ng.
Bes des, the nc nerat on of these landf ll substances that are free from
norgan c matters produces less res due and tox c gases. It contr butes to less
amount of landf ll leachate w th rare ex stence of organ c matters n landf lls.
In add t on to the env ronmental protect on, the proposed segregat on process
nd rectly enables the econom cal growth from b ogas and potent al b omass
energy generat on.
Page 25
1.3 Object ves of the Thes s
Electrostat c separat on of food waste st ll lacks of bas c research, and
t s cruc al to carry out th s study so as to ncrease the eff c ency of the
process. Th s project seeks to contr bute to the fundamental knowledge that s
requ red for future ut l sat on of pract cal sort ng system. Therefore, the

project s d v ded nto the follow ng steps:
) develop ng an env ronmentally fr endly way for waste segregat on,
) character s ng the performance of an electrostat c separator n terms of
the separat on eff c ency and pur ty,
) des gn ng a robust electrostat c separator wh ch m n m ses random
error, and
v) modell ng the separat on process by determ n ng the s gn f cant factors
and the opt mal operat onal cond t ons.
1.4 Outl ne of the Thes s
Th s chapter s the f rst of f ve chapters, wh ch ntroduces the research
background, problems and purposes of th s study. Chapter 2 reports the
general des gn of the electrostat c separator and the l terature rev ews. The
electrostat c theory, var ous types of gas d scharges, the separator appl cat ons
and the descr pt ons of some related formulae are also descr bed. The mater al
and method requ red to run the exper ment are qual tat vely narrated n
Chapter 3. The spec f cat ons of exper mental equ pment and the des gn
parameters are eluc dated us ng Taguch ’s method. The mechan sms of the
Page 26
corona format on process are ntroduced and the mpact of the d scharge of
corona on var ous matters s descr bed n th s chapter.
Chapter 4 evaluates and d scusses the electr cal performances of the

electrostat c separator w th respect to ts recovery eff c ency and pur ty. The
performance var es for d fferent exper mental des gns. Th s phenomenon s
thus stud ed n the same chapter by exam n ng a number of potent al
nfluenc ng parameters. The opt mal operat onal cond t ons are thus
summar sed by employ ng the opt m sat on method. F nally, Chapter 5 s
devoted to conclus ons and recommendat ons. The overall f nd ngs of th s
study are summar sed and some recommendat ons are suggested for future
mprovement work.
Page 27
CHAPTER TWO

LITERATURE REVIEW
2.1 Introduct on
Th s rev ew s ntended to prov de an up-to-date account of research on
the electrostat c separat on of wastes, w th part cular emphas s placed on the
corona charg ng techn que. The fundamentals of electrostat c such as
d scharge phenomena are ntroduced. In add t on, nformat on on typ cal
des gns and appl cat ons of electrostat c separator are rev ewed. The former s
becom ng more mportant because of the knowledge requ red to pred ct
charged granular waste dynam cs. The des gn and construct on of a robust
separat on system us ng Taguch ’s method s d scussed. F nally, the
nformat on essent al for response surface methodology s br efly rev ewed, n
term of the select on of the des gn of the exper ment and ts appl cat ons.
2.2 Food Waste Overv ew and Treatment
Throughout the food supply cha n (FSC), food loss can occur dur ng
the product on stage and post-harvest ng processes. Food waste s def ned as
the food loss at the reta l and f nal consumpt on stages of the food cha n,
wh ch relates to the behav our of the reta lers and consumers (Parf tt, Barthel
and Macnaughton, 2010). In the reta l stage, the foods nclude vegetables and
Page 28
fru ts wh ch w ll be prov ded to wet markets, grocers and supermarkets.

Before reach ng the shelves, about 10-15% of them w ll be d scarded for the
reasons of mproper handl ng, e.g. nsuff c ent cool ng storage. A large port on
of crops s rejected before be ng d str buted, due to the r gorous qual ty
standards on the s ze, shape and appearance (Stuart, 2009). Upon reach ng the
consumpt on stage, wastage s once aga n generated from household,
restaurants, hosp tal ty sector, pr sons, cafes and so on. Vegetables and fru ts
contr bute the h ghest port on of food waste, f compared to cereal, roots and
tubers, o lseeds and pulses, meat, f sh and seafood, and m lk. The waste of
food does not only represent the waste of econom c value, but also the waste
of the l m ted natural resources such as water, nutr ents, land and energy.
Bes des, the em ss on of greenhouse gases such as methane and carbon
d ox de, due to the waste of food, t can cause global warm ng (We tz et al.,
2002).
Food waste n the consumpt on stage can be class f ed nto two
categor es, namely pre-consumer food waste and post-consumer food waste.
The pre-consumer waste gets ts name for never be ng appeared n front of the
consumer. For nstance, overcooked, exp red, contam nat on and tr m waste
contr bute to th s type of waste. Post-consumer waste, on the other hand, s
ma nly caused by the lack of awareness from both caterers and guests. The
port on s ze and the behav our of guests n the self-serv ce buffet would lead
to the waste of food (Garnett, 2006). Some author t es have started to put
regulat ons on food waste management, before the food waste s sent for
nc nerat on or landf ll. In Ireland, the Env ronmental Protect on Agency and
Page 29

the Clean Technology Centre publ shed Waste Management (Food Waste)
regulat ons 2009 to ncrease the recovery amount of food waste (Galway
Country Counc l, 2013). Food waste from household must be source
segregated, before be ng collected by an author sed waste collector. Source
segregat on refers to the waste segregat on at source by the producers to avo d
spec f ed waste from be ng contam nated or m xed w th others. Nevertheless,
t strongly depends on the awareness of c t zens and the cont nual enforcement
from the pol cy author t es.
Pre-treatment of waste has recently rece ved momentous nterests. To
date, researchers have stud ed the pre-treatment of organ c waste w th a
number of processes. These nclude mechan cal (L ndmark et al., 2012),
thermo-chem cal (Vavourak , Vol ot and Kornaros, 2014) and enzymat c
(Taherzadeh and Kar m , 2008) treatments. However, very few stud es were
conducted us ng the electrostat c techn que. Although electrostat c separat on
s not a common treatment for organ c waste or food waste, t has rece ved
cons derable attent on for metal segregat on from electron c waste (Ve t et al.,
2005).
Electrostat c separat on prov des an effect ve approach n recover ng
the reusable matter from sol d wastes. It has been w dely employed n
appl cat ons nvolv ng dry separat on process, e.g. to recover conductors from
non-conduct ng m xtures (Lawver and Dyrenforth, 1973). The separat on
process sorts the charged bod es from the uncharged under an ntens ve
electr c f eld. It serves as an env ronmentally fr endly way for recycl ng and
10

Page 30
reus ng the resources w thout g v ng negat ve mpact to the surround ng
(K ew et, Bergougnou and Brown, 1978).
2.3 Fundamental of Electr cal D scharges
Upon energ s ng by the h gh voltage electr cal equ pment, electr cal
d scharge through the gaseous med um, or known as gas d scharge, can be
v s ble and aud ble. A gas d scharge can be generated when the electr cal
energy passes through gas med um. A cons derable amount of electr cal
charge should be created and stored. Stud es show that gas d scharges are
formed by neutral and part ally on sed part cles. The negat ve charged
part cles (free electrons) dr ft n an oppos te d rect on w th the electr c f eld.
Ow ng to the elast c coll s ons w th the molecules, the speed of the electrons s
l m ted. When the f eld strength becomes larger and the coll s ons become
nelast c, the on sat on effect happens and leads to an avalanche of charged
part cles. Th s process s known as gas d scharg ng, where the electr c f eld
strength s h gher than the electr c breakdown of amb ent gas at about 3 MV
per meter. The const tut on of electr c current depends on the number of
charges, the polar ty and the speed w th wh ch they move. The nteract ons
between part cles have made the gas d scharge a complex system wh ch
requ res deta led nvest gat ons. Gas d scharge phenomena can be class f ed
typ cally nto three categor es, namely dark d scharges (e.g. Townsend
d scharge), glow d scharges and arc d scharge. Var ous low pressure d scharge
modes between the parallel flat electrodes are schemat cally shown n F gure
2.1.

11
Page 31
F gure 2.1: The voltage-current character st c between parallel plate electrodes
n a low pressure env ronment (Wagenaars, 2006)
There are s x d fferent d scharge reg mes, namely non-self-susta n ng
d scharge (Reg me I), Townsend d scharge (Reg me II), subnormal glow
d scharge (Reg me III), normal glow d scharge (Reg me IV), abnormal glow
d scharge (Reg me V) and arc d scharge (Reg me VI), wh ch are descr bed n
the follow ng subsect ons. It s clear that the observed voltage-current
character st c s h ghly non-l near.
2.3.1 Dark d scharges
Dark d scharge reg ons refer to the d scharge reg mes I and II as
llustrated n F gure 2.1. The d scharge s generally nv s ble to the eye, except
for the corona d scharges and the breakdown. By study ng the voltage-current
character st c, d fferent d scharge modes or reg mes can be thereafter

recogn sed.
12
Page 32
) Reg me I: Non-self-susta n ng d scharge
−10
An extremely small current below 10 A can be measured when a
low voltage of 200 to 500 V passes through an electrode gap w th a few
m ll metres wh ch conta n ng gases such as oxygen and n trogen. Th s s due
to the fact that the cosm c rays or nearby UV lamp that generates the electrons
n the gap. These few electrons produce a very small current and accelerate
towards the anode by the potent al d fference. The appl ed voltage s not
suff c ent to on se the atoms as observed n h gher voltages. Therefore, t s
not self-susta n ng s nce the d scharge requ res external sources for the
electrons generat on. The d scharge w ll cease when the electron source s
removed.
) Reg me II: Townsend d scharge
The Townsend d scharge s also named as dark d scharge. Th s s
because there s no substant al l ght em ss on from the d scharge. Increas ng n
the appl ed voltage causes a changeover from a non self-susta n ng d scharge
to a self-susta n ng d scharge. The electr c f eld between the d scharge gap
enhances w th the ncreas ng of voltage. Electrons n the d scharge gap on se
the neutral atoms and result n a mult pl cat on of charged ons and electrons
w th n the gap. Due to the mpact on sat on, new electrons at the surface of

the cathode exper ence a secondary em ss on nto the gas. A susta nable
current through the d scharge gap s thus produced. The requ red voltage for
13
Page 33
trans t on from a non self-susta n ng to a self-susta n ng d scharge s called the
breakdown voltage.
For Townsend d scharge, the appl ed voltage of approx mately 700 V
s sl ghtly h gher than the breakdown voltage w th a large res stance and a low
-10 -6
current of 10 to 10 A. The space charge effect n the d scharge gap s not
s gn f cant as there are a l m ted number of charged part cles. As t can be seen
n F gures 2.1 (Reg me II), the voltage-current curve for the Townsend
d scharge s cons stently constant. Th s s due to the fact that the avalanche
process takes ts place n the gap. The ncreased voltage d rects h gher electron
mult pl cat on and produces more secondary em ss on of electrons at the
cathode. The process results n a further mult pl cat on of charges and
electrons n the gap. In other words, the current r ses cons derably w th a small
ncrease n voltage.
2.3.1.1 Corona d scharge
Corona d scharge s a relat vely weak lum nous electr cal d scharge
wh ch takes place at or near the atmospher c pressure. The corona s created
by a strong electr c f eld us ng small needles or sharp edge on the electrode
(Chang et al., 1995). Corona d scharge may be cons dered as a Townsend

d scharge depend ng on f eld and potent al d str but on (Schütze et al., 1998).
Th s d scharge wh ch em ts from the electrode appears as a fa nt blue-v olet
f lamentary d scharge, d ffers from the appl ed f eld polar ty and the
geometr cal conf gurat on of electrodes. The coronas can be pos t ve or
14
Page 34
negat ve, depend ng on the potent al polar ty of the electrode. For nstance,
electrode w th pos t ve charge generates pos t ve corona and v ce versa.
Corona d scharge has a relat vely slower energy release compared w th
other gas d scharges. The d scharge does not leave any def n te traces, but
secondary effects such as wettab l ty mprovement of the d rected mater al
surface. Corona d scharge can ex st n two ways, .e. pass ve and act ve, as
llustrated n F gure 2.2. Pass ve corona n F gure 2.2 (a) refers to the
conduct ng needle electrode connected to ground and exposed to an electr c
f eld by the sphere conductor. When the needle s moved towards the electr c
f eld unt l the f eld strength reaches the breakdown of amb ent a r, corona
d scharge s em tted n w th n the gap. On the other hand, the use of h gh
voltage power supply denotes the act ve corona as shown n F gure 2.2 (b).
The process s revers ble to the pass ve corona, where the power supply
appl es a h gh potent al to the needle for produc ng corona d scharge. Act ve
corona s w dely used to charge objects electrostat cally, such as powder
coat ng, electrostat c copy ng and separat on appl cat ons.

F gure 2.2: Type of corona d scharges (a) pass ve corona, (b) act ve corona
15
Page 35
As the on-em tt ng plasma of one polar ty w ll accumulate n the nter
electrode space, the corona d scharge s mostly l m ted by the space charge.
Consequently, the corona has a pos t ve res stance character st c that a h gher
voltage s requ red w th ncreas ng current. As soon as the current n the
d scharge s adequately ra sed, add t onal current-carry ng spec es w ll be
produced and thus the spark d scharge w ll be generated.
2.3.1.2 Spark d scharge
Generally, f the source for a d scharge s l m ted, the electr cal
d scharge tends to man fest tself nto a rap d mpulse type f lament d scharge
form known as spark. The ex stence of the spark d scharge symbol ses a
complex phys cal phenomenon that rel es on plent ful of var ables such as
pressure, electrode gap and electrode geometry. Spark can be formed when the
appl ed electr c f eld strength s h gher than the d electr c f eld strength of a r.
Th s sporad cally d scharge red str butes charge and form reg ons of excess
charge, wh ch may create a h ghly conduct ve path from the electrode to the

surround ng conductor. In electrostat c separat on process, spark d scharge s
cons dered an undes red phenomenon as t may br ng damage to the
equ pment and harm the operator.
2.3.2 Glow d scharges
D scharges n reg mes III, IV and V as llustrated by F gure 2.1 are
class f ed as glow d scharges, ow ng to the generat on of lum nous glow. Glow
16
Page 36
d scharge occurs once the breakdown voltage s reached. The d scharges em t
l ght due to the h gh enough electron energy of 1 to 5 eV and dens ty, typ cally
6 13 -3
of the order of 10 to 10 cm , to generate exc ted gas atoms by coll s ons.
Glow d scharge s w dely appl ed n var ous appl cat ons wh ch nclude
fluorescent l ght ng and plasma telev s on.
) Reg me III: Subnormal glow d scharge
Space charge effects are apparent n the d scharge gap when the
voltage further ncreases from that of prev ous reg me II. The space charge s
most l kely pos t vely due to the b g mob l ty d fference between the electrons
and ons (Chang et al., 1995). As the pos t ve space charge accumulates n
front of the cathode, a cathode fall reg on s thus formed. Cathode fall, or
known as cathode dark space, refers to the relat ve dark reg on near the
cathode. The voltage drop n the cathode fall s almost equ valent to the

voltage d fference across the electrodes. Typ cally, the electron mult pl cat on
across the cathode fall enhances when the electron mult pl cat on ncreases
under a h gher electr c f eld. Th s has resulted n a lower requ red voltage n
order to susta n the d scharge. Hence, the voltage decreases w th the
ncreas ng current as shown n the voltage-current curve n F gures 2.1
(Reg me III). Th s mode s not stable and eas ly changed to the glow d scharge
mode.
17
Page 37
) Reg me IV: Normal glow d scharge
In normal glow d scharge mode, the m n mum of susta n ng voltage
can be atta ned as the development of the cathode fall s completed. As there s
a contact between the plasma w th the cathode surface, the electrode current
dens ty rema ns w th the current change. In other words, the ncreased current
has no effect on the voltage change, but to spread the d scharge over the
surfaces of the electrode. The glow d scharge reg me stops when the ent re
electrode surfaces are covered by the d scharge. The d scharge voltage n th s
-3 -1
reg me s cons stent over a large var at on range of current (10 -10 A).
) Reg me V: Abnormal glow d scharge

The d scharge covers ent re electrodes n th s mode. The cathode fall
ncreases w th the ncreas ng current and the voltage across the electrodes
ascends sharply. Consequently, the average on energy ncreases w th
enhancement of cathode current dens ty. The on bombard ng the cathode
surface generates therm on c em ss on and turns the glow d scharge nto the
arc d scharge.
2.3.3 Arc d scharge
Arc d scharge s the d scharge reg me VI shown n F gure 2.1. The
trans t on of glow nto arc d scharge s observable when the current s further
ncreased. An electr c arc s a type of electr c d scharge wh ch has the h ghest
18
Page 38
current dens ty, extend ng from an order of 0.1 to 1 A to a very large (10 kA)
upper l m t. Arc d scharge s commonly used for ndustr al appl cat ons such
as weld ng and plasma cutt ng. However, undes red arc ng can br ng harmful
damages to the power stat ons, electr cal transm ss on systems and equ pment.
2.4 Electrostat c Separator
Electrostat c d scharge (ESD) s one of the electr cal d scharges. It
occurs when two electr cally charged tems contact ng each other and
generates a flow of electr c ty. The charge s typ cally appl ed n the
electrostat c separator. Electrostat c separator sorts two tems or substances
based on the d fferences n electr cal character st cs such as fr ct on charges

and surface result ng work funct on. Th s separat on s a cruc al manufactur ng
process n the ore benef c at on ndustry to remove the mpur t es. In general,
the separat on system does not only rely on electrostat c force, but also on the
grav tat onal and centr fugal forces. As the electrostat c force s nversely
proport onal to the surface area, the separat on works eff c ently for the small
and l ght-we ght substances such as th n sheet and short w res.
2.4.1 Typ cal separat on techn ques
Electrostat c separator typ cally sorts two d fferent types of substances
(w th d fferent electrostat c character st cs) at once. Pr or to the electrostat c
separat on, the substances are typ cally charged by fr ct on charge, nduced
charge or corona charge before subject ng to the electrostat c and grav ty
19
Page 39
forces. Implementat ons of var ous charg ng methods are llustrated n F gure
2.3.

F gure 2.3: Typ cal electrostat c separat on techn ques (a) tr boelectr c (b)
nduct on (c) corona charg ng (Chang, Crowley and Kelly, 1995)
Tr boelectr c charg ng equ pment as shown n F gure 2.3 (a) cons sts of
a v brat on feeder, collect on tanks and parallel plate electrodes connect ng to
h gh voltage supply. The charg ng act v ty takes place n the v brat ng feeder
20
Page 40
and turn ng the substances nto e ther pos t vely charged or negat vely
charged. The charg ng tendenc es can be referred n the tr boelectr c ser es
Table 2.1. The pos t vely charged substance tends to move toward the negat ve
electrode, whereas the negat vely charged tends to move n an oppos te
d rect on.
Table 2.1: Tr boelectr c ser es (Harper, 1967)
Human hands Pos t vely charged

Glass
Nylon
Human ha r
Wool
Alum n um
Food
Paper
Cotton Neutral
Steel
Wood
Hard rubber
Brass
Polyester
Polyethylene
PVC (V nyl)
Teflon Negat vely charged
21
Page 41
Th s separat on techn que rel es on the magn tude of suppl ed voltage
and the grav tat onal force. Thus, t s su table for appl cat ons w th large
grav ty d fference, such as remov ng mpur t es (ashes) from ore part cles. In
add t on, t can be comb ned w th the cyclone type separator, as shown n
F gure 2.4, to prov de better separat on eff c ency.

F gure 2.4: Cyclone electrostat c separator (Toraguch and Haga, 1982)
Induct on charg ng separator can be appl ed n the food process ng
ndustry to remove the mpur t es such as ha r, plast cs and waste straw. The
food s placed on a v brat ng conveyor belt under mult ple h gh voltage
electrodes, as shown n F gure 2.5. The charged mpur t es are to be moved
toward the electrodes and carr ed away by a suct on pump. Th s techn que
ensures the foods are not n contact w th the electrodes due to hyg en c
purposes.
22
Page 42

F gure 2.5: Induct on type electrostat c separator (Masu , 1982)
In the waste process ng ndustry, the corona charg ng separator can be
used to recover copper from used electr cal w re, or to separate scrap papers
from the paper-plast c m xture. Th s techn que s free from pollut on as t does
not nvolve burn ng or chem cal react on. As demonstrated n F gure 2.6, the
substances on the roller surface are subjected to an electr c charge, on sed
from the needle corona electrode and eventually separated due to the
d fferences n conduct v ty and electrostat c propert es.
F gure 2.6: Corona charg ng type electrostat c separator (Masu , 1982)
23
Page 43
2.4.2 Appl cat ons and des gn cons derat on
Electrostat c separat on has long been appl ed n the m n ng ndustry.

For nstance, the ores w th good conduct v ty such as ron and magnes um can
be separated from quartz and s l ca wh ch have poorer conduct v ty. Back n
1982, Murata et al. (1982) has ut l sed the method to separate copper part cles
rang ng from 37 to 840 μm n d ameter. The part cles were put on an ncl n ng
plate electrode under a non-un form electr c f eld. The results revealed that the
part cles could be sorted to the collect ng boxes, and the sort ng eff c ency
ncreases w th ncreas ng part cle s ze d fference as the forces exerted on the
smaller part cle could be eas ly d fferent ated from that of larger part cle.
In the past decade, Iuga et al. (2001) appl ed the techn que n
process ng granular wastes of chopped electr cal w res to remove polyv nyl
chlor de (PVC) w res nsulat on from the copper conductor. Moreover, they
ut l sed the electrostat c force for feldspar extract on from pegmat te wh ch
conta ns quartz and muscov te m ca. It was summar sed that electrostat c
separat on s a better way for m neral benef c at on techn ques, as compared to
flotat on method and magnet c sort ng, espec ally for small granular s zes ( .e.
< 0.5 mm) (Iuga et al, 2004).
Recently, Rav shankar and Kolla (2009) stated that the separat on
eff c ency s not only affected by part cle s zes, but also other factors such as
hum d ty and temperature. Inl ne w th the arguments, several stud es had been
made to analyse the d fferent electrostat c separat on processes. Elder et al.
24
Page 44
(2003) concluded the separat on eff c ency of m neral sand rel es on the

rotat on speed, electrodes conf gurat on, temperature of granule and other
parameters. Bes des, Aman et al. (2004) dent f ed that the supply potent al
and electrode pos t on to be the key factors for metal recovery. A study from
Cal n et al. (2008) revealed that the d fferent compos t ons may affect the
separat on results n separat ng plast c m xtures. In short, electrostat c
separat on s a mult -factor al process wh ch requ res s multaneous control of
both mechan cal and electr cal forces on the granular m xtures (Samu la et al.,
2005). Var ous parameters could affect the performance of the separat on
process.
2.5 Taguch ’s Method
In order to make the process less sens t ve to the effects of random
var ab l ty, one may class fy var ous des gn factors nto two groups, namely
system factor and random (or no se) factor. A robust des gn should be made to
reduce the var at ons of the process cond t ons caused by the random factors,
such as manufactur ng var at on and component deter orat on. Robust des gn
refers to a proper exper mental arrangement that makes the process nsens t ve
to the sources that are hard to control n pract cal cond t ons. Th s des gn,
mplemented by us ng Taguch ’s method (1986), s a stat st cal techn que n
enhanc ng the manufactured goods’ qual ty. It dent f es the dom nant process
parameters and determ nes the appropr ate operat onal env ronment n wh ch
an exper ment s to be performed. G ven a general model of control system as
shown n F gure 2.7, Taguch ’s method can be employed to:
25
Page 45

( ) determ ne wh ch factors are nfluent al on the response output y
( ) determ ne where to set the nfluent al controllable system factor x so
that the var ab l ty n y s small
( ) determ ne where to set the nfluent al controllable x so that the effects
of the uncontrollable no se factor z are m n m sed
F gure 2.7: General model of control system.
Taguch ’s method has been appl ed n the ndustr es and t has been
proven to be a cr t cal success n controll ng the qual ty of process
(Dascalescu, 2008). Basavarajappa et al. (2008), Dav dson et al. (2008), Hsu
et al. (2009) and Mahapatra et al. (2008) have ut l sed Taguch ’s method n
analys ng the mpacts caused by var ous parameters such as speed of the
exper mental tools. The nteract ons between the emp r cal factors and
responses were d scussed. Ch ang and Hs eh (2009), Comakl et al. (2009),
Keles (2009) and L n et al. (2009) have employed the same approach n
process opt m sat on and performance evaluat on. The a m was to dent fy the
opt mal operat onal cond t ons. Taguch ’s exper mental des gn, wh ch s based
only on a m n mum number of exper ments, prov des a rel able model for
cr ter on select on and dec s on mak ng (Chou, Ho and Huang, 2009).
26

Page 46
2.5.1 Array des gn
Taguch ’s method s a robust parametr c analys s techn que w th
reduced the number of exper ments (Senth lkumar, Senth kumaar and
Sr n vasan, 2013). Taguch ’s methodology of exper mental des gn s based on
an orthogonal array (OA) arrangement, where the nner array and outer array
are respect vely the system factors and random factors. In a typ cal des gn, a
total of four factors were var ed by three d fferent levels. In the full factor al
4
des gn, a total of 3 or 81 set of exper ments are requ red to be performed and
stud ed. However, only n ne exper ments (L-9 orthogonal array) were requ red
for th s four-factor-three-level system by us ng Taguch ’s method. Th s has
reduced the t me and cost to acqu re the same necessary results. G ven a
process w th four factors (A, B, C and D) and three levels (L1, L2 and L3), the
respond ng L-9 OA was thus formed, as tabulated n Table 2.2.
Table 2.2: L-9 orthogonal array des gn
Exper ment Factors
No. A B C D
1 L1 L1 L1 L1
2 L1 L2 L2 L2
3 L1 L3 L3 L3
4 L2 L1 L2 L3
5 L2 L2 L3 L1
6 L2 L3 L1 L2
7 L3 L1 L3 L2
8 L3 L2 L1 L3
9 L3 L3 L2 L1

27
Page 47
The L-9 orthogonal array s equ valent to a 2-(v, k, 1) array, or a set of
(k-2) mutually Lat n squares of order v, where v and k respect vely denote the
number of levels and number of factors. In other words, the L-9 orthogonal
array n th s study s a 2-(3, 4, 1) array, result ng n a 24 3− fract onal factor al
des gn. The f rst ( ndex ng) column of the table s an array on v-set and other
columns appeared as square arrays of order k. The f rst square array s a
transpose array of the ndex ng column, whereas the result ng square s the
Lat n squares of order v. Lat n square s a v × v array n wh ch each entry
occurr ng exactly once n each row and each column.
2.5.2 S gnal-to-no se rat o
A s gnal-to-no se rat o (SNR) s used as the object ve funct on to
determ ne the robustness. It rel es on the orthogonal array output, wh ch
spec f es the effects of var ous factors on the response format on. There are
three d fferent types of object ve funct ons, namely nom nal- s-best, larger- s-
better and smaller- s-better. Nom nal- s-best type of object ve funct on can be
used to determ ne the character st cs that are needed to be drawn as close as
poss ble to a nom nal response value, wh ch s shown n equat on (2.1):
2
y
SNR = 10 log (2.1)
s
where y s the mean of responses and s s the standard dev at on. Larger- s-
better type of object ve funct on dent f es the character st cs for the response

28
Page 48
to have ts max m sed value. S m larly, smaller- s-better type targets for the
m n m sed value of the response.
Larger- s-better type of object ve funct on s def ned as:
1 n 1
SNR −= log10 ∑ (2.2)
n 12=
y
whereas smaller- s-better type of object ve funct on s calculated by:
1 n 2
SNR −= 10 log ∑ (2.3)
n =1 y
where n s the number of sample and y s the response collected n each
evaluat on. The relat ve s gn f cance of the factors on the response can be
analysed by study ng the percentage (%) mpact that determ ned by:
(SNR − SNR )
2
% mpact = (2.4)
∑ (SNR − SNR )
2
where SNR s the mean value of SNR.
2.6 Response Surface Methodology
Response surface methodology (RSM) s an effect ve way to use w th
mult var able system to determ ne the nteract ons among system factors and
to pred ct the response. It has been successfully employed to several

opt m sat on processes such as n ckel removal (Arav nd et al., 2013) and
wastewater den tr f cat on (Sr nu Na k and Pyd Setty, 2014).
29
Page 49
2.6.1 Operat onal des gn
A number of factors can be analysed s multaneously w th proper
des gn of exper ments (Rezouga et al., 2009). Bes des, the opt m sat on results
deduced from the stat st cal analys s reduce the comput ng effort and cost
(Vlad et al., 2014).
In order to dent fy the cr t cal po nts n RSM, the response s
expressed as a quadrat c model accord ng to the follow ng polynom al
funct on:
p p p
y =β + ∑ β + ∑ β 2+ ∑ β + (2.5)
0 =1 X =1 X 1≤≤j jX X j e
where β0 s a constant coeff c ent, X , Xj are ndependent var ables, β , β , β j
are the coeff c ents of l near, quadrat c and nteract on equat ons, and e s the
error. Equat on (2.5) can be rewr tten as:
2 +β 2 + β 2
y =β +β X +β X + β X ++... β X +β X X X
0 11 22 33 p p 111 22 2 33 3
++... β 2+β +β +β
X XX XX XX
pp p 2112 3113 3223
++... β X X +e (2.6)
p −1p p −1 p

2.6.2 Analys s of var ance
Analys s of var ance (ANOVA) was employed to assess the f tted
qual ty of model and stat st cal s gn f cance of regress on coeff c ents.
ANOVA compares the change of var able levels and the var at on due to
30
Page 50
random errors of response measurement (Lee and Lee, 2012). The data
d spers on (d) for each observat on (x) s obta ned us ng the equat on:
2)
d = (xx − (2.7)
The total sum of square (SStot) adds all observat on d spers ons:
SStot = SSreg + SSlf + SSpe (2.8)
~ 2
SSreg = ∑ m
∑ n x x− (2.9)
j
~ 2
SSlf = ∑ m
∑ n x x− (2.10)
j
SSpe = ∑ m
∑ n
()
j x jx
−
2
(2.11)
where SSreg, SSlf, SSpe, m, n and x are the sum of error due to regress on, sum
of error due to loss of f t, sum of error due to pure error, number of level,
number of observat on and est mated value, respect vely. The model qual ty s
evaluated by values of the s gn f cance of regress on test (F-value,reg) and the
lack of f t test (F-value,lf). A s gn f cant regress on and a non-s gn f cant lack
of f t mply that the model could be f tted well to emp r cal data. The values of
the ment oned tests can be determ ned by us ng equat ons (2.12) and (2.13):
kn− SS reg
regF = ⋅ (2.12)
k −1 SS + peSS
lf
SS
mn− lf
F = ⋅ (2.13)
lf km− peSS
31
Page 51
where k s the number of parameters of the model.
Accuracy of the model can be measured by the coeff c ent of determ nat on, or
2
known as R :
+ peSS
lfSS
R 12−= (2.14)
totSS
2
A larger value of R s des rable as t means h gher accuracy. The
opt mum cond t ons of the quadrat c model can be determ ned by calculat ng
the cr t cal po nts. The quadrat c funct on, equat on (2.6) for three var ables
can be descr bed as the f rst grade system n equat ons (2.15) to (2.17):
∂y
= ββ + X +β X + β X (2.15)
∂X 1 12 2 1112313
1
∂y
=β + β X + β X + β X (2.16)
∂X 2 112 23 2223 2
2
∂y
= β +β X +β X + β X (2.17)
∂X 3 113 23 3322 3
3

In order to obta n the opt mum values, part al d fferent at ons of output
response y w th respect to XX and2 3 X are set to zero. The cr t cal po nt, .e.
1,
the max mum and m n mum coord nates of XX and2 3 X , can be obta ned by
1,
sat sfy ng equat ons (2.15) to (2.17) and solv ng the system.
32
Page 52
2.7 Force Model
Electrostat c separat on process rel es on the forces act on the part cles,
allow ng to be sorted out from a m xture. A number of forces ex st n the
process. Th s ncludes the p nn ng force nduced by the corona electrode and
the mechan cal centr fugal force due to rotat on. F gure 2.8 llustrates the forces
that act on a part cle dur ng the separat on process.

F gure 2.8: Forces act on part cles (magn tude not accord ng to scale)
The f ve forces were summar sed as ( ) centr fugal force due to the
rotat on, ( ) l ft ng force due to the attract on by electrostat c electrode, ( )
grav ty force, ( v) p nn ng force due to the on generat on from the corona
electrode on the nsulat ve part cles, and (v) a r drag force due to a r fr ct on.
The depos t on of part cles from the feeder was set to 12 g/m n n order to form
a monolayer on the surface of the roller.
33
Page 53
The part cles that pass through the corona on s ng zone are subjected to
on s ng and p nn ng effects generated by the corona electrode. Accord ng to
Lu et al. (2008), the nduced p nn ng force, or known as mage force, F
correlates to the s ze of the part cles and s def ned as:
2
Q
F = 2
(2.18)
4επ⋅ t 1
where Q s part cle charge, t s the part cle th ckness and ε1 s d electr c
constant of part cle. The electrostat c electrode located downstream the corona
electrode nduces a l ft ng force, the strength of wh ch rel es on the electr c
f eld. The l ft ng force, also known as electrostat c force, Fe, appl ed on the
part cles was calculated by:
ρε1 2
Fe = QE = EK c (2.19)

where K s a constant, E s the electr c f eld strength and ρc s the surface
charge dens ty. The grav ty force, Fg, acted on the mass, m, of the part cles was
def ned as:
Fg = mg = ρ gAt⋅ (2.20)
where A s the surface area of part cle and ρ s the dens ty of the part cle.
34
Page 54
The centr fugal force appl ed on the part cles s generated by the roller
wh ch rotates n counter-clockw se angular d rect on. Th s force, Fct, s always
at an oppos te d rect on to that of the p nn ng force and computed as (Younes et
al., 2007):
2
Fct = ωρ
At R (2.21)
where R s the rad us of separator roller, ω s the angular veloc ty and ρ, A, t are
the dens ty, surface area, th ckness of the part cle respect vely. An a r drag
force, Fad n an oppos te d rect on from the rotat onal trajectory prov des
fr ct on to the part cles. It can be determ ned by:
1 2
F ad = C AD ⋅ vAρ r
(2.22)

where CAD s the a r drag coeff c ent and vr s the relat ve speed of part cle.
2.7.1 Food waste
Due to the d ss m lar ty of surface res st v ty, the charges acted on food
waste part cles d ffered from that of on non-food part cles. For nstance, when
both food and non-food part cles were moved nto the corona on s ng zone, the
more conduct ve food part cles d scharged rap dly to the grounded roller f
compared to the non-conduct ve plast c and glass part cles. F gure 2.9
llustrates the comb nat on of forces appl ed on the food part cles.
35
Page 55

F gure 2.9: Forces exerted on food part cles n (a) feed ng, (b) on s ng and (c)
detach ng stages (L et al., 2007)
F gure 2.9 (a) shows the food part cle be ng depos ted onto the surface
of the roller rotat ng n a counter-clockw se d rect on. When the food part cle
passed by the on s ng zone as shown n F gure 2.9(b), t exper enced a
comb nat on effect of a r-drag force, centr fugal force, electrostat c force,
p nn ng force and grav ty force. However, s nce the food part cle s
36
Page 56
d ss pat ve, the charges on t were d scharged through the grounded electrode
and the effect of p nn ng force was weakened. D ss pat ve part cle allows the
charges to flow to ground more slowly than the conduct ve part cle. The
result ng force equ l br um state can be hence determ ned as:
F g cos γ = Fct + Fe (2.23)
and
F g s n γ = F ad (2.24)

In order to detach from the ground rotat ng roller towards the food
collect on tank as shown n F gure 2.9(c), the mechan cal and electr cal forces
exert ng on the conduct ve food part cle have to be sat sf ed,
Fct + Fe ≥ F g cos γ (2.25a)
or
2 2
ωρ
At + ρε1
EkR ρ Atg cos γ (2.25b)
c ≥
It s not ceable n equat on (2.25b) that the process rel es on the
propert es of the part cle ( .e. dens ty, area, and th ckness), roller conf gurat on
( .e. rotat on speed and roller rad us), electr c f eld strength, part cle surface
charge and pos t on of corona electrode (angle γ). In th s study, propert es of
the part cles were set to have small var ab l ty (further d scuss on n chapter
3). Impact of these var ables on the separat on performance was ns gn f cant
f compared to the rotat onal speed and electr c f eld strength. The two
parameters were then def ned as the system factors n th s study for the
parametr c stud es and the assessment of system robustness n chapter 4.
37
Page 57
2.7.2 Non-food waste
Unl ke d ss pat ve part cles, electrostat c force s absent n nsulat ve
non-food part cle. Non-food part cle exper ences only a r-drag force,
centr fugal force, p nn ng force and grav ty force, as shown n F gure 2.10.

F gure 2.10: Forces exerted on non-food part cles n (a) feed ng, (b) on s ng
and (c) detach ng stages (L et al., 2007)
38
Page 58
As shown n F gure 2.10(b), the force equ l br um state can be calculated by:
Fct+= FF g cos γ (2.26)
and
F g s n γ = F ad

In contrast w th food part cle, the non-food part cle was to be kept
attached to the roller surface before reach ng the brush. Th s ensures the
part cle to drop nto the non-food collect on tank. Thus, the follow ng
cond t ons have to be sat sf ed:
FF+ g cos γ ≥ F ct (2.27a)
or
2 2
ε2
EK ρ c 2
2
+ ρ Atg cosωργ
≥ At R (2.27b)
t
where K s a constant and ε2 s d electr c constant of the non-food part cle. As
the non-food part cle travelled to a po nt that γ = π, equat on (2.27a) can be
rewr tten as:
F ≥ Fct + F g , when γ = π (2.28a)
or
2 2
ε2
EK ρ c
≥ ωρ R + ρ Atg
2
2
At (2.28b)
t
It was apparent that a h gh electr c f eld strength, E was requ red to
generate large p nn ng force on the non-food part cle. Alternat vely, the slow
rotat on speed of the separator could ensure that the cond t on stated n
39
Page 59
equat on (2.28) can be sat sf ed and t ma nta ns a m n m sed usage of
electr c ty. However, slow separat on process was undes rably for product ve
outputs and hence the moderate range of speed should be therefore selected.

By observ ng equat ons (2.25) and (2.28), t seems that electr c f eld
demonstrated the synerg st c effect to the separat on eff c ency for both food
and non-food part cles. Effect of rotat on speed, on the other hand, appeared as
synerg st c to food part cle separat on and antagon st c to non-food separat on.
Nevertheless, some essent al evaluat ons ought to be performed and analysed
n order to conf rm the r nfluences on the performance of waste separat on.
2.8 Summary
Electrostat c separat on has been w dely used for mpur ty removal and
waste recycl ng. Its comparat vely s mple conf gurat on has ga ned a w de
acceptance. Charg ng of the substances can be ach eved by means of
tr boelectr c charge, nduct on charge and corona d scharge. The spark
d scharge can be generated f the current n the d scharge s adequately ra sed.
However, spark d scharge produces eros on and degrad ng traces at the
electrodes. The energy released by the d scharge should be ma nta ned to
prevent the undes red spark effect.
The separat on process s sens t ve to the effects of random var ab l ty.
Taguch ’s method wh ch based on an orthogonal array arrangement can be
employed to prov de robustness to the process. S gnal-to-no se rat os are used
n the robust parametr c analys s techn que. The robust process reduces the
40
Page 60
var at ons of the process cond t ons caused by the random factors such as
manufactur ng var at on and component deter orat on.

The eff c ency of the separat on rel es on the un form electr c f eld and
the separator conf gurat on. As the separat on process relates to the system
parameters such as appl ed voltage and pos t on of electrodes, the evaluat on
of these parameters should thus be performed. Research on the opt mum
electr c f eld and equ pment construct on for a complete separat on system
should also be carr ed out by employ ng respond surface methodology along
w th ANOVA analys s. The rev ew on development of waste separator and
rev ew on des gn cons derat on were summar sed n Table 2.3 and 2.4,
respect vely.
41
Page 61

Table 2.3: Summary of rev ew on development of waste separator
Authors and Descr pt on Major F nd ngs
Year
Lawver and • Recover conductor from • Sort ng of sol d part cles by
Dyrenforth non-conduct ng means of electr cal forces that

(1973) m xtures exerted on charged bod es
K ew et, • Sort the very f ne • Recovery of ron of 60 % can
Bergougnou part cles n flu d sed be ach eved.
and Brown base w th electrostat c • Recovery ncreases w th
(1978) f elds ncreas ng voltage
Masu (1982) • Remove mpur t es n • Non- nvas ve separat on w th
food process ng h gh voltage, together w th
ndustry w th nduct on e ther suct on pump
type and corona ( nduct on) or neutral ser
charg ng type separators (corona charg ng)
Toraguch and • Remove ashes from ore • Su table for appl cat ons w th
Haga (1982) part cles w th cyclone large grav ty d fference
type separator
Iuga et al. • Propose an mprove • Compared var ous electrodes
(1993) des gn over the • Un form ty of space-charge
trad t onal w th corona d str but on on the surface of
electrodes electrode
Chang, • Separat on us ng • Charg ng tendenc es of
Crowley and tr boelectr c separator substances are referred n
Kelly (1995) w th parallel electrodes tr boelectr c ser es
Ve t et al. • Recover metal from • Comb ne magnet c and
(2005) pr nted c rcu t boards electr cal forces for metal
scraps recovery
Wu, L and • Propose an mprove • Improved recovery f
Xu (2008) des gn w th double compared to s ngle roller
rotat ng roller des gn separator
42

Page 62
Table 2.4: Summary of rev ew on des gn cons derat on
Authors and Parameters and Major F nd ngs
Year Descr pt on
Murata et al. • Part cle s ze: 37 - • Sort ng eff c ency ncreases
(1982) 840µm w th ncreas ng s ze d fference
Iuga et al. • Part cle s ze decreases • Separat on eff c ency ncreases
(2001)
Elder et al. • Electr c f eld ncreases • Separat on eff c ency ncreases
(2003) • Un form part cle s ze • Better separat on eff c ency
• Rotat on speed • Affects the eff c ency
Aman et al. • Corona electrode angle • Separat on eff c ency
(2004) ncreases decreases (conductor only)
• Effect ve when angle > 25°
Iuga et al. • DC electr c f led • Better than AC electr c f eld
(2004) • M neral s ze decreases • Separat on eff c ency ncreases
Samu la et al. • Corona electrode angle • Separat on eff c ency
(2005) ncreases decreases
L et al. • Part cle s ze • Opt mum value of 0.6 – 1.2
(2007) mm for ndustr al appl cat on
Bendaoud et • Corona electrode • Separat on eff c ency
al. (2008) d stance decreases decreases
Cal n et al. • M xture compos t on • Affects the separat on
(2008) eff c ency
Lu et al. • Part cle shape • Affects the fall ng trajectory
(2008)
Rav shankar • Part cle s ze, hum d ty • Affects the separat on
and Kolla and temperature eff c ency

(2009)
Xue, L and • Rotat onal speed • Opt mum value of 50 – 70 rpm
Xu (2012)
43

Page 63
CHAPTER THREE
MATERIALS AND METHODS
3.1 Introduct on
Th s chapter descr bes the propert es of test sample and the operat onal
concept of the electrostat c separator, as well as, the exper mental techn ques
performed for val dat on. It also summar ses the electrode construct on,
separator des gn and calculat on, cal brat on and quant f cat on of errors for
the exper mental work. The evaluat on analys s presented n chapter 4 ncludes
the separat on character sat on, the robustness assessment of the separator and
the process opt m sat on. The exper mental work started w th waste samples
preparat on. Character sat on of the separat on system was then carr ed out by
analys ng the parameters of the separator under var ous operat onal cond t ons,
followed by the robustness assessment us ng Taguch ’s method, where the L-9
orthogonal array des gn was appl ed to l m t the random errors caused by no se
factors. Larger- s-better and smaller- s-better types of object ve funct ons were
employed for the y eld of food waste and m ddl ng respect vely. System
performance opt m sat on techn que was then appl ed to determ ne the opt mal
work ng cond t ons by ut l s ng response surface methodology. The proposed
methods n th s study are summar sed nto a flowchart as shown n F gure 3.1.

44
Page 64
F gure 3.1: Flowchart of the proposed method
3.2 Waste Granule Preparat on
The object ves of the project are to electrostat cally treat the waste
m xture. Thus, the commonly ava lable wastes were used. The sample of test
was typ cally a m xture of food waste part cles, FW (fru ts peel, 40 wt%) and
non-food waste part cles, NF (glass, 10 wt% and plast c, 50 wt%) per each
exper mental run. The fru t peel refers to the peel from carrot and apple w th
water content ranged from 10% to 30%, whereas the plast c refers to the
polystyrene (PS) food box and polyethylene terephthalate (PET) m neral water
bottles. All part cles were cut manually nto s zes w th n 1.0 mm to 4.0 mm.
The phys cal and electr cal propert es of the samples are g ven n Table 3.1.
45
Page 65
Table 3.1: Typ cal propert es of test samples
Propert es Symbol Un t FW Glass Plast c

-3
Dens ty p kgm 1200 2000 1050
- 1400 - 2500 - 1360

-4 -3 -4
Part cle t m 2.0x10 1.0x10 3.0x10
-4 -3 -3
th ckness - 5.0x10 - 2.0x10 - 3.0x10
-1 -4 -11 -20
Electr cal σ Sm > 1x10 1x10 < 1x10
-15
Conduct v ty - 1x10
4 11 20
Res st v ty pr Ωm < 1x10 1x10 > 1x10
15
- 1x10
The dens ty p was calculated by we ght ng a known area of sample.
The volume was obta ned by mult ply ng the measured area w th the measured
th ckness. In order to val date the calculat on of volume, a known mass of
sample was put nto a conta ner f lled w th l qu d med um and the l qu d
he ght was measured. A total of 10 measurements were taken and the range of
dens ty was determ ned. The res st v ty was measured w th a megaohm meter
and the electr cal conduct v ty was the rec procal of res st v ty. All samples
were not reused n the electrostat c character sat on as the electr cal propert es
may change once be ng used. The corona n the v c n ty of the electrode t p
may cause large molecules n the samples to break down to molecular
fragments, wh ch produces smaller fragments w th b gger on c mob l ty and
thus chang ng the electr cal conduct v ty. Note that the sample of test was
m xed w th fru t peel w th the h ghest electr cal conduct v ty, plast c w th the

largest th ckness and glass w th the largest dens ty, thus t s mpract cal to
employ tr boelectrostat c or cyclone type separator. The rotat ng roller type
separator was suggested nstead, for ts capab l ty n treat ng the glass and
46
Page 66
plast c ndependently. However, there s lack of l teratures to treat these
samples at the same t me and w th m xture of food waste. To the best of our
knowledge, th s s the f rst report of food waste-glass-plast c separat on
process us ng roll-type separator.
F gure 3.2 shows the d fference n conduct v t es of the waste part cles,
when be ng put on a ground plate electrode exposed to a corona needle
electrode. It s apparent that the fru t peel s conduct ve as the corona
d scharge penetrates through the part cle to the ground ng, whereas the glass s
relat vely non-conduct ve (d ss pat ve) and plast c (PS) s absolutely
nsulat ve.
F gure 3.2: D fferent conduct v t es of part cles (a) food, (b) plast c and (c)
glass
3.3 Electrostat c Separator Des gn and Setup

In th s study, a roll-type electrostat c separator has been ut l sed to
segregate the food waste from a m xture of plast c and glass env ronmentally
fr endly. The food appears more conduct ve for ts h gh water content (Wang
et al., 2008), mak ng t detachable from others dur ng the rotat onal separat on
process. The separator sorts the more conduct ve matter and less conduct ve
47
Page 67
matter to d fferent locat ons. However, some matter may fall n between as
m ddl ng product, result ng n a decl ne n separat on eff c ency (Wu, L and
Xu, 2008). The ma n object ves of present study focus on determ n ng the
eff c ency of waste recovery by electrostat c process, dent fy ng the dom nant
emp r cal factors and bu ld ng up the equat ons for opt mal separat on results
w th respect to operat onal cond t ons.
3.3.1 Test r g
The test r g was constructed w th the object ve of produc ng
cont nuous corona d scharges and to exam ne qual tat vely and quant tat vely
ts eff c ency for waste separat on. A schemat c of the exper ment layout and
electr cal connect ons of the exper mental work s shown n F gure 3.3.

F gure 3.3: D agram of electrostat c separator
48
Page 68
The separat on process ut l sed an earth-grounded roller type
electrostat c separator rotat ng n counter-clockw se d rect on. The samples n
the feeder were depos ted onto the roller surface. An on s ng needle electrode
powered by a h gh voltage d rect current (DC) power supply
(Informat onUnl m ted, USA) w th rated output up to 35 kV was placed at one
s de of the separator for corona d scharge generat on. The connect on from
power supply to the electrode was housed ns de a polytetrafluoroethylene
(PTFE) tube for electr cal nsulat on purpose. A set of nsulated retort stand
was used to hold the needle electrode and prov de flex b l ty to change the
electrode d splacement. An electrostat c plate electrode was connected
downstream, prov d ng the non-d scharg ng electrostat c f eld. A non-
conduct ve brush was placed at another s de to remove part cles left on the
roller and to prov de a clean roller surface for the next rotat on cycle.
The output of h gh voltage power supply was cal brated us ng a
purpose bu lt voltage d v der w th a 1: 301 rat o, form ng by three 10 MΩ

res stors and a 100 kΩ res stor. The actual value of the res stors was measured
us ng a d g tal Mult meter model 8145 (GW-Instek, Ta wan), w th rated
accuracy of ± 0.25% n the 20 MΩ res stance range and ± 0.10% n the 200
kΩ range. The res stors were connected n ser es and the output voltage across
the 100 kΩ res stor was measured, w th rated accuracy of ± 0.03% n the 200
V range. The measurements were conducted one hour after the un ts were
sw tched on so that the equ pment have suff c ent t me to warm up and were
carr ed out n tr pl cate as to assess the magn tude of random error sources.
The magn tude of the percentage error for the cal brated voltage due to the
49
Page 69
accuracy of measur ng nstrument was calculated to be 0.38%.
The speed of the roller was controlled by a geared motor w th a power
consumpt on of 40 W (Pee Moger, Ta wan). Output of the motor was
controlled by an electron c speed controller to prov de a speed range of 0 –
120 rpm. In order to avo d electromagnet c wave nterference, the motor
controller was sh elded w th a metal plate and placed away from the power
source. The exper ment was run n a vent lat on room and the amb ent was
recorded to be 24 – 28 °C w th relat ve hum d ty (RH) of 55 – 65 %.
3.3.2 The separator des gn
A lab scale roll type separator, used as the start ng po nt of the des gn
of the corona charge type separator as descr bed by F gure 2.3 (c), was

fabr cated. The separator n th s study has a better control of rotary speed as
dr ven by the electron c speed controller, wh ch marks the key mprovements
over the trad t onal des gn. The roller surface, corona electrode and
electrostat c electrode were made of sta nless steel. Sta nless steel was chosen
over other cheaper electr cal conductors such as alum n um s nce the latter
tends to form an ox de layer that s h ghly electr cal res stant. The photo and
schemat c of the separator des gn are shown n F gures 3.4 and 3.5,
respect vely. Note that the metal plate that used as the sh eld has been
removed wh le prepar ng F gure 3.4.
50
Page 70
F gure 3.4: Photograph of the separator

F gure 3.5: Des gn schemat c of the separator
Dur ng the electrostat c separat on process, var ous factors may affect
the separat on performance. Referr ng to the sect on 2.7 d scussed prev ously,
the factors ncluded the rotat on speed and electrode pos t on. The system
parameters nvolved n th s study are summar sed n Table 3.2 and the effects
w ll be d scussed n the follow ng chapter.
51
Page 71
Table 3.2: System parameters and the r range
Var ables M n. value Max. value
Voltage level, U (kV) 20 30
Rotat on speed, N (rpm) 60 90
Corona electrode angle, α1 (°) 25 50
Corona electrode d stance, d1 (mm) 50 60
Electrostat c electrode d stance, d2 (mm) 40 50
Note: Electrostat c electrode angle, α2 = 90°, feed rate = 12 g/m n.
The range of voltage level was selected based on the constra nt of the
h gh voltage supply, wh ch prov des a max mum rated voltage of 35 kV. The
range of rotat on speed was summar sed n Table 3.2 by referr ng to the

l teratures. The max mum speed of the motor can ach eve to be 120 rpm. The
corona electrode must not be placed too close to the feeder outlet as the charge
may affect the feed ng rate. D rect corona may also d spel the part cles to
another s de of the roller before they can travel along the counter-clockw se
rotat onal d rect on, thus result ng n a poor separat on performance.
The electrostat c electrode was set at an angle of 90 degrees. The
nterval between two electrodes was var ed n a range of 25 – 50 degrees, or
approx mately 31 – 61 mm on surface of the roller w th a rad us of 70 mm.
The d stance of the corona electrode s preferred to be further away from the
roller surface n order to generate a w der p nn ng zone. It s to be noted,
however, that plac ng t too far may lead to a weak space charge on the
granules (Samu la et al., 2005). Thus, the range of corona electrode d stance d1
was set as shown n Table 3.2. On the other hand, the d stance for electrostat c
electrode d2 should be set w th n a range of 1.0 to 2.5 t mes of voltage level U
52
Page 72
(Xu et al., 2009). These values were calculated to be suff c ent for fall ng
trajectory dev at on resulted from electrostat c force and prevent on of h gh
r sk spark d scharge. Deta led calculat on work w ll be descr bed n Chapter 4.
3.4 Analyt cal Procedures
Dur ng the electrostat c separat on process, the a r surround ng the
on s ng electrode was on sed by a h gh ntense corona d scharge, thus
form ng an on s ng zone. When the roller del vered the part cles through the

zone, the more conduct ve food part cles d ss pated the r charge through the
grounded roller, avo d ng them from be ng p nned for a longer t me than the
less conduct ve one. W th the cont nuous rotat on from the roller, the more
conduct ve part cles were subjected to h gher d fference between centr fugal
force and p nn ng force, and thrown off the roller. The electrostat c electrode
nduced an evenly d str buted electr c f eld and dev ated the more conduct ve
part cles from the r natural fall ng trajectory. The less conduct ve part cles
rema ned p nn ng onto the roller due to the larger p nn ng force appl ed on
them and they subsequently fell off at d fferent locat ons. In chapter 4, the
separat on mechan sms were analysed by referr ng to the force model. The
potent al system factors were dent f ed and effects of these factors on the
separat on performance were determ ned dur ng the character sat on process.
Nevertheless, some low mo sture (<10% mo sture content) food
part cles may not able to detach from the roller to food tank due to poor
conduct v ty. S m larly, some overs ze or overwe ght non-food part cles may
53
Page 73
drop earl er before fall ng nto the non-food tank, when the centr fugal force
surpassed the p nn ng force. These hard-to-control factors, e.g. un form
s ze/area of granule due to mach nery constra nt, l m t the system performance
and affect the accuracy of the analys s. The unsorted part cles tend to fall as
m ddl ng products (F gure 3.3). Thus, n add t on to the character sat on, a
robust des gn created by employ ng Taguch ’s method s ut l sed to analyse
the robustness of the separat on process control and to m n m se the unwanted

effects of random factors. The character sed results were further evaluated by
employ ng response surface methodology as to dent fy the opt mal cond t ons.
3.4.1 Eff c ency and pur ty determ nat on for OVAT evaluat ons
In electrostat c separat on, food waste and non-food waste were
separated by electrostat c processes. However, m sclass f cat on may occur
undes rably as d scussed earl er n sect on 3.4. Thus, OVAT evaluat ons were
necessary to analyse the eff c ency and pur ty of the separat on results. The
test sample of waste m xture was depos ted onto the roller surface and
eventually dropped nto three collect on tanks, namely food waste tank, non-
food waste tank and m ddl ng (M) tank. The mass from each tank was
measured by a d g tal prec s on balance w th a resolut on of 0.01 g after each
run. The separat on eff c enc es, S (%) of food waste, SFW and non-food waste,
SNF were calculated by:
m FW
% SFW , S FW = × %100 (3.1)
FW 0
m NF
% SNF , S NF = × %100 (3.2)
NF 0
54
Page 74
where mFW, FW0, mNF and NF0 are the mass n FW tank, n t al food waste
mass n the feeder, mass n NF tank and n t al non-food waste mass n the
feeder, respect vely. S m larly, the pur t es of the food waste tank (PFW) and
non-food waste tank (PNF) were determ ned by:
m ' FW
% PFW , PFW = × %100 (3.3)
m FW

m ' NF
% PNF = PNF = × %100 (3.4)
m NF
where m’FW and m’NF are the food waste mass n FW tank and non-food waste
mass n NF tank, respect vely. Both m’FW and m’NF were determ ned by
manuall sort ng.
3.5 Robust des gn w th Taguch ’s method
Robust des gn s an eng neer ng approach for rel able product ons
under var ous controllable c rcumstances. The des gn enhances the qual ty and
d m n shes the negat ve no se mpact on the products. Robust des gn may not
produce the most opt mal result. However, mplementat on of the des gn
mproves the process and qual ty of the products (Camposeco-Negrete, 2013).
Th s study employed Taguch ’s method to assess the robustness of
food waste segregat on process from the sol d waste m xtures. The four
factors, namely voltage level, rotat on speed, water content and part cle s ze
were cons dered based on reports found n the l terature. A s gnal-to-no se
rat o (SNR) was used as the object ve funct on to determ ne the robustness.
55
Page 75
SNR s v able for repl cated exper ments to fac l tate the data f ltrat on and
no se reduct on process. It has been w dely employed n stochast c screen ng
and opt m sat on problems as a performance measure (Besser s, 2014). In th s
study, larger- s-better type of object ve funct on was appl ed for max mal food

waste recovery, whereas smaller- s-better type of object ve funct on was used
to y eld m n mal m ddl ng product.
In the robust des gn, voltage level and roller rotat on speed were
selected as system factors (Dascalescu et al., 2005), whereas the water content
and s ze of food waste part cles were def ned as hard-to-control random
factors. Water content n food waste dec des ts conduct v ty and l kel hood to
be moved to the food collect on tank. However, th s parameter s hard to
control n pract cal cond t ons (Hou et al., 2010). The s ze of part cles, on the
other hand, var ed due to the mach nery l m tat on. Th s no se factor may
eventually lead to low recovery eff c ency (Wu, L and Xu, 2008). Robust
des gn us ng Taguch ’s method n present study targeted to l m t the negat ve
mpact of these hard-to-control random factors ( .e. water content and part cle
s ze) on the separat on process.
Four factors wh ch var ed w th n three levels have been selected as:
part cle s ze (1.0 mm, 2.5 mm and 4.0 mm), mo sture or water content (10 %,
20 % and 30 %), roller rotat on speed (60 rpm, 70 rpm and 80 rpm) and
suppl ed voltage level (20 kV, 25 kV and 30 kV). Table 3.3 tabulates the
factors and the r correspond ng levels. The recovered waste and m ddl ng
product were selected as the responses. The max m sat on of waste recovery
56
Page 76
and meanwh le the m n m sat on of m ddl ng product are des rable n
enhanc ng the separat on eff c ency of the electrostat c separator.

Table 3.3: Factors and the r levels
Factor Level
1 2 3
Part cle s ze, D (mm) 1.0 2.5 4.0
FW water content, WC (%) 10 20 30
Rotat on speed, N (rpm) 60 70 80
Voltage level, U (kV) 20 25 30
As th s s a des gn of four factors w th three levels, an orthogonal array
was bu lt as shown n F gure 3.6. The orthogonal array was des gned to
understand the mpact of each factor by assum ng that there s no nteract on
between any two factors. A total of n ne des gn po nts were selected for the
Taguch ’s method to form an L-9 array, as prev ously d scussed n Table 2.2.
In general, the system factors (rotat on speed, voltage level) have
s gn f cant effects on the average output of the process, whereas the random
factors (part cle s ze, water content) may or may not have nfluences on the
values around the average of output. Effect of these factors was evaluated by
analys ng the mean value of each array and the respect ve s gnal-to-no se rat o
(SNR) value. The parameters of the orthogonal array des gn were shown n
Table 3.4. Analyt cal d scuss on based on these parameters was performed n
the chapter 4.
57
Page 77

F gure 3.6: Inner and outer arrays by Taguch des gn
Table 3.4: L-9 orthogonal array and the factors
Exp. Part cle s ze FW water content Rotat on speed Voltage level
no. (mm) (%) (rpm) (kV)
1 1.0 10 60 20
2 1.0 20 70 25
3 1.0 30 80 30
4 2.5 10 70 30
5 2.5 20 80 20
6 2.5 30 60 25
7 4.0 10 80 25
8 4.0 20 60 30
9 4.0 30 70 20
58
Page 78

3.6 Separat on Process Opt m sat on us ng Response surface
methodology
Des gn of exper ments and analys s of the var ables were performed by
employ ng RSM us ng Des gn Expert 8.0.5 software (Stat-Ease Inc., USA). A
Plackett-Burman (PB) factor al des gn was appl ed as a screen ng
exper mental des gn approach to nav gate an n-d mens onal des gn space w th
n + 1 exper ments. In present study, 11 factors w th h gh/low values were
selected and led to a total of 12 exper ments. E ght of them were pract cal
factors nclud ng appl ed voltage (F1), rotat on speed (F2), corona electrode
angle (F3), corona electrode d stance (F4), electrostat c electrode d stance
(F5), %FW content n m xture (F6), m xture volume (F7), amb ent
temperature (F8), and three of them were dummy factors (D1, D2 and D3).
The h gh and low values of each pract cal factor were determ ned by referr ng
to the d scuss on n Table 3.2 and OVAT (one var able at t me) parametr c
analys s n the follow ng chapter.
Based on the PB des gn outputs, central compos te des gn (CCD) was
employed for modell ng. CCD s a factor al des gn w th a central po nt and an
α-d stance (α = 1.68) star topology. As CCD requ res the extens on of range
due to the α-d stance, t comes as a cons derat on n choos ng the appropr ate
operat onal range and not exceed ng the max mal rated output of the mach nes.
Separat on eff c ency and pur ty of both food and non-food waste, as d scussed
prev ously n sect on 3.4, were used as responses n evaluat ng the nfluence of
the ment oned pract cal factors. In order to dent fy the ma n effects, the Pareto
59

Page 79
analys s was appl ed to calculate the v tal percentage of each s gn f cant factor
towards the response as follows:
2
β
B = × 100 (3.5)
∑β 2
where B represents the mportance level of each var able and β represents
each polynom al coeff c ent.
Opt m sat on process was started w th the mathemat cal and stat st cal
analyses of the exper mental data. The polynom al models and the f tness of
the models were evaluated, n add t on to the tests of s gn f cance. The opt mal
operat onal cond t ons would thus be determ ned w th respect to the opt mum
values for each factor (Aggarwal et al., 2008).
3.7 Summary
The test r g and the electrostat c separator have been descr bed and the
character st cs of the waste were def ned n th s chapter. The h gh voltage
power supply was cal brated by us ng a purpose bu lt voltage d v der. Dur ng
the separat on process, an electrostat c f eld was produced by the corona and
electrostat c electrodes. Conduct ve food part cles n th s f eld are h ghly
subjected to electrostat c charge, whereas non-food part cles are ma nly
subjected to corona charge.

60
Page 80
The select on of factors and the r range were just f ed n th s chapter.
The range of the appl ed voltage, U (20 - 30 kV) was selected based on the
output constra nt of the h gh voltage supply at 35 kV. The range of rotat on
speed (60 - 90 rpm) was chosen on the same reason. D stance of corona
electrode from the roller surface was set w th n a range of 50 and 60 mm, so as
to prov de a w de p nn ng zone and meanwh le ma nta n ng a strong space
charge on the part cles. A s mple rule was appl ed n determ n ng the d stance
of the electrostat c electrode, .e. 1.0*U < d stance (mm) < 2.5*U. The
d stance was calculated to be suff c ent for fall ng trajectory dev at on and for
sparks prevent on.
Performance of the separator was affected by the system factors, e.g.
appl ed voltage. However, the performance may also be nfluenced by some
hard-to-control random factors. A robust des gn employ ng an L-9 orthogonal
array was employed to m n m se the negat ve mpact from these random
factors. The system factors were analysed to dent fy the r effects on the
separat on eff c ency and pur ty. Response surface methodology was then
employed to pred ct the opt mal separat on performance.

61
Page 81
CHAPTER FOUR
RESULTS AND DISCUSSION
4.1 Character sat on of Recovery Eff c ency
The theoret cal d scuss on n the l terature rev ew suggested a number
of factors that could nfluence the separat on performance of the separator. It
s equ valently cr t cal to recogn se the r effects wh ch govern the separat on
process. The one-var able-at-t me (OVAT) analyt cal method was employed to
assess the mpacts of these effects. In an attempt to make a comprehens ve
understand ng on the process, th s sect on explores the nfluence of the
operat ng cond t ons n terms of the appl ed voltage, roller rotat on speed,
pos t on of the electrodes and compos t on of the m xture. The waste samples
throughout the parametr c study were prepared n 100 g and the s ze of each
part cle was n average 4.0 mm.
4.1.1 Effect of appl ed voltage
The extent of electrostat c separat on caused by appl ed voltage could
be quant f ed n terms of food waste recovery and y eld of m ddl ng product.

The corona d scharge from the corona electrode s generated cont nuously
once the power supply s turned on, and ceased mmed ately as soon as the
power s cut off. As llustrated n F gure 4.1, the separat on and recovery of
62
Page 82
food waste were related s gn f cantly to the appl ed voltage. A drast c
ncrement of pur ty (%PFW) along w th a d st nct r se of separat on eff c ency
(%SFW) was noted at the voltage level of above 15 kV. For nstance, the
pur ty was mproved from 51% to 89% when the voltage ncreased from 5 kV
to 15 kV. The observat on suggests that the voltage can be set ( .e. 15 kV) as a
threshold for the parametr c stud es throughout sect on 4.1. Although h gh
separat on eff c ency can be observed at 5 kV, t n fact resulted n a very low
pur ty and consequently poor food waste recovery as shown n F gure 4.2,
ma nly due to the weak electr c f eld.

F gure 4.1: Effect of appl ed voltage on food waste recovery and pur ty
(rotat on speed = 70 rpm; feed content FW:NF = 40:60)
To better understand the extent of appl ed voltage on separat on
process, the we ght of m ddl ng products and the recovered food waste were
63
Page 83
measured. Recovered food waste (FW) refers to the mult pl cat on product of
separat on eff c ency (%SFW) and pur ty (%PFW). As shown n F gure 4.2, t
can be observed that the recovered food waste ncreases w th the ncrease of
appl ed voltage and reduct on of m ddl ng products. Th s can be nterpreted
that a stronger electr c f eld enhances the detachment of food waste to the FW
tank and as well the separat on pur ty, thus reduc ng the mass n the m ddl ng
collect on tank. It has been establ shed that the separat on process s attr buted
to the electrostat c force. As expressed by equat on (2.19), the electrostat c
force Fe s d rectly proport onal to the square of electr c f eld. It can be
deduced that the h gher the electr c f eld, the h gher the electrostat c force
acted to dev ate food part cles from fall ng trajectory wh ch leads to better
separat on eff c ency and food recovery (Bendaoud et al., 2008).

F gure 4.2: Effect of appl ed voltage on mass of recovered food waste and
m ddl ng (rotat on speed = 70 rpm; feed content FW:NF = 40:60)
64
Page 84
F gure 4.3 llustrates the effect of appl ed voltage on the non-food
part cles. It shows that h gh pur ty of non-food part cles, more than 90% n
th s case, was found ns de the NF collect on tank w th ns gn f cant
mpur t es. However, the recovered mass of NF, wh ch der ved from the
mult pl cat on product of NF separat on eff c ency (%SNF) and pur ty
(%PNF), exh b ted a comparat vely low output when the appl ed voltage was
below 15 kV. Th s was due to the fact that the low electr c f eld s nsuff c ent
to generate strong p nn ng force to move NF part cles nto NF tank, as
prev ously expressed by equat on (2.28). Bes des, the results ver f ed the
prev ous d scuss on that the m n mal requ red voltage level should be set to
obta n effect ve separat on results.

F gure 4.3: Effect of appl ed voltage on pur ty and recovered mass of non-food
waste (rotat on speed = 70 rpm; feed content FW:NF = 40:60)
65
Page 85
4.1.2 Effect of roller rotat on speed
The relat onsh p between the voltage and the separat on eff c ency at
d fferent rotat on speed s dep cted n F gure 4.4. Interest ngly, the separat on
of food waste at slower rotat on speed was found to be more effect ve. For
nstance, at the same appl ed voltage of 30 kV, the FW separat on eff c ency
can reach about 87% at a speed of 20 rpm f compared to 77% at 80 rpm. It
appeared that the lowest level of rotat on speed can prolong the t me of waste
part cles to stay w th n the on s ng zone. Nevertheless, the slow process ng
speed s mpract cal for energy cost opt m sat on. As such, the speed range of
40 rpm to 80 rpm could be cons dered for a rap d separat on process on one
hand, and for ma nta n ng a h gh recovery of waste on the other hand.

F gure 4.4: Effect of roller rotat on speed on FW separat on eff c ency (feed
content FW:NF = 40:60)
66
Page 86
As shown n F gure 4.5, n general, the m ddl ng mass reduced w th
faster roller rotat on, nd cat ng an enhancement to the separat on eff c ency. It
s n contrast to the prev ous d scuss on. From equat on (2.21), the centr fugal
force Fct relates pos t vely to the rotat on speed. A h gher rotat on speed
mpl es a larger centr fugal force to dev ate the food part cles to FW tank.
Nevertheless, results n th s sect on exh b ted that a slower rotat on process
can also contr bute to an effect ve separat on. The separat on performance n
th s case can be regarded as a compet t on between centr fugal force and the
on s ng t me (Elder, Yan and Ra ser, 2003; Xue, L and Xu, 2012). There s
l kely a cr t cal roller rotat on speed level, and t should be determ ned
spec f cally v a an opt m sat on methodology, wh ch would be d scussed n the
sect on 4.3.

F gure 4.5: Effect of roller rotat on speed on m ddl ng (feed content FW:NF =
40:60)
67
Page 87
4.1.3 Effect of angular pos t on of electrodes
The corona electrode and the electrostat c electrode are the essent al
elements n the process of electrostat c separat on. In most appl cat ons, these
electrodes are regularly character sed by d fferent angles and pos t ons, ow ng
to the d fferent natures and propert es of the waste part cles. As prev ously
shown n F gure 3.5, placement of the electrodes depends on a number of
parameters and nvest gat on on the r nteract ons could be an ntr cate task. In
order to s mpl fy the complex ty, the d stances of the corona electrode and
electrostat c electrode from the roller surface was set as 50 mm and 40 mm,
respect vely. Th s study also suggested a def n t on of electrodes gap GE,
wh ch reflects the relat onsh p between the two electrodes, was termed. As the
corona electrode angle was set by referr ng to the f xed electrostat c electrode
angle (α2 = 90°), th s parameter measured the nterval (mm) between the two
electrodes accord ng to the follow ng equat on:
2π R
GE = (αα
2 − 1 ) = 222.1 (90 − α )
1 (4.1)
360
The value of the electrodes gap corresponds to var ous angles of the
corona electrode (see F gure 3.5). These values were calculated us ng equat on
(4.1) and t was summar sed n Table 4.1.
F gure 4.6 shows the results for the change of the FW separat on
eff c ency w th d fferent electrodes gap des gns under var ous appl ed
voltages. A s gn f cant separat on was not ced for a small gap of 50 mm,
whereas a large gap of 80 mm exh b ted marg nal mprovement of eff c ency.
68
Page 88
Influence of electrodes gap on the separat on process s also reflected from the
measurement of m ddl ng mass. As dep cted n F gure 4.7, great reduct on n
m ddl ng was accompl shed w th the electrodes gap of 50 mm at the same
appl ed voltage compared to that of 65 mm and 80 mm. When the corona
electrode was moved further from the electrostat c electrode, the gap ncreased
to 80 mm and the eff c ency deter orated below that w th a smaller gap.
Table 4.1: Corona electrode angle and correspond ng electrodes gap
α2 (°) Corona electrode angle, α1 (°) Electrodes gap, GE (mm)
90 25 80

90 37 65
90 50 50
F gure 4.6: Effect of electrodes gap on separat on eff c ency (rotat on speed =
70 rpm; feed content FW:NF = 40:60)
69
Page 89
It appears that the separat on eff c ency would enhance n term of
reduct on of m ddl ng w th small electrodes gap. Set on a bas s that the
ntens ty of space charge between the electrodes ncreases w th a closer
d stance would y eld h gher eff c ency. However, t s also poss ble that f the
electrodes are located too close, charge would gather around the electrodes
and to cause local sed h gh electr c f eld and pressure. Th s can lead to
operat onal nterrupt on and a decl ne n mach ne l fespan. It s hence des rable
to establ sh an opt m sat on model for effect ve separat on.

F gure 4.7: Effect of electrodes gap on m ddl ng (rotat on speed = 70 rpm;
F gure 4.8 shows the pur ty and mass of samples for FW and NF along
w th the mass of m ddl ng under an appl ed voltage of 25 kV. It s not ceable
that the pur ty and separat on eff c ency decreased w th the ncrement of gap
70
Page 90
d stance between the electrodes. It could be due to the fact that the corona
electrode s placed too close to the feeder output. Based on the exper mental
observat on when the angle of corona electrode was set at 25 degrees, the large
coverage of the nduced on s ng charge nterfere the part cles com ng out
from the feeder. The charge repels the part cle undes rably nstead of p nn ng
t onto the surface of the roller. In other words, w th electrodes gap of 80 mm,

the part cle could be eventually fallen nto the NF tank or class f ed as
m ddl ng f fallen outs de any tank, thereby ncreas ng the mass of m ddl ng
and affect ng the pur ty of the NF collect on.
F gure 4.8: Mass and pur ty w th d fferent electrodes gap (rotat on speed = 70
rpm; feed content FW:NF = 40:60, appl ed voltage = 25 kV)
71
Page 91
4.1.4 Effect of m xture compos t on
Bes des the separator parameters d scussed prev ously, the eff c ency
of separat on s also affected by the type of m xture n terms of m x ng rat os.
M xed samples der ved from m x ng the food (mo sture 10% - 30%) and non-
food part cles at d fferent rat os as tabulated n Table 4.2 were nvest gated.
The groups G1, G2, G3, G4 and G5 for the FW:NF m xtures cons sted of
m x ng rat os of 50:50, 40:60, 30:70, 20:80 and 10:90 by volume, respect vely.
Table 4.2: M xture w th d fferent m x ng rat os (part cles were n same s ze)
Group M xture
FW (%) NF (%)
G1 50 50
G2 40 60
G3 30 70
G4 20 80
G5 10 90
F gure 4.9 llustrates the separat on eff c ency and pur ty of the m xed
FW-NF samples. The outcomes nd cated that the m xture n G5 exh b ted the
most prevalent NF separat on n term of pur ty. On the contrary, the FW
separat on pur ty turned out to be h ghest. The result could be expla ned by the
fact that the separat on pur ty s h ghly dependent on the m x ng rat o. For
nstance, the h gher the m x ng rat o of part cular part cles, the h gher the
respect ve separat on pur ty, and v ce versa. On the other hand, small var ance
of the separat on eff c enc es, for both FW and NF part cles, s gn f ed the
tr v al effect of the m x ng rat o n th s study.
72
Page 92

F gure 4.9: Effect of m x ng rat os on separat on eff c ency and pur ty (rotat on
speed = 70 rpm; appl ed voltage = 25 kV)
Effect of the m x ng rat o on the mass of m ddl ng and recovery s
dep cted n F gure 4.10. As d scussed prev ously, the nfluence of m x ng rat o
on the separat on eff c ency of FW part cle was ns gn f cant. However, w th
the drast cally decl ne of pur ty from m xture G1 to G5, the recovered FW
part cles were between the mass of 37.5 and 5.5 g out of the FW collect on
tank w th mass rang ng from 40.6 to 7.8 g, respect vely. S m larly, the NF
mass between 42.4 g (out of 44.2 g) and 80.1 g (out of 80.9 g) can be
extracted. The mass of m ddl ng l es n the range between 15.2 and 11.3 g. It s
l kely that the h gh amount of NF part cles n m xture G5 contr butes to
frequent tr bocharg ng act v t es, caus ng a great repellent effect wh ch
mproves the separat on eff c ency.
73
Page 93

F gure 4.10: Effect of d fferent m x ng rat os (rotat on speed = 70 rpm; appl ed
voltage = 25 kV)
4.1.5 Summary
The appl ed voltage, roller rotat on speed, angular pos t on of
electrodes and compos t on of the m xture were exam ned on the r mpacts to
the separat on process. It can be summar sed from the parametr c stud es and
theoret cal cons derat on that the most super or effect of electrostat c
separat on was assoc ated w th test sample under a h gh voltage level. The
separat on process can be enhanced by determ n ng the opt mum values of
74
Page 94

rotat on speed and electrodes gap for creat ng an ntens ve charg ng coverage,
wh le keep ng the process t me short to save energy. The opt m sat on model
should be establ shed as t would prov de a comprehens ve analys s on
separat on eff c ency and pur ty.
4.2 Robust Des gn
In add t on to the nfluenc ng factors prev ously d scussed, the
electrostat c separat on process was sens t ve to the effects of random
var ab l ty. Character st c var at on of samples and any source of natural
occurr ng were cons dered as random no se factors (Dascalescu et al., 2005).
Taguch ’s method prov des an approach to assess the robustness of the
separat on processes. The ma n purpose of robust des gn us ng Taguch ’s
method s to l m t the var at ons of the hard-to-control random errors. The L-9
array des gn n th s study cons sted of two types of factors, namely dom nant
des gn factors ( .e. appl ed voltage and rotat on speed) and random no se
factors ( .e. s ze and water content of the part cles).
4.2.1 Exper mental results
The mass of the recovered food waste and m ddl ng products collected
from the tanks were used as the exper ment responses of the L-9 array des gn.
Th s study was to dec de the effect ve separat on conf gurat ons for max mal
food waste recovery and meanwh le m n mal m ddl ng products. Thus, larger-
s-better type and smaller- s-better type of object ve funct ons were appl ed,
75

Page 95
respect vely. The former type s generally al gned w th a pos t ve s gn,
whereas the latter type has a negat ve SNR. The recorded responses were
tabulated n Table 4.3 w th the r correspond ng SNRs calculated us ng
equat ons (2.1) and (2.2).
Table 4.3: Exper mental response and the correspond ng SNR.
Exp. Food % SFW SNR M ddl ng, % SNR
no. waste, (FW) M (g) Recovery (M)
FW (g)
1 28.9 72 29.22 16.0 27 -24.08
2 32.3 81 30.18 15.1 25 -23.58
3 30.1 75 29.57 14.5 24 -23.23
4 31.4 79 29.94 13.7 23 -22.73
5 30.0 75 29.54 16.8 28 -24.51
6 32.5 81 30.24 14.9 25 -23.46
7 31.5 79 29.97 15.1 25 -23.58
8 34.0 85 30.63 13.4 22 -22.54
9 31.3 78 29.91 15.3 26 -23.69
Table 4.4 shows the percentage mpacts of the des gn factors. The
mpact of each factor on the food waste recovery and the m ddl ng was
calculated us ng equat on (2.4). It s not ceable that the voltage level has the
max mum effect (39.44 %) on effect ve recovery of food waste, followed by
the part cle s ze (26.93 %), water content (17.90 %) and rotat on speed (15.73
%) w th n the des gn range. On the other hand, the m n mal m ddl ng product
was mostly nfluenced by the voltage level (79.78 %). The water content has a
negl g ble effect (0.78 %) compared to rotat on speed (12.07 %) and part cle
s ze (7.37 %).

76
Page 96
Table 4.4: Percentage mpact of d fferent factors on food waste and m ddl ng.
Factor Level Food waste, FW M ddl ng, M
SNR SNR % mpact SNR SNR % mpact
D (mm) 1 29.67 -23.64
2 29.91 29.92 26.93 -23.60 -23.51 7.37
3 30.18 -23.29
WC (%) 1 29.71 -23.48
2 30.13 29.92 17.90 -23.58 -23.51 0.78
3 29.91 -23.46
N (rpm) 1 30.05 -23.39
2 30.01 29.92 15.73 -23.35 -23.51 12.07
3 29.69 -23.79
U (kV) 1 29.56 -24.10
2 30.13 29.92 39.44 -23.54 -23.49 79.78
3 30.06 -22.84
It s not surpr s ng at all that the appl ed voltage appears as a key factor
for electrostat c separat on. H gh level of appl ed voltage mpl es that a large
number of food waste part cles can be attracted towards food waste collect on
tank. Meanwh le, the non-food waste part cles were effect vely p nned and
moved to the non-food waste tank. Therefore, the m ddl ng product dropped
w th n was consequently m n m sed. S m larly, h gh recovery rate and low
m ddl ng were observed on the part cles w th large s ze of 4.0 mm as shown n
Tables 3.4 and 4.3. Th s was due to the fact that part cles w th large surface
area were subjected to more attract on and p nn ng effects by on sat on
process, thus mak ng them to drop to the r respect ve collect on tanks
effect vely (Ruan and Xu, 2012).

77
Page 97
Bes des, t s worth not ng that water content mproves the conduct v ty
of food waste part cles. Hence, th s factor has much larger mpacts on the
results of food waste recovery than that on m ddl ng. The rotat on speed of the
separator also demonstrated percept ble nfluences on both responses. Th s can
be expla ned by the centr fugal force generated by the rotat ng roller.
Centr fugal force plays a certa n role to detach the wastes at the r ght po nt, n
add t on to the attract on and p nn ng forces nduced by the electrodes.
4.2.2 SNR analys s
The graph cal representat on of the SNR effect of des gn factors on the
max mal food waste recovery was shown n F gure 4.11. It showed that the
best s tuat on happened at large part cle s ze (4.0 mm) w th 20 % water
content under a 60 rpm rotat on speed and 25 kV of appl ed voltage. S m larly,
by analys ng the graph cal representat on of the effect of des gn factors on the
m n mal m ddl ng product shown n F gure 4.12, t was observed that large
part cle s ze s preferable for ts h ghest SNR value w th n the exper mental
range. However, the water content of 20 % s not good for m ddl ng
m n m sat on, though t contr butes to the max mal recovery of food waste.
Cons der ng the water content does not have much nfluence on the m ddl ng
product as shown n Table 4.4, 20 % water content of food waste s an
opt mum select on.

78
Page 98
F gure 4.11: Effect of factors on SNR for max mal food waste recovery.
By study ng the system factors n F gures 4.11 and 4.12, t can be seen
that a appl ed voltage of 25 kV and rotat on speed of 60 rpm y eld the best
max mal food waste recovery. L kew se, an appl ed voltage of 30 kV produces

the m n mal m ddl ng product at rotat on speed of 70 rpm. Therefore, a robust
des gn of food waste separat on process can be ach eved at an appl ed voltage
of 25-30 kV at the roller rotat on speed of 60-70 rpm w th part cles of 4.0 mm
n s ze and 20% of water content. Hav ng th s set of cond t on ach eved us ng
79
Page 99
Taguch ’s method, the var at ons from random factors can be m n m sed and
the separat on eff c ency can be eventually enhanced.
F gure 4.12: Effect of factors on SNR for m n mal m ddl ng product.

4.2.3 Summary
The robust des gn targets to l m t the negat ve mpact of the hard-to-
control random factors. The mpacts of four factors, namely appl ed voltage,
rotat on speed, water content and part cle s ze d agonal were determ ned. It
80
Page 100
was not ceable that the appl ed voltage has a s gn f cant mpact on effect ve
recovery of food waste and reduct on of m ddl ng product w th n the des gn
range. Bes des, the SNR analys s revealed that the abovement oned funct ons
can be atta ned w th 4.0 mm part cle s ze of food waste and 20 % water
content, wh ch to result n less error from the var at ons of no se.
4.3 Opt m sat on and Modell ng
The parametr c stud es descr bed n prev ous sect on have been used to
exam ne a number of system parameters of the electrostat c separator along
w th d fferent m x ng rat o. Nevertheless, t s mpract cal to study every
potent al parameter as t would be t me-consum ng and not cost-effect vely.
Opt m sat on plays a role for such purposes n th s sect on so as to exam ne the
parameters w th great reduct on of requ red exper mental work.
A screen ng procedure by apply ng Plackett-Burman (PB) factor al
des gn was performed to m n m se the number of exper ments for

opt m sat on, and to nd cate the s gn f cant parameters wh ch nfluence the
eff c ency of the responses. The parameters of e ght pract cal factors, namely
potent al level (F1), rotat on speed (F2), corona electrode angle (F3), corona
electrode d stance (F4), electrostat c electrode d stance (F5), %FW content n
m xture (F6), m xture volume (F7), amb ent temperature (F8), and three
dummy factors (D1, D2 and D3) were dep cted n Table 4.5. The responses of
the FW separat on eff c ency (R1), m ddl ng product (R2) and NF separat on
eff c ency (R3) were also tabulated n the table.
81
Page 101
Table 4.5: Screen ng w th PB factor al des gn
F1 F2 F3 F4 F5 F6 F7 F8 D1 D2 D3 R1 R2 R3
(kV) (rpm) (°) (mm) (mm) (%) (g) (°C) (%) (g) (%)
20 90 50 60 40 40 100 28 -1 1 1 74.5 18.5 86.2
30 60 50 60 50 40 100 24 1 -1 1 83.8 13.1 89.0
30 90 30 50 40 60 100 28 1 -1 1 59.7 36.7 68.8
20 90 30 60 50 40 200 28 1 -1 -1 57.5 37.1 66.6
20 90 50 50 50 60 200 24 -1 -1 1 74.8 20.6 86.4
30 60 50 60 40 60 200 28 -1 -1 -1 83.6 14.2 89.1
30 60 30 50 50 40 200 28 -1 1 1 64.5 33.2 68.4
20 60 30 50 40 40 100 24 -1 -1 -1 61.3 32.9 71.0
20 60 30 60 40 60 200 24 1 1 1 61.1 35.1 70.6
20 60 50 50 50 60 100 28 1 1 -1 76.5 18.3 89.5
30 90 50 50 40 40 200 24 1 1 -1 77.9 15.4 89.2
30 90 30 60 50 60 100 24 -1 1 -1 59.5 36.6 69.3
In order to dent fy the ma n effects on the responses, Pareto analys s
was used to determ ne the nfluence of the ndependent factors (Zare et al.,
2010). As can be seen from the Pareto chart shown n F gure 4.13a, three

factors were found to be stat st cally s gn f cant and above the t-value l m t of
the FW separat on eff c ency, .e. the corona electrode angle, rotat on speed
and potent al level. L kew se, the results for the m ddl ng product and NF
separat on eff c ency were deduced. As shown n F gure 4.13b, the Pareto
chart showed that the corona electrode angle, rotat on speed and potent al level
account for major ty of the total effect on the m ddl ng mass. The Pareto chart
for the NF separat on eff c ency demonstrated that only two factors, namely
the corona electrode angle and rotat on speed, were s gn f cant and l ed above
the t-value as shown n F gure 4.13c. The ANOVA tables (Table 4.6 – 4.8)
conf rm the s gn f cance of the model for each response and ts s gn f cant
82
Page 102
factors at p-value less than 0.05 (Bash r et al., 2010). No outl ers were found
n the normal probab l ty test.

F gure 4.13: Pareto charts for (a) FW separat on eff c ency; (b) m ddl ng; (c)
NF separat on eff c ency.
83
Page 103
Table 4.6: ANOVA table for FW separat on eff c ency n PB des gn.
Source Sum of Degree of Mean square value F-value p-value
squares freedom
Model 1068.56 3 356.19 279.73 <0.0001
F1 45.24 1 45.24 35.53 0.0003
F2 60.30 1 60.30 47.36 0.0001
F3 963.02 1 963.02 756.30 <0.0001
Table 4.7: ANOVA table for m ddl ng product n PB des gn.
squares freedom
Model 1078.06 3 359.35 169.51 <0.0001
F1 14.74 1 14.74 6.95 0.0299
F2 27.30 1 27.30 12.88 0.0071
F3 1036.02 1 1036.02 488.69 <0.0001

Table 4.8: ANOVA table for NF separat on eff c ency n PB des gn.
squares freedom
Model 1106.61 2 553.30 359.96 <0.0001
F2 10.27 1 10.27 6.68 0.0295
F3 1096.34 1 1096.34 713.24 <0.0001
In part cular, the potent al level and corona electrode angle prov ded
pos t ve effect whereas the rotat on speed gave the negat ve effect to the
separat on eff c ency as shown n F gure 4.13a. However, n F gure 4.13b, the
three factors exh b ted an overall oppos te effect on the m ddl ng product, and
t nd cated that the separat on eff c ency can be enhanced by m n m s ng the
m ddl ng product. It s n close agreement w th our exper ments wh ch show
84
Page 104
that the recovery of food waste ncreases w th decreas ng amount of unsorted
m ddl ng for a g ven amount of each test sample. Therefore, the m ddl ng
response s not ncluded for further d scuss on, whereas separat on eff c ency
s focused n the next sect on for performance opt m sat on.
4.3.1 Operat onal process des gn analys s and opt m sat on
The PB factor al des gn dent f ed three s gn f cant factors to be
opt m sed. The f ve factors that were not cons dered s gn f cant n the PB
des gn were set to low level when perform ng the exper ments for

opt m sat on. A total of four responses was measured, namely FW separat on
eff c ency (S1), NF separat on eff c ency (S2), FW separat on pur ty (P1) and
NF separat on pur ty (P2). Th s study focused on opt m s ng the separat on
cond t ons by employ ng the central compos te des gn (CCD) approach.
Three ndependent var ables, namely, electr cal potent al (A), rotat on
speed (B) and electrodes gap (C) were selected for the opt m sat on process.
The CCD des gn of the exper ment was used to evaluate the nfluence of the
var ables n 20 exper ments, wh ch cons sted of 8 factor al po nts, 6 repl ca
center po nts and 6 ax al po nts. Table 4.9 shows the levels and ranges of the
three ndependent var ables nvolved. The CCD des gn matr x and the
correspond ng exper mental results were summar sed n Table 4.10.
85
Page 105
Table 4.9: Exper mental levels of ndependent process factors.
Var able
Levels A: Potent al B: Rotat on C: Electrodes
level (kV) speed (rpm) gap (mm)
-α 16.59 49.77 39.77
Low (-1) 20.00 60.00 50.00
Center (0) 25.00 75.00 65.00
H gh (+1) 30.00 90.00 80.00
+α 33.41 100.23 90.23

The CCD approach extends the α-d stance to both low and h gh levels.
These cr t cal values of the potent al level and rotat on speed are st ll w th n
the capab l ty of exper mental apparatus, whereas the electrodes gap s
adjusted accord ngly. Table 4.10 l sts the exper mental des gn matr x w th the
correspond ng evaluated values for the separat on eff c ency and pur ty. The
ANOVA results for quadrat c models of S1, S2, P1 and P2 were tabulated n
Tables 4.11 – 4.14, respect vely, whereas the polynom al equat ons were
deduced as equat ons 4.13 – 4.16.
86
Page 106
Table 4.10: CCD for var ous exper mental cond t ons.
Run Factor Response
A B C S1 S2 P1 P2
(kV) (rpm) (mm) (%) (%) (%) (%)
1 20.00 60.00 50.00 76.5 89.5 90.0 97.1
2 20.00 90.00 50.00 74.7 86.4 88.7 97.0
3 30.00 60.00 50.00 84.0 89.2 92.7 97.6

4 30.00 90.00 50.00 77.9 89.2 90.1 97.7
5 20.00 60.00 80.00 58.6 68.5 68.9 74.3
6 20.00 90.00 80.00 57.2 66.6 68.0 74.3
7 30.00 60.00 80.00 64.3 68.3 71.0 74.8
8 30.00 90.00 80.00 59.7 68.9 69.3 74.7
9 25.00 49.77 65.00 80.4 82.5 87.7 93.4
10 25.00 100.23 65.00 72.9 83.9 85.7 93.2
11 16.59 75.00 65.00 69.6 85.5 85.4 93.7
12 33.41 75.00 65.00 75.9 89.1 88.9 94.3
13 25.00 75.00 39.77 71.6 81.4 82.5 88.5
14 25.00 75.00 90.23 41.6 47.3 47.9 51.4
15 25.00 75.00 65.00 76.7 87.1 88.4 94.8
16 25.00 75.00 65.00 75.4 85.7 86.8 93.2
17 25.00 75.00 65.00 74.7 84.8 86.0 92.3
18 25.00 75.00 65.00 76.7 87.2 88.0 94.9
19 25.00 75.00 65.00 74.9 84.8 85.9 92.1
20 25.00 75.00 65.00 75.5 85.7 86.5 93.2
In order to demonstrate the emp r cal correlat on between response S1 and the
ndependent factors, a second-order equat on was deduced as follows:
S1(%) = − 912.41 + 324.3 A − 244.0 B + 206.3 C − 013.0 AB
2 −3 2 2
− 034.0 A + 387.2 × 10 B − 029.0 C (4.13)
87
Page 107
where S1 was the separat on eff c ency of FW n terms of actual factors.
Bes des, a polynom al equat on was deduced for response S2:
S2(%) = 1.364 − 1.715 A 3.599

+ C
2 −3 2 2
+ 0.024 A - 3.745 × 10 B - 0.033 C (4.14)

where S2 was the separat on eff c ency of NF n terms of actual factors.
The pur ty of the FW separat on, P1 was represented by a quadrat c
polynom al n terms of the actual factors:
2
P1(%) = - 9.914 + 0.130 A - 0.0362 B + 3.647 C - 0.034 C (4.15)
Concern ng the pur ty of the NF separat on, P2, the suggested model was
quadrat c and the f nal equat on n terms of actual factors was determ ned as:
2
P2(%) = - 1.784 + 3.981 C - 0.036 C (4.16)
The factors w th e ther synerg st c or antagon st c effect on the
response were dec ded based on the pos t ve or negat ve s gns of the
regress on coeff c ents. Tables 4.11 to 4.14 summar sed the ANOVA results
for the eff c ency and pur ty of the models. The model F-values of 258.17,
242.32, 373.69 and 332.84 for responses S1, S2, P1 and P2 respect vely,
mpl ed all the models were s gn f cant. The p-values were <0.0001 nd cated
the adequacy of the models. The pred cted R-Squared value of 0.9796 was n
close agreement w th the adjusted R-Squared value of 0.9919 for S1. The
s gnal to no se rat o (adequate prec s on) of 65.257 nd cated that the model
88
Page 108
can be appl ed to nav gate the des gn space. L ke S1, the pred cted R-Squared
of 0.9805 for S2 was close to the adjusted R-Squared of 0.9913. Adequate
prec s on of 58.178 (greater than 4) mpl ed that the model S2 could also be

used n the des gn space nav gat on (Olmez, 2009). Regard ng pur ty
responses P1 and P2, not much d fference was observed between the R-
Squared values for both models.
Table 4.11: ANOVA results for quadrat c model of S1.
Source Sum of Degree Mean F-value p-value Remarks
squares of square
freedom value
Model 1885.29 9 209.48 258.17 <0.0001 S gn f cant
A 63.70 1 63.70 78.51 <0.0001 S gn f cant
B 51.47 1 51.47 63.44 <0.0001 S gn f cant
C 1121.41 1 1121.41 1382.10 <0.0001 S gn f cant
AB 7.03 1 7.03 8.67 0.0147 S gn f cant

2
A 10.21 1 10.21 12.59 0.0053 S gn f cant
2
B 4.15 1 4.15 5.12 0.0471 S gn f cant
2
C 618.60 1 618.60 762.40 <0.0001 S gn f cant
Res due 8.11 10 0.81
Lack of 4.36 5 0.87 1.16 0.4370 Not
ft s gn f cant
Pure 3.75 5 0.75
error
Cor 1893.41 19
Total
R-squared: 0.9957; Adj R-Squared: 0.9919; Pred R-square: 0.9796; Adeq

prec s on: 65.257.
89
Page 109
Table 4.12: ANOVA results for quadrat c model of S2.

squares of square
freedom value
Model 2274.32 9 252.70 242.32 <0.0001 S gn f cant
A 8.31 1 8.31 7.97 0.0181 S gn f cant
C 1421.86 1 1421.86 1363.45 <0.0001 S gn f cant

2
A 5.31 1 5.31 5.09 0.0477 S gn f cant
2
B 10.23 1 10.23 9.81 0.0106 S gn f cant
2
C 812.17 1 812.17 778.80 <0.0001 S gn f cant
Res due 10.43 10 1.04
Lack of 4.80 5 0.96 0.85 0.5672 Not
ft s gn f cant
Pure 5.63 5 1.13
error
Cor 2284.75 19
Total
prec s on: 58.178.
90
Page 110

Table 4.13: ANOVA results for quadrat c model of P1.

squares of square
freedom value
Model 2351.64 9 261.29 373.69 <0.0001 S gn f cant
A 13.12 1 13.12 18.76 0.0015 S gn f cant
B 7.12 1 7.12 10.19 0.0096 S gn f cant
C 1486.68 1 1486.68 2126.16 <0.0001 S gn f cant

2
C 823.34 1 823.34 1177.49 <0.0001 S gn f cant
Res due 6.99 10 0.70
Lack of 1.56 5 0.31 0.29 0.9016 Not
ft s gn f cant
Pure 5.43 5 1.09
error
Cor 2358.63 19
Total

prec s on: 74.855.
91

Page 111
Table 4.14: ANOVA results for quadrat c model of P2.

squares of square
freedom value
Model 2718.60 9 302.07 332.84 <0.0001 S gn f cant
C 1729.68 1 1729.68 1905.90 <0.0001 S gn f cant

2
C 959.35 1 959.35 1057.09 <0.0001 S gn f cant
Res due 9.08 10 0.91
Lack of 1.89 5 0.38 0.26 0.9158 Not
ft s gn f cant
Pure 7.19 5 1.44
error
Cor 2727.68 19
Total

prec s on: 68.008.
2
The good correlat ons R value for the eff c ency and pur ty models
shown n F gures 4.14 and 4.15, respect vely, nd cated that there are only
m nor d fference between the real values and the theoret cal values. The p-
values nd cated the s gn f cance effect of the model, part cularly when p-
value s less than 0.05. It s apparent that all the models were h ghly s gn f cant
as the p-values were <0.0001 (Tables 4.11 to 4.14). In add t on, the
quant tat ve mpact of three ndependent factors was determ ned by the r p-
values. In the S1 model, the s gn f cant model terms were dent f ed as A, B, C,
2 2 2
AB, A ,B ,C w th p-values <0.05 (Table 4.11). All the ndependent factors,
.e. A-potent al level, B- rotat on speed, C-electrode gap and second-order
effect of C-electrode gap have h gh s gn f cance effect on the separat on
2 2 2
eff c ency for p <0.0001. On the other hand, the parameters of A, C, A ,B ,C
92
Page 112
are s gn f cant var ables for the S2 model. Th s model was ma nly nfluenced
by the second-order but less s gn f cantly by the f rst-order effect of B-rotat on
2
speed. The s gn f cant model terms for the P1 model were found as A, B, C, C ,
2
whereas the s gn f cant terms for the P2 model were C, C . It s worth to note
that the pur ty models were mostly nfluenced by the C-electrode gap and ts
second-order effect, but not the nteract on effects amongst A-potent al level,
B-rotat on speed and C-electrode gap.

93
Page 113
F gure 4.14: Pred cted values versus actual values. (a) FW separat on
eff c ency; (b) NF separat on eff c ency.

94
Page 114
F gure 4.15: Pred cted values versus actual values. (a) FW separat on pur ty;
(b) NF separat on pur ty.

4.3.2 Surface plot analys s
Three-d mens onal surface plots were generated to further study the
nteract ons between the ndependent factors A, B, C and the separat on
95
Page 115
cond t ons of the waste. The comb ned effects of factors A and B on the FW
separat on eff c ency was shown n F gure 4.16a. It showed that at an electrode
gap of 65 mm the separat on eff c ency ncreases w th an ncreas ng
magn tude of potent al level and decreases w th an ncreas ng speed of roller
rotat on. The surface plot nd cated that the potent al level has a synerg st c
effect on corona charg ng of the waste granules. For nstance, the separat on
eff c ency enhances from 69.6% to 76.7% w th the ncreas ng magn tude of
potent al level from 16.59 kV to 25.0 kV when the rotat on speed s
ma nta ned at 75 rpm. The reason of th s observat on s that the s ze of the
part cle test sample s not un form, and thus a larger coulomb force s requ red
to treat the larger part cles. It s generally known that Coulomb force relates
pos t vely w th potent al suppl ed. Compared w th the potent al level, the
rotat on speed demonstrated an antagon st c effect on the separat on
performance. It was noted that the separat on eff c ency decreased from 80.4%
to 72.9% when the rotat on speed ncreased from 49.77 rpm to 100.23 rpm and
the potent al level was ma nta ned at 25 kV. The reason s that the value of the
m ddl ng product ncreased w th the ncreas ng rotat on speed (see F gure
4.13). The observed relat on between the m ddl ng and rotat on speed was n a
good agreement w th the work from other researchers (Wu et al., 2008). The
h ghest value of separat on eff c ency of waste was obta ned at the m n mum
rotat on speed and potent al level closed to 30.0 kV.
F gure 4.16b demonstrates the correlated effects of factors A and B on
the FW separat on eff c ency w th a constant speed of 75 rpm. It was obv ous
that the separat on eff c ency ncreased w th ncreas ng potent al level,
96
Page 116
nd cat ng a stronger electrostat c f eld enhances the nduct on charg ng of
conduct ve part cles. For the electrode gap, the percentage of separat on
eff c ency f rstly rose w th ncreas ng gap from 50 mm to 60 mm, and then
dropped progress vely when the gap was beyond 60 mm. It nd cated that 50
mm to 60 mm was the proper spac ng to y eld a un formly extens ve charge
d str but on to enhance the separat on eff c ency (Xu et al., 2009). Therefore,
the eff c ency dropped once t exceeded the proper range. F gure 4.16b also
demonstrates that the potent al level and electrode gap have the synerg st c
effect on the separat on eff c ency, and the h ghest value of eff c ency could
ach eve at the max mum potent al level w th a proper spac ng of electrodes
approx mately closed to 60 mm.
W th a f xed potent al level of 25 kV, the comb ned effects of factors B
and C on the FW separat on eff c ency was shown n F gure 4.16c. It was
apparent that the separat on eff c ency of waste decreased w th ncreas ng
rotat on speed when the electrode gap was less than 60 mm. The reason s that
the unsorted m ddl ng ncreased w th the rotat on speed as d scussed above.
Moreover, F gure 4.16c shows that the separat on eff c ency decreased w th

ncreas ng electrode gap after reach ng the ment oned proper gap of 60 mm. It
s attr buted to the weaker charge on the test part cles. The h ghest value of the
FW separat on eff c ency was ach eved at the m n mum rotat on speed w th an
electrode gap of approx mately closed to 60 mm.
Comparat vely, surface plots for models S2, P1 and P2 were shown as
respect ve three-d mens onal d agrams n F gures 4.17 - 4.19. The d agrams
97
Page 117
exh b ted how each of the models var ed as a funct on of two-factor (A-
potent al level and B-rotat on speed, A-potent al level and C-electrodes gap, B-
rotat on speed and C-electrodes gap) nteract on. It was apparent n F gure
4.17a that the NF separat on eff c ency d d not ncrease cont nually w th
ncreas ng magn tude of potent al level and rotat on speed. Th s mpl es that
the opt mal des gn po nts should be dent f ed n order to determ ne the
opt mal separat on cond t ons.

98
Page 118

F gure 4.16: Surface plots for comb ned effects of two ndependent factors on
FW separat on eff c ency. (a) Potent al level and rotat on speed (electrodes
gap = 65 mm); (b) Potent al level and electrodes gap (rotat on speed = 75
rpm); (c) Rotat on speed and electrodes gap (potent al level = 25 kV)
99
Page 119

NF separat on eff c ency (a) Potent al level and rotat on speed (electrodes gap
= 65 mm); (b) Potent al level and electrodes gap (rotat on speed = 75 rpm); (c)
Rotat on speed and electrodes gap (potent al level = 25 kV)
100
Page 120

FW separat on pur ty (a) Potent al level and rotat on speed (electrodes gap =
65 mm); (b) Potent al level and electrodes gap (rotat on speed = 75 rpm); (c)
101
Page 121

NF separat on pur ty (a) Potent al level and rotat on speed (electrodes gap = 65
mm); (b) Potent al level and electrodes gap (rotat on speed = 75 rpm); (c)
102
Page 122
4.3.3 Model opt m sat on and val dat on
Opt m sat on was performed to dent fy the opt mum operat onal
cond t ons for electrostat c separat on. Th s study def ned the opt mal
separat on cond t ons so as to have max m sat on models of the separat on
eff c ency (S1, S2) and pur ty (P1, P2). Th s allows the m n m sat on of
unsorted and m sclass f ed products. Based on the des gn range tabulated n

Table 4.9, the opt mal cond t ons were found at the potent al level of 30 kV,
roller rotat on of 60 rpm and electrode gap of 54 mm. The m n mum m ddl ng
percentage of 12.7% (volume 100g) was also obta ned. In order to val date the
stat st cal exper mental strateg es, a tr pl cate set of exper ments was
performed at the spec f ed opt mum cond t ons and an average FW separat on
eff c ency of 84.2% was ach eved. A compar son between the observed and
pred cted values was summar sed n Table 4.15. The exper mental value
agreed well w th the pred cted value. It nd cated the val d ty of the model and
the success of RSM n opt m s ng the FW separat on from waste m xture.
Table 4.15: Compar son results of observat on and pred ct on.
Model Observed Pred cted Percentage
value (%) value (%) of error (%)
FW separat on eff c ency 84.20 83.88 0.38
NF separat on eff c ency 88.70 89.51 0.91
FW separat on pur ty 93.00 92.89 0.12
NF separat on pur ty 98.50 97.97 0.54
103
Page 123
4.3.4 Summary
The response surface methodology was used to opt m se the separat on
cond t ons and to dec de on the s gn f cant terms. Pr or to opt m sat on, the PB
screen ng procedure was appl ed to conclude that the separat on cond t ons
were ma nly nfluenced by three factors ( n the order of stat st cal

s gn f cance): electrode gap, rotat on speed and electr cal potent al level. Thus,
the determ nat on of opt mal cond t ons was performed by employ ng the CCD
approach. The CCD des gn shown that the three factors, namely A-potent al
level, B-rotat on speed and C-electrode gap were stat st cally s gn f cant to the
separat on eff c ency, whereas the C-electrode gap and ts second-order effect
played a key role for the separat on pur ty. The opt mum separat on cond t ons
were potent al level / rotat on speed / electrode gap = 30 kV / 60 rpm / 54 mm,
and the max mum FW eff c ency / NF eff c ency / FW pur ty / NF pur ty was
84.2% / 88.7% / 93.0% / 98.5% w th a m n mum m ddl ng product of 12.7%.
The qual ty control concern ng the pred ctab l ty of the model s very
sat sfactory.
104
Page 124
CHAPTER FIVE
CONCLUSION AND FURTHER WORKS

5.1 Introduct on
The work has extended the knowledge from electrostat c
separator and prov ded d scuss on on mproved separator
conf gurat ons, wh ch enhances the separat on qual ty of food waste
and non-food part cles. The force model analys s wh ch ncludes the
centr fugal force, electrostat c force and corona mage force was
eluc dated. The nfluent al effects of the parameters on the separat on
eff c ency have been descr bed, and the var at ons of undes red random
factors have been nvest gated. A mathemat cal surface model has been
der ved n th s study, and the opt mal operat onal cond t on was
calculated from the model. Overall speak ng, the results of th s study
have contr buted to the fundamental knowledge requ red for future
ut l sat on of pract cal electrostat cally dr ven food waste separator.
5.2 Conclus on
The ma n f nd ngs of the study are summar sed as follows:
1. Based on the force model, the appl ed voltage and rotat on speed
are the dom nant factors of the separat on process. Theoret cally,
105
Page 125
the ncrement of e ther or both factors enhances the separat on
eff c ency of food waste at a predeterm ned angle of the corona

electrode, γ. For non-food separat on, the change of the appl ed
voltage s assoc ated w th the value of the rotat on speed, assum ng
the surface charge dens ty and corona electrode angle are f xed.
2. Above the appl ed voltage of 15 kV, the separat on eff c ency and
pur ty of both food and non-food part cles enhanced w th the
ncrement of voltage. The recovered mass of food part cles
pos t vely correlates to h gher voltage and shorter angular pos t on
of electrodes, whereas the m ddl ng mass decreases n the same
manner. The effect of the rotat on speed s not able to pred ct from
the parametr c study.
3. The s ze and mo sture of the part cles affect the dens ty,
conduct v ty and surface res st v ty, wh ch n turn nfluent the
separat on results. Thus, the values of hard-to-control parameters of
the part cles should be f xed so as to m n m se the r mpacts on the
analys s of the system des gn factors.
4. The pred ct on from the response surface methodology conf rms
that the max mal separat on result could be ach eved at h gh
voltage level under the opt mal speed and angle. Although one may
be able to further enhance the eff c ency by ut l s ng h gher
voltage, the opt m sat on model prov des an alternat ve to reduce
106
Page 126

the dependent on the costly power supply for generat ng a
comparable outcome.
5.3 Further Work
The recommendat ons for further works are concluded as follows:
1. The electrostat c separator research could be extended to evaluate
and recover other waste samples, e.g. food gra n n order to reduce
the waste d sposal nto the landf ll. The relevant ntegrator system
shall thus be stud ed.
2. The separat on eff c ency may be mproved by employ ng a double
roller des gn separator, or re-depos t ng the separated part cles onto
the roller for a second round of process.
3. The L fe-Cycle Assessment could be performed to assess the
balance po nt w th n the energy usage of the h gh voltage, energy
generat on of the b ogas from the separated food waste, and also the
return on nvestment of develop ng a large scale of separator based
on the electrostat cal method as descr bed.
107
Page 127
LIST OF REFERENCES
Aggarwal, A., S ngh, H., Kumar, P. and S ngh, M., 2010. Opt m z ng power
consumpt on for CNC turned parts us ng response surface methodology and
Taguch ’s techn que – A comparat ve analys s. Journal of Mater als
Process ng Technology, 200, pp. 373-384.
Aman, F., Morar, R., Kohnlechner, R., Samu la, A. and Dascalescu, L., 2004.
H gh-voltage electrode pos t on: a key factor of electrostat c separat on
eff c ency. IEEE Transact ons on Industry Appl cat ons, 40(3), pp. 905-910.
Arav nd, J., Len n, C., Nancyflav a, C., Rash ka, P. and Saravanan, S., 2015.
Response surface methodology opt m zat on of n ckel (II) removal us ng
p geon pea pod b osorbent. Internat onal Journal of Env ronmental Sc ence
and Technology, 12, pp. 105-114.
Badg e, D., Samah, M.A.A., Manaf, L.A. and Muda, A.B., 2011. Assessment
of mun c pal sol d waste compos t on n Malays a: Management, pract ce and
challenges. Pol sh Journal of Env ronmental Stud es, 21, pp. 539-547.
Basavarajappa, S., Chandramohan, G. and Dav m, J.P., 2008. Some stud es on

dr ll ng of hybr d metal matr x compos tes based on Taguch techn ques.
Journal of Mater als Process ng Technology, 1(3), pp. 332-338.
Bash r, M.J.K., Az z, H.A., Yusoff, M.S., Az z, S.Q. and Mohajer , S., 2010.
Stab l zed san tary landf ll leachate treatment us ng an on c res n: Treatment
opt m zat on by response surface methodology. Journal of Hazardous
Mater als, 182, pp. 115-122.
Bash r, M.J.K, Az z, H.A., Yusoff, M.S. and Adlan, M.N., 2010. Appl cat on of
response surface methodology (RSM) for opt m zat o of ammon acal n trogen
removal from sem -aerob c landf ll leachate us ng on exchange res n.
Desal nat on, 254, pp. 154-161.
Bendaoud, A., T lmat ne, A., Medles, K., Rahl , M., Huzau, M. and
Dascalescu, L., 2008. Charater zat on of dual corona-electrostat c electrodes
for electrostat c processes appl cat ons. IEEE Transact ons on Industry
Appl cat ons, 44(3), pp. 692-698.
108

Page 128
Besser s, G.J., 2014. Mult -response non-parametr c prof l ng us ng Taguch ’s

qual metr c eng neer ng and neurocomput ng methods: Screen ng a foam ng
process n a solar collector assembly. Appl ed Soft Comput ng, 22, pp. 222-
237.
Cal n, L., Neamtu, V., Morar, R., Iuga, A., Samu la, A. and Dascalescu, L.,
2008. Tr bocharg ng of granular plast c m xtures n v ew of electrostat c
separat on. IEEE Transact ons on Industry Appl cat ons, 44(4), pp. 1045-1051.
Camposeco-Negrete, C., 2013. Opt m zat on of cutt ng parameters for

m n m z ng energy consumpt on n turn ng of AISI6061 T6 us ng Taguch
methodology and ANOVA. Journal of Cleaner Product on, 53, pp. 195-203.
Chang, J.S., Crowley, J.M. and Kelly, A.J., 1995. Handbook of Electrostat c
Processes, Marcel Dekker Inc.: New York.
Ch an, E.S. and De Walle, F.B., 1976. San tary landf ll leachates and the r
treatment. Journal of Env ronmental Eng neer ng (ASCE), 102, pp. 411-431.
Ch ang, M.Y. and Hs eh, H.H., 2009. The use of the Taguch method w th grey
relat onal analys s to opt m ze the th n-f lm sputter ng process w th mult ple
qual ty character st c n color f lter manufactur ng. Computers & Industr al
Eng neer ng, 2, pp. 648-661.
Chou, C.S., Ho, C.Y. and Huang, C.I., 2009. The opt mum cond t ons for
comb nat on of magnet c part cles dr ven by a rotat ng magnet c f eld us ng
the Taguch method. Advanced Powder Technology, 20(1), pp. 55-61.
Chr stensen, T.H., Kjeldsen, P., Bjerg, P.L., Jensen, D.L., Chr stensen, J.B.,
Baun, A., Albrechtsen, H.J. and Heron, G.., 2001. B ogeochem stry of landf ll
leachate plumes. Appl ed Geochem stry, 16, pp. 659-718.
Chua, K.H., Sah d, E.J.M., Leong, Y.P., 2011. Susta nable mun c pal sol d
waste management and GHG abatement n Malays a. Green and Energy
Management, 04-02, pp. 1-8.
Comakl , K, S msek, F., Comakl , O. and Sah n, B., 2009. Determ nat on of
opt mum work ng cond t ons R22 and R404A refr gerant m xtures n heat-
pumps us ng Taguch method. Appl ed Energy, 86(11), pp. 2451-2458.

109
Page 129
Dascalescu, L., Samu la, A., M halc o u, A., Bente, S. and T lmat ne, A., 2005.
Robust des gn of electrostat c separat on process. IEEE Transact ons on
Industry Appl cat ons, 41(3), pp. 715-719
Dascalescu, L., 2008. Robust corona-electrostat c separat on method for sol d

waste recycl ng. Internat onal Journal of Env ronment and Waste
Management, 2(4/5), pp. 423-435.
Dav dson, M.J., Balasubraman am, K. and Tagore, G.R.N., 2008. Exper mental
nvest gat on on flow-form ng of AA6061 alloys a Taguch approach. Journal
of Mater als Process ng Technology, 1(3), pp. 283-287.
Des der , U., D Mar a, F., Leonard , D. and Pro ett , S., 2003. San tary landf ll
energet c potent al analys s: A real case study. Energy Convers on and
Management, 44, pp. 1969-1981.
Elder, J., Yan, E. and Ra ser, G., 2003. eForce - Newest generat on of
electrostat c separator for the m neral sands ndustry. In: Heavy M nerals
Conference. South Afr ca Inst tute of M n ng and Metallurgy.
FAOSTAT, 2012. Food Balance Sheets 2009. [onl ne] Ava lable at:
<http://faostat.fao.org/s te/354/default.aspx> [Accessed 20 August 2012].
Galway Country Counc l, 2013. Waste Management. [onl ne] Ava lable at:
<http://www.galway. e/en/Serv ces/Env ronment/WasteManagement/>
[Accessed 22 March 2013].
Garnett, T., 2006. Fru t and vegetables and UK greenhouse gas em ss ons:
Explor ng the relat onsh p. UK: Food and Cl mate Research Network,
Un vers ty of Surrey.
Gustavsson, J., 2010. The Cl mate Change Impact of Reta l Waste from
Hort cultural Products. MSc. Un vers ty of Gothenburg, Sweden.
Harper, W.R., 1967. Contact and fr ct onal electr f cat on. Monographs on the
Phys cs and Chem stry of Mater als. Clarendon Press: Oxford.

Holm-N elsen, J.B., Al Sead , T., Oleskow cz-Pop el, P., 2009. The future of
anaerob c d gest on and b ogas ut l zat on. B oresource Technology, 100, pp.
5478-5484.
110
Page 130
Hou, S., Wu, J., Q n, Y and Xu, Z., 2010. Electrostat c separat on for recycl ng
waste pr nted c rcu t board: a study on external factor and a robust des gn for
opt m zat on. Env ronmental Sc ence & Technology, 44, pp. 5177-5181.
Hsu, W.H., Chao, C.K., Hsu, H.C., L n, J. and Hsu, C.C., 2009. Parametr c
study on the nterface pullout strength of the vertebral body replacement cage
us ng FEM-based Taguch method. Med cal Eng neer ng & Phys cs, 31(3), pp.
287-294.
Iuga, A., Morar, R., Dascalescu, L., Samu la, A. and Raf ro u, D., 1993. A new
type of corona electrode for h gh-tens on separators. Magnet c and Electr cal
Separat on, 4, pp. 75–90.
Iuga, A., Morar, R., Samu la, A. and Dascalescu, L., 2001. Electrostat c
separat on of metals and plast cs from granular ndustr al wastes. IEE
Proceed ngs - Sc ence, Measurement and Technology, 148(2), pp. 47–54.
Iuga, A., Cuglesan, I., Samu la, A., Blajan, M., Vadan, D. and Dascalescu, L.,
2004. Electrostat c separat on of muscov te m ca from feldspath c pegmat cs.
IEEE Transact ons on Industry Appl cat ons, 40(2), pp. 422-429.
Jaffr n, A., Bentounes, N., Joan, A.M. and Makhlouf, S., 2003. Landf ll b ogas
for heat ng greenhouses and prov d ng carbon d ox de supplement for plant
growth. B osystems Eng neer ng, 86, pp. 113-123.
Kath rvale, S., Muhd Yunus, M.N., Sop an, K. and Samsudd n, A.H., 2004.
Energy potent al from Mun c pal Sol d Waste n Malays a. Renewable Energy,
29(4), pp. 559-567.
Keles, O., 2009. An opt m zat on study on the cementat on of s lver w th

copper n n trate solut ons by Taguch des gn. Hydrometallurgy, 95, pp. 333-
336.

K ew et,onC.,ofBergougnou,
separat f ne part cle M.A. and Brown,
n v brates J.D.,
flu d zed 1978.
beds. IEEEElectrostat
Transactcons on
Industry Appl cat ons, 6, pp. 526-530.
Kjeldsen, P., Barlaz, M.A., Rooker, A.P., Baun, A., Led n, A. and Chr stensen,
T.H., 2002. Present and long-term compos t on of MSW landf ll leachate: A
111
Page 131
rev ew. Cr t cal Rev ews n Env ronmental Sc ence and Technology, 32, pp.
297-336.
Kofoworola, O.F., 2007. Recovery and recycl ng pract ces n mun c pal sol d
waste management n Lagos, N ger a. Waste Management, 27, pp. 1139-1143.
Lawver, J.E. and Dyrenforth, W.P., 1973. Electrostat c separat on n

electrostat cs and ts appl cat ons. New York: W ley.
Le Man, H., Behera, S.K. and Park, H.S., 2010. Opt m zat on of operat onal
parameters for ethanol product on from Korean food waste leachate.
Internat onal Journal of Env ronmental Sc ence and Technology, 7, pp. 157-
164.
Lee, C.-Y. and Lee, Z.-J., 2012. A novel algor thm appl ed to class fy
unbalanced data. Appl ed Soft Comput ng, 12(8), pp. 2481-2485.
L , J., Lu, H., Guo, J., Xu, Z. and Zhou, Y., 2007. Recycle technology for
recover ng resources and products from waste pr nted c rcu t boards.
Env ronmental Sc ence and Technology, 41, pp. 1995-2000.
L n, C.S.K., Pfaltzgraff, L.A., Herrero-Dav la, L., Mufobu, E.B., Abderrah m,

S., Clark, J.H. et al., 2013. Food waste as a valuable resource for the
product on of chem cals, mater als and fuels. Current s tuat on and global
perspect ve, Energy & Env ronmental Sc ence, 6(2), pp. 426-464.
L n, Y.C., Chen, Y.F., Wang, D.A. and Lee, H.S., 2009. Opt m zat on of
mach n ng parameters n magnet c force ass sted EDM based on Taguch
method. Journal of Mater als Process ng Technology, 7, pp. 3374-3383.
L ndmark, J., Leksell, N., Schnürer, A. and Thor n, E., 2012. Effects of
mechan cal pre-treatment on the b ogas y eld from ley crop s lage. Appl ed
Energy, 97, pp. 498-502.
Lu, H., L , J., Guo, J., Xu, Z., 2008. Movement behav or n electrostat c
separat on: Recycl ng of metal mater als from waste pr nted c rcu t board.
Journal of Mater als Process ng Technology, 197, pp. 101-108.
Magdalena, V. and Dana, A., 2014. Can vegetat on nd cate a mun c pal sol d
waste landf ll’s mpact on the env ronment? Pol sh Journal of Env ronmental
112
Page 132
Stud es, 23(2), pp. 501-509.
Mahapatra, S.S., Patna k, A. and Satapathy, A., 2008. Taguch method appl ed
to parametr c appra sal of eros n behav our of GF-re nforced polyester
compos tes. Wear, 1(2), pp. 214-222.
Manaf, L.A., Samah, M.A.A., Zukk , N.I.M., 2009. Mun c pal Sol d Waste
management n Malays a: Pract ce and challenges. Waste Management, 29, pp.
2902-2906.
Masu , N., 1982. Electrostat c separat on for removal from green tea of stems
and from food of mpur t es. Proceed ngs of the IEJ, 6(3), pp.159-162.
McM lan, J.D., 1997. B oethanol product on: status and prospects. Renewable
Energy, 10, pp. 295-302.
Meen-Chee, H. and Narayanan, S., 2006. Restor ng the sh ne to a pearl:

Recycl ng behav our n Penang, Malays a. Development and Change, 37, pp.
1117-1136.
Mohabuth, N. and M les, N., 2005. The recovery of recyclable mater als from
waste electr cal and electron c equ pment (WEEE) by us ng vert cal v brat on
separat on. Resources, Conservat on and Recycl ng, 45(1), pp. 60-69.
Moukamnerd, C., Kawahara, H. and Katakur, Y., 2013. Feas b l ty study of

ethanol product on from food Wastes by consol dated cont nuous sol d-state
fermentat on. Journal of Susta nable B oenergy Systems, 3, pp. 143-148.

Murata, Y., Masuc separator.
f eld electrostat , N., M yazawa, S. and Inculet,
In: Conference paper I.I., 1982.
of the IEJ,Non-un form AC
pp. 293-294.
Nat onal Sol d Waste Management Department, 2013. [onl ne] Ava lable at:
<http://www.kpkt.gov.my/jpspn_en_2013/ma n.php> [Accessed 11 June
2013].
Parf tt, J., Barthel, M. and Macnaughton, S., 2010. Food waste w th n food
supply cha ns: quant f cat on and potent al for change to 2050. Ph losoph cal
Transact ons of the Royal Soc ety, 365, pp. 3065-3081.
113
Page 133
Per athamby, A., Ham d, F.S. and Kh dz r, K., 2009. Evolut on of sol d waste
management n Malays a: Impacts and mpl cat ons of the sol d waste b ll,
2007. Journal of Mater al Cycles and Waste Management, 11, pp. 96-103.
Rav shankar, S.A. and Kolla, H., 2009. Chem cal enhanced electrostat c
separat on. In: 7th Internat onal Heavy M neral Conference 'What next', The
South Afr can Inst tute of M n ng and Metallurgy, pp.203-206.
Renou, S., G vaudan, J.G., Poula n, S., D rassouyan, F. and Moul n, P., 2008.
Landf ll leachate treatment: Rev ew and opportun ty. Journal of Hazardous
Mater als, 150, pp. 468-493.
Rezouga, M., T lmat ne, A., Ou dd r, R. and Medles, K., 2009. Exper mental
modell ng of the breakdown voltage of a r us ng des gn of exper ments.
Advances n Electr cal and Computer Eng neer ng, 9, pp. 41-45.
Ruan, J. and Xu, Z., 2012. Approaches to mprove separat on eff c ency of
eddy current separat on for remov ng alum n um from waste toner cartr dges.
Env ronmental Sc ence and Technology, 46, pp. 6214-6221.
Saeed, M.O., Hassan, M.N. and Mujeebu. M.A., 2009. Assessment of

mun c pal sol d waste generat on and recyclable mater als potent al n Kuala
Lumpur, Malays a. Waste Management, 19, pp. 2209-2213.
Samsud n, M.D.M. and Don, M.M., 2013. Mun c pal sol d waste management
n Malays a: Current pract ces, challenges and prospect, Jurnal Teknolog ,
62(1), pp. 95-101.
Samu la, A., Urs, A., Iuga, A., Morar, R., Aman, F. and Dascalescu, L., 2005.
Opt m zat on of corona electrode pos t on n roll-type electrostat c separators.
IEEE Transact ons on Industry Appl cat ons, 41(2), pp. 527-534.
Schütze, A., James, Y.J., Steven, E.B., Park, J.Y., Gary, S.S. and Robert, F.H.,
1998. The atmospher c pressure plasma jet: A rev ew and compar son to other
plasma sources. IEEE Transact ons on Plasma Sc ence, 26(6), pp. 1685-1694.
Senth lkumar, K., Senth kumaar, J.S. and Sr n vasan, A., 2013. Reduc ng
surface roughness by opt m s ng the turn ng parameters. The South Afr can
Journal of Industr al Eng neer ng, 24(2), pp. 78-87.
114
Page 134
Sr nu Na k, S. and Pyd Setty, Y., 2014. Opt m zat on of parameters us ng

response surface methodology and genet c algor thm for b olog cal
den tr f cat on of wastewater. Internat onal Journal of Env ronmental Sc ence
and Technology, 11, pp. 823-830.
Stuart, T., 2009. Waste - Uncover ng the global food scandal. New York: W.W.
Norton and Company.
Tagar s, E., Sot ropoulou, R.E.P., P l n s, C. and Halvadak s, C.P., 2003. A

methodology to est mate odors around landf ll s tes: The use of methane as an
odor ndex and ts ut l ty n landf ll s tt ng. Journal of the A r and Waste
Management Assoc at on, 53, pp. 629-634.
Taguch , G., 1986. Introduct on to qual ty eng neer ng: des gn ng qual ty nto
products and processes. Tokyo: As an Product v ty Organ zat on.
Taherzadeh, M.J. and Kar m , K., 2008. Pretreatment of l gnocellulos c wastes

to mprove ethanol and b ogas product on: a rev ew. Internat onal Journal of
Molecular Sc ences, 9, pp. 1621–1651.
Tammemag , H.Y. and Tammemag , H., 1999. The waste cr s s: landf lls,
nc nerators, and the search for a susta nable future. New York: Oxford
Un vers ty Press.

Toraguch , M. and Haga, K., 1982. Electrostat c separat on of coal.
Proceed ngs of the IEJ, 6(3), pp.139-147
Van Wyk, J.P.H, 2001. B otechnology and the ut l zat on of b owaste as a

resource for b oproduct development. Trends n B otechnology, 19(5), pp. 172-
177.
Vavourak , A.I., Vol ot , V. and Kornaros, M.E., 2014. Opt m zat on of thermo-
chem cal pretreatment and enzymat c hydrolys s of k tchen wastes. Waste
Management, 34, pp. 167-173.
Ve t, H.M., D ehl, T.R., Salam , A.P., Rodr gues, J.S., Bernardes, A.M. and
Tenor o, J.A.S., 2005. Ut l zat on of magnet c and electrostat c separat on n
the recycl ng of pr nted c rcu t boards scrap. Waste Management, 25, pp. 67-
74.
Vlad, I., Campeanu, A., Enache, S. and Petropol, G., 2014. Operat on
115
Page 135
character st cs opt m zat on of low power three-phase asynchronous motors.

Advances n Electr cal and Computer Eng neer ng, 14, pp. 87-92.
Wagenaars, E., 2006. Plasma breakdown of low-pressure gas d scharges. PhD,

Techn sche Un vers te t E ndhoven, E ndhoven.
Wang, Q., Ma, H., Xuc, W., Gong, L., Zhang, W. and Zou, D., 2008. Ethanol
product on from k tchen garbage us ng response surface methodology.
B ochem cal Eng neer ng Journal, 39, pp. 604-610.
We tz, K.A., Thorneloe, S.A., N shtala, S.R., Yarkosky, S. and Zannes, M.,
2002. The mpact of mun c pal sol d waste management on greenhouse gas
em ss ons n the Un ted States. Journal of the A r and Waste Management
Assoc at on, 52(9), pp. 1000-1011.
Whalen, S.C., Reeburgh, W.S. and Sandbeck, K.A., 1990. Rap d methane
ox dat ons n a landf ll cover so l. Appl ed and Env ronmental M crob ology,
56, pp. 3405-3411.
Wu, J., L , J. and Xu, Z., 2008. Electrostat c separat on for mult -s ze granule
of crushed pr nted c rcu t board waste us ng two-roll separator. Journal of
Hazardous Mater als, 159, pp. 230-234.
Xu, C., Chen, W. and Hong, J., 2014. L fe-cycle env ronmental and econom c
assessment of sewage sludge treatment n Ch na. Journal of Cleaner
Product on, 67, pp. 79-87.
Xu, Z.M., L , J., Lu, H.Z. and Wu, J., 2009. Dynam cs of conduct ve and
nonconduct ve part cles under h gh-voltage electrostat c coupl ng f elds.
Sc ences n Ch na Ser es E: Technolog cal Sc ences, 52(8), pp. 2359-2366.
Xue, M., L , J. and Xu, Z., 2012. Env ronmental fr endly crush-magnet c
separat on technology for recycl ng metal-plated plast cs from end-of-l fe
veh cles. Env ronmental Sc ence and Technology, 46, pp. 2661-2667.
Yamane, L.H., de Moraes, V.T., Esp nosa, D.C.R. and Tenór o, J.A.S., 2011.
Recycl ng of WEEE: Character zat on of spent pr nted c rcu t boards from
mob le phones and computers. Waste Management, 31(12), pp. 2553-2558.
Younes, M., T lmat ne, A., Medles, K., Rahl , M., and Dascalescu, L., 2007.
116
Page 136
Numer cal model ng of conduct ve part cle trajector es n roll-type corona-

electrostat c separators. IEEE Transact ons on Industry Appl cat ons, 43(5),
pp. 1130-1136.
Zare , M., N ae , A., Salar , D. and Khataee, A., 2010. Appl cat on of response
surface methodology for opt m zat on of perox -coagulat on of text le dye
solut on us ng carbon nanotube-PTFE cathode. Journal of Hazardous
Mater als, 173 (1–3), pp. 544–551.

117
Page 137
LIST OF ORIGINAL PUBLICATIONS
Journal publ cat ons:
1. KoonChun La , SooK ng L m, PehCh ong Teh, K mHo Yeap (2015).

Model ng and opt m zat on of b owaste segregat on from MSW: An
approach of landf lls treatment and reduct on, Research Journal of
Appl ed Sc ences, Eng neer ng and Technology. (Scopus ndexed, n
press)
2. KoonChun La , SooK ng L m, PehCh ong Teh (2015). Opt m zat on

of electrostat c separat on process for max m z ng b owaste recovery
us ng Taguch method and ANOVA, Pol sh Journal of Env ronmental
Stud es, 24(3). DOI: 10.15244/pjoes/30927 (ISI ndexed, n press)

3. KoonChun La , SooK ng L m, PehCh ong Teh, K mho Yeap (2014).

Key factors of the electrostat c separator for sol d waste segregat on,
Sc enceAs a. (ISI ndexed, n press)
4. KoonChun La , SooK ng L m, PehCh ong Teh (2014). Electrostat c

separat on n recover ng the food from sol d waste, World Appl ed
Sc ences Journal, 32(7): 1296-1303. (Scopus ndexed)
Conference papers:
1. KoonChun La , SooK ng L m, PehCh ong Teh (2014). Opt m zat on

of the electrostat c separator for food waste segregat on, Internat onal
Conference on Mathemat cs, Eng neer ng and Industr al Appl cat ons
(ICoMEIA 2014), Penang, Malays a, 28 – 30 May 2014. (Paper ID 211)
2. KoonChun La , SooK ng L m, PehCh ong Teh (2013). Appl cat on

and opt m zat on of a plasma d scharge bubble jet for water treatment,
Proceed ngs of 6th Internat onal Eng neer ng Conference, Energy and
Env ronment (ENCON 2013), Sarawak, Malays a, 1 – 4 July 2013,
ISBN: 978-981-07-6059-5, pp. 187 – 192.
3. KoonChun La , SooK ng L m, PehCh ong Teh, K mHo Yeap,

YunThung Yong (2013). Study of a plasma d scharge bubble jet:
Des gn and opt m zat on n water treatment, 8th Internat onal
118
Page 138
Conference on Mult phase Flow (ICMF), Jeju, South Korea, 26 – 31

May 2013.
4. La Koon Chun, L m Soo K ng, A ssa Boudjella and Yeap K m Ho

(2011). Protect on aga nst human body electrostat c d scharge n
manufactur ng, Advanced Process and System n Manufactur ng
(APSIM), Kuala Lumpur, Malays a, 14 – 15 December 2011, Paper
AP62.

119

Design and Development of An Electrostatic Separator For Waste Segregation Lai Koon Chun Doctor of Philosophy (Engineering) Facu

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Design and Development of An Electrostatic Separator For Waste Segregation Lai Koon Chun Doctor of Philosophy (Engineering) Facu

Uploaded by

Copyright:

Available Formats

29.11.

http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf dosyasının HTML sürümüdür. Googlet aradığı belgeler n HTML

DESIGN AND DEVELOPMENT OF AN ELECTROSTATIC

LAI KOON CHUN

DOCTOR OF PHILOSOPHY (ENGINEERING)

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 1/147

FACULTY OF ENGINEERING AND GREEN

DESIGN AND DEVELOPMENT OF AN ELECTROSTATIC

LAI KOON CHUN

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 2/147

A thes s subm tted to the Department of Electron c Eng neer ng,

DESIGN AND DEVELOPMENT OF AN ELECTROSTATIC

Household sol d waste s commonly d sposed of n landf lls, generat ng

hazardous by-products such as leachate and methane gas from ts organ c

contents. Thus, organ cs n landf lls wh ch cons st ma nly of food matter

should be reduced. In th s study, an electrostat c separator was des gned and

of the electrostat c separat on were d scussed by referr ng to force models.

both system and no se factors. The hard-to-control no se factors, .e. s ze and

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 3/147

factors are key factors that should be properly selected so as to avo d

separat on process. A stat st cal analys s w th central compos te des gn was

nteract ve effects of ndependent factors on separat on performance were

be 60 rpm rotat onal speed, 30 kV appl ed voltage and 54 mm electrodes

nterval on part cles w th 4.0 mm s ze and 20 % water content. Under these

study concluded that the electrostat c separat on could be an effect ve pre-

treatment alternat ve n deal ng w th leachate and methane problems caused by

landf lled organ c wastes.

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 4/147

Acknowledgements are due to my superv sors, Assoc ate Professor Dr. L m

th s study. I am part cularly grateful to Assoc. Prof. Dr. L m, w thout whose

Green Technology, Un vers t Tunku Abdul Rahman for be ng support ve and

understand ng w th the d ff cult es n work ng and study ng at the same t me

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 5/147

throughout all these years.

I also would l ke to ded cate th s thes s to my parents and colleagues. Many

thanks are due to Professor Andrew Raga Henry R g t for h s encouragement

Th s d ssertat on/thes s ent tled “DESIGN AND DEVELOPMENT OF AN

ELECTROSTATIC SEPARATOR FOR WASTE SEGREGATION” was

Un vers t Tunku Abdul Rahman.

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 6/147

FACULTY OF ENGINEERING AND GREEN TECHNOLOGY

UNIVERSITI TUNKU ABDUL RAHMAN

SUBMISSION OF FINAL YEAR PROJECT /DISSERTATION/THESIS

Development of an Electrostat c Separator for Waste Segregat on” under the

I understand that Un vers ty w ll upload softcopy of my f nal year project /

be made access ble to UTAR commun ty and publ c.

*Delete wh chever not appl cable

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 8/147

Name LAI KOON CHUN N

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 9/147

2.0 LITERATURE REVIEW 8

3.0 MATERIALS AND METHODS 44

4.0 RESULTS AND DISCUSSION 62

5.0 CONCLUSION AND FURTHER WORK 105

LIST OF REFERENCES 108

webcache.googleusercontent.com/search?q=cache:http://epr nts.utar.edu.my/1661/1/XXX-2015-1004870-1.pdf 11/147

1.1 Food waste n d fferent reg ons of the world 1

1.2 Sol d waste compos t on (wt %) n Malays a 3

2.1 The voltage-current character st c between parallel 12

2.2 Type of corona d scharges (a) pass ve corona (b) 15

2.3 Typ cal electrostat c separat on techn ques (a) 20

2.4 Cyclone electrostat c separator 22