&EPA
United States
Environmental Protection
Agency
Municipal Environmental Research EPA-600/2-80 1 15
Laboratory August 1980
Cincinnati OH 45268
Research and Development
Significance of Size
Reduction in Solid
Waste Mangement
Volume 2
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-80-115
August 1980
SIGNIFICANCE OF SIZE REDUCTION IN
SOLID WASTE MANAGEMENT
Volume 2
by
George M. Savage
George J. Trezek
University of California
Berkeley, California 94607
Grant No. R-805414-010
Project Officer
Carl ton C. Wiles
Solid and Hazardous Waste Research Division
Municipal Environmental Research Laboratory
Cincinnati, Ohio 45268
MUNICIPAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
DISCLAIMER
This report has been reviewed by the Municipal Environmental Research
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection Agency* nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
ii
-------
FOREWORD
The Environmental Protection Agency was created because of increasing public
and government concern about the dangers of pollution to the health and
welfare of the American people. Noxious air, foul water, and spoiled land
are tragic testimony to the deterioration of our natural environment. The
complexity of that environment and the interplay between its components
require a concentrated and integrated attack on the problem.
Research and development is that necessary first step in problem solution
and it involves defining the problem, measuring its impact, and searching
for solutions. The Municipal Environmental Research Laboratory develops
new and improved technology and systems for the prevention, treatment, and
management of wastewater and solid and hazardous waste pollutant discharges
from municipal and community sources, for the preservation and treatment
of public drinking water supplies, and to minimize the adverse economic,
social, health, and aesthetic effects of pollution. This publication is
one of the products of that research; a most vital communications link
between the researcher and the user community.
This report is the second in a three-volume series of comprehensive studies
dealing with the size reduction of solid waste. In this report (Volume 2),
raw municipal solid waste, air-classified light fraction, and screened
light fraction were used in a number of tests simulating single- and
multiple-stage size reduction. The basic test data are generalized so
that the characteristic particle size and energy consumption could be
predicted. Various hardfacing materials and their ability to perform
with the different solid waste materials are also tested.
Francis T. Mayo, Director
Municipal Environmental Research
Laboratory
iii
-------
ABSTRACT
This report is the second in a three-volume series of comprehensive
studies dealing with the size reduction of solid waste. Tests dealing with
solid waste shredding were conducted at the University of California, Berke-
ley Solid Waste Processing Laboratory located at the Richmond Field Station
in Richmond. The objects of these tests were (1) to gain a better under-
standing of the factors controlling the interrelated variables of feed rate,
power consumption, and hammer wear, and (2) to determine how to produce the
final product with a minimum expenditure of energy and machine wear.
Volume 1 identifies and quantifies the variables governing refuse size
reduction and then gives a preliminary formulation and evaluation of these
relationships. Volume 2 addresses the basic questions that were unanswered
in the first report: What size grates, how many hammers, what type of hard-
facing, and how many stages of shredding should be used for optimum machine
wear and energy consumption.
Volume 3 expands the discussions presented in the first two volumes and
incorporates the data obtained in the large-scale shredder tests reported by
Cal Recovery Systems, Inc.
In this report (Volume 2), raw municipal solid waste, air-classified
light fraction, and screened light fraction were used in a number of tests
simulating single-and multiple-stage size reduction. Tests were performed
with a nominally rated, 10-ton/hr swing hammermill and a small, high-speed
fixed hammer shredder. The basic test data are generalized so that the
characteristic particle size and energy consumption could be predicted.
Various hardfacing materials and their ability to perform with the different
solid waste materials are also tested. Furthermore, a comprehensive, compu-
terized comminution model of a size-discrete, time-continuous nature has
been developed to predict size distribution and energy consumption.
This report was submitted in fulfillment of Contract No. R-805414-010
by the University of California, Berkeley, under the sponsorship of the U.S.
Environmental Protection Agency.
iv
-------
CONTENTS
Foreword iii
Abstract 1v
Figures vii
Tables xl
Abbreviations and Symbols xiii
1. Introduction 1
2. Materials, Equipments and Methods 3
Characteristics of Raw Municipal Solid Waste 3
Size Reduction Equipment . . 3
Air Classifier and Screen 6
Moisture Content and Size Distribution Analysis 9
Measurement of Power and Energy Consumption 11
Data Collection Procedure 13
Procedure for Obtaining Residence Time . 14
3. Shredding Studies 16
Municipal Solid Waste (Gruendler Hammermill) Studies .... 16
Air-Classified Light Fraction Studies 27
Screened Light Fraction Studies 38
4. Wear Studies 48
Hardfacing Materials and Application 48
Experimental Procedure 49
Hardness Measurements ........ 53
Wear Measurements 53
Results 54
Conclusions 62
5. interrelation of the Comminution Parameters 68
Representative Particle Size 68
Degree of Size Reduction 70
Energy 70
Residence Time 79
Mill Holdup 81
Hammer Wear 84
-------
6. Comminution Models 91
Size-discrete, Time-discrete Models 92
Size-discrete, Time-continuous Model 95
Implementation of the Model 100
Residence Time Distribution 106
Simulation Results 109
Concluding Remarks 114
7. Unit Operation and System Design Considerations 121
Grate Spacing Considerations 121
Torque Relations 123
Unit Process Optimization 125
System Optimization 132
References 137
Appendix. Comminution Computer Model 139
vi
-------
FIGURES
Number Page
1 Size distribution of raw MSW, Richmond, Calif 4
2 Test facility for the Gruendler swing hammer-mi 11 shredder . . 5
3 W-W fixed hammermill positioned to shred either air-
classified light fraction or screened light fraction .... 7
4 Rotor and hammers of W-W hammermill 8
5 Grate of W-W hammermill 8
6 Schematic of power monitoring systems 12
7 Energy consumption for size reduction of raw MSW as a
function of grate opening size (Gruendler hammermill) .... 22
8 Energy required to reduce raw MSW to a given characteristic
product size with a Gruendler hammermill 23
9 Energy required to reduce raw MSW to a given nominal product
size with a Gruendler hammermill 24
10 Relation of shredded raw MSW product size to size of grate
openings (Gruendler hammermill) . . 25
11 Sample of raw MSW shredded in a Gruendler hammermill without
grate bars 26
12 Size distribution of raw MSW after multiple shredding with
grate openings of 2.5 and 1.8 cm in the Gruendler hammermill 29
13 Size distribution for ACL.F feed after primary and secondary
shredding in the Gruendler hammermill 32
14 Size distribution of the ACLF after single-stage shredding
with the W-W hammermill (Run 1) . 34
15 Size distribution of the SLF after progressive shredding with
the W-W hammermill 42
vii
-------
16 Size distribution of the SLF after multiple shredding with
2.5-cm (1.0-in.) grate openings in the Gruendler hammermill . 45
17 Size distribution of the SLF after multiple shredding with
1.8-cm (0.7-in.) grate openings in the Gruendler hammermill . 46
18 Bare manganese steel hammers before hardfacing 51
19 Hammers used for shredding raw MSW . . . 51
20 Hammer wear as a function of grate opening: raw MSW, Gruendler
hammermill 56
21 Hammer wear as a function of average characteristic product
size: raw MSW, Gruendler hammermill 57
22 Hammer wear as a function of average nominal product size: raw
MSW, Gruendler hammermill 59
23 Hammer wear as a function of hardness for several different
grate openings: raw MSW, Gruendler hammermill 60
24 Evidence of chipping on hammers tipped with hardfacing alloys
exceeding Rc 50 61
25 Closeup of hammers showing chipping of hardfacing weld
deposits on hammer tips 61
26 Hammer wear as a function of grate opening: screened light
fraction, W-W hammermill 64
27 Hammer wear as a function of hardness for several different
grate openings: SLF, W-W hammermill 65
28 Hammer wear as a function of average characteristic product
size: SLF, W-W hammermill 66
29 Mass fraction representation of the size distribution for raw
MSW . 69
30 Nominal size as a function of characteristic size for shredded
raw MSW, Gruendler hammermill 71
31 Nominal size as a function of characteristic size for shredded
ACLF and SLF, W-W hammermill 72
32 Degree of size reduction as a function of the ratio of grate
spacing to characteristic feed size for raw MSW, ACLF and SLF 73
33 Specific energy as a function of degree of size reduction for
raw MSW, ACLF, and SLF 74
viii
-------
34 Degree of size reduction as a function of the ratio of grate
opening to characteristic feed size for raw MSW 76
35 Degree of size reduction as a function of the ratio of grate
opening to characteristic feed size for ACLF 77
36 Degree of size reduction as a function of the ratio of grate
opening to characteristic feed size for SLF 78
37 Schematic representative of milling process . 79
38 Residence time distribution for size reduction of raw MSW . . 80
39 Normalized residence time distribution for size distribution
for size reduction of raw MSW 82
40 Flow rate as a function of mill holdup for raw MSW 83
41 Characteristic particle size as a function of flow rate for
raw MSW 85
42 Net power as a function of holdup for raw MSW 86
43 Net power as a function of moisture content for raw MSW ... 87
44 Hammer wear as a function of degree of size reduction for raw
MSW and SLF 89
45 Hammer wear as a function of D0/F0 for raw MSW and SLF .... 90
46 Diagram of open circuit milling 99
47 Diagram of closed circuit milling 99
48 Nonlinear optimization flow chart for parameter estimation . . 104
49 Flow chart for mill simulation calculations 105
50 Residence time distribution 107
51 Normalized residence time distribution . . 108
52 Complementary residence time function 110
53 Characteristic size and variation with mean residence time . . Ill
54 Variation of K* flow rate 116
55 Variation of initial selection function with flow rate .... 117
56 Variation of initial breakage function with flow rate .... 118
IX
-------
57 Comparison of predicted and actual measured size distributions 119
58 Size reduction of raw MSW extended to several different grate
openings 124
59 Degree of size reduction and specific energy requirement for
shredding raw MSW as a function of D0/F0 127
60 Possible configurations for resource recovery unit operations . 133
61 Energy pathway for hypothetical RDF processing trains ...... 135
-------
TABLES
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Maximum Mass of Shredded Refuse For Efficient Screening Using
Using the Sweco Vibratory Screens (4.8-cm grate openings) . . .
Feed Rates Used in the Size Reduction of Raw MSN with the
Size Reduction of Raw MSW Using 4.8-cm (1.9-in. ) Grate Openings
Size Reduction of Raw MWS using 2.5-cm (1.0-in. ) Grate Openings
with the Gruendler Hammermill
Size Reduction of Raw MSW using 1.8-cm (Q.7-in. ) Grate Openings
Multiple Shredding of Raw MSW with the Gruendler Hammermill . . .
Multiple Shredding of ACLF with the Gruendler Hammer-mill ....
Single-Stage Shredding of ACLF with the W-W Hammermill
Multiple Shredding of the ACLF with the W-W Hammermill
Progressive Shredding of ACLF with the W-W Hammermill
Single-Stage Shredding of the SLF for a Given Feed Size and Dif-
ferent Grate Openings with the W-W Hammermill
Progressive Shredding of SLF with the W-W Hammermill
Multiple Shredding of SLF with the Gruendler Hammermill
Comparison of Hammer and Hardfacing Materials Used for Shredding
Raw MSW ......
Comparison of Hardfacing Materials Used for Shredding SLF ....
Wear Data for Various Hardfacing Materials Used for Shredding Raw
MSW
Page
11
17
18
20
21
28
31
33
36
37
39
41
44
50
52
55
xi
-------
17 Wear Data for Various Hardfacing Materials Used for Shredding
SLF 63
18 Comparison of Degree of Size Reduction and Specific Energy
for Different Solid Waste Fractions 75
19 Summary of Optimization Results 115
20 Single- and Double-Stage Shredding of Raw MSW in the Gruendler
Hammermill 128
21 Single- and Double-Stage Shredding of ACLF in the W-W Hammermill 130
22 Single- and Double-Stage Shredding of SLF in the W-W Hammermill . 131
23 Mass and Energy Balance on Processing Trains for Recovery of
RDF 134
xii
-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
ACLF — air-classified light fraction
amp — ampere
cm — centimeter
fps — feet per second
ft — foot
g — gram
hp — horsepower
h — hour
J — joule
kg — kilogram
kW — kilowatt
kwh — kilowatt-hour
lb — pound
m ~ meter
MC — moisture content, reported as quotient of the difference of the
initial weight minus the dry weight divided by the initial
weight
min — minute
MJ — megajoules (106 joules)
MSW — municipal solid waste
RC — Rockwell C hardness
RDF — refuse derived fuel
rpm — revolutions per minute
s — second
SLF — screened light fraction
W — watt
Wh — watt-hour
SYMBOLS
D0 — characteristic linear dimension of grate opening or grate spac-
ings; the minimum dimension for rectangular grate openings and
the diameter for circular openings
E0 — specific energy or net energy required for processing, generally
reported as megajoules per metric ton or kilowatt hours per ton
of material processed on a dry-weight basis
F0 — characteristic size of the feed material
— nominal size of the feed material
X0 — characteristic size of the product material
XQQ — nominal size of the product material
Z0 — degree of size reduction
xili
-------
SECTION 1
INTRODUCTION
This report is the second in a three-volume series of comprehensive
studies dealing with the size reduction of solid waste. Results are based
on shredder testing conducted at the University of California, Berkeley,
Solid Waste Processing Laboratory at Richmond, California. The motivation
for studying the behavior of shredders began in the early 1970's, when
facilities were being considered, and in some cases constructed, to process
solid wastes by a number of alternative methods such as volume reduction for
landfilling and for material and energy recovery. In each of these process
schemes, one or more shredders became an integral part of the system. Typi-
cally, machines designed for comminuting homogeneous, brittle materials were
being used to reduce the size of refuse that was heterogeneous and approxi-
mately 75 percent nonbrittle. From some initial field experience, it became
apparent that the factors controlling the interrelated variables of feed
rate, power consumption, and wear had to be understood if this equipment was
to be successfully operated.
Volume 1 of this report addressed the above issues. Basically, the
first results indicated how the variables governing refuse size reduction
were initially identified and then quantified. Preliminary relationships
governing the behavior of these variables were then formulated and evaluated
consistent with available data. However, additional questions still re-
mained unanswered and in effect provided the need for continued study.
The additional needs can be summarized as follows. The initial work
could not completely answer certain questions such as why the flow rate
affected the product size distribution, why moisture content affects the
energy consumption, and how all these variables actually affected size dis-
tributions, energy consumption, and wear. Furthermore, though the once-
through comminution model could predict the product size distribution, it
could not provide an adequate physical distribution of the physical pro-
cess. Other considerations such as the residence time distribution, mill
holdup, and mill geometry needed to be identified, evaluated, and formulated
into a working model. Another area of concern was the energy consumption
required in the so-called fine-grinding process. Essentially no information
was available for design and evaluation of processing that required particle
sizes finer than 2.5 cm (1 in.). Some evidence indicated that excessive
energy and wear would result if a unit operation producing a fine particle
size were used. Also lacking was the ability to extrapolate information to
other processing regimes. For example, information obtained for various
ranges of flow rates, grate spacings, and power consumptions needed to be
generalized to the point that it would be useful to predict the comminution
events in a significantly larger unit operation. The controversy involved
-------
in staging various shredder unit operations also needed to be addressed.
For example, it was not clear whether it would be more prudent to have a
single-stage shredding operation produce the final particle size or to
employ two stages to achieve the end result. Furthermore, the problems
involved in knowing how to structure each stage needed to be considered —
that is, what size of grates, how many hammers, etc., were needed for each
stage in order to produce the final product,with the minimum expenditure of
energy and machine wear.
The present report addresses all of these basic questions. Raw refuse,
air-classified light fraction (ACLF), and screened air-classified light
fraction (SLF) were used in a number of tests simulating single- and
multiple-stage size reduction. Tests were performed with a nominally rated,
10-ton/hr swing hammermill and a small, high-speed fixed hammer shredder.
The basic test data have been arranged and generalized enough to predict the
characteristic particle size and energy consumption of other unit operation
arrangements. In addition, the wear problem has been treated through the
investigation of various hardfacing materials and their ability to perform
with the different solid waste materials tested. Furthermore, a comprehen-
sive, computerized comminution model of a size-discrete, time-continuous
nature has been developed to predict size distribution and energy
consumption.
The information reported here consists mainly of basic data obtained
from testing. A discussion of an improved and analytical model is given in
Section 6, and Section 7 assembles the data so that predictions can be
made. These discussions are expanded in Volume 3 with the incorporation of
the data obtained in the large-scale shredder tests reported by Cal Recovery
Systems, Inc., Richmond, California.
-------
SECTION 2
MATERIALS, EQUIPMENT, AND METHODS
Scientific studies must be able to modify and control systematically
the conditions and variables germane to a particular investigation. Consis-
tent with this requirement, a size-reduction facility capable of operating
on a scale resembling a commercial installation and yet able to produce
laboratory-quality data was developed to conduct the various comminution
studies sponsored by the U.S. Environmental Protection Agency (EPA) since
1972. The waste processing facility is located at the University of Cali-
fornia (Berkeley) Richmond Field Station. Although the initial emphasis of
the laboratory was on size reduction, the facility was later expanded to
include complete front-end dry processing as well as certain types of
back-end processing for material and energy recovery. The type of proces-
sing equipment and instrumentation contained in this facility, and the char-
acteristics of most of the raw waste processed, are discussed in this
section.
CHARACTERISTICS OF RAW MUNICIPAL SOLID WASTE
Most of the solid waste processed is residential packer truck material
obtained through the courtesy of the Richmond Sanitary Service. The company
delivered municipal solid waste (MSW) and removed unwanted processed materi-
als at no cost. Basic composition of the raw MSW is as follows, as a per-
cent of dry weight:
Fe . 7.3
Glass 10.2
Aluminum 0.7
Paper 43.2
Plastic 4.5
Other 33.5
Figure 1 shows the size distribution of the raw MSW. Most of the size
distribution curve occupies between 2.5 and 25.4 cm (1.0 and 10.0 in.).
The range of characteristic particle sizes extends from 11.7 to 15.2 cm
(4.6 to 6.0 in.). The corresponding nominal particle sizes are in the
interval between 18.8 and 33 cm (7.4 and 13.0 in.).
SIZE REDUCTION EQUIPMENT
The data reported in the present series of comminution studies were
obtained from either the Gruendler swing hammermill shredder (Figure 2)
or a W-W Grinder Company, high-speed fixed hammermill shredder (Figures
-------
100 f-
O Richmond, Calif., 1973
D Richmond, Calif., 1377
Screen Size
j I
1 in.
i i til
10
cm
10
i 1
40
Figure 1. Size distribution of raw MSW,
Richmond, Calif.
-------
Figure 2. Test facility for the Gruendler
swing hammermill shredder.
-------
3, 4 and 5) commonly used for compost size reduction. This machine is dri-
ven at 3500 rpm by a 5.6 kW (7.5 hp) motor and has a hammer tip speed of 56
m/s (183 fps). Twenty 0.27-kg (0.59-lb) hammers are attached to the rotor
in four rows of five so that approximately 18 to 45 kg/h (40 to 100 lb/h) of
light fraction can be processed for the feed size and grate opening condi-
tions investigated.
Most of the shredding was conducted using the Gruendler machine, which
is a model 48-4 horizontal hammermill equipped with 44 9-kg, free-swinging
manganese steel hammers capable of processing as much as 0.014 mg (15 tons)
of refuse per hour. Refuse is introduced into the grinding chamber through
the top and removed at the bottom. Power is provided by a 186-kW (250-hp),
480-volt, three-phase, 300-amp motor running at 1200 rpm, and it is trans-
ferred to the grinder by way of a variable-speed drive providing discrete
ratios of 1:1, 1.22:1, 1.52:1, and 2.16:1, which correspond to shredder
rotor speeds of 1200, 985, 790, and 555 rpm. Product size may be controlled
by a series of grates placed in the lower portion of the grinding chamber.
The range of the grate spacings is from 1.8 cm to a maximum of 4.8 cm.
Raw refuse is fed to the grinder by means of a 1-m (3.2 ft) piano-
hinged, steel belt conveyor which is 1.2 m (4 ft) wide, 8.5 (27.8 ft) long,
and inclined 20" from the horizontal. The feed conveyor is driven through a
variable-pitch drive sheave and one of two interchangeable drive sheaves.
The arrangement permits a variation of belt speed from 0.6 to 2.3 m/min and
thus provides a means of controlling the feed rate to the grinder. In addi-
tion, the drive system is equipped with a sensor that monitors the current
drawn by the grinder and automatically stops the feed conveyor in the event
of an overload.
The reduced product is removed by way of a second piano-hinged steel
belt conveyor. The belt is 0.9 m (3 ft) wide and has a 2.5 (8.2 ft) hori-
zontal section positioned directly under the grinding chamber. The remain-
der of the belt is 6.8 m long and is inclined at 30 to the horizontal.
AIR CLASSIFIER AND SCREEN
An air classifier and trommel screen were used to prepare the various
forms of light fraction that were used in the size reduction experiments.
The air classifier has a capacity of approximately 4 metric tons/h and
is of the vertical type consisting of a straight rectangular column. Shred-
ded refuse enters near the top of the column through a gate that acts as an
air seal. Lights are moved from the column through a transition piece and
pneumatic ducting, which leads to the centrifugal cyclone collector. Heav-
ies fall through the column and are removed by a belt conveyor.
The overall dimensions of the air classifier are 1.1 m (3.6 ft) wide,
0.9 m (3 ft) deep and 4.6 m (15 ft) high. The rectangular cross section of
the column has a width of 1.1 m (3.6 ft) and a variable depth of 0.2 to 0.5
m (0.65 to 1.6 ft). Average column air velocities can be controlled within
the approximate range of 0.2 to 10 m/s.
-------
Figure 3. W-W hammermill positioned to shred
either air-classified light fraction
or screened light fraction.
-------
Figure 4. Rotor and hammers of W-W hammermill
Figure 5. Grate of W-W hammermill
8
-------
The light fraction trommel screen has a capacity of 2.8 metric tons/h
and consists of a cylindrical screen with 1.3-cm perforations. The screen
has a diameter of 0.9 m (3 ft) and an overall length of 4.6 m (15 ft). In-
clination of the trommel can be varied within the range of 5" to 20" from
the horizontal. In addition, rotational frequency of the trommel can be ad-
justed within the range of 23 to 38 rpm.
Both undersized and oversized fractions are collected and removed from
the trommel screen by belt conveyors. The frictional drive system for the
trommel consists of a 1.5 kW motor, speed reducer, chain drive, and rubber
drive wheel.
The ACLF used in the experiments was recovered using the Gruendler ham-
mermill for size reduction and the laboratory air classifier. The typical
composition of the ACLF on a dry-weight basis was 60.2 percent paper, 5.7
percent plastics, and 34.1 percent other material (namely, organics, dirt,
and glass fines).
The SLF was recovered from the ACLF using the trommel screen in our
facility. The typical composition of the SLF on a dry weight basis was 78.8
percent paper, 7.8 percent plastics, and 13.4 percent other material, most
noticeably organics.
MOISTURE CONTENT AND SIZE DISTRIBUTION ANALYSIS
The characteristic and nominal sizes cited in this study were deter-
mined from the size distribution of samples collected during the experimen-
tal runs. These samples were dried and subsequently screened in a 46-cm
(18-in.) SWECO vibratory screening device for 20 min using various screen
sizes. Data were plotted as the cumulative percent passing versus each
screen size in order to represent the size distribution of each sample. The
characteristic sizes (i.e., 63.2 percent cumulative passing) and the nominal
sizes (i.e., 90 percent cumulative passing) for each sample were determined
from the graphical size distributions.
The samples of shredded refuse collected from the shredder discharge
were dried to constant weight at 22"C (72"F) and 65 percent relative humid-
ity. Consequently, all quantities in the report that use material weights
(e.g., specific energy, cumulative weight percent passing, etc.) are based
on the above criteria for dry weight, commonly referred to as air-dry mois-
ture content. Moisture content 1s reported as the quotient of the differ-
ence of the initial wet weight minus the air-dry weight divided by the ini-
tial wet, weight:
MP (Wet weight) - (A1r dried weight)
NL „ wet weight
The air-dried moisture content of shredded refuse 1s typically 10 percent
less than the bone-dry moisture content.
Samples of shredded refuse for screening analysis were chosen from the
air-dried samples after thorough mixing and sectioning. Samples selected
for drying ranged in size from 2 kg (4.4 Ib) for experimental runs using
-------
4.8-cm (1.9-in.) grate spacings to 0.5 kg (1.1 Ib) for experimental runs us-
ing 0.64-cm (0.25-in.) grate spacings. The minimum sample size for screen-
ing was determined for each grate spacing by successively screening smaller
and smaller sample sizes and plotting their size distributions. A sample
was deemed too small if its size distribution deviated by more than 10 per-
cent from the average value obtained at each screen size when the aggregate
sample was combined and analyzed for an overall size distribution. The
trend that was established for sample sizes showed that they could be de-
creased as the grate spacing decreased.
In this report, the shredded refuse size is described on the basis of
characteristic size, X0 (63.2 percent cumulative passing) and nominal size,
Xgo (90 percent cumulative passing). Each of these values can be determined
from the size distribution curve.
Because of the motion of the rotary screen, it is possible for a large
piece of material to "move" around the screen without tumbling and thus act
as a raft for many smaller particles. For this reason, only one screen at a
time was placed on the screen base. Not only did this step have the effect
of lowering the total inertia of the system and lead to a more vigorous
screening action, but it also allowed the operator to observe the process
and to stir the particles occasionally to avoid development of the rafting
effect. It was also noted that the efficiency of the screen, that is the
rapidity and thoroughness with which all undersized material was passed
through a screen, was highly dependent on the mass of material held on the
screen at any given time. Therefore careful attention was paid to the
amount of material that could initially be placed on a screen before its
efficiency was seriously impaired. The maximum amounts shown to be handled
efficiently at each screen size are given in Table 1. Each batch of mater-
ial was screened for 15 min, after which all oversized material was removed
and a new batch was placed on the screen. This procedure was repeated until
the entire sample had been processed.
Since screening is the most time-consuming of the operations in the
experimental program, and since screen time is directly proportional to the
weight of the sample, considerable effort was expended to determine the min-
imum sample size that would be truly representative of the total product.
To make this determination, a segment of the refuse stream was removed from
the outlet conveyor and was divided into a series of smaller samples in mul-
tiples of 500 g. The samples thus obtained were then dried and screened to
arrive at a size distribution for each. The segments of the size distribu-
tion curve around the 63.2 percent passing mark next were linearized and an
equation was found for each. From the equations, corresponding values for
X0 were calculated and the mean and standard variation found. The calcula-
ted values of X0 were compared to the mean, and the .absolute value of the
errors were converted to percentages and noted. In this manner, it was pos-
sible to show that a sample size of roughly 2000 g was all that was required
to provide a size distribution whose characteristic size was within a few
percentage points of that of parent distribution when 100 percent of the
sample was finer than 3 cm.
10
-------
TABLE 1. MAXIMUM MASS OF SHREDDED REFUSE FOR EFFICIENT
SCREENING USING THE SWECO VIBRATORY SCREENS
(4.8 cm grate openings)
Screen size Maximum mass
(cm) (g)
+ 2.54 250
- 2.54 + 1.5875 250
- 1.5875 + 0.9525 150
-0.9525 + 0.5138 150
- 0.5138 + 0.2743 100
- 0.2743 + 0.1295 50
MEASUREMENT OF POWER AND ENERGY CONSUMPTION
The power monitoring equipment consists of a Scientific Columbus watt/
watt-hour transducer, a divider, a chart recorder, and a counter (Figure
6). The transducer supplies (with ±0.10 percent.accuracy) an analog cur-
rent signal directly proportional to the power (watts) and a digital signal
directly proportional to energy (watt-hours). Because the transducer is
designed to operate from 120-volt, 5-amp inputs, instrument transformers are
used to transform the larger primary currents and voltages to these values.
The analog dc current signal is passed through a precision 1000-ohm resis-
tor, and the resulting voltage drop is recorded using a chart recorder.
Because the transducer is designed to provide exactly 1 milliamp of current
for a secondary power of 1500 watts, the instantaneous power in the primary
circuit is easily found by
P . KI (1500 watts/mill lamp)
where: K = reduction ratio, and
I = dc output current in mi 11 lamps.
The voltage drop recorded by the chart recorder is related to the current
signal by Ohm's Law, namely V = IR. Since 1 milliamp conducted through 1000
ohms produces a voltage drop of 1 volt, the instantaneous power may be
written as
P = KV (1500 watts/volt)
where: P = instantaneous power in watts
K = reduction ration (product of current and potential transformer
ratios), and
V = recorded voltage.
11
-------
VOLTAGE
TAPS
h
MOTOR
TRANSDUCER
CURRENT
TAPS
CHART
RECORDER
DIVIDER
COUNTER
Figure 6. Schematic of power monitoring system,
-------
The digital watt-hour signal is also based on a full load power draw of
1500 watts in such a way that a specified pulse rate is produced for
full-load power. In this particular case, 2160 pulses/h for a full load of
1500 watts was chosen. It corresponds to a value of 0.696 watt-hours (Wh)
per pulse for the secondary circuit. The numerical value for the primary
circuit is found by multiplying 0.696 by the reduction ratio. For the
Gruendler grinder, the product of the tranformer ratios is 144; so each dig-
ital pulse provided by the transducer represents 100 Wh in the primary cir-
cuit. This signal is then passed to a divider, which counts the incoming
pulses and for every n pulse (where n is a specified number between 1 and
99,999) provides two output pulses. One of the pulses (representing lOOn
watts) is passed to the counter, and thereby a record of the total energy
(kWh) is kept. The second pulse triggers an event marker on the chart
recorder and thus provides a permanent record of the energy consumption.
Specific energy, i.e., energy consumption per unit of "dry weight of
reduced material, was determined from the energy data provided by the trans-
ducer and from measurement of the actual tonnage*of material shredded during
the energy measurement interval. That is, specific energy (megajoules
(MJ)/metric ton or kWh/ton) was calculated as the quotient of energy
measured (MJ or kWh) divided by the dry weight of the material reduced dur-
ing the test period. Net energy was used for specific energy calculations
and was determined by subtracting the freewheeling energy from the gross
energy that the transducer measured. Freewheeling power was measured and
recorded before and after all experimental runs*.
DATA COLLECTION PROCEDURE
The shredder and all electronic equipment are turned on at least a half
hour before any data are taken to allow steady-state operating conditions to
be reached. At this time, the first check on freewheeling power consumption
is made. A second check is made at some time during the run, and a third
when all the refuse has been shredded, just before shutting down the
system. The average of the three then is used as the freewheeling power for
the day.
Raw refuse is desposited on the laboratory floor by a packer truck and
loaded onto the feed conveyor by a front-end loader. Although it is not
essential that the feed rate be held constant, the loader operator tried to
maintain an even flow rate once the feed conveyor is turned on. Once load-
ing begins, both feed and product conveyors are started, and shredding is
commenced. Refuse is shredded for approximately 10 min before data collec-
tion is initiated. This time interval allows sufficient time for material
to bufld up in the shredder and for the shredder to reach steady-state
operation.
To obtain a sample of shredded refuse, the output conveyor is stopped
at a specified position, At the same time, the chart recorder pen position
is noted. Since both chart speed and conveyor speed are known, accurate
correlation of energy consumption and particle size is possible. The refuse
sample is obtained by removing all of the shredded refuse from the specified
conveyor section. From the weight of this sample the flow rate, Q, is found
from:
13
-------
Q • cvw
where: v = belt speed on m/min
w = sample mass In kg/m section
of conveyor, and
c a conversion factor from kg/min to metric tons/h = 0.06.
A portion of the sample taken from the output conveyor is weighed,
dried, and reweighed to determine the air-dry moisture content of the sample
before being analyzed for size distribution information. Size distributions
are obtained with the use of two sets of screens. For particle sizes above
2.5 cm, large, manually vibrated screens having a fixed ratio of 2 between
successive screen sizes are used. Material placed on the screens is shaken,
stirred, and shaken again until no further material is observed to fall from
the screen. For sizes below 2.5 cm, a SWECO Vibro-Energy Rotary Screen is
utilized. The SWECO screens make it possible to screen particles as small
as 0.02 cm, although for all practical purposes, screening to 0.13 cm is
usually the lower limit. Because of the somewhat lower degree of control
that could be exerted on the screening action of the SWECO screens, much
closer attention was paid to the mass of material originally placed on the
screen and to the duration of the screening.
PROCEDURE FOR OBTAINING RESIDENCE TIME
Data on the residence time of material within the grinder was obtained
through the introduction of a small amount of tracer into the feed material
and subsequent examination of the output. Because the grinder is equipped
with a hood that extends slightly more than 2 m down the feed conveyor, it
was difficult as well as dangerous to attempt to observe the entrance of the
tracer into the grinder itself. To surmount this problem, tracer material
was weighed and placed in a small bag to which a length of string was
fastened. The bag of tracer was then buried in the feed material. While
the tracer was traveling with the conveyor, a steady pull could be felt on
the string. The time of the sharp tug (i.e., the signal of entry) was duly
noted.
The output was monitored for the appearance of the tracer, which
consisted of strips of distinctively colored paper that allowed it to be
distinguished from the remainder of the waste stream. As soon as tracer
material was observed in the output, its position was marked on the moving
bed of shredded refuse. Likewise, the point of depletion of the tracer
material was marked on the bed of refuse. After depletion of the tracer,
both input and output conveyors were halted. The shredded refuse containing
the tracer on the ouput conveyor was then completely removed in 0.9 m seg-
ments, and each segment was weighed. Each segment was then thoroughly and
completely mixed and divided into 2000-g samples, one of which was retained
for analysis. The retained sample was dried and weighed again to ascertain
its moisture content before manually sorting tracer material from the ref-
use. In this manner, the complete time history of tracer material from
introduction into the grinder to discharge could be determined.
14
-------
Mill holdup was determined by stopping both conveyors at the time a
steady-state condition was reached in the grinder and collecting the mate-
rial that then emerged from the grinder. To determine when a steady-state
condition had been reached, the chart recording of power consumption was
monitored for a period of relatively constant power draw. When such a point
of constant power draw was observed, the feed and discharge conveyors were
stopped, and a large sheet of plywood was placed under the grinder to catch
and retain the mill holdup for weighing. Output flow rate data were then
collected as previously described, and if manpower requirements permitted,
energy consumption was determined as well.
15
-------
SECTION 3
SHREDDING STUDIES
Study data were gathered systematically using various forms of single,
multiple, and progressive shredding of raw and processed refuse. In the
present context, (a) single-stage shredding implies a single pass through
the shredder at a specified feed rate, moisture content, grate spacing,
speed, etc.; (b) multiple shredding implies that the product becomes the
feed for the next stage of shredding, which occurs for the same shredder
conditions of grate spacing and speed; and (c) progressive shredding also
implies that the product becomes the feed for the next stage of shredding,
which in general occurs in a situation where the grate spacing has been
reduced. To reflect adequately the various processing configurations that
could be used in developing commercial waste processing facilities, three
material forms were studied: Raw packer truck MSW, light fraction obtained
after air classsification of shredded MSW product (ACLF), and the light frac-
tion obtained after air classification and screening (SLF). The results are
presented in terms of material groups for the^various machine arrangements.
MUNICIPAL SOLID WASTE STUDIES (GRUENDLER HAMMERMILL)
Single-Stage Shredding With Three Grate Openings
Raw refuse was shredded in a single pass manner in the Gruendler ham-
mermill for three separate grate openings of 4.8, 2.5, and 1.8 cm (1.9, 1.0,
and 0.7 in.). Various mass flow rates were shredded at each,grate spacing.
Energy measurements and product samples for size distribution analysis were
gathered for each experimental run.
For these tests, a range of feed rates was tested so that an overall
average product size and energy consumption per ton could be established.
For the experiments conducted with 1.8-cm (0.7-in.) grate spacings, the max-
imum hourly feed rate was limited to between 0.7 and 5,3 metric tons (0.8
and 6.0 tons) on a dry-weight basis to avoid overloading the grinder motor.
The range of dry arid wet feedrates used in these experiments is given in
Table 2.
The results of the tests conducted with 4.8-cm (1.9-in.) grate openings
showed average characteristic and nominal product sizes of 1.83 and 4.23 cm
(0.72 and 1.68 in.) (Table 3). The standard deviation was 0.41 cm (0.16
in.) for the former and 0.89 cm (0.35 in.) for the latter. The average spe-
cific energy consumption (E0) was 54.6 MJ/metric ton (13.8 kWh/ton), with a
standard deviation of 13.9 MJ/metric ton (3.5 kWh/ton). The average mois-
ture content (MC) of the shredded material was 29.8 percent, with a standard
deviation of 7.6 percent.
16
-------
TABLE 2. FEEDRATES USED IN THE SIZE REDUCTION OF
RAM MSW WITH THE GRUENDLER HAMMERMILL
Grate Minimum feed rate Maximum feed rate
opening
Dry Met Dry Met
metricmetricmetricmetric
cm (in.) tons/h (tons/h) tons/h (tons/h) tons/h (tons/h) tons/h (tons/h)
4.8 (1.9) 1.0 (1.1) 1.5 (1.7) 5.8 (6.4) 7.4 (8.1)
2.5 (1.0) 0.5 (0.6) 0.9 (1.0) 5.1 (5.6) 7.4 (8.2)
1.8 (0.7) 0.7 (0.8) 1.14 (1.3) 3.7 (4.1) 5.3 (5.8)
17
-------
Table 3. SIZE REDUCTION OF RAW MSW USING 4.8-CM
(1.9-IN) GRATE OPENINGS WITH THE GRUENDLER HAMMERMILL
Product size
cm
1.12
1.60
1.88
1.75
1.57
2.16
2.54
1.55
2.31
1.68
Avg. 1.83
Std. Dev. 0.41
Xo
(in.)
(0.44)
(0.63)
(0.74)
(0.69)
(0.62)
(0.85)
(1.00)
(0.61)
(0.91)
(0.66)
(0.72)
(0.16)
cm
2.97
2.92
4.70
3.94
5.59
4.95
4.78
3.30
4.57
4.95
4.23
0.89
X90
(in.)
(1.17)
(1-15)
(1.85)
(1.55)
(2.20)
(1.95)
(1.88)
(1.30)
(1.80)
(1.95)
(1.68)
(0.35)
Eo
MJ/
metric ton
40.4
58.6
62.6
49.5
69.3
41.6
34.1
64.2
80.0
36.4
54.6
13.9
Air-dry MO
(kWh/ton) (percent)
(10.2)
(14.8)
(15.8)
(12.5)
(17.5)
(10.5)
(8.6)
(16.2)
(20.2)
(9.2)
(13.8)
(3.5)
40.4
39.1
36.5
20.5
23.1
21.-
29.3
27.3
22.4
38.1
29.8
7.6
*Moisture content of shredded material.
18
-------
Energy consumption using a grate spacing of 2.5 cm (1.0 in.) averaged
99.4 MJ/metric ton (25.1 kWh/ton), with a standard deviation of 28.1 MJ/
metric ton (7.1 kWh/ton) (Table 4). The average characteristic and nominal
product sizes were, respectively, 0.74 and 1.52 cm (0.29 and 0.60 in.). The
standard deviations were, respectively, 0.18 and 0.25 cm (0.07 nd 0.10
in.). The average moisture content of the shredded material was 28.8 per-
cent, with a standard deviation of 7.9 percent.
Size reduction at the smallest grate openings, namely 1.8 cm (0.7 in.)
(Table 5) required an average specific energy of 167.1 MJ/metric ton (42.2
kWh/ton), with a standard deviation of 45.1 MJ/metric ton (11.4 kWh/ton).
The average characteristic and nominal product sizes were, respectively,
0.53 and 1.17 cm (0.21 and 0.46 in.), with standard deviations of 0.13 and
0.18 cm (0.05 and 0.07 in.), respectively. The shredded material had an
average moisture content of 27.2 percent, with a standard deviation of 9.0
percent.
Size reduction of raw MSW using a Gruendler hammermill has shown the
increase in energy consumption and the decrease in product size as a conse-
quence of decreasing the size of the grate openings. There was a signifi-
cant increase in energy requirement as the grate spacing was decreased from
2.5 to 1.8 cm (1.0 to 0.7 in.) (Figure 7). Grate openings in the range of
1.8 and 2.5 cm (0.7 and 1.0 in.) appear to represent the knee of the curve,
with the curve becoming exceedingly steep for values smaller than 1.8 cm
(0.7 in.).
The same data from another viewpoint, namely that of characteristic and
nominal product sizes (Figures 8 and 9), show the knee of the energy curve
to occur in the range of 0.74 and 0.53 (0.29 and 0.21 in.) for the former
and in the range of 1.52 and 1.17 cm (0.60 and 0.46 in.) for the latter.
The relation of shredded product size to size of grate openings is shown in
Figure 10.
In the size reduction of raw MSW, given a specific grate spacing, a
considerable variation in sample-to-sample data can be expected, as evi-
denced by the values of the standard deviation for energy consumption, char-
acteristic product size, and nominal product size. In general, the ratio of
standard deviation to average value was approximately 26 percent for energy
data, 24 percent for characteristic sizes and 18 percent for nominal sizes.
Consequently, to secure representative data with regard to energy consump-
tion and shredded product size, perhaps 8 to 10 samples should be the mini-
mum collected.
Single-Stage Shredding With No Grates
An additional series of tests was performed with the Gruendler hammer-
mill operating without any grate bars at a feed rate of 6.9 metric tons/h
(7.6 tons/h) on a wet basis and 5.9 metric tons/h (6.5 tons/h) on a dry
basis. As expected, this manner of shredding produces a very coarse product
size (Figure 11) when compared to raw refuse. The characteristic and nomi-
nal sizes are 15.2 and 25.4 cm (6.0 and 10.0 in.), respectively, because of
the relatively small residence time of the material within the mill. The
physical character of the material consists of torn grocery bags, plastic
19
-------
TABLE 4. SIZE REDUCTION OF RAW MSW USING 2.5-OM
(1.0-IN.) GRATE OPENINGS WITH THE GRUENDLER HAMMERMILL
cm
0.56
0.76
0.99
0.48
0.71
0.79
0.53
0.46
0.58
0.38
0.79
0.66
0.84
0.69
0.76
1.04
0.56
0.94
1.60
0.76
0.94
0.86
1.02
0.89
Avg. 0.74
Std.Dev. 0.18
Product
Xo
(in.)
(0.22)
(0.30)
(0.39)
(0.19)
(0.28)
(0.31)
(0.21)
(0.18)
(0.23)
(0.15)
(0.31)
(0.26)
(0.33)
(0.27)
(0.30)
(0.41)
(0.22)
(0.37)
(0.26)
(0.30)
(0.37)
(0.34)
(0..40)
(0.35)
(0.29)
(0.07)
size
cm
1.37
1.70
1.80
1.12
1.57
1.63
1.22
1.02
1.37
1.04
1.73
1.42
1.63
1.35
1.57
1.93
1.42
1.83
1.45
1.70
1.70
1.70
1.80
1.52
1.52
0.25
X90
(in.)
(0.54)
(0.67)
(0.71)
(0.44)
(0.62)
(0.64)
(0.48
(0.40)
(0.54)
(0.41)
(0.68)
(0.56)
(0.64)
(0.53)
(0.62)
(0.76)
(0.56)
(0.72)
(0.57)
(0.67)
(0.67)
(0.67)
(0.71)
(0.60)
(0.60)
(0.10)
Eo
MJ /metric ton
82.0
85.1
112.1
86.3
67.7
118.0
137.4
80.0
99.4
108.9
97.4
74.8
115.6
95.0
77.6
99.0
85.9
92.7
116.4
158.0
106.1
168.3
75.6
49.9
99.4
28.1
(kWh/ton)
(20.7)
(21.5)
(28.3)
(21.8)
(17.1)
(29.8)
(34.7)
(20.2)
(25.1)
(27.5)
(24.6)
(18.9)
(29.2)
(24.0)
(19.6)
(25.6)
(21.7)
(23.4)
(29.4)
(39.9)
(26.8)
(42.5)
(19.1)
(12.6)
(25.1
(7.1)
Air-dry MC*
(percent)
15.1
8.8
11.7
38.0
35.8
36.8
32.4
35.0
32.2
35.1
35.1
29.6
30.6
35.1
30.0
21.3
37.4
25.3
32.3
28.8
32.2
.27.8
22,2
23.5
28.8
7.9
*Moisture content of shredded material.
20
-------
TABLE 5. SIZE REDUCTION OF RAW MSW USING
1.8-CM (0.7-IN) GRATE OPENINGS WITH THE GRUENDLER HAMMERMILL
Product
cm
0.76
0.58
0.84
0.41
0.38
0.46
0.61
0.51
0.53
0.48
0.66
0.53
0.58
0.51
0.43
0.48
0.51
Avg. 0.53
Std.Oev. 0.13
Xo
(in.)
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
.30)
.23)
.33)
.16)
.15)
.18)
• 24)
.20)
• 21)
.29)
.26)
.21)
.23)
.20)
.17)
.19)
.20)
.21)
.05)
size
X
cm
1.37
1.32
1.63
0.97
1.07
1.04
1.27
1.02
1.19
1.02
1.35
1.12
1.22
1.17
0.94
1.14
1.02
1.17
0.18
90
Eo
(in.) MJ /metric ton (kWh/ton)
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
(0
.54)
.52)
.64)
.38)
.42)
.41)
.50)
.40)
.47)
.40)
.53)
.44)
.48)
.46)
.37)
.45)
.40)
.46)
.07)
288.
184.
131.
153.
161.
163.
140.
119.
230.
145.
249.
157.
134.
173.
144.
131.
133.
167.
45.
7
5
9
3
2
9
2
6
5
3
1
6
2
4
1
5
5
1
1
(72.
(46.
(33.
(38.
(40.
(41.
(35.
(30.
(58.
(36.
(62.
(39.
(33.
(43.
(36.
(33.
(33.
(42.
(11.
9)
6)
3)
7)
7)
4)
4)
2)
2)
7)
9)
8)
9)
8)
4)
2)
7)
2)
4)
Air-dry MC*
(percent)
37.
33.
22.
30.
25.
27.
15.
20.
23.
23.
21.
30.
29.
31.
28.
33.
26.
27.
9.
1
0
5
6
0
6
9
8
0
0
8
9
6
8
7
9
5
2
o •
*Moisture content of shredded material.
21
-------
ro
ro
50
40
30
~ 20
o
u
10
• o
l_
1.0
Size of Grate Opening
I
1.0
in.
I
2.0 3.0
cm
4.0
200
160 £
u
120
80
40
o
UJ
2.0
5.0
Figure 7. Energy consumption for size reduction of raw MSW as a
function of grate opening size (Gruendler hammermin).
-------
no
CO
70
60
50
2 40
o30
20
10
O.I
Grate Openings
cm (in.)
A 1.8 (0.7)
o 2.5 (1.0)
n 4.8 (1.9)
• No Grates
Characteristic Product Size
J I ill
_L
1.0
in.
j i
0.4 0.6 0.8 1.0
8 10
240
200
160
120
80
40
a>
E
10
20
cm
Figure 8. Energy required to reduce raw MSW to a given characteristic
product size with a Gruendler hammermiTI.
-------
70
60
50
r 40
of
30
20
10
O.I
Grate Openings
cm (in.)
A 1.8 (0.7)
o2.5 (1.0)
a4.8 (1.9)
• No Grates
Nominal Product Size
I
1.0
in.
240
200
160
o
•«-
o
80
40
10
04 0.6 0.8 I
cm
6 8 10
20
Figure 9. Energy required to reduce raw MSW
to a given nominal product size
with a Gruendler hamrnermlll.
24
-------
2.0
w
N
1.0
O
£
D Nominal
O Characteristic
Size of Grate Openings
i i i i ii r
1.0
in.
5.0
4.0
3.0
2.0
1.0
a)
N
V)
o
3
•o
O
2.0
1.0 2.0 3.0 4.0 5.0
cm
Figure 10. Relation of shredded raw MSW product size to size
of grate openings (Gruendler hammermlll).
25
-------
Figure 11. Sample of raw MSW shredded in a Gruendler
hammermill without grate bars.
26
-------
bags, and cardboard containers. Other than glass containers, objects smal-
ler than approximately 30 cm (12 in.) undergo little size reduction. The
energy consumption amounted to 4.4 MJ/metric ton (1.1 kW/h/ton), with res-
pective characteristic and nominal product sizes of 9.6 and 19.8 cm (3.8 and
7.8 in.).
Multiple Shredding
In this series of experiments, raw MSW was shredded three consecutive
times in the Gruendler hammer-mill at each of three different grate spacings,
namely 4.8, 2.5, and Ii8 cm (1.9, 1.0, and 0.7 in.). Three runs were con-
ducted for primary, secondary, and tertiary size reductions so that average
values for energy consumption and product sizes could be developed.
Use of the 4.8-cm (1.9-in.) grate spacing for the multiple shreddings
reduced the characteristic and nominal sizes from 15.2 and 25.4 cm (6.0 and
10.0 in.) to 0.53 and 1.12 cm (0.21 and 0.44 in.) respectively (Table 6).
The specific energy ranged from 54.6 MJ/metric ton (13.8 kWh/ton) for pri-
mary size reduction to 17.4 MJ/metric ton (4.4 kWh/ton) for tertiary size
reduction.
Multiple shredding using grate spacings of 2.5 cm (1.0 in.) decreased
the characteristic and nominal sizes from 15.2 and 25.4 cm (6.0 and 10.0
in.) to 0.46 and 1.02 cm (0.18 and 0.40 in.) respectively (Table 6). The
energy required decreased from 143.7 MJ/metric ton (36.3 kWh/ton) for pri-
mary size reduction to 16.2 MJ/metric ton (4.1 kWh/ton) for tertiary size
reduction. The shift in the size distribution as a consequence of multiple
shredding using grate spacings of 2.5 cm (1.0 in.) is shown in Figure 12.
Last, multiple size reduction at the smallest grate spacing tested,
namely 1.8 cm (0.7 in.), decreased the characteristic and nominal sizes from
15.2 and 25.4 cm (6.0 and 10.0 in.) to 0.36 and 0.74 cm (0.14 and 0.29 in.)
respectively (Table 6). As in the two previous cases, energy consumption
decreased. In this case the decrease was from an initial value of 167.1
MJ/metric ton (42.2 kwh/ton) for primary shredding to 31.3 MJ/metric ton
(7.9 kWh/ton) for tertiary shredding. The movement of the size distribution
towards the finer end of the size spectrum as the number of size reductions
using 1.8 cm (0.7 in.) grate spacings increased is shown in Figure 12.
AIR-CLASSIFIED LIGHT FRACTION STUDIES
The results of tests dealing with the shredding of ACLF, whose charac-
teristics are given in Section 2, are presented first in terms of multiple
shredding studies performed using the Gruendler hammermill and then for a
sequence of single, multiple, and progressive shredding tests using the W-W
machine.
Multiple Shredding (Gruendler Hammermill)
In this test, raw MSW was shredded using the Gruendler hammermill
equipped with 2.5-cm (1.00-in.) grate openings and air classified to obtain
an ACLF. The ACLF was then shredded using the same mill and grate openings
27
-------
TABLE 6. MULTIPLE SHREDDING OF RAW MSW WITH THE GRUENDLER HAMMERMILL
ro
CO
Feed size
cm
4.8
4.8
4.8
2.5
2.5
2.5
1.8
1.8
1.8
Do
(in.)
(1.9)
(1.9)
(1.9)
(1.0)
(1.0)
(1.0)
(0.7)
(0.7)
(0.7)
cm
15.2
1.83
0.86
15.2
0.84
0.58
15.2
0.53
0.41
Fo
(in.)
(6.0)
(0.72)
(0.34)
(6.0)
(0.33)
(0.23)
(6.0)
(0.21)
(0.16)
cm
25.4
4.27
1.73
- 25.4
1.60
1.24
25.4
1.19
0.91
F90
(in.)
(10.0)
(1.68)
(0.34)
(10.0)
(0.63)
(0.49)
(10.0)
(0.47)
(0.36)
cm
1.83
0.86
0.53
0.84
0.58
0.46
0.53
0.41
0.36
Product size
Xo
(in.)
(0.72)
(0.34)
(0.21)
(0.33)
(0.23)
(0.18)
(0.21)
(0.16)
(0.14)
cm
4.27
1.73
1.12
1.60
1.24
1.02
1.19
0.91
0.74
X90
(in.)
(1.68)
(0.68)
(0.44)
(0.63)
(0.49)
(0.40)
(0.47)
(0.36)
(0.29)
Eo
MJ/
metric ton
54.6
21.8
17.4
143.7
31.7
16.2
167.1
53.1
31.3
kWh/
ton
(13.8)
(5.5)
(4.4)
(36.3)
(8.0)
(4.1)
(42.2)
(13.4)
(7.9)
-------
ro
vo
Screen Size
1
0.05
O.IO
0.25
cm
0.5 O
• Primary
A Secondary
• Tertiary
Grate Opening
cm (in.)
- 2.5 (I.O)
- 1.8 (07)
0.02
I I
O.I
in.
t 1 t I
1.0
i
I.O
1.5 2.0
Figure 12. Size distribution of raw MSW after multiple shredding
with grate openings of 2.5 and 1.8 cm (1.0 and 0.7 in.)
in the Gruendler hammer-mill.
-------
(primary size reduction of ACLF). An additional shredding of the primary
size reduction product was also conducted (secondary size reduction). In
essence, this experiment simulates size reduction of ACLF using two ham-
mer-mills with identical grate openings operating in series. Energy consump-
tion and material flow rates were monitored for each step of size reduc-
tion. In addition, samples were collected for size distribution analyses of
feed and product materials. The characteristic and nominal feed sizes of
the ACLF for this set of tests were 1.22 and 1.93 cm (0.48 and 0.76 in.),
respectively.
The characteristic size of the ACLF was reduced to 0.94 and 0.89 cm
(0.37 and 0.35 in.) after primary and secondary shredding, respectively
(Table 7). The energy consumption as a consequence of these size reductions
decreased from 30.1 MJ/metric ton (7.6 kWh/ton) for primary size reduction
to 19.4 MJ/metric ton (4.9 kWh/ton) for secondary size reduction. There was
only a small decrease in material size as a consequence of secondary shred-
ding: FQ was 0.94 cm (0.37 in.) and X0 was 0.89 cm (0.35 In.). The shift
in the size distribution toward the finer size ranges is shown in Figure 13.
Single-Stage Shredding (W-W Hammermill)
In this set of experiments, the ACLF with characteristic feed sizes in
the W-W hammer-mill of 1.14 and 1.40 cm (0.45 and 0.55 in.) was reduced in
one step using grates with openings of 2.54, 1.91, 1.27, and 0.64 cm (1.00,
0.75, 0.50, and 0.25 in.) to determine the energy required for single-stage
size reduction. Energy consumption and the weight of the material shredded
were measured so that the specific energy could be calculated. At the same
time, samples of feed and product material were collected for size distribu-
tion analyses.
The results of this set of experiments show the rapid increase in
energy consumption as the size of the grate openings are decreased (Table
8). In Run 1, energy consumption increased more than ninefold — 44.7 to
407.1 MJ/metric ton (11.3 to 102.8 kWh/ton) — as the characteristic product
size decreased from 1.07 to 0.30 cm (0.42 to 0.12 in.). Similarly, in Run 2
energy consumption increased more than twelvefold — 26.5 to 324.7 MJ/metric
ton (6.7 to 82 kWh/ton) as the characteristic size of the product decreased
from 0.39 to 0.28 cm (0.35 to 0.11 in.).
The size distribution of the shredded ACLF shifted toward the finer end
of the size spectrum as the size of the grate openings was reduced from 2.54
to 0.64 cm (1.00 to 0.25 in.) (Figure 14). This figure illustrates the test
conditions and results of Run 1.
Multiple Shredding (W-W Hammer-mill)
In this set of experiments, the ACLF of a particular feed size was
reduced three times through the same grate openings of the W-W hammermill to
study the effect of multiple size reductions. The tests consisted of pri-
mary shredding of the ACLF, secondary shredding of the primary product, and
tertiary shredding of the secondary product. In all, four separate tests
30
-------
TABLE 7. MULTIPLE SHREDDING OF ACLF WITH THE GRUENDLER HAMMERMILL
Feed size Product size
Do r*FN xo X90 Eo
MJ/ kWh/
cm (in.) cm (in.) cm (in.) cm (in.) cm (in.) metric ton ton
CO
I-*
2.54 (1.00) 1.22 (0.48) 1.93 (0.76) 0.94 (0.37) 1.42 (0.56) 30.1 (7.6)
2.54 (1.00) 0.94 (0.37) 1.42 (0.56) 0.89 (0.35) 1.40 (0.55) 19.4 (4.9)
-------
CO
ro
roo
o»
in
-------
TABLE 8. SINGLE-STAGE SHREDDING OF THE ACLF WITH THE W-W HAMMERMILL
u»
Test
run
cm
1 2.54
1.91
1.27
0.64
2 2.54
1.91
1.27
0.64
Feed size
Do
(In.)
(1.00)
(0.75)
(0.50)
(0.25)
(1.00)
(0.75)
(0.50)
(0.25)
cm
1.14
1.14
1.14
1.14
1.40
1.40
1.40
1.40
Fo
(in.)
(0.45)
(0.45)
(0.45)
(0.45)
(0.55)
(0.55)
(0.55)
(0.55)
F
cm
2.08
2.08
2.08
2.08
3.68
3.68
3.68
3.68
90
(In.)
(0.82)
(0.82)
(0.82)
(0.82)
(1.45)
(1.45)
(1.45)
(1.45)
cm
1.07
0.79
0.61
0.30
0.89
0.76
0.61
0.28
Product size
xo
(in.)
(0.42)
(0.31)
(0.24)
(0.12)
(0.35)
(0.30)
(0.24)
(0.11)
cm
1.98
1.78
1.55
0.91
1.63
1.37
1.27
1.02
X90
(in.)
(0.78)
(0.70)
(0.61)
(0.36)
(0.64)
(0.54)
(0.50)
(0.40)
Eo
MJ/
metric ton
44.7
78.0
133.1
407.1
26.5
45.1
77.2
324.7
kWh/
ton
(11.3)
(19.7)
(33.6)
(102.8)
(6.7)
(11.4)
(19.5)
(82.0)
-------
100 »—
80
CO
-pi
o»
•S 60
i/>
o
Q_
0)
.> 40
E
3
O
20
0
ACLF Feed
Grate Opening
A
B
C
cm
2.54
1.91
1.27
Screen Size
D 0.64
(in.)
(1.00)
(0.75)
(0.50)
(0.25)
O.I
in.
I
1.0
0.10
0.25
0.50
cm
1.0
2.0
Figure 14.
Size distribution of the ACLF after single-stage
shredding with the W-W hammermill.
-------
were conducted at grate openings of 2.54, 1.91, 1.27, and 0.64 cm (1.00,
0.75, 0.50, and 0.25 in.). In each case, energy consumption was recorded,
and samples for size distribution analysis of feed and product material were
gathered. The characteristic and nominal feed sizes of the ACLF were 1.24
and 1.98 cm (0.49 and 0.78 in.), respectively (Table 9).
Within each test run (i.e., all size reductions with a given grate
opening) energy consumption decreased as the number of shreddings increased
(Table 9). This decrease resulted from the relative decrease in the ratio
of feed size to product size as the number of shreddings increased.
This relative decrease in specific energy within each test run was gen-
erally smaller as the grate openings became smaller. This observation must
be qualified, however, because the data collected for the 0.64-cm (0.25-in.)
grate openings are of dubious quality owing to the difficulty of screening
such fine, fluffy material. Consequently, the 0.64 cm (0.25 in.) data did
not follow the indicated trend. However, for test runs 1, 2, and 3 (Table
9), the trend is apparent. Specific energy decreased 24.5 percent in the
first run, 48.9 percent in the second, and 59.0 percent in the third (2.54,
1.91, and 1.27 cm grate openings, respectively).
Progressive Shredding (W-W Hammermill)
In this series of tests, the ACLF with an initial characteristic feed
size of 1.02 cm (0.40 in.) and a nominal size of 2.08 cm (0.82 in.) was
reduced in the W-W hammer-mill by consecutive shreddings through sequentially
decreasing grate openings. The procedure was as follows:
1) The ACLF was shredded in the W-W hammer-mill fixed with 2.54-cm
(1.00-in.) grate openings.
2) The product from the first shredding was shredded subsequently
through 1.91-cm (0.75-in.) grate openings.
3) The product from the second shredding was then shredded through
1.27-cm (0.50-in.) grate openings.
4) Last, the product from the third shredding was shredded through
0.64-cm (0.25-in.) grate openings.
In this manner, there was both a decrease in feed size and a decrease in
grate opening for each shredding step. Measurements of energy consumption,
quantities of material shredded, and samples for size distribution analyses
of feed and product sizes were gathered at each step of size reduction.
Energy consumption increased from 24.2 to 177.4 MJ/metric ton (6.1 to
44.8 kWh/ton) as the size of the grate openings decreased from 2.54 to 0.64
cm (1.00 to 0.25 in.) (Table 10). Over the course of the test, the material
decreased in characteristic and nominal size from 1.02 and 2.08 cm (0.40 and
0.82 in.) to 0.15 and 0.46 cm (0.06 and 0.18 in.), respectively.
35
-------
TABLE 9. MULTIPLE SHREDDING OF THE ACLF WITH THE W-W HAMMERMILL
CO
o>
Test
run
1
2
3
4
cm
2.54
2.54
2.54
1.91
1.91
1.91
1.27
1.27
1.27
0.64
0.64
Do
(in.)
(1.00)
(1.00)
(1.00)
(0.75)
(0.75)
(0.75)
(0.50)
(0.50)
(0.50)
(0.25)
(0.25)
cm
1.24
0.76
0.71
1.24
0.76
0.66
1.24
0.53
0.41
1.24
0.43
Feed size
Fo
(in.)
(0.49)
(0.30)
(0.28)
(0.49)
(0.30)
(0.26)
(0.49)
(0.21)
(0.16)
(0.49)
(0.17)
Product size
F
cm
1.98
1.19
1.09
1.98
1.14
1.12
1.98
1.24
1:14
1.98
1.19
90
(in.)
(0.78)
(0.47)
(0.43)
(0.78)
(0.45)
(0.44)
(0.78)
(0.49)
(0.45)
(0.78)
(0.47)
cm
0.76
0.71
0.76
0.66
0.56
0.53
0.41
0.36
0.43
0.15
Xo
(in.)
(0.30)
(0.28)
(0.23)
(0.30)
(0.26)
(0.22)
(0.21)
(0.16)
(0.14)
(0.17)
(0.06)
cm
1.19
1.09
1.14
1.12
1.09
1.24
1.14
1.04
1.19
0.76
X90
(in.)
(0.47)
(0.43)
(0.37)
(0.45)
(0.44)
(0.43)
(0.49)
(0.45)
(0.41)
(0.47)
(0.30)
Eo
MO/
metric ton
28.9
25.3
21.8
52.7
34.1
28.5
85.9
50.7
35.2
272.1
147.7
kWh/
ton
(7.3)
(6.4)
(5.5)
(13.3)
(8.6)
(7.2)
(21.7)
(12.8)
(8.9)
(68.7)
(37.3)
-------
TABLE 10. PROGRESSIVE SHREDDING OF THE ACLF WITH THE W-W HAMMERMILL
CO
"xl
Feed
size
D0 F0
cm
2.54
1.91
1.27
0.64
(in.)
(1.00)
(0.75)
(0.50)
(0.25)
cm
1.02
0.76
0.69
0.28
(in.)
(0.40)
(0.30)
(0.27)
(0.11)
cm
2.08
1.65
1.52
0.86
Product size
F90
(in.)
(0.82)
(0.65)
(0.60)
(0.34)
)
cm
0.76
0.69
0.28
0.15
(0
(in.)
(0.30)
(0.27)
(0.11)
(0.06)
)
cm
1.65
1.52
0.86
0.46
(90
(in.)
(0.65)
(0.60)
(0.34)
(0.18)
MJ/
metric
24.2
27.2
45.1
177.4
Eo
kWh/
ton ton
(6.1)
(9.4)
(11-4)
(44.8)
-------
SCREENED LIGHT FRACTION STUDIES
Single-Stage Shredding (W-W Hammennin)
The SLF was reduced in four separate tests using grate openings of
2.54, 1.91, 1.27, and 0.64 cm (1.00, 0.75, 0.50, and 0.25 in.) in the W-W
hammermill. The characteristic and nominal sizes of the SLF before each
size reduction were 3.36 and 4.57 cm (1.15 and 1.80 in.), respectively. All
tests involved single-stage size reduction — that is, SLF of the aforemen-
tioned size was used as the feed material for each grate opening.
Samples for size distribution analyses of the reduced products and of
the SLF feed material were collected for each set of experiments. Energy
measurements and quantities of material shredded were recorded to determine
the specific energy required for size reduction at each grate opening.
The energy requirements for size reduction of the SLF at each grate
opening ranged from 63.8 MJ/metric ton (16.2 kWh/ton) for producing charac-
teristic and nominal sizes of 1.68 and 2.37 cm (0.66 and 0.92 in.), respec-
tively, to 356.4 MJ/metric ton (90.0 kWh/ton) for producing characteristic
and nominal sizes of 0.94 and 1.42 cm (0.37 and 0.56 in.)» respectively
(Table 11). The ratio of characteristic product size to characteristic feed
size ranged from 0.57 for the 2.54-cm (1.00-in.) grate opening to 0.32 for
the 0.64-cm (0.25-in.) grate opening. As this ratio of product to feed size
decreased, the specific energy required for size reduction climbed rather
steeply. To illustrate, a ratio of 0.57 resulted in an energy consumption
of 63.8 MJ/metric ton (16.2 kWh/ton), whereas a ratio of 0.32 required 356.4
MJ/metric ton (90.0 kWh/ton). The energy requirement increased by a factor
of five, and the ratio of product to feed size decreased by approximately 56
percent.
Progressive Shredding (W-W Hammermill)
In this series of size reduction experiments, the SLF was shredded con-
secutively through grate openings of decreasing size in the W-W hammermill.
The characteristic size of the SLF was 3.18 cm (1.25 in.), and the corres-
ponding nominal size was 6.10 cm (2.40 in.). The test consisted of a pro-
gressive decrease in feed size and grate opening and consisted of the fol-
lowing procedure:
1) The SLF was shredded using grates with 2.54-cm (1.00-in.) openings.
2) The grate size was then decreased to 1.91 cm (0.75 in.) and the
reduced product from the previous run was shredded through it.
3) The grate size was next decreased to 1.27 cm (0.50 in.), and the
reduced product from the previous run was shredded through it.
4) Last, the grate openings were changed to 0.64 cm (0.25 in.), and the
reduced product from the previous run was shredded through it.
The product from shredding at each larger grate opening consequently became
the feed for size reduction at the next smaller grate opening. Samples for
38
-------
CO
to
TABLE 11. SINGLE-STAGE SHREDDING OF THE SLF FOR A GIVEN FEED SIZE
AND DIFFERENT GRATE OPENINGS WITH THE W-W HAMMERMILL
Feed size
cm
2.54
1.91
1.27
0.64
Do
(in.)
(1.00)
(0.75)
(0.50)
(0.25)
cm
3.36
3.36
3.36
3.36
Fo
(in.)
(1.15)
(1.15)
(1.15)
(1.15)
cm
4.57
4.57
4.57
4.57
F90
(in.)
(1.80)
(1.80)
(1.80)
(1.80)
Product size
Xo
cm
1.68
1.37
1.12
0.94
(in.)
(0.66)
(0.54)
(0.44)
(0.37)
X90
cm
2.37
1.98
1.60
1.42
(in.)
(0.92)
(0.78)
(0.63)
(0.56)
MJ/
metric
63.8
115.6
226.1
356.4
Eo
kWh/
ton ton
(16.2)
(29.2)
(57.1)
(90.0)
-------
size distribution analyses of feed and product sizes were taken after each
size reduction step. Energy measurements and shredded quantities were
recorded to determine the specific energy for each step of size reduction.
Results of the progressive shredding of SLF show a relatively small
increment in energy requirement for each of the first three steps of size
reduction (Table 12). The relatively small increments of size reduction are
responsible for this. Typically, the ratios of characteristic product size
to characteristic feed size for each shredding step were in the range of
0.50 to 0.80. The gradual movement of the size distribution toward the
finer end of the spectrum was a consequence of each incremental step of size
reduction (Figure 15). More than a twofold increase in energy consumption
was observed in the fourth step of size reduction.
The cumulative energy requirement (sum of each E0 increment) was 302.0
MJ/metric ton (76.7 kWh/ton) to achieve a characteristic size reduction of
3.18 to 0.58 cm (1.25 to 0.23 in.), or correspondingly, a nominal size re-
duction of 6.10 to 1.09 cm (2.40 to 0.43 in.). The ratio of final charac-
teristic product size to initial characteristic feed size, 0.58 and 3.18 cm
(0.23 and 1.25 in.), respectively, was 0.18.
Multiple Shredding (Hammermi11)
The shredding experiments conducted with the Gruendler hammermill in-
volved multiple shredding of SLF previously obtained through processing in
our resource recovery facility. The overall scope of the experiment consis-
ted of the following steps:
1) Primary size reduction of raw MSW.
2) Air classification of shredded MSW.
3) Trommel screening of the ACLF to recover the SLF.
4) Storage of the SLF.
5) Shredding of the SLF using the Gruendler hammermill (primary size
reduction of the SLF).
6) Collection and storage of the shredded SLF.
7) A final shredding of the previously shredded SLF (secondary size
reduction of the SLF).
Consequently, the SLF was actually shredded three times — once during pri-
mary size reduction of the MSW and twice as the SLF. The latter two shred-
dings are the ones of concern here.
To avoid any confusion with primary, secondary, and tertiary size
reduction of raw MSW (the terms generally applied to multiple size reduction
of raw MSW), one should realize that a previous size reduction as MSW is
implicit in the definition of that term, "screened light fraction". In this
study, the SLF is taken as the base or feed material.
40
-------
TABLE 12. PROGRESSIVE SHREDDING OF THE SLF WITH THE W-W HAMMERMILL
Feed size
cm
2.54
1.91
1.27
0.64
Do
(in.)
(1.00)
(0.75)
(0.50)
(0.25)
cm
3.18
1.68
1.30
0.97
Fo
(in.)
(1.25)
(0.66)
(0.51)
(0.38)
cm
6.10
2.26
1.73
1.30
F90
(in.)
(2.40)
(0.89)
(0.68)
(0.51)
cm
1.65
1.30
0.97
0.58
Product size
xo
(in.)
(0.66)
(0.51)
(0.38
(0.23)
cm
2.26
1.73
1.30
1.09
X90
(in.)
(0.89)
(0.68)
(0.51)
(0.43)
MJ/
metric
42.1
52.0
62.6
145.3
Eo
kWh/
ton ton
(10.7)
(13.2)
(15.9)
(36.9)
-------
100
90
in
o
Q.
60
> 50
*-
o
| 40
3
0 30
20
10
0
O.I
I
0.5
Grate Opening
cm (in.)
A 2.54 (1.00)
B 1.91 (0.75)
C 1.27 (0.50)
D 0.64 (0.25)
in.
_L
1.0
_L
_L
3.0
1.0
1.5 2.0
cm
3.0 4.0 5.06D7.0
Figure 15. Size distribution of the SLF after progressive
shredding with a W-W hammermill.
42
-------
Primary and secondary size reductions of the SLF were studied for two
different conditions of feed size — namely, the SLF recovered after primary
size reduction of raw MSW using 2.5-cm (1.0-in.) grate openings and 1.8-cm
(0.7-in.) grate openings. The characteristic sizes of the SLF corresponding
to shredding with these grate openings and subsequent air classification and
trommel screening were 1.45 and 1.02 cm (0.57 and 0.40 in.), respectively.
Feed rates for the primary and secondary size reduction of the SLF were 1.8
and 2.7 metric tons/h (2.0 and 3.0 tons/h) for grate openings of 1.8 and 2.5
cm (0.7 and 1.0 in.), respectively.
Energy recordings and samples for size distribution analyses were taken
for each stage of size reduction. The particle size data are reported as
the average of three samples each of the SLF feed, the primary shredded pro-
duct, and the secondary shredded product.
The results of the shredding tests show the degree of size reduction
and energy requirement for various feed sizes and grate openings (Table
13). In the shredding experiments using 2.5-cm (1.0-in.) grate openings,
the SLF was reduced from a feed with a characteristic particle size of 1.45
cm (0.57 in.) to a product with a characteristic particle size of 1.17 cm
(0.46 in.) while undergoing primary size reduction. These characteristic
particle sizes correspond to nominal particle sizes of 2.29 and 1.60 cm
(0.09 and 0.63 in.), respectively. The energy associated with this degree
of size reduction was 49.1 MJ/metric ton (12.4 kWh/ton).
Secondary size reduction of the SLF using 2.5-cm (1.0-in.) grate open-
ings further decreased the characteristic particle size of 1.02 cm (0.40
in.) and the nominal size to 1.5 cm (0.59 in.). This additional degree of
size reduction required 28.5 MJ/metric ton (7.2 kWh/ton). The shift in the
size distribution toward the finer end of the size spectrum as a consequence
of multiple shreddings at a grate opening of 2.5 cm (1.0 in.) is shown in
Figure 16.
Results of the experiments conducted with 1.8-cm (0.7-in.) grate open-
ings showed the following degree of size reduction and energy requirement.
Primary size reduction decreased the characteristic particle size of the SLF
from 1.02 cm (0.40 in.) to 0.99 cm (0.39 in.). These characteristic parti-
cle sizes correspond to nominal sizes of 1.55 cm (0.61 in.) and 1.45 cm
(0.57 in.), respectively (Table 13). The energy required to achieve this
degree of size reduction was 35.2 MJ/metric ton (8.9 kwh/ton).
Secondary size reduction of the SLF using 1.8-cm (0.7-in.) grate open-
ings resulted in a decrease in characteristic size from 0.99 to 0.91 cm
(0.39 to 0.36 in.), or alternatively, from 90 percent passing sizes of 1.45
to 1.40 cm (0.57 to 0.55 in.). This degree of size reduction required 49.9
MJ/metric ton (12.6 kWh/ton). The shift in size distributions as a result
of multiple shreddings at a grate opening of 1.8 cm (0.7 in.) is shown in
Figure 17.
The cumulative energy consumed by the Gruendler hammermill for both
primary and secondary size reduction is the total net energy required to
shred SLF to the size resulting from two consecutive shreddings with the
43
-------
TABLE 13. MULTIPLE SHREDDING OF THE SLF WITH THE GRUENDLER HAMMERMILL
Shredding
stage
Feed size
Product size
'90
cm (in.)
cm
kWhT
(in.) cm (in.) cm (in.) cm (in.) metric ton ton
2.5 (1.0) Primary
2.5 (1.0) Secondary
1.8 (0.7) Primary
1.8 (0.7) Secondary
1.45 (0.57) 2.29 (0.90) 1.17 (0.46) 1.60 (0.63) 49.1 (12.4)
1.17 (0.46) 1.60 (0.63) 1.02 (0.40) 1.50 (0.59 28.5 (7.2)
1.02 (0.40) 1.55 (0.61) 0.99 (0.39) 1.45 (0.57) 35.2 (8.9)
0.99 (0.39) 1.45 (0.57) 0.91 (0.36) 1.40 (0.55) 49.9 (12.6)
-------
100
80
o»
60
o
GL
~ 40
_o
3
E
o Feed
D Primary
Secondary
Screen Size
O.I
I
in.
0.25
0.5
1.0
I I I
cm
1.0 1.5 2.0
Figure 16. Size distribution of the SLF after multiple
shredding with 2.5-cm (0.7-1n.) grate
openings 1n the Gruendler hammermlll.
45
-------
100 r-
0»
c
"eft
«/>
o
O.
9)
>
a
•3
E
3
-------
same grate openings. For example, in the case of 2.5-cm (1.0-in.) grate
openings, a cumulative energy requirement of 77.6 MJ/metric ton (19.6 kWh/
ton) was needed to reduce the SLF from a characteristic size of 1.45 to 1.02
cm (0.57 to 0.40 in.). The corresponding reduction in nominal size was 2.29
to 1.50 cm (0.90 to 0.59 in.).
Similarly, the cumulative energy requirement for the 1.8-cm (0.7-in.)
grate openings was 85.1 MJ/metric ton (21.5 kWh/ton) for consecutive shred-
ding of the SLF from a characteristic size of 1.02 to 0.91 cm (0.40 to 0.36
in.), or a nominal size of 1.55 to 1.40 cm (0.61 to 0.55 in.).
47
-------
SECTION 4
WEAR STUDIES
To characterize the size reduction process completely, the hammer wear
was evaluated along with other comminution parameters. Detailed investiga-
tions were performed of the type and degree of hammer wear resulting from
the size reduction of raw MSW and SLF. The raw MSW measurements were con-
ducted with the Gruendler hammermill operating at 1200 rpm using a rotor
complement of 44 hammers in conjunction with three different sets of grate
bar spacings and five hardfacing materials. A previous raw MSW shredding
study [3] considered the effects of rpm on hammer wear. The investigations
with SLF were conducted with the W-W hammermill using four different grate
sizes and four hardfacing materials. The goals of the wear experiments were
to determine the best type of hardfacing alloy to be used and to establish
the rate of material loss from the hardfaced hammers as a consequence of
size reduction of these particular solid waste forms.
Hammer wear is generally characterized as wear resulting from abrasion
or impact. Soft hardfacing materials generally perform well against abra-
sive materials. Raw municipal solid waste is a composite material contain-
ing both abrasive and hard materials; consequently, materials with a full
spectrum of hardness values (14 to 56 Rockwell C hardness) were tested to
establish the optimum hardfacing material. Because of the abrasive nature
of the cellulosic materials in the SLF, relatively hard welding alloys (42
to 63 Rockwell C hardness) were chosen for testing. One point of the
research was to test whether or not extremely hard alloys would result in
better hammer wear than other softer alloys. In addition, wear rates were
examined to quantify the trend in wear as the grate openings were reduced.
HARDFACING MATERIALS AND APPLICATION
Raw MSW Tests
The hardfacing materials tested were alloys in welding electrode form
and were readily available from local distributors. The alloys were
deposited on the hammer surface using manual electric arc welding. Hard-
facing was done in such a manner as to minimize localized heating of the
base hammer material (cast manganese steel), which causes embrittling
transformations.
In all, five different hardfacing alloys were applied and tested:
Stoody 2110, Stoody 31, Coated Tube Stoodite (all manufactured by Stoody
Company), Amsco Super 20 (manufactured by Abex Corporation), and Alloy X
48
-------
(manufacturer unknown, but the composition was determined metallurgically).
the hardfacing materials contained various alloy contents and elements
(Table 14). The cost of the hardfacing electrodes tested (50-lb lots,
5/32-in. diameter) varied from ?123 to ?193 (Table 14).
Al1 electrodes tested were easy to apply except for the Coated Tube
Stood He, which had a tendency to drip and splatter. Considerable practice
was necessary to obtain a good weld deposition with this electrode. The
hammers used are shown in Figures 18 and 19.
Screened Light Fraction
The hardfacing materials chosen were Stoody 31 and Coated Tube Stoodite
manufactured by the Stoody Company, Amsco Super 20 manufactured by the Abex
Corporation, and Haystellite Tungfine manufactured by the Cabot Corpora-
tion. The hardfacing materials contained various alloy contents and ele-
ments (Table 15). The approximate costs of the hardfacing materials tested
(50-lb lots, 5/32-in. diameter) varied from $123 to J668 (Table 15).
All hardfacing materials (except Haystellite Tungfine) were in welding
electrode form and were applied to the hammers by manual electric arc weld-
ing. The Haystellite Tungfine was applied to the hammer tips by the hammer-
mill manufacturer using oxy-acetylene welding. The base hammer material was
1045 carbon steel.
EXPERIMENTAL PROCEDURE
Raw MSN Test (Gruendler Harnnermill)
The general experimental procedure called for three sets of wear exper-
iments. Each set was accomplished at a specific grate opening, namely 4.8,
2.5, and 1.8 cm (1.9, 1.0, and 0.7 in.). Before each data set commenced,
the hammers were hardfaced and weighed to establish the initial weight for
that test set. After shredding a specific tonnage of MSW, the hammers were
cleaned with a wire brush to remove all electrode slag, dust, and dirt, then
weighed again to establish the final weights. The net weight loss was then
calculated for each hammer.
The experimental procedure also called for determination of the hard-
ness of each hardfacing material as well as that of the base material (cast
manganese steel). Consequently, test beads of each hardfacing alloy were
run on a hammer and tested for hardness.
Screened Light Fraction (W-H Hammermill)
The experimental procedure called for wear determinations using: (1) a
characteristic feed size of 1.73 cm (0.68 In.) and a nominal feed size of
2.72 cm (1.07 in.), (2) grate openings of 2.54, 1.91, 1.27, and 0.64 cm
(1.00, 0.75, 0.50, and 0.25 in.), and (3) the four hardfacing materials des-
cribed in the preceding section. The as-deposited hardness of each hard-
facing material was determined before Initiating the wear experiments.
49
-------
Table 14. COMPARISON OF HAMMER AND HARDFACING MATERIALS
USED FOR SHREDDING RAW MSW
Material
Composition
Hardness (R )
Approximate cost
As deposited Work-hardened of electrode
en
O
Cast manganese steel
Hardfacing alloy:
Stoody 2110
Alloy Xc
Stoody 31
Coated Tube1 Stoodite
Amsco Super 20
C = 1.11 percent
Mn = 12.7 percent
Si = 0.54 percent
Alloy content = 37 percent
(Cr, Mn, Nc, Si, C)
iron base
Alloy content = 20 percent
(Cr, Mn, Ni, Mo, C)
iron base
Alloy content = 35 percent
(Cr, Mn, Si, Mo, C)
iron base
Alloy content = 39 percent
(Cr, Mn, Si, Mo, Zr, C)
iron base
Alloy content = 40 percent
(Cr, Mo, C, W)
iron base
14L
21
38
46
55
56
38
NA"
45
NA
NA
NA
S123
NA
123
129
193
50-lb lot, 5/32-in. diameter.
As cast.
Not available.
Manufacturer unknown; composition determined by metallurgical analysis.
-------
Figure 18. Bare manganese hammers before hardfacing,
Figure 19. Hammers used for shredding raw MSW:
Left, new manganese steel hammer;
Center, worn hammer; right, re-
surfaced hammer.
51
-------
Table 15. COMPARISON OF HARDFACING MATERIALS
USED FOR SHREDDING SLF
Material
Composition
Hardness
as deposited
Approximate cost
of electrode*
01
Stoody 31
Coated Tube Stoodite
Amsco Super 20
Haystellite Tungfine
Alloy content = 35 percent
(Cr, Mn, Si, Mo, C),
iron base
Alloy content = 39 percent
(Cr, Mn, Si, Mo, Zr, C),
iron base
Alloy content = 40 percent
(Cr, Mo, C, W),
iron base
61 percent tungston carbide
filler, balance steel tube
41.9
49.5
56.6
63.0
$123
129
193
668
* 50-1b lot, 5/32-in. diameter.
-------
Before beginning experiments at each grate opening, all tipped hammers
were brushed clean with a wire brush and weighed. After shredding a speci-
fic weight of SLF, the hammers were removed from the grinder, brushed clean,
and weighed. The difference between the initial and final weights equalled
the net weight loss. The wear rate is reported here as kg of hammer weight
lost per metric ton of material shredded (kg/metric ton) or as pounds per
ton (Ib/ton).
HARDENESS MEASUREMENTS
The Rockwell C hardness of each hardfacing material was measured using
a standard Rockwell hardness tester. To measure the hardness, a sample bead
of each hardfacing material was welded on a hammer and ground to provide a
flat testing surface. Care was taken during grinding to minimize localized
temperature buildups that could result in low hardness values. During the
hardness measurements, the hammers with the sample welds were mounted
securely on the tester to insure a valid hardness test.
Raw MSW Test
Four hardness tests were conducted for MSW on each hardfacing alloy as
well as on the surface of the cast manganese steel hammer. The average
hardness values (Rockwell C) ranged from 14 for the hammer itself to 56 for
the hardest alloy tested (Amsco Super 20) (Table 14). These values are re-
ported as deposited, and therefore they do not include any work-hardening.
The effect of work-hardening was measured in only two instances, since hard-
ness determinations after work-hardening require sectioning of the hammers
to obtain test specimens. In the two cases tested, the base hammer material
(manganese steel) increased in hardness from 14 to 38, and Alloy X increased
in hardness from 38 to 45 (Table 14).
Screened Light Fraction
Four hardness measurements were conducted on each hardfacing material.
The average hardness values (Rockwell C) ranged from 41.9 for the softest
material tested (Stoody 31) to 63.0 for the hardest material tested (Hay-
stellite Tungfine (Table 15). These hardness values are as deposited and
therefore do not include any work-hardening.
WEAR MEASUREMENTS
MSW Tests
Wear measurements were conducted at three different grate openings —
namely 4.8, 2.5, and 1.8 cm (1.9, 1.0, and 0.7 in.). Experiments at 4.8-cm
(1.9-in.) grate openings used bare hammers of cast manganese steel, hammers
tipped with Alloy X (manufacturer unknown), and hammers tipped with Amsco
Super 20. The alignment of the tipped and untipped hammers consisted of two
rows of bare hammers (12 hammers in each row), one row of 10 hammers hard-
faced with Alloy X, and one row of 10 hammers hardfaced with Super 20, Con-
sequently, each row encountered exactly the same material and quantities
seen by the other rows. This approach eliminated tonnage and composition as
factors within each data set.
53
-------
In the second data set, four materials were tested at a grate spacing
of 2.5 cm (1.0 in.). The hammer alignment consisted of one row of bare ham-
mers and one row each of hammers tipped with Stoody 31, Coated Tube
Stoodite, and Amsco Super 20.
To test an additional alloy (Stoody 2110) in the wear experiments con-
ducted at a grate spacing of 1.8 cm (0.7 in.), the row normally containing
bare hammers was tipped with Stoody 2110, except for two hammers along the
center of the rotor. Each of the other three rows consisted of hammers
hardfaced with Stoody 31, Coated Tube Stoodite, and Amsco Super 20, respec-
tively. The two bare hammers were needed to maintain a control for the
1.8-cm (0.7-in.) data set and also for later correlation with the data ob-
tained in the 4.8- and 2.5-cm (1.9- and 1.0-in.) data sets. Likewise, Amsco
Super 20 was used in each data set as yet another control.
Screened Light Fraction
These wear measurements were conducted with grate opening diameters of
2.54, 1.91, 1.27, and 0.64 cm (1.00, 0.75, 0.50, and 0.25 in.). For each
grate opening, there were four rows of hammers. Each row contained five
hammers tipped with one of the four alloys — Stoody 31, Coat Tube Stoodite,
Amsco Super 20, or Haystellite Tungfine. Consequently, each row encountered
exactly the same material and quantities experienced by the other rows.
This technique eliminated tonnage and composition as variables for experi-
ments carried out at each grate opening.
Wear data are reported as kg/metric ton and Ib/ton. The data have been
normalized for a full complement of 20 hammers.
RESULTS
Raw MSW Tests
The wear experiments conclusively showed that wear increased dramati-
cally as the grate opening was reduced to 1.8 cm (0.7 in.). The two control
cases (i.e., bare hammers and hammers hardfaced,with Amsco Super 20) used in
each data set provide a record of increased wear as the grate spacing was
reduced. If the hammer wear found in the experiments is normalized to a
full set of hammers (i.e., 44 hammers) and expressed as kilograms of weight
lost per metric ton shredded (kg/metric ton), or as pounds per ton (Ib/ton),
the bare manganese hammers exhibited wear rates of 0.042, 0.057, and 0.140
kg/metric ton (0.084, 0.114, and 0.279 Ib/ton) at grate spacings of 4.8,
2.5, and 1.8 cm (1.9, 1.0 and 0.7 in.), respectively. For the same grate
openings, the wear rates for hammers hardfaced with Amsco Super 20 were
respectively, 0.043, 0.054, and 0.089 kg/metric ton-(0.086, 0.107, and 0.178
Ib/ton) (Table 16). The rapid increase in wear is particularly apparent in
Figure 20. The curve becomes very steep when grate openings are smaller
than 2.5 cm (1.0 in.).
Another perspective of the wear rates can be gained by considering wear
as a function of the average characteristic shredded product size (Figure
21). The average characteristic product sizes of MSW shredded through grate
openings of 4.8, 2.5, and 1.8 cm (1.9, 1.0 and 0.7 in.) were 1.83, 0.74, and
54
-------
TABLE 16. WEAR DATA FOR VARIOUS HARDFACING MATERIALS
USED FOR SHREDDING RAW MSW
en
en
Material
Manganese steel
base material
Stoody 2110
Alloy Xc
Stoody 31
Coated Tube
Stood ite
Am sco Super 20
Hardness
-------
in
en
0.30
0.25
0.20
0.15
S o.io
0.05
0.5
1.0
o Manganese Steel Hammers
OManganese Steel Hammers
Tipped with Amsco Super 20
Grate Opening
I
1.0
in.
1.5
0.15
0.10
c
o
a>
£
0.05 -
o
o
2.0
2.0
cm
3.0
4.0
5.0
Figure 20, Hammer wear as a function of grate opening: Raw MSW, Gruendler hammermill
(characteristic feed size, 15 cm or 6 in., nominal feed size, 33 cm or 13 in.)
-------
Ul
c
o
o
0.05
O.I 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
in.
Figure 21.
Hammer wear as a function of average characteristic product size:
Raw MSW, Gruendler hammermill (characteristic feed size, 15 cm
or 6 in.; nominal feed size, 33 cm or 13 in.).
-------
0.53 cm (0.72, 0.29, and 0.21 in.), respectively. Figure 21 shows the wear
rates from Table 16 plotted versus the characteristic product size. The
steepness of this curve for product sizes smaller than 0.74 cm (0.29 in.)
shows the penalty that was paid to produce the finer particle sizes.
These same wear data can also be correlated with the average nominal
particle size (Figure 22). These curves are similar to those of Figure 21
and show a rapid increase in wear when nominal sizes are smaller than 1.5 cm
(0.6 in.).
Figure 23 shows the relationship between hammer wear and alloy hardness
for several grate openings after weld deposition. The fact that hardness
can be used as a parameter to characterize hardfacing alloys is noteworthy,
since hardness is a quantity that can be easily measured. Despite differ-
ences in manufacturer and composition, alloys that exhibited similar hard-
ness values also underwent similar degrees of wear in our tests.
The functional relationship of wear and hardness shows that for certain
hardness ranges, a minimum of wear occurs. For instance, in experiments
conducted with a grate opening of 4.8 cm (1.9 in.), a minimum of wear
occurred in the range 36 £ Rc < 44. Similarly, in the experiments conducted
at a grate spacing of 2.5 cm (T.O in.), a minimum occurred in the range 40 <
Rc £ 46.
For the harder alloys, Coated Tube Stoodite (Rc = 55) and Amsco Super
20 (Rg = 56), chipping of the weld deposits represented the primary mode of
material wear. For the tests at grate spacings of 4.8 and 2.5 cm (1.9 and
1.0 in.), these harder materials wore no better than the softer bare manga-
nese hammers (Figure 23).
The results of the experiments conducted with a grate spacing of 1.8 cm
(0.7 in.) were slightly different from those obtained at the larger grate
openings. At grate spacings of 1.8 cm (0.7 in.), wear decreased as the
hardness increased to a value of about 46 Rc (Figure 23). Although the ham-
mers tipped with Coated Tube Stoodite (55 Rc) showed considerably less wear
than either Stoody 31 (46 Rc) or Amsco Super 20 (56 Rc), this result should
be viewed with skepticism. The results for Coated Tube Stoodite should not
be viewed as representative of improved wear resistance, but as indicative
of the haphazard nature of impact loadings on the hammer tips during shred-
ding of MSW. This point of view is supported by the identical wear exhibi-
ted by Coated Tube Stoodite and Amsco Super 20 in experiments carried out at
grate openings of 2.5 cm (1.0 in.) (Figure 23). The fact that there was a
noticeable lack of hard, difficult material to reduce in the experiments
conducted at 1.8-cm (0.7-in.) grate openings lends further support to this
viewpoint.
Visual inspection of the weld deposits after shredding showed that
chipping was the predominant mode of material loss for the harder alloys,
namely Amsco Super 20 and Coated Tube Stoodite (Figures 24 and 25). The
softer hardfacing alloys presented a smooth, worn appearance as a result of
abrasion and did not show signs of chipping. Consequently, chipping was
considered to be the limiting factor regarding the wear characteristics of
the harder alloys. Accelerated material loss as a result of chipping occurs
when the Rc is greater than 50 (Figure 23).
58
-------
ID
C
o
o
0)
0.30
0.25
0.20
0.15
0.10
0.05
0
I
o Manganese Steel Hammers
a Manganese Steel Hammers
Tipped with Amsco Super 20
Nominal Product Size (90% Cumulative Passing)
I i i
0.5
1.0
in.
1.5
I
1.0
2.0 3.0
cm
4.0
0.15
0.10
c
o
a>
E
^
o»
0.05 g
2.0
5.0
Figure 22. Hammer wear as a function of average nominal product size:
Raw MSW, Gruendler hammermin (characteristic feed size,
15 cm or 6 in., nominal feed size, 33 cm or 13 in.).
-------
0.30
0.25
- 0.20
E
o
x
0.15
0.10
0.05
Cost
Mongonese
Steel stdody
2110
I
Alloy Stoody
X 31
I
Grate Opening
cm (in.)
• 1.8 (0.7)
D 2.5 (1.0)
o 4.8 (1.9)
lAmsco Super 3D
Coaled
Tube
Stoodite
0.15
0.10
o
o
w
«
E
*x
a*
0)
0.05 E
o
x
10 20 30 40 50
Rockwell C Hardness (Rc)
60
70
Figure 23. Hammer wear as a function of hardness for several different
grate openings: Raw MSW, Gruendler hammermill (characteristic
feed size, 15 cm or 6 in., nominal feed size, 33 cm or 13 in.).
-------
Figure 24. Evidence of chipping on hammers tipped
with hardfacing alloys exceeding Rc 50
(before removal for weighing).
Figure 25. Closeup of hammers (showing chipping of
hardfacing weld deposits on hammer tips).
61
-------
Screened Light Fraction
The wear experiments with SLF for various hardfacing materials show the
general trend of increased wear with smaller grate openings (Table 17). As
seen from Figure 26, these wear rates increase from an average of approxi-
mately 0.003 to 0.007 kg/metric ton (0.005 to 0.014 Ib/ton) as the diameters
of the grate openings were decreased from 2.54 to 0.64 cm (1.0 to 0.25 in.)
(Figure 26).
Rather surprisingly, curves for the best fit of the experimental data
show relatively little difference in wear rates among the different hard-
facing materials for experiments conducted at various grate opening sizes
(Figure 26). If the unusually high wear rate for Haystellite Tungfine at
grate openings of 0.64 cm (0.25 in.) is considered unrepresentative of that
material, an argument could possibly be made that the softest alloy (Stoody
31) exhibited the highest degree of wear. This argument can be further
supported by examining the functional dependence of wear rates on alloy
hardness (Figure 27). Here again, discounting the high wear rate of
Haystellite Tungfine at a grate opening diameter of 0.64 cm (0.25 in.), a
slightly greater degree of wear could be argued for the softest alloy,
Stoody 31. Practically, however, given the limitations and uncontrollable
variables in the size reduction of SLF, all hardfacing materials tested here
were considered to have similar wear characteristics.
The relationship between hammer wear and characteristic product size
(Figure 28) is similar to that depicted for hammer wear and grate opening
diameter (Figure 27). In general, wear rates increased from approximately
0.003 to 0.007 kg/metric ton (0.005 to 0.014 Ib/ton as the characteristic
product size decreased from 1.52 to 0.43 cm (0.60 to 0.17 in.).
Visual inspection of the hammers after removal from the hammermill
showed the wear to be abrasive in nature, as evidenced by a shiny, smooth
appearance of the weld deposits. No signs of chipping were found.
CONCLUSIONS
Several noteworthy findings resulted from the hammer wear research on
ihrprlriinn nf raw MSW. First, hammer wear under ontimum rrmrH *•!««,. «*
8
or
several notewurtny • muiiiys resun-eu irum trie ridiiinier wear research on
the shredding of raw MSW. First, hammer wear under optimum conditions of
hardfacing selection can be expected to be roughly two to three times as
great for grate openings of 1.8 cm (0.7 in.) than for grate openings of 4.
cm (1.9 in.) in the size reduction of solid waste by hammermill shredding."
This fact should be weighed when considering the use of 2.5-cm (l.o in.) o
smaller grate openings for shredding raw MSW.
Second, using a scientific procedure (such as that followed here), an
optimum hardfacing alloy can be found that will result in a minimum of'wear
for specific operating conditions. The tests reported here indicate that
hardfacing alloys with as-deposited hardness in the range 36 _< Rc < 45
should lose a minimum of material during the shredding of MSW (at Teast in
the range of grate spacings tested).
Third, alloys that exhibit an as-deposited hardness in excess of 50
Rc may show high degrees of wear because of chipping of the weld deposits
62
-------
TABLE 17. WEAR DATA FOR VARIOUS HARDFACING MATERIALS
USED FOR SHREDDING SLF
Wear rate, kg/metric ton (Ib/ton)
Material Hardness* 2.54-cm (1.0-in.) 1.91-cm (0.75-in.) 1.27-cm (0.50-in.) 0.64-cm (0.25-in.)
grate grate grate grate
Stoody 31 41.9 0.0039 (0.0078) 0.0030 (0.0061) 0.0066 (0.0132) 0.0060 (0.0120)
Coated Tube 49.5 0.0022 (0.0043) 0.0029 (0.0058) 0.0055 (0.0110) 0.0041 (0.0081)
Stoodite
Amsco 56.6 0.0025 (0.0049) 0.0033 (0.0065) 0.0051 (0.0101) 0.0058 (0.0115)
Super 20
Haystellite 63.0 0.0027 (0.0054) 0.0023 (0.0046) 0.0044 (0.0088) 0.0085 (0.0170)
Tungfine
*H rdness values are initial values as deposited without work-hardening.
-------
c
o
o
01
0.020
0.015
0.010
0.005
0
L
O Stoody 31
O Coated Tube Stoodite
©Amsco Super 20
AHoystellite Tungfine
Grate Opening
0.25
JL
_L
0.50
in.
_L
0.5
1.0 1.5
cm
0.75
2.0
0.010
o
0
4)
0.005
1.00
J I
2.5
O
a
Figure 26.
Hammer wear as a function of grate opening: SLF, W-W hammermill
(characteristic feed size, 1.73 cm or 0.68 in.; nominal feed
size, 2.72 cm or 1.07 in.}.
-------
en
01
0.020
0.015
0.010
0.005
Grate Opening
cm
Stoody 31
I i I
j I
Coated Tube
Stoodite
I I I I I I
Amsco
Super 20
I I I i
I I i i
Haystellite
Tungfine
0.010
c
o
0.005
40
45
50 55
Rockwell C Hardness
60
65
Figure 27. Hammer wear as a function of hardness for several different grate
openings: SLF, W-W hammermill (characteristic feed size, 1.73 cm
or 0.68 in.; nominal feed size, 2.72 cm or 1.07 in.).
-------
0.020
c
o
0.015
S o.oio
0.005
A Haystellite Tungfine
a Stoody 31
e Coated Tube Stoodite
o Amsco Super 20
Characteristic Product Size
j | i I
0 O.I
I
j_
J_
0.5
1.0
1.5
0.010
0.005
0.2 0.3 0.4 0.5 0.6 0.7
in.
j
o
*-
o
w
tt)
E
o
o
cm
Figure 28. Hammer wear as a function of average characteristic product size:
SLF, W-W hammermill (characteristic feed size, 1.73 cm or 0.68
in.; nominal feed size, 2.72 cm or 1.07 in.).
-------
The degree of chipping will depend on the nature of the MSW being shredded.
MSW that contains large quantities of hard objects such as white goods,
steel bar stock, hardened gears, etc. will cause more severe chipping than
will MSW that contains little of such material. In the latter case, alloys
in the range of 48 to 52 Rc could provide the optimum hardfacing material.
Screened Light Fraction
This research on hammer wear as a consequence of the size reduction of
the SLF has uncovered several important findings. First, hammer wear in-
creased approximately 250 percent as the grate openings were decreased from
2.54 to 0.64 cm (1.0 to 0.25 in.). These results are for SLF with a charac-
teristic feed size of 1.73 cm (0.68 in.) and a nominal size of 2.72 cm (1.07
in.).
Second, hammer wear was found to be independent of the type of hard-
facing applied. This is an interesting finding, especially since a rela-
tively wide range of alloy hardness values (41.9 <_ R <_ 63.0) was tested.
Most hardfacing manufacturers contend that the harder alloys should wear
better under abrasive conditions. The results of our tests did not uphold
this contention.
Last, the shiny, smooth appearance of the weld deposits after shredding
the SLF indicated that this material was abrasive in nature. None of the
chipping behavior that characterized the harder alloys (Rc > 50) during the
size reducion of raw MSW was evident during the size reduction of the SLF.
67
-------
SECTION 5
INTERRELATION OF THE COMMINUTION PARAMETERS
Variables such as flow rate, moisture content, shredder holdup, resi-
dence time for material within the mill, degree of size reduction, etc. have
been identified as controlling the size reduction process of refuse. For
example, product particle size distribution is a function of the feed part-
icle size and the mean residence time, and the power requirements of the
shredder are shown to be a function of the mill holdup and the moisture con-
tent of the refuse. Also, if the grate opening size and characteristic feed
size are specified, the degree of size reduction can be computed. This
determination in turn enables the prediction of characteristic product size,
specific energy, and hammer wear. Consideration will be given to the quan-
titative relationship between the governing variables and their use for pre-
dictive purposes.
REPRESENTATIVE PARTICLE SIZE
The product resulting from a size reduction or shredding operation typ-
ically exhibits a range of sizes that may span three to four orders of mag-
nitude. It has been shown that this size distribution Y(x) can be repre-
sented by the Rosin-Rammler equation, as follows:
Y(x) = 1 - exp[-(x/XQ)n] (1)
where n is a numerical constant, x is the screen size, and X0 is a charac-
teristic particle size. When x is substituted for X0 in the above equation,
namely
Y(x) . (1 - 1/e) = 0.632 (2)
it follows that the percent passing is 63.2 percent. Thus the character-
istic particle size, X0, is the size that corresponds to 63.2 percent pas-
sing. The exponent n may be found from the slope of the line
ln[ln(l-Y(x))] . n (In x - In XQ) (3)
and is a measure of the uniformity of the size distribution. In other
words, n is a measure of the width of a size distribution if plotted as in
Figure 29, which i's a typical size distribution for raw MSW presented in
terms of the mass fraction (m-j) within a given class (i) instead of the more
commonly used percent passing (Y(x)).
To develop useful predictive relationships, it becomes necessary to
correlate the comminution variables in terms of some particle size, since
correlations in terms of the entire distribution are usually not tractable.
68
-------
0.751—
0.50 -
u
o
(£>
tf>
c
0.25 —
Size Closs, i Screen Size (cm)
I
2
3
4
5
6
7
8
9
10
40 x20
20x 10
10 x 5
5 x 2.54
2.50x1.25
1.25 x 0.63
0.63 * 0.31
0.31X0.16
O.I6XQ.08
0.08 x 0.00
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
Size Closs, i
Figure 29. Mass fraction representation of the size distribution for raw MSW.
10.0
-------
The two representative particle sizes that are used are: (a) the previously
discussed characteristic size corresponding to 63.2 percent passing, and (b)
the so-called nominal size, which corresponds to 90 percent passing. The
nominal size is the most prevalent parameter used in the industry, particu-
larly in refuse-derived fuel situations. Nonetheless, it should be noted
that the characteristic size has proven to be the most useful variable for
denoting particle size. Attempts to use nominal size have resulted in sig-
nificantly more data scatter when studying energy and size relationships.
Perhaps one reason for the good correlation of characteristic size and en-
ergy consumption is that this size is more representative of the average
size than is the nominal size. Furthermore, the characteristic size is
derived from a fundamental consideration of the size distribution represent-
ing the entire size spectrum of the material. Consequently, the character-
istic size is used here as the primary size for correlation purposes. The
relationship between these two sizes for shredded raw MSW and for two forms
of light fraction are given in Figures 30 and 31, respectively.
DEGREE OF SIZE REDUCTION
The degree of size reduction (Z0) is introduced as a means of general-
izing the size reduction data obtained over a wide range of operating and
mill conditions. This term is defined as
Z0 = (F0 - XO)/FO (4)
where F0 and X0 are the characteristic feed size and characteristic product
size, respectively. Thus the degree of size reduction is a dimensionless
quantity approaching unity as X0 approaches zero, and approaching zero as X0
approaches F0. To complete the generalization, an additional non-dimensional
parameter is required and can be formed by the ratio of the grate opening
size (D0) to the characteristic feed size (F0) — that is, D0/F0. The para-
metric nature of the comminution process with grate opening size is shown in
Figure 32 when the degree of size reduction (Z0) is represented in terms of
the DQ/FO ratio. Essentially, for a particular value of the ratio D0/F0,
the value of Z0 increases as the size of the grate openings increase. In
other words, a smaller characteristic particle size is produced as the grate
openings increase in size. As shown in Figure 32, this trend holds for the
ACLF forms as well as the raw MSW. A full explanation for this seemingly
contradictory effect requires the introduction of other parameters affecting
the size reduction process.
ENERGY
Another example of the parametric nature of grate opening size is
demonstrated by the relationship between specific energy (E0) and degree of
size reduction (Z0). For the three material forms studied the Eft increases
as the D0 decreases for a given value of Z0 (Figure 33). In the Sase of raw
MSW, E0 values for the interval 0.88 < Z0 <_ 1.00 correspond to the average E0
values for single-stage size reduction using the grate spacinqs soecified
But EQ values in the region 0.00 < Z0 <. 0.55 correspond to the E« values ex-
perienced during multiple size reductions using the grate spacinas indicated
Though each mill has a distinctive family of curves, for the ACLF the
70
-------
6.0 r
5.0
4.0
2.0
1.0
2.0
1.5
_ c
1.0
0.5
L 0
0>
N
o
6
o
Z
0
L
Grate Opening
cm (in.)
o 4.8 (1.9)
a 2.5 (1.0)
A 1.8 (0.7)
Characteristic Size
I I I I I
0.5
.0
in.
_L
JL
J
1.0 2.0
cm
3.0
Figure 30. Nominal size as a function of characteristic size
for shredded raw MSW, Gruendler hammer-mill.
71
-------
3.0 r
2.0
u
1.0
A SLF
o ACLF
Characteristic Size
i t i i i i
i i
0
I
1
0.5
in.
i
1.0
i
1.0
cm
2.0
3.0
Figure 31. Nominal size as a function of characteristic size
for shredded ACLF and SLF, W-W hammermlll.
72
-------
1.0
0.8
M° 0.6
C
.2
o
"S 0.4
2 *
•S 0.2
0)
01
Row MSW
Air Classified Lights
Screened Light Fraction
\ 2.5cm 'x
0.6 em (1.0 in.)
(0.3 in.)
-O.Scm
(0.3 in.)
1.0
2.0
3.0
4.0
»0/Fo
Figure 32. Degree of size reduction as a function of the ratio of grate spacing
to characteristic feed size for raw MSW, ACLF, and SLF.
-------
140
c
o
120
100
80
Q)
C
Ld
u
u
Q)
Q.
£/>
60
40
20
•
/ /
Screened Light 0 ( / /
Fraction K * « '0.
Fraction
Air Classified
Lights
Raw MSW
6cm
0.2 0.4 0.6 0,8 1.0
Degree of Size Reduction (Z0), (Fo-Xo)/Fb
Figure 33. Specific energy as a function of degree of
size reduction for raw MSW, ACLF, and SLF.
74
-------
similarity of energy consumption data obtained from both the Gruendler and
W-W mills implies a similarity in comminution mechanism for this material.
In the case of SLF, the data indicate that the Gruendler mill requires more
energy to attain a given degree of size reduction. For example, at 2.5-cm
(1.0-in.) grate openings for both the Gruendler and W-W hammermills, the
specific energy required to attain a degree of size reduction of 0.25 would
be 72 and 24 MJ/metric ton (18 and 6 kWh/ton), respectively. The difference
in the location of the curves for both mills when fitted with the same grate
openings represents a departure from the data presented for the size reduc-
tion of ACLF, where similar grate openings resulted in approximately the
same energy consumption for both hammermills.
To gain a further appreciation of the interrelated nature of the energy
and size variables (E0, Z0, and D0/F0), several diagrams (Figures 34, 35,
and 36) were developed for the three fractions of solid waste studied. The
motivation for combining these three variables stems from the fact that the
ratio of grate opening to characteristic feed size specifies both the degree
of size reduction (which characterizes the product size) and the energy con-
sumption. Consequently, one figure provides both the energy required for
size reduction and the resulting size of the product when the type of mate-
rial, feed size, and grate opening size are specified. A comparison of
these figures shows that for a given value of D0/F0» the steepness of both
Z0 and E0 curves is generally the greatest for the SLF, followed by the ACLF
and raw MSW. Consequently, from the standpoint of minimum energy consump-
tion and maximum particle size reduction, raw MSW is the least difficult
material to grind, followed closely by the ACLF, and more distantly by the
SLF. The reason for the relatively high energy requirement and small degree
of size reduction for SLF is probably that fiber makes up more than 75 per-
cent of this fraction.
This situation may also be viewed in terms of the following two cases.
In the first case, 2.5-cm (1.0-in.) grate openings and a D0/F0 ratio of 0.2
result in a decrease in the degree of size reduction from 0.93 for raw MSW
to 0.86 for SLF (Table 18). In addition, the specific energy increases from
Table 18. COMPARISON OF DEGREE OF SIZE REDUCTION AND
SPECIFIC ENERGY FOR DIFFERENT SOLID WASTE FRACTIONS
°0
cm
2.5
2.5
2.5
2.5
2.5
2.5
(in.)
(1.0)
(1.0)
(1.0)
(1.0)
(1.0)
(1.0)
VFo
0.2
0.2
0.2
0.8
0.8
0.8
Material
Raw MSW
ACLF
SLF
Raw MSW
ACLF
SLF
Zo
0.93
0.92
0.86
0.72
0.71
0.48
Eo
MJ/metric ton
67
79
154
37
40
59
(kWh/ton)
(17)
(20)
(39)
(9)
(10)
(15)
75
-------
O>
uf
x°
1.0
80(20)
90.8
c
_0
£ 0.6
3
•o
0.2
C3»
0>
Q 0
iaO(30)
20(5)
1OO(2S)
Specific Energy (E0)
MJ /MT (kwh/ton)
80(20)
60(15)
l.8cm(0.7ln) 40(10)
1.0
2.0
3.0
4.0
Do/Fo
Figure 34. Degree of size reduction as a function of the ratio of
grate opening to characteristic feed size for raw MSW.
-------
o
X
I
£
o
tvj
c
JO
o
3
•o
a:
0
N
55
L.
a»
a>
O
W-W
Grate Openings
2.54cm (I.OOin.)
1.91 cm (0.75in.)
1.27cm (0.50in.)
0.68cm (0.25 in.)
Gruendler
Grate Opening
2.5cm (I.Oin.)
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 I
.6 1.8 2.0 2.2 2.4 26 2.8 3.O 3.2 3.4 3.6 3.8
Do / FO
Figure 35. Degree of size reduction as a function of the ratio
of grate opening to characteristic feed size for ACLF.
-------
-J
00
o
N
c
o
o
3
•o
O>
0>
o
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
O.I
W-W
Grate Openings
2.54cm. {1.00in.)
D 1.91 cm.(0.75in.)
1.27cm. (O.SOin.)
0.64cm(0.25in.)
Gruendler
Grate Openings
D 2.5cm. (I.Oin.)
F) 1.8cm. (0.7in.)
..!_ 1
0 0.2 0.4 0.6 0,8 1.0 1.2 1.4 1.6 1.8 2.0 2.1 2.4 2.6
Figure 36. Degree of size reduction as a function of the ratio of
grate opening to characteristic feed size for SLF.
-------
67 MJ/metric ton (17 kWh/ton) for raw MSW to 154 MJ/metric ton (39 kWh/ton)
for SLF. The second case assumes the same grate spacing, 2.5 cm (1.0 in.),
and a D0/F0 ratio of 0.8. For these conditions, raw MSW and ACLF yield sim-
ilar degrees of size reduction and energy requirements (namely, Z0 values of
0.72 and 0.71 and E0 values of 37 and 40 MJ/metric ton (9 and 10 kWh/ton),
respectively (Table 18). However, the SLF yields considerably lower Z0 val-
ues and considerably higher EQ values than the other two materials.
RESIDENCE TIME
Although the ability of a refuse shredder to affect size reduction may
be partly described by a study of the feed and product size distributions,
such a study does not truly characterize the actual process of comminution.
One important variable affecting the degree of size reduction is the amount
of time a particle resides within the shredder. This residence time may be
best expressed as a residence time distribution (i.e., a distribution of the
probable time a particle will remain within the comminution device). Most
commercial refuse shredders are some form of hammermill with an integral set
of grates, the spacing of which may be varied to control the product parti-
cle size. It may reasonably be assumed that because of the action of the
hammers, the material within the shredder is uniformly mixed and that there
is no stratification of particles on the basis of particle size. If this
assumption is valid, then the residence time distribution may in turn be
assumed to be the same regardless of the size of the particle. The grates
thus act as a classifier, passing those particles smaller than the grate
spacing and returning the rest to the grinding zone where the same residence
time distribution that is applied to the feed is again applied to the recy-
cled material. This process is shown schematically in Figure 37. In this
figure, Figure 37, f and £ are the vectors describing the circuit feed and
product size distributions, r is the recycled material size distribution,
and mf and nip are the size distributions of the material entering and exit-
ing The actual grinding zone. Figure 38 shows the form of the residence
time distribution R(t), for the test shredder as determined from unit im-
pulse tracer tests. The mean residence time may be determined from either
the residence time distribution or the equation
Scf
D
=
I-C
s =:
Figure 37. Schematic representation of milling process,
79
-------
30
60 90
Time, T (sec )
120
150
Figure 38. Residence time distribution for
size reduction of raw MSW.
80
-------
T - {j 3.6 (5)
where: T is the mean residence time (sec), H is. the mill holdup (kg), which
is the instantaneous mass of material within shredder; and Q is the flow
rate (tons/hr). There is some evidence that the form of a normalized
residence time distribution, R(o), is invariant with flow rate where R(t)
and R(e) are related by
R(t) = R(o)/r (6)
where: is the mean residence time (sec) and 9 = t/r. Figure 39 shows the
normalized residence time distribution for the test shredder.
MILL HOLDUP
Both the dynamics of hammermilling and the highly compressible nature
of refuse lead to the belief that there may be some relationship between
flow rate and holdup. To test for this relationship, the Gruendler hammer-
mill was fitted with 1.3- and 2.5-cm grates, and holdup/flow rate data were
collected for a wide range of operating conditions. The results of regres-
sion analysis on the data show that for the 1.3-cm grate spacing
Qj 3 « 0.039 H0'95 (7)
and for the 2.5-cm grates
0.086 H°*94 (8)
where: Q is the output flow rate (tons/hr) and H is the mill holdup (kg).
Flow rate is shown as a function of holdup because laboratory (as well
as most commercial) conditions are such that it is the output rather than
the input flow rate that is measured. Of course, under optimum steady state
conditions, the converse of Equations 7 and 8 also hold, i.e.,
H - 30.4 QL31*05 (9)
H = 13.6 Qg.s1'06 (10)
Figure 40 shows the curves representing Equations 7 and 8 (or 9 and 10).
Since flow rate is more readily measured than holdup, substitution of
Equation 9 into Equation 5 and correcting for the different units of Q and H
yields:
T - 109.4 QL30'05
and
T - 49.0 Q2.5
81
-------
I 2
Normalized Time, 9
Figure 39. Normaltzed residence time distribution
for size reduction of raw MSW.
82
-------
2.5-cm Grotes^
10
0>
o
0>
*-
o
w
O
5 —
50 100
Holdup, H (kg)
ISO
200
Figure 40. Flow rate as function of mill
holdup for raw MSW.
83
-------
Thus an increase in flow rate leads to increased holdup and a longer mean
residence time.
The effect of the grates is twofold. Not only do the grates act as an
internal classifier for the mill, but they also act as an impedance to flow
of material through the shredder. The resistance to material flow is a
function of the actual physical characteristics of the grates; i.e.,
1 H0.95
A' Atotal>
where: gs is the grate spacing, A is the space between the grate bars, and
Atot = A + total cross sectional area of the grate bars. Examination of
Equations 7 and 8 show that for constant holdup, the output flow rate
roughly doubles as the grate spacing is doubled.
Calculation of the cross sectional areas of the grate bars and the
spaces between the bars show that the area through which material could flow
is effectively doubled in the test mill by using the larger grate bars. In
other words, changing from the 1.3- to the 2.5-cm grates doubled the effec-
tive area through which material could flow, thus leading to the propor-
tional increase in product flow rate.
Through the use of X0 as a convenient measure of the coarseness of a
size distribution, it is possible to demonstrate, as in Figure 41, that as
flow rate increases, the shredder product size distribution becomes steadily
smaller. This effect is explained by recalling that Equations 11 and 12
indicate that the mean residence time increases with flow rate. Not unex-
pectedly then, a longer residence time within the mill leads to a smaller
characteristic particle size.
It was also found that power requirements increase with holdup (Figure
42). On the other hand, it was noted that increasing the moisture content
for constant holdup led to a decrease in power requirements. This effect is
shown in Figure 43 and may be explained by the fact that fibrous material
(which is the major component of refuse) tears more readily as it becomes
wetter. Since the test mill had a known fixed volume, power draw can be
related to the actual density of material within the mill. Alternatively,
since flow rate is a readily measured quantity and flow rate and holdup are
related by Equations 9 and 10, power requirements can also be explained 1n
terms of flow rate. Analysis of data obtained with the 2.5 cm grates shows
the data to be fitted by the equation
P = 18.6 Q1'1 (1-MC)0'36 (14)
where: P is the instantaneous power draw (kW), MC is the fractional mois-
ture content, and Q is the flow rate (tons/hr).
HAMMER WEAR
The parameters Z0 and D0/F0 were used to present an overall picture of
hammer wear as it pertains to the size reduction of raw MSW and SLF using
84
-------
1.2
1.0
0.8
o
X
4)
N
O
-------
200
150
A 0-
O
O
V
a
I -2
2-3
3-4
4-5
<3> 5-6
* 6-7
$37-8
O 8-9
0 10-11
0)
5
O
a.
100
50
1
1
10 20 30
Moisture Content (%)
40
50
Figure 42. Net power as a function of
holdup for raw MSW.
86
-------
200 r
150 -
- 100 —
Q>
z
A Experimental
Predicted
50
Holdup, H (kg)
100
150
Figure 43. Net power as a function of moisture
content for raw MSW.
87
-------
the Gruendler and W-W hammermills. Each particular material was found to
occupy a characteristic region of wear when expressed as a function of
degree of size reduction (Z0) or the ratio D0/F0.
The functional dependence of hammer wear on the degree of size reduc-
tion shows a greater degree of wear for raw MSW than for SLF (Figure 44).
Although the actual data for SLF and raw MSW do not overlap in the same
areas of degree of size reduction, extrapolation of the curves does allow
some comparison of the wear characteristics of each material. In particu-
lar, for a specified value of Z0, hammer wear resulting from shredding of
raw MSW in the Gruendler mill can be seen to be approximately five times as
great as that experienced for shredding SLF with the W-W mill.
The smaller degree of wear experienced for SLF is paradoxical when one
recalls that from an energy consumption standpoint, SLF required signifi-
cantly more energy than raw MSW to achieve the same degree of size reduc-
tion. This paradox seems to confirm the hypothesis that energy consumption
and wear are independent of each other.
The information contained in Figure 44 can be expressed in terms of the
ratio of grate opening size (D0) to feed size (F0) (Figure 45). This figure
indicates wear experienced as a result of size reduction of raw MSW and SLF
for particular conditions of grate opening and feed size. The curves occupy
distinctive regions of the diagram, and the greater degree of wear for raw
MSW compared to that of SLF is again apparent.
-------
0.10
0.01
JO
•k
w
O
0.001
Raw MSW
Gruendler Hammermill
Extrapolated.
Regions
Screened Light Fraction
W-W Hammermill
• Cast Manganese Steel
A Haystellite Tungfine
o Stoody 31
>,D Amsco Super 20
O Coated Tube Stoodlte
_l 1 L
0.2 0.4 0.6 0.8
Degree of Size Reduction (Z0)
0.10
0.05
0.0010
Figure 44. Hammer wear as a function of degree of
size reduction for raw MSW and SLF.
89
-------
TT~I—r
O.I -
o
^s.
.0
O
a>
0.01
i—i—i—r~
Raw MSW
Gruendler
Hammermill
T 1 1 1 1 T
Extrapolated
Regions
SLF
W-W Hammermill
O.I
0.05
c
o
0.025 o
0.01
0.005
o
E
>*
o>
o
0*
0.0025
0.2 0.4 0.6 0.8 1,0 1.2 1.4 1.6
Figure 45. Hammer wear as a function of
for raw MSW and SLF.
90
-------
SECTION 6
COMMINUTION MODELS
The concept of employing matrix modeling to simulate the comminution of
refuse in a swing hammermill was introduced in previous publications
[1,2,3]. In general, matrix models are phenomenological and macroscopic in
nature. They have been developed as a means of overcoming the inability of
the earlier energy-yersus-degree-of-reduction models to provide insight into
the nature of the size reduction process. Although some microscopic models
that deal with crack formation and propagation are physical in nature, they
deal with single events and as such are not suitable for describing complete
processes in a mill. The basic feature of the matrix model is its ability
to predict product size through the use of the selection and breakage func-
tion. In effect, these functions describe what portion of the material will
be selected for breakage in each pass, and over what size spectrum the
breakage will occur.
Most studies dealing with the application of time-discrete, size-
discrete models described by Callcott [4], Callcott and Lynch [5], and Lynch
and Draper [6] have been applied to homogeneous brittle materials. Later,
this approach was extended by Obeng [1] and Obeng and Trezek [2] to hetero-
geneous refuse, a solid waste that contains only on the order of 25 percent
brittle materials. On the whole, this time-discrete, size-discrete method
of modeling the comminution process in a swing hammermill has been only par-
tially successful because these models are essentially black-box representa-
tions of the comminution process, in which material transport is implicitly
assumed to occur by plug flow. Also, because time is not an explicit vari-
able in the governing equations, a detailed description of mill dynamics is
not possible. In these models, the mill matrix coefficients are not easy to
interpret physically. Though these models can be used to predict size dis-
tributions and illustrate qualitative feed rate and moisture effects, cer-
tain basic questions such as why and how the feed rate affects the product
size distribution cannot be answered in a fundamental manner.
Motivated by the successful extension of the size-discrete, time-
continuous model used for describing mineral batch ball milling to contin-
uous laboratory mills, [7] the matrix approach to the analysis of refuse
comminution pursued in the initial studies [1,2,3] was similarly modified to
include a residence time distribution. This distribution, the mean value of
which is essentially the mill holdup divided by the feed rate, is a major
factor in determining the product size distribution. In the case of the
overflow discharge ball or rod mills, the mass of holdup is assumed constant
so that the mean residence time is solely a function of the feed rate.
Unfortunately, this simplification cannot be extended to the hammermi11ing
91
-------
of refuse because of its highly compressible nature. Thus the mean resi-
dence time will vary with both holdup and feed rate.
The model presented here provides a means of simulating the hammermill
size reduction of refuse based on the actual physical variables of the proc-
ess. It addresses the specific issues of: (a) how the feed rate and mois-
ture content affect the mill holdup, (b) how the mill holdup affects the
energy consumption and residence time, and (c) how the selection and.break-
age functions can be estimated for different operating conditions.
SIZE-DISCRETE, TIME-DISCRETE MODELS
A brief review of the initial application of matrix modeling to refuse
comminution is given to aid those readers not familiar with this area and to
illustrate the improvement in the method through the use of the size-
discrete, time-continuous approach. By representing the particle size range
of a particulate assembly by n discrete size intervals
i = 1,2, ... n , (15)
(16)
the particulate assembly may be described by a vector M, i.e.,
M = Hm = H
nv
nr
where: H = total mass of the assembly, and
mj = fraction of JJ1 within the iH size class.
Lumping the various breakage mechanisms that may exist in the mill
introduces two additional parameters — the selection function S and the
breakage function £. The selection function is a diagonal matrTx whose
elements represent the fractional rate at which material in each size class
is broken, i.e.,
0 ; i 4 J
s; 1 - J
(17)
where:
broken.
fractional rate at which material in the i Ulsize class is
The breakage function is a lower triangular matrix whose elements rep-
resent the fraction of material that enters a higher size class after being
.broken from a lower one. Note that in this case, the term "higher" indi-
cates a particle size that actually is smaller, whereas the converse 1s-
92.
-------
indicated by the term "lower". The assumption also is made that all parti-
cles, once broken, leave the original size class and that there is no
agglomeration (clumping of particles). The breakage function thus becomes:
0 ; i = j
BJJ - 0 ; 1 < j (18)
blj51 >J
where: bii = the fraction broken out of the j*Jl class, which reports to the
th
the lass.
In the time-discrete formulation, the breakage process may be regarded
as a series of discrete breakage events in which change of the fractional
amount of the particulate assembly within a given size class is the differ-
ence between that which is broken out of that class and that which enters
from other size classes in the course of a single breakage event. The
formulation is:
-
A(Hm.) = -s. Hm. + E s.b..m.h
i ii j * j ij j
(19)
where: H « total mass of material in the mill at the time of the breakage
event.
Under steady-state conditions, H may be factored out, and the result for a
single size class is
i-1
&mi - ~ siV ^ bijsjmj
The n difference equations representing all n size classes may be described
by the single matrix expression
Am = - S mn + B S mn
" = == (21)
or Mm- (± - Q) £ TOO
where I • n x n identity matrix, and
m m original size distribution.
—o 3
Since Ann is the change in size distributions resulting from a single
breakage event, Equation 21 may be rewritten as
rnj - njo » -(WJ.JS r^
or --- (22)
21 - II - (JL-i)H!k>
where: mj = size distribution after a single breakage event.
93
-------
If it is assumed that ^ and J3 are invariant with respect to the number
of breakage events, Equation 22, may be extended to apply to any number of
breakage cycles by replacing m, with fn^+i, and ITL with rn. such that
Hi+1 = [l_ - (.L-B_)_S]mj
Multiple use of this equation, replacing rn-j each time with mj+i from the
previous cycle, shows that for more than one cycle of breakage, Equation 22
becomes
where N = the number of cycles of breakage.
The preceding method is the one employed by Obeng [1] in his research
in which both single and multiple cycles of breakage were studied. For mul-
tiple cycles, N was estimated from
c
Nr = m-1 (24)
C
where: m-1 = n(xj/xm)/ln(x1/x2),
xi = maximum feed size, and
xm = maximum product size.
Subject to the important restriction that the eigenvalues must be positive
[8,9] for
1- (I-|)l» M - 1 - d-bii)si, i = 1,2 ..... n
Equation 23 may be reduced by a similarity transform to [8]
TJ (Nc) T-l mo (25)
in which £ is the n x n lower triangular matrix composed of the column
eigenvectors of 1 = (,!.-£)£, such that:
0 1 < j
1 1 - J (26)
j](Nc) is an n x n diagonal matrix with elements
0 i 4 j
Nc (27)
c
Equation 25 is obviously more amenable to use than .is Equation 23 because it
eliminates the need to compute the NcJ-!l power of [^ - (L-BJSJ. Implicit in
the use of Equations 23 and 25 is the existence of plug fTow conditions. jn
94
-------
the mill. Such an assumption has not been substantiated by any experi-
mental evidence [10,11,12].
SIZE-DISCRETE, TIME-CONTINUOUS MODEL
In the second approach, the particle size range of any particulate
assembly may again be divided into n discrete intervals
x. x.i i =
where x. is some size usually associated with a set of standard screen sizes.
A particulate assembly whose individual particle sizes are within this range
is then represented by the vector M^ such that
jjl = H ni =|H m,mp ... m. ... ml
where: H is the total mass of the assembly and m,. is the fraction within
the x. x.+, size class.
It follows that if f4 represents the material within the grinder at any par-
ticular time, H represents the mass of material held in the mill (the mill
holdup).
A mass balance on the i— size class at time t yields
i-1
fa [H m.(t)] = -s..(H m.(t)) + I b. .s.(Hm.(t)) (28)
j=l
where: H « mill holdup,
m-j = mass fraction in the i£ll interval,
s-j » fractional rate at which material is-broken out of the
iM interval (the i*il selection function), and
b-jj = fraction of material which, once broken, reports to the
iiil interval from the jlll interval (the breakage function).
If it is assumed that Sj and b-jj are independent of both x-j and time, then
the model may be linearized and the breakage process represented by n simul-
taneous differential equations, which may be written in matrix form as
[H m (t)] = -[J.-BJ JS H m (t) (29)
where: J, = n x n identity matrix,
S. = n x n diagonal matrix of selection functions, and
£ = the n x n lower triangular matrix of breakage functions.
For a batch process with an initial size distribution 515 (o), H is
constant and Equation 29 is reduced to
95
-------
St I3>(t)] --U-B] Sr^ (t) (30)
the solution of which is
ra^t) = exp [-(1-|) |t] 1115(0) (31)
Equation 31 may be evaluated through the iterative solution for the
exponential
OQ
exp[-(I-B)S t] = jS0 [-(I-B) S t]J/ji (32)
However, when s-j 4 Sj for all i and j, a similarity transform leads to an
easier solution.
Through the use of a constant n x n matrix T, a linear transformation
may be made between 015 and a new state vector m^*", such that:
nfc(t) - Tn»5*(t), B>*(t) = T'1 m^t) (33)
Replacing (55 by m^* in Equation 30 gives the equation
^ [T mb*(t)] = -a-|]S T mb*(t) (34)
°r £t Cmb*(t)] - T-l[-(r-|)S]T mb*(t) (35)
Assume that
T-l [-(L-Bl) |] J = £ (36)
where: £ is an n x n diagonal matrix with diagonal elements pj, P2, •••> Pn<
It follows that the eigenvalues of the, left hand side of Equation 36
are equal to the diagonal elements, PJ, i = 1,2, ..., n. Solving for the
roots of the characteristic equation shows
I\L - I"1 C-l-i) |] T| - 0 (37)
However
Hence the characteristic equation may be written as
IT"1 [jJ + (I-B) |] J| = 0 (39)
from which T~* and T^ may be factored, leaving
IxJ. + (I-B) S.| - 0 (40)
and thus showing that the eigenvalues of -Q.-B.) S. are equal to the eigen-
96
-------
values of J_~ [-(I_-B_)$JT, which in turn are the diagonal elements of £. In
addition, Tince I and" ];[ «re diagonal and J3 is lower triangular, -(i-T)^
becomes lower trTangular, the main diagonal of which are the elements s-j,
i = l,2...n. Standard matrix algebra tells us that the eigenvalues of a
lower triangular matrix are simply the elements of the main diagonal. From
this it is obvious that
£.-•-!•
Substituting Equation 41 into Equation 36 shows that
i- [mb* (t)] . -S.mb* (t)
which has the solution
* (t) = mD *(o)
(41)
(42)
(43)
where
-sit
e
o
o
o
-snt
By using Equation 33, Equation 43 may be transformed to give
mb(t) - T J.T-1 mb(o)
(44)
Because of the decoupling that occurs when the original differential
equation is transformed, the £-matrix is known as the modal matrix. Return-
ing to Equation 36 and premultiplying both sides by T[ results in
(45)
I • I
The right side of this equation becomes
T P -
P1T11
PlT21
PI Tni
P2 T12
P2 T22
P2 Tn2
• • • PnTln
. . . pnT2n
*
(46)
and the left side may be rewritten as
[-(1-B )S] T . [-(I-l)S] [uj
(47)
97
-------
where: u^ = a column vector representing the ith. column of T_.
Equating, column by column, Equations 46 and 47 results in
-[I-B] ! uj = p^ (48)
or
[Pi I+ (I-|) |] Hi - ° (49)
which establishes a set of n equations for which Uj may be solved when s^
is substituted for p-j. Solving for the values of the eigenvectors u-j yields
the modal matrix J. In this case, a set of eigenvectors that satisTies
Equation 49 is such that
0 i < j
T - 1 i = J (50)
z
k=j
bik sk
T
kj
1 > J
Equation 44 is the solution of the linearized, lumped-parameter,
size-discrete, time-continuous equation for batch grinding. To extend the
equation to continuous grinding, it is necessary to introduce the concept of
residence time. Residence time is defined as the length of time a particle
within a size interval remains in the mill. Since no single value for the
residence time will hold in all cases, a residence time distribution is pro-
posed that is the probable time a particle will remain within the mill. If
it is assumed that all size intervals are characterized by a single
residence time distribution, then the open circuit mill is the average of
the batch response weighted with respect to residence time, i.e.,
R(t)dt (51)
o "
where: R(t) is the residence time distribution. Substituting the relation
given in Equation 44 for m^ gives the solution for continuous grinding
mpp . T[ / ^J(t) R(t)dt] T-1 mcf (52)
~" o ~~ ~
where: m^Q = continuous product size distribution, and
= continuous feed size distribution.
Now, letting
where: t = time and
T = mean residence time = ratio of holdup to feedrate, H/M ..
The result, by substituting, is:
-------
(53)
where: _JC(T) = / exp[(-SiT)«] R(a)de
~" 0
J
J
A block diagram of the open circuit mill is shown in Figure 46.
2 represents the mill matrix,
2 = 1 OP (T) r1
(54)
If, after passing through the mill, the product is classified and a
portion recycled to the mill feed, a closed circuit is formed that signifi-
cantly alters the final product. The closed circuit schematic is shown in
Figure 47. The classifier matrix £ of Figure 47 is an n x n diagonal
matrix, the entries of which represent the fraction within each size class
that is recycled to the mill.
For the closed circuit [13,14]
(55)
(56)
Figure 46. Diagram of open circuit milling.
Mfmf
M
cf
= I\F
D
r
v/
'
I-C
Figure 47. Diagram of closed circuit milling.
99
-------
£Mcp (57)
Mf+R (58)
Substituting Equation 57 into Equation 58 gives
Mcf = Mf * i Mcp (59)
which in turn may be substituted into Equation 55 to show that
By rearranging terms to get
(i-D£) M
the left side may be premultiplied by (^.-2^.)"
Mcp = Or]) C)-1 DMf (60)
Finally, if Equation 60 is substituted into Equation 56, the result is
D-1 Mf (61)
or
Mpmp = (UC) j> (Irgg)-1
Since under steady-state conditions Mp = Mf, the solution to the continuous
closed circuit mill product equation is:
IMPLEMENTATION OF THE MODEL
In hammermill size reduction of MSW studies, it is virtually impossible
to determine selection and breakage functions for even the coarsest size
classes by an examination of zero order production relationships as des-
cribed by Herbst [15], Herbst and Mika [16], and Grandy and Fuerstenau
[17]. On the other hand, nonlinear, optimal parameter estimation has been
used with some success in the past [18,19] and was chosen in this instance
as the technique most likely to prove useful.
Parameter estimation itself is based on two fundamental assumptions:
1) The size reduction process is linear, i.e..,
si ^ s-j(m(t)); ial, ..., n
and
4 bij(m(t)); j=l n-1; 1 = j+1, .... n
100
-------
or, in other words, si and bij, and not functions of the size distribution
existing in the mill (under steady-state conditions, m-j(t) is relatively
constant).
2) The size-discrete breakage function can be normalized. This
implies that for a constant ratio between successive sizes,
and
B-JJ a BI-J+I, i; J=l» •••» n-2; i = j+1, ..., n-1
where
n
Bij = 2 bkj
k=i
The second assumption means that the breakage function depends only on the
difference, i-j, and not independently on i and j. It also implies that if
B.JJ is plotted against (x^/Xj Xj+j)l/2, a single curve will result for all
i,j. Finally, since the oH size class contains all material finer than xn,
it is apparent that
bnj - E- bki; J-2» .... n-1 (63)
k=n-j+l
With single-size class feed and short grind times, some batch size-
reduction systems are characterized by what is known as zero-order kinetics
of appearance of material in size classes much smaller than the feed size.
If this is true it follows that
where: Y-j(t) » cumulative fraction finer than x^, and
«i a zero order production rate constant.
In the limit as t » 0
and therefore Equation 64 reduces to
Bi,l sl * Ki (65)
It has been experimentally noted [20] that the zero-order production rate
constants, K., are of the form
a/2
K, = K" *i ; i«2, .... n (66)
xl X2
101
-------
where: K* and o are time-invariant constants.
When Equation 66 is applicable and is substituted for K-j in Equation 65,
then the feed size cumulative breakage functions may be found from
a/2
(67)
xl X2
The more general constraint
shows that
i i "* i ~ i* J > •••> l—
B . , B . .
il ij
(68)
which, in light of Equation 67, becomes
x.
.J.
xk
kj
(69)
If Equation 68 is to be satisfied, the breakage function must be subject to
being normalized in the sense of Epstein [21], i.e.,
'1
x. x
j j+l
a/2
(70)
The remaining selection functions may be established from Equation 68
with i=n. Rearranging yields
...M
Substituting the expressions for B . and K in Equations 70 and 66 gives the
~«r.,H- RJ n
result
sj - sl
XJXJ+1
X1X2
(71)
Finally, the individual breakages may be obtained from the difference
n'
(72)
from the normalized equation
102
-------
and from Equation 63.
In the actual use of the model, it is necessary to find a means of
evaluating the integral of Equation 53. One way to do this is to use a
numerical scheme. One of the simplest is Simpson's approximation
b h
/ f(x)dx = £ [f0(x) + 4fj(x) + 2f2(x) + 4f3(x) + ... + 2fn_2(x) +
d
4fn_1(x)+fn(x)] (74)
where: h = (b-a)/n, and
n = number of intervals, 2,4,6,...
Since six values for the residence time distribution can be obtained
experimentally, this approximation to the integral is relatively easy to
implement.
The procedure followed for the parameter estimation by means of a
nonlinear optimization scheme is shown in Figure 48. The general procedure
is as follows: *
1. Initial guesses for s^, K*, and a are supplied.
2. Equality, inequality, and tolerance constraints are supplied.
3. Experimental values of mf, mp, R(e), T and £ are entered.
4. A predicted m is calculated.
5. The percent error between the predicted and experimental values of
m is calculated.
-P
6. S-], K*, and a are adjusted to reduce the objective function, and
steps 2 through 5 are followed again.
7. If AS-), AK*, and Aa are less than the tolerance criterion, then
exit; otherwise go back to step 6.
The procedure followed for mill simulation uses the results from the
parameter estimation to predict the mill product for untested operating
conditions. The steps are only slightly different from those employed in
parameter estimation. Figure 49 is the flow chart of the steps followed.
They are:
1. Enter the maximum size and number of size classes;
2. Enter size distribution, residence time, flow rate, s-|, K* and a
data.
103
-------
Start
Read
optimization
parameters
Read
experimental
data
Compute
required
matrices and
predict
product
Compare
with experimental
product and com-
pute percent error
no
adjust s.|,K i
and a to
minimize sum
of squared
errors
J
Figure 48. Nonlinear optimization flow chart
for parameter estimation.
104
-------
Start
Read
-f Xmax' >R-*N'-
Compute
x, |, BI b
I. I"1. Jc> B
5f
Yes
Compute
m
-P
Write
2» !f, Yp, mf, mp, T, T
Jc, D, S, B, b, H
No
Closed Circuit
Model ?
Compute
V -YP
Write
.-1
1
Stop
Stop
Figure 49. Flow chart for mill simulation calculations;
105
-------
3. If a closed loop model is to be used, enter the experimental
classifier matrix.
4. Calculate the size classes (a fixed ratio x-j/x-j+i = 2 was used).
5. Calculate the remaining selection functions by Equation 71.
6. Calculate the cumulative and individual breakage functions from
Equations 63, 70, 72, and 73.
7. Calculate the modal matrix, _T, and its inverse.
8. Calculate the matrix, j£, by either Simpson's approximation or
the use of a functional model for the normalized residence time
distribution.
9. Calculate the open circuit mill matrix.
10. If an open circuit model is desired, calculate the product and
exit; otherwise, calculate the closed circuit matrix.
11. Calculate the product and exit.
RESIDENCE TIME DISTRIBUTION
The residence time distribution model implies neither a fully mixed nor
a plug flow, but rather a system whose product lies somewhere between the
two limits. Obviously, two of the major parameters involved are the shape
of the normalized residence time distribution curve and the mean residence
time. With most systems, it may reasonably be assumed that the form taken
by the normalized form of this curve may be easily found from
R(t) = R(G)/T (75)
where: R(e) = normalized residence time, and
T = mean residence time.
The experimentally determined form of the residence time distribution
as determined from unit impulse tests is shown in Figure 50 and the normal-
ized form in Figure 51. The data from which these graphs were constructed
are taken from Shiflett [22]. The results of the tests were plotted at the
midpoints of each sample time interval, which in turn was determined by the
speed of the product conveyor and the distance between conveyor cleats. A
complementary function, R'(t), was.formed from Figure 50 by selecting values
of R(t) every 10 sec for the interval 0 < t £ 140, summing the values, and
dividing each by the sum (since the totaT must be no greater than unity).
The complementary function thus becomes
R'(t) =l-z DUt)/ z lUt)]
106
-------
60 90
Time.T (sec.)
Figure 50. Residence time distribution,
(Data from Shlflett (22))
107
-------
I 2
Normalized Time , 9
Figure 51. Normalized residence time distribution,
(Data from SMflett (22))
108
-------
and is plotted in Figure 52. It is interesting to note that this represen-
tation shows the grinder to approximate a pure delay and a perfect mixer in
series, the time constant for which is found from the slope of the straight
portion of the curve.
The mean residence time, T, may be found in several ways:
1. From flow rate/holdup data
(76)
where: H = mill holdup (kg),
Q = flow rate (tons/hr), and
k a conversion factor between kg-hr/ton to sec • 3.6
2. From the residence time distribution curve
T = £ R,-(t)t./ £ R(t) (77)
1=1 1 ] 1.1
3. From the complementary function R'(t)
T = delay + time constant (78)
As shown by the data listed below, the mean residence times calculated by
each method are all in fairly close agreement:
Method T
1 51.4
2 45.7
3 44.7
Figure 53 shows the variation in the characteristic size with the
change in mean residence time. As the figure indicates, the charactertistic
size shifts toward the finer sizes with an increase in T. This fact again
demonstrates the general validity of the assumptions on which the model is
based.
SIMULATION RESULTS
Optimization Technique
To use Equations 55 or 62 for mill simulation, it is first necessary to
determine values for the selection and breakage functions. The selection
function for large size classes normally is determined by way of batch
grinding tests with single-size feed. Examination of the rate of disappear-
ance of the feed (i.e., first order kinetics of disappearance) after incre-
mental amounts of grinding yields sufficient information to determine the
initial selection function through the use of
109
-------
10
^ -\
I- 10
c
o
u
c
0»
£
at
Q.
e -2
< 10
-3
10
I
I
50 100
Time, T (sec)
150
Figure 52. Complementary residence time function.
110
-------
1 . 2
1, 1
1.0
. 9
. 8
.7
8 .6
o
X
. S
SO.O 55. 0
Mean Residence Time (sec)
60.0
Figure 53. Characteristic size and variation
with mean residence time.
Ill
-------
- 3£ [In
The remaining selection functions may be obtained by employing a functional
relationship such as Equation 71. Feed-size breakage functions may be ob-
tained through a study of the zero-order production rates of material in
size classes that are much smaller than the feed after short grinding
times. The remainder of the breakage functions are found through the as-
sumption of "normalizability" of the breakage function.
Because batch tests could not be conducted with the system discussed
here, an alternative means of determining selection and breakage functions
had to be found. If the techniques described above are followed, the
unknown selection and breakage functions are determined from the initial
selection function, the particle size under question, and the zero order
production constants. A log-log plot of the zero order production rate con-
stants against particle size yields values for K* and a in Equation 67.
Since the particle sizes are known, it follows that once the feed size se-
lection function and K* and » from Equation 67 are determined, the system is
completely described. Therefore, rather than having to determine the n + 1
unknowns (n zero order production constants and the feed size selection
function) of the n-size class system, it is necessary only to determine the
values of the three parameters s-j, K*, and o. For systems with which the
batch tests cannot be made, a possible technique for determining S], K* and
a is the use of a nonlinear optimization approach.
Austin and Klimpel [19] did some work along such lines, though they
determined experimentally values for the breakage function and were seeking
alternative means of obtaining selection functions. The objective function
chosen to show how well the predicted product matched that of the experimen-
tal was the sum of the squares of the errors between predicted and experi-
mental values. An alternative objective function included a weighting fac-
tor for those size classes for which experimentally determined selection
functions were available. The optimization technique employed by the two
researchers was the conjugate gradient technique proposed by Goldfarb and
Lapidus [23]. The conjugate gradient technique can be used in solving (min-
imizing) the general nonlinear, n dimensional objective function, subject to
certain univariant constraints, i.e.,
0 1 *i 1 ci » 1-1. 2,000, n (79)
where x^ = a principal variable of the objective function, and
C-j = a limiting value.
In the present study, the assumption that Equation 68 holds for all
values of i imposes an additional constraint, namely that
(80)
B2,l
K*
" sl
x 2
X2
X1X2
o/2
= 1
112
-------
Further equality constraints that may be placed on the problem are
£ '1 predicted • 1 '81>
and
mi experimental - "l predicted - 0; i ' X'2 n (82)
It Is possible to reduce the number of equality constraints involved by not-
ing that satisfaction of the last set of constraints automatically satisfies
the second (although the converse is generally not true). An objective
function to be minimized may then be formed through a combination of the
first and third equality constraints.
The optimization technique selected for use in the present work is
based on Powell's nonderivative search method [24] for unconstrained prob-
lems. Powell's basic technique was modified by Fiacco and McCormick [25] to
include the use of penalty functions for constraints and was further modi-
fied and adapted for use on a CDC 6400 by Radcliffe and Suh [26]. Complete
descriptions of the techniques and modifications are available in the source
articles and texts and hence are not presented here. The constrained objec-
tive function to be minimized took on the exterior penalty function form
ma mh
u(x) - F(x) + f- E [g^x)]2*!- E [h.(x)]2 (83)
rg 1.1 n rh 1.1 1 "
where: F(x) » unconstrained objective function,
9iT\.) * inequality constraint,
hjjxj . equality constraint, and
rg,rn • inequality and equality constraint weighting factors that
decrease as each suboptimum is found.
While operating in a feasible region, the g-j(x) are set equal to zero, leav-
ing only the weighted equality constraints and the unconstrained objective
function. It is to be expected that the choice of unconstrained objective
function and equality constraints would have a bearing on the efficiency of
the solution.
For the current problem, it was found that in the initial stages of the
search, convergence was facilitated by forming the unconstrained objective
function from
F(x) = 1.0 - B21| (84)
and by allowing the squares of the errors to be added as equality con-
straints. Once a feasible region was established by this method for the
variables sj, K*, and a, it was possible to solve the problem as an uncon-
strained minimization problem using the values of S], K* and a previously
found as initial guesses and the sum of the squares of the errors as an
objective function.
113
-------
Results
Several minimizations were made in correspondence with different flow
rates and experimental size distributions. The results are presented in
Table 19. The sum of the squares of the errors ranged from 0.0614 to
0.0028, with an average of 0.0200. The exponent a from Equation 67 remains
relatively constant (aave = 0.835, whereas the constant K* ranges between
0.19 and 0.41. Values of K* are plotted against feed rate in Figure 54.
Although K* decreases with flow rate before achieving a constant value, the
ratio of K* to S], is, in almost all cases, close to unity. The initial
selection function is plotted in Figure 55 and shows a trend to decrease
with flow rate from 1 to 4 metric tons/hr, and thereafter to maintain a con-
stant value of approximately 0.29. The implication therefore is that for
low flow rates, the mass of material within the mill (how densely it is
packed) affects the selection function; but as the flow rate increases fur-
ther, the change in mill holdup has little or no effect.
The model suffers from its failure to give a value for 621 equal to one.
At the lowest flow rate, 831 is equal to 0.35, and at times it even exceeds
1.0 (which, according to the definition, is physically impossible). The
difficulty is caused by attempting to apply Equation 68 to all values of i,
when in general it has been noted [17] that the zero order production rate
constants for the coarse size classes cannot be characterized by the same
equation as for the fine sizes. The cumulative breakage function presented
in this section must then be taken to be the best compromise between the
characteristic equations for both the coarse and fine size ranges.
Figure 56 shows the relationship between the initial cumulative break-
age function, B2i, and flow rate. As might be expected from the fact that
K*/S] generally is close to unity (particularly at higher flow rates) and a
is constant, 621 remains relatively constant.
A listing of the optimization program is presented in the Appendix.
Since the values of S], K*, and 0 tend to remain constant for higher flow
rates, a flow rate of 7.4 metric tons/hr was chosen as a test of the ability
of the model to predict the output of the mill. With S] = 0.290, K* = 28,
and a = 0.804, a product size distribution was predicted and compared to the
experimental results. Figure 57 graphically demonstrates differences be-
tween the predicted and actual measured size distributions. Except for the
first less-than-zero value for mj, the predicted m^ values follow.the exper-
imental values fairly closely. The sum of the squares of the errors in the
simulation trial was under 0.02. This fact demonstrates that the compromise
breakage functions obtained as a result of parameter estimation permit rea-
sonably good results to be obtained with simulation. A listing of the simu-
lation program is also provided in the Appendix.
CONCLUDING REMARKS
Despite the fact that 821 is not predicted to equal unity, the size-
discrete, time-continuous model adapted to describe the swing hammermill
size reduction of refuse is shown to be a close approximation to the real
system. Significant differences between this system and typical ball mill
114
-------
Table 19. SUMMARY OF OPTIMIZATION RESULTS
sl
.405
.297
2.95
.300
.106
.407
.336
.514
.306
.332
.071
.073
.287
.286
.285
.289
.290
.290
.291
K*
.190
.347
.347
.336
.311
.411
.253
.396
.294
.252
.292
.300
.314
.308
.308
.290
.284
.281
.276
K*/s]
.469
1.168
1.176
1.120
2.934
1.010
.753
.770
.961
.759
4.113
4.110
1.094
1.077.
1.081
1.003
.979
.969
.948
a
.826
.804
.804
.804
.809
.977
.805
1.057
.804
.814
.841
.898
.804
.804
.804
.804
.804
.804
.804
F(x)
.0119
.0271
.0614
.0504
.00597
.00458
.0161
.00166
.0215
.00745
.00964
.00277
.0196
.0549
.0249
.0168
.0135
.0123
.0168
B21
.351
.884
.890
.849
2.220
.720
.569
.553
.726
.573
3.090
3.010
.829
.815
.817
.759
.742
.732
.717
Q
(tons/hr)
.9
1-4
1.5
1.5
2.1
2.4
2.4
2.4
3.0
3.4
3.8
3.8
3.9
4.6
5.0
6.2
6.6
7.0
7.4
115
-------
0.4 —
0.3
0.2
A
A
O.I
A A
I
I
I
2 46 8
Flowrote, Q (tonnes/hr)
Figure 54. Variation of K* flow rate.
116
10
-------
0.4
_ A
0.3
u
0)
v>
A A A
0.2
O.I
I
I
468
Flow rote, Q (tonnes/hr)
Figure 55. Variation of Initial selection
function with flow rate.
117
10
-------
A
A
CO
2
11 A ** tl A
A A
A
I | | | p
2 4 6 8 10
Flowrote.Q (tonnes/hr)
Figure §6'. Variation of Initial breakage
function with flow rate.
118
-------
0.5
0.4
0.3
c
o
u
o
«. 0.2
in
O
0.1
1.0
A Experimental
0 Predicted
7.4 tonnes/hr
2.0
3.0
5.0 6.0
Size Gloss, I
7.0
8.0
9.0
10.0
Figure 57. Comparison of predicted and actual
measured size distributions.
-------
applications are that (1) the mean residence time increases with refuse flow
rate rather than decreasing as in overflow discharge ball mills, and (2) it
is not possible to determine experimentally any of the selection or breakage
functions. The optimization technique used has been found to be an effec-
tive means of finding values for the selection and a compromise breakage
function.
The fractional rate of breakage decreases with flow rate, while the
zero-order production rate remains constant. Therefore, from a knowledge of
the mean residence time and the relation between S] and Q, the product size
distribution may be predicted for any set of mill operating conditions. An
alternative to the single stage of breakage model for predicting size dis-
tribution has been found that takes into account the fact that the residence
time is appreciable.
120
-------
SECTION 7
UNIT OPERATION AND SYSTEM DESIGN CONSIDERATIONS
This section deals with the manner in which the size reduction data
obtained from the experimental protocol can be used and extended. These
generalizations pertain to performance predictions of both unit operations
and system configurations.
GRATE SPACING CONSIDERATIONS
To develop the capability of predicting energy requirements and degree
of size reduction for various grate spacings, a functional relationship
between specific energy (E0) and degree of size reduction (Z0) as well as a
a relationship between Z0 and the grate spacing parameter Dn/F0 was estab-
lished. These relations assumed the general form of y « axB, namely
Eo - alzo • al[(Fo - V/FoJ <85>
and
b.
Z
'2
o
a2(D0/F0) * (86)
Using the raw MSW data with grate openings of 4.8 cm {1.9 in.), 2.5 cm (1.0
in.), and 1.8 cm (0.7 in.), specific energy can be expressed in terms of
ZQ as follows:
D0 = 4.8 cm (1.9 in.) E0(MJ/metric ton) = 61.4 Z01*40 (87)
E0(kWh/ton) « 15.5 Z01«40 (88)
DO m 2.5 cm (1.0 in.) E0(MJ/metric ton) « 109.3 Z01-17 (89)
E0(kWh/ton) - 27.6 Z01*17 (90)
DO « 1.8 cm (0.7 in.) E0(MJ/metric ton) « 172.3 Z0°-82 (91)
E0(kWh/ton) « 43.5 Z0°'82 (92)
The correlation coefficient of each of the above equations exceeded 0.95.
To complete the generalization, the coefficients a}, a?, ti\ and b2 in
Equations 85 and 86 can be expressed in terms of D0 in the form such as
121
-------
&l = (CONST) D0(CONST) etc. The expressions for these coefficients take the
form
ai = 295.23 Do(cm)-1-01 for Equations 87, 89, and 91 (93)
ai = 29.21 D0 (in.)'1-01 for Equations 88, 90, and 92 (94)
and
bi = 0.66 DO (cm)0-50 for Equations 87, 89, and 91 (95)
bi = 1.05 DO (in.)0-50 for Equations 88, 90, and 92 (96)
with correlation coefficients of each equation exceeding 0.87.
Substitution of Equations 91 and 93 into Equation 85 yields
n nn
, m 0.66 Dft(cm)u*3U
EQ(MJ/metric ton) = 295.23 DQ (cm)"1*^ ° (97)
Similarly, substitution of Equations 92 and 94 into Equation 85 yields
i Q1 1.05 D (in.)°'5°
EQ(kWh/ton) = 29.21 DQ (1n.) ZQ ° (98)
These relations can be used to predict E0 for any grate spacing in the mill,
the highest confidence being within the range of the data.
Applying a similar procedure for expressing Equation 86 yields the
following:
D0 = 4.8 cm (1.0 in.): Z0 = 0.65 (D0/Fo)-°*28 (99)
D0 = 2.5 cm (1.0 in.): Z0 = 0.45 (D0/F0)-°-43 (100)
DO = 1.8 cm (0.7 in.): Z0 = 0.34 (D0/F0)-°-51 (101)
The correlation coefficients for these equations were greater than 0.94 in
all cases. Likewise, the coefficient a2 and b2 can be expressed as
a2 = 0.24 D0 (cm)0-65 (102)
a2 = 0.43 DO (in.)0-65 (103)
and
b2 = -0.72 D0 (cm)-0-60 (104)
b2 = -0.42 D0 (in.)-0-60 (105)
with correlation coefficients of each equation exceeding 0.98.
122
-------
Substitution of Equations 102 and 104 into Equation 86 yields
Z0 = 0.24 DO (cm)0.65 (D0/F0)-°'72 Do (cm)-0'60 (10fi)
Substitution of Equations 103 and 105 into Equation 86 yields
Z0 = 0.43 Do (in.)°-65(Do/F0)-°-42 Do (in.)-0**0 (1Q7)
Equations 106 and 107 can now be used to predict the characteristic
product size (X0) through the definition of degree of size reduction (Z0)
when F0 and D0 are specified. As is the case with the energy expressions,
the empirical nature of the equations limits their useful range.
It is interesting to note that the general form of Equations 97 and 98
is
» n 0-5
-1 * Do
EO = «DO zo doe)
where the parametric influence of D0 enters the equation both as a reciprocal
and also as an exponent of Z0. The parametric influence of D0 is also
apparent in the equations expressing Z0 in terms of the ratio D0/F0, Equa-
tions 106 and 107.
The relationships among E0, Z0, and D0/F0 (ie., Equations 97, 98, 106
and 107) can be used to develop curves that describe size reduction using
the Gruendler hammermill (Figure 58). This figure presents analytical re-
sults for size reduction of raw refuse as a function of various hypothetical
grate spacings.
The data shown in Figure 58 are the result of mathematical calculations
based on the manipulative procedure outlined previously in this section as
opposed to the actual test data presented in Figure 34. Comparison of the
analytical computations (Figure 58) and actual test results (Figure 34)
shows good agreement.
The primary purpose and utility of the analytical approach is to allow
prediction and evaluation of shredder performance without resorting to
actual testing. For example, in the following section on unit process opti-
mization, different processing options will be evaluated analytically to
ascertain the optimum processing sequence for size reduction of raw MSW.
TORQUE RELATIONS
In a rotating system subject to drag, the torque necessary to overcome
the drag force is given by
R2
T - / rdFd (109)
o
123
-------
ro
1.0
0.8
0.6
0.4
x°
0.2
O.I
Reliability
Limit for __
Empi rical ~"
Expressions
I i 1 I
I I
I I I
0.5
1.0
10
Figure 58. Size reduction of raw MSW extended to
several different grate openings.
-------
In addition, it has often been noted that the force of drag on a body of
cross-sectional area A is given by
Fd - cd —TT (110)
Replacing v^ by (rw)2 in Equation 110, taking the derivative with respect to
r, and substituting the resulting expression into Equation 109 gives
R2
T =| C.Nh pa,2 / r3dr (111)
o
in which ACS has been replaced by Nhr where N is the number of hammers, h is
the hammer width, and r is the total hammer length. Since most hammermills
are designed so that hammers are offset from the center of rotation and only
a portion of the hammer actually collides with the refuse in the mill, the
lower limit of integration in Equation 111 may be changed from zero to Rj,
the radius at which the hammer first becomes exposed to refuse. Solving
Equation 111 then yields
CNh P R-RCR* RR + R + R] (112)
d
The effect of varying the physical characteristics of the mill (holding
refuse density constant) may be found by taking the ratio of Equation 112
for different sets of operating conditions, i.e.,
Ta Na ha \ ~ \ RPMa R2
'h "k "k KO ~ "1 ""PI. 0 J+ 0 ^0 +00 *-*O J
b b b 2b *b b R2b * R2b Rlb * R2bRlb ^Ib
UNIT PROCESS OPTIMIZATION
The determination of the optimum processing sequence to minimize energy
consumption during size reduction is one of the perplexing problems facing
the resource recovery industry. Both single-stage and two-sta'ge shredding
of raw MSW have been practiced for a number of years, with the proponents of
either camp maintaining that theirs is the optimum method of size reduc-
tion. If indeed there is an optimum manner in which to reduce raw MSW as
well as other fractions of processed refuse, considerable energy savings
could be realized, since the typical resource recovery plant may process as
much as 1000 metric tons/day.
The present size reduction data obtained for raw MSW, ACLF, and SLF are
applicable to the controversy involving the optimal processing sequence for
size reduction. To shed light on this problem, the results of our studies
will be extended to hypothetical resource recovery systems to ascertain the
optimum processing sequence. Although the actual numerical values for spe-
cific energy and product size may be different for some grinders shredding
the same material and using identical grate opening sizes, the trends
125
-------
established here are general enough to apply to most size reduction proces-
ses. The hypothetical processes to be studied will involve the size reduc-
tion of ACLF as well as raw MSW and will be limited to the consideration of
single- and two-stage size reduction.
Raw MSW
To determine the relative merits of single- and two-stage size reduc-
tion of raw MSW, two cases are examined. The first case assumes a material
with a characteristic feed size of 15.2 cm (6.0 in.) (a typical raw MSW) and
a desired characteristic product size of approximately 0.60 cm (0.24 in.).
Two different processing sequences for two-stage size reduction can be
developed for the preceding conditions with the use of Figure 59. This fig-
ure was developed using the analytical approach described in this section
under Grate Spacing Considerations. It contains curves for the grate spac-
ings of interest — namely, 7.6, 4.8, 2.5, and 1.8-cm (3.0, 1.9, 1.0, and
0.7 in.).
To obtain the desired output size, an iterative procedure must be used
to determine the appropriate grate opening sizes (Table 20). The first
processing sequence requires a grate opening size (D0) of 2.5 cm (1.0 in.)
for both primary and secondary size reduction. Cumulative energy consump-
tion is 107 MJ/metric ton (27 kWh/ton), with 75 and 32 MJ/metric ton (19 and
8 kWh/ton) required for primary and secondary shreddings, respectively.
The second processing sequence requires primary and secondary size re-
duction utilizing 7.6- and 1.8-cm (3.0- and 0.7-in.) grate openings, respec-
tively (Table 20). The cumulative energy consumption is 124 MJ/metric ton
(31 kWh/ton) with 37 and 87 MJ/metric ton (9 and 22 kWh/ton) required for
primary and secondary size reduction, respectively.
Single-stage size reduction to the desired product size is affected by
utilizing a grate spacing of 1.8 cm (0.7 in.) (Table 20). This method re-
quires 158 MJ/metric ton (40 kWh/ton).
Minimum energy consumption to meet the assumed specifications is real-
ized using two-stage size reduction with both stages having 2.5-cm (1.0-in.)
grate spacings. Notice that even though the two-stage size reduction se-
quences are more energy efficient than the single-stage system, there is
also an optimum between the different two-stage systems.
The second case investigated assumes the same feed size as used previ-
ously. However, the desired product size is now approximately 0.90 cm (0.35
in.). To arrive at the desired product size, two-stage size reduction using
4.8-cm (1.9-in.) grate openings in each shredder is required (Table 20).
The cumulative energy required by this sequence is 75 MJ/metric ton (19
kWh/ton), with 55 MJ/metric ton (14 kWh/ton) required for the first stage
and 20 MJ/metric ton (5 kWh/ton) required for the second stage.
Single-stage size reduction to meet the product size specification can
be accomplished using a grate spacing of 2.5 cm (1.0 in.), the energy con-
sumption for single-stage size reduction is 91 MJ/metric ton (23 kWh/ton),
126
-------
ro
MJ/metric ton (kWh/Ton)
.40(10)
IOOU5)
Specific £nergy(Ee)
MJ/Metricton (kWh/ton)
3.0
4.0
Figure 59. Degree of size reduction and specific energy requirement
for shredding raw MSW as a function of D0/F0.
-------
Table 20. SINGLE- AND DOUBLE-STAGE SHREDDING OF RAW MSW IN THE GRUENDLER HAMMERMILL
ro
oo
Stages
of size
reduction
Case 1:
Two
Two
One
Case 2:
Two
One
Fo
cm (in.)
15.2 (6.0)
0.91 (0.36)
15.2 (6.0)
1.21 (0.48)
15.2 (6.0)
.15.2 (6.0)
1.83 (0.72)
15.2 (6.0)
cm
2.5
2.5
7.6
1.8
1.8
4.8
4.8
2.5
Do
(in.)
(1.0)
(1.0)
(3.0)
(0.7)
(0.7)
(1.9)
(1.9)
(1.0)
VFo
0,17
2.78
0.50
1.46
0.12
0.32
2.64
0.17
Zo
0.94
0.34
0.92
0.53
0.96
0.88
0.52
0.94
VFo
0.06
0.66
0.08
0.47
0.04
0.12
0.48
0.06
Xo
cm (in.)
0.91 (0.36)
0.61 (0.24)
1.22 (0.48)
0.58 (0.23)
0.64 (0.25)
1.83 (0.72)
0.89 (0.35)
0.91 (0.36)
MJ/
mt*
75
32
37
87
158
55
20
91
Eo
(kWh/
ton)
(19)
(8)
(9)
(22)
(40)
(14)
(5)
(23)
Cumulative
Eo
MJ/
mt*
107
124
158
75
91
(kWh/
ton)
(27)
(31)
(40)
(19)
(23)
*mt = metric ton
-------
or approximately 21 percent greater than that for the two-stage sequence
described previously.
The trend evolved here indicates that two-stage size reduction is the
preferred mode of processing with regard to raw MSW. However, because of
the complicated relationships among E0, D0, and Fg, there may also be an
optimal sequence among various two-stage alternatives. Consequently, care
must be taken to consider a number of possible combinations of grate spac-
ings to achive the desired product size with a minimum expenditure of energy.
Air Classified Light Fraction
Single- and double-stage size reductions of ACLF were investigated to
determine if an optimal processing sequence exists for this material. This
section examines three different cases with the help of Figure 35 — namely,
reduction of material from 10.16 to 0.71 cm (4.00 to 0.28 in.), from 5.08 to
0.70 cm (2.00 to 0.28 in.), and from 2.54 to 0.35 cm (1.00 to 0.14 in.)
(Table 21).
As with raw MSW, double-stage size reduction of the ACLF requires less
energy than single-stage reduction for all cases examined. The discrepancy
between cumulative energy consumption of single-stage and double-stage
sequences in each case diminishes as the initial feed size decreases. The
cumulative energy figures in Table 21 show the Increases in energy require-
ments of single-stage processing as opposed to double-stage processing as
124 percent, 54 percent, and 46 percent for feed sizes of 10.16, 5.08, and
2.54 cm (4.00, 2.00, and 1.00 in.), respectively.
The double-stage shredding process thus offers the optimal method for
size reduction of the ACLF. The discrepancies in total energy expenditure
between single- and double-stage processing sequences are generally larger
than those outlined previously for raw MSW.
Screened Light Fraction
The consequences of single- and double-stage size reduction of SLF are
evaluated for three different cases (Figure 36) assuming feed sizes of
10.16, 5,08, and 2.54 cm (4.00, 2.00, and 1.00 in.), respectively, and a
desired product size of approximately 1.27 cm (0.50 in.) (Table 22).
In each case, double-stage size reduction requires less total energy
than single-stage. As with ACLF, the difference between energy required for
double- versus single-stage size reduction for each case tends to diminish
as the initial feed size decreases.
The information developed here on the relative merits of single- and
double-stage size reduction of raw MSW, ACLF, and SLF indicates that double-
stage size reduction yields the minimum expenditure of energy. Definite
energy savings can result, particularly for large overall degrees of size
reduction (i.e., X0/F0 approaching zero). The data presented here are
especially pertinent to the size reduction of ACLF and SLF.
129
-------
Table 21. SINGLE- AND DOUBLE-STAGE SHREDDING OF ACLF IN THE W-W HAMMERMILL
C*>
o
Stages
of size
reduction
Case 1:
Two
One
Case 2:
Two
One
Case 3:
Two
One
F
cm
10.16
1.02
10.16
5.08
0.91
5.08
2.54
0.86
2.54
0
(in.)
(4.00)
(0.40)
(4.00)
(2.00)
(0.36)
(2.00)
(1.00)
(0.34)
(1.00)
°o
cm (in.)
2.54 (1.00)
1.91 (0.75)
1.27 (0.50)
1.91 (0.75)
1.91 (0.75)
1.91 (0.50)
1.91 (0.75)
0.64 (0.25)
0.64 (0.25)
VFo
0.25
1.88
0.13
0.38
2.08
0.25
0.75
0.74
0.25
Zo
0.90
0.28
0.93
0.82
0.23
0.87
0.66
0.62
0.86
VFo
0.10
0.72
0.07
0.18
0.77
0.13
0.34
0.38
0.14
Xo
cm (in.)
1.02 (0.40)
0.74 (0.29)
0.71 (0.28)
0.91 (0.36)
0.71 (0.28)
0.66 (0.26)
0.86 (0.34)
0.33 (0.13)
0.36 (0.14)
MJ/
mt*
71
44
257
91
40
202
63
265
479
Eo
(kWh/
ton)
(18)
(11)
(65)
(23)
(10)
(51)
(16)
(67)
(121)
Cumu
MJ/
mt*
115
257
131
202
328
479
lative
Eo
(kWh/
ton)
(29)
(65)
(33)
(51)
(83)
(121)
*mt = metric ton
-------
Table 22. SINGLE- AND DOUBLE-STAGE SHREDDING OF SLF IN THE W-W HAMMERMILL
Stages
of size
reduction
Case 1:
Two
One
Case 2:
Two
One
Case 3:
Two
One
F
cm
10.16
1.52
10.16
5.08
1.78
5.08
2.54
1.57
2.54
0
(in.)
(4.00)
(0.60)
(4.00)
(2.00)
(0.70)
(2.00)
(1.00)
(0.62)
(1.00)
Do
cm (in.)
1.91 (0.75)
1.91 (0.75)
0.64 (0.25)
2.54 (1.00)
1.91 (0.75)
1.27 (0.50)
2.54 (1.00)
1.91 (0.75)
1.91 (0.75)
V'o
0.19
1.25
0.06
0.50
1.07
0.25
1.00
1.21
0.75
V
0.85
0.19
0.88
0.65
0.27
0.75
0.38
0.21
0.51
VFo
0.15
0.81
0.12
0.35
0.73
0.25
0.62
0.79
0.49
Xo
cm (in.)
1.52 (0.60)
1.24 (0.49)
1.22 (0.48)
1.78 (0.70)
1.30 (0.51)
1.27 (0.50)
1.57 (0.62)
1.24 (0.49)
1.24 (0.49)
MJ/
mt*
269
32
606
91
48
297
44
36
100
Eo
(kWh/
ton)
(68)
(8)
(153)
(23)
(12)
(75)
(ID
(9)
(25)
Cumulative
Eo
MJ/
mt*
301
606
139
297
80
100
(kWh/
ton)
(76)
(153)
(35)
(75)
(20)
(25)
*mt = metric ton
-------
SYSTEM OPTIMIZATION
This section considers extending optimization of the unit operation to
a resource recovery processing train to minimize the overall energy expendi-
ture. Through the case study approach, a technique is demonstrated for
evaluating different processing configurations with regard to product size
and energy consumption. The system chosen for evaluation is a resource
recovery process that will recover SLF as fluff refuse-derived fuel (RDF).
The fuel specification is assumed to call for a characteristic size of 0.71
cm (0.28 in.). The general processing train is assumed to contain the
sequence of shredding, air classification, trommel screening of the light
fraction, and subsequent shredding of the SLF to meet the size specification
of the RDF. The characteristic size of raw MSW is taken as 15.2 cm (6.0
in.).
Two different processing trains are evaluated (Figure 60). The lower
train employs single-stage size reduction for both raw MSW and SLF. Initial
shredding of raw MSW is followed by air classification, trommel screening,
and final shredding to arrive at the specified product size.
Opposing the above, a train is employed using double-stage size reduc-
tion for both raw MSW and SLF. Other than an increase in the number of
shredders, the two trains are identical with respect to equipment.
By using Figures 35, 36, and 59 and an iterative procedure, the neces-
sary grate spacings were determined for each train (Figure 60). Note that
air classification and screening increase the size of the material after
processing with each piece of equipment.
The energy balance on the processing trains is also determined through
the use of Figures 35, 36, and 59. The quantity of interest is the total
energy required for processing expressed as MJ/metric ton (kWh/ton).
Air classifier and trommel screen splits are assumed to be 70/30 and
71/29, respectively. Since we have previously developed specific energy
relationships based on the material being reduced, a mass balance on each
train must be calculated to determine the energy contribution of each opera-
tion to the total energy requirement (Table 23).
Values for the total energy consumption show that the double-stage
train requires 214.6 MJ/metric ton (54.4 kWh/ton) of MSW, whereas the
single-stage train requires 244.1 MJ/metric ton (61.9 kWh/ton) of MSW, or 14
percent more than the double-stage train.
An energy pathway diagram (Figure 61) shows each processing sequence,
with the interval between letters referring to the various processing opera-
tions (primary and secondary shredding, air classification; etc.). The
greater energy requirement of the single-stage train 1s evident in this
diagram.
The procedure developed here lends itself to evaluations of other proc-
essing sequences. The technique requires that data on energy consumption
132
-------
Primary Secondary
Shredder Shredder
ACLF SLF DO ^ 00
Raw
MSW
FO
15.2 cm
(6.0 in.)
Primary Second
Shredder Shred o
Do~ Y DO
A
4.8cm |.83
(1.9 in.) (0.7
cm 4.8cm
F2in.T(l.9irt)
ary
er
Xo
0.91cm h
(0.36ia)
A.
C.
AP j *a . m
1.45cm ^ Qrreen l.73cm^ '-9I CT
0«;7 -.n^ , ....... J tn rn-.i ^'nTSini i
1.27cm ^ 0.64cm 0.71 cm ^ Fluff
0.50 in.) (025 in.) (0.2 8 in.) RDF
DOUBLE-STAGE SIZE REDUCTION
CJ
CO
Shredder
Raw Shredder
MSW on v
FQ . . *•*
15. 2 cm 2.5
(6.0in.) (t.C
cm 0. 9 1 cm ^
lin^ (0.36 in.)
JL
A.
C.
ACLF SLF D
lU.o / in.ji lU.uuia; (U2
o Xft
4cn 0.7Icmb Fluff
5inJ (0.28 W RDF
SINGLE-STAGE SIZE REDUCTION
Figure 60. Possible operating configurations for resource recovery units.
-------
Table 23. MASS AND ENERGY BALANCE ON PROCESSING TRAINS FOR RECOVERY OF RDF
(A) (B)
Item Operation ^av^MSW^ EQ/operation
MJ/ (kWh/
mt* ton)
(A)x(B)
E /metric ton
(ton) raw MSW
MJ (kWh)
Energy
Pathway
Input material for double-stage train:
CO
Raw MSW
Raw MSW
Raw MSW
ACLF
SLF
SLF
Total energy
Primary shredding
Secondary shredding
Air classification
Trommel screening
Primary shredding
Secondary shredding
100
100
100
70
50
50
55
20
15
8
44
194
(14)
(5)
(4)
(2)
(ID
(49)
55
20
15
5.6
22.0
97.0
(14)
(5)
(4)
(1.4)
(5.5)
(24.5)
AA'
A'B'
BC
CO
DO1
D'E
214.6 (54.4)
Input material for single-stage train:
Raw MSW
Raw MSW
ACLF
SLF
Total energy
Single-stage shredding
Air classification
Trommel screening
Single-stage shredding
100
100
70
50
91
15
8
265
(23)
(4)
(2)
(67)
91
15
5.6
132.5
244.1
(23)
(4)
(1.4)
(33.5)
(61.9)
AB
BC
CD
OE
-------
co
en
3
o
a:
>
3
C3
80
60
40
> 20 -
O.I
FLUFF
RDF
Single-stage train
Double-stage train
0.5
B
RAW MSW
A
Screen Size ,
in. |
i
10
cm
320
160
10
240 .^
a>
80 |
B
3
|
n o
Figure 61. Energy pathway diagram for hypothetical
RDF processing trains.
-------
for reducing different fractions of MSW be available as a function of the
fundamental parameters Z0 and D0/F0. If such energy consumption information
is available in this form, generalizations can be made regarding the optimi-
zation of resource recovery processing lines.
136
-------
REFERENCES
1. Obeng, D.M. Comminution of a Heterogeneous Mixture of Brittle and Non-
Brittle Materials.Ph.D. thesis, University of California, Berkeley,
Calif., 1974.
2. Obeng, D.M. and Trezek, G.J. Simulation of the Comminution of a Heter-
ogeneous Mixture of Brittle and Non-brittle Materials in a Swing Ham-
mermill. Ind. Eng. Chem. Process Des. Dev., 14, No. 2, 1975.
3. Trezek, G.J. Significance of Size Reduction in Solid Waste Manage-
ment. EPA-600/2-77-131, U.S. Environmental Protection Agency, July
1577.
4. Calcott, T.G. Solution of Comminution Circuits. Trans. Inst. Min.
Metall., Section C, 76, 1967.
5. Calcott, T.G. and Lynch, A.J. An Analysis of Breakage Processes within
Rod Mills. Proc. Austral. Inst. Min. and Met. 209, 1964.
6. Lynch, A.J. and Draper, N. An Analysis of the Performance of Multi-
stage Grinding and Cyclone Classification Circuits. Proc. Austral.
Inst. Min. and Met. 213, 196.
7. Herbst, J.A., Grandy, G.A., and Mika, T.S. On the Development and Use
of Lumped Parameter Models for Continuous Open- and Closed-Circuit
Grinding Systems. Trans. Inst. Min. Met., Section C, 80, 1971.
8. Pease, M.C. Methods of Matrix Algebra. Academic Press, New York,
N.Y., 1965.
9. Hildebrand, F.B. Advanced_Calculus for Applications. Prentice Hall,
Englewood Cliffs, N.J., 1962.
10. Stewart, P.S.B. and Restarick, C.J. A Comparison of the Mechanism of
Breakage in Full Scale and Laboratory Scale Grinding Mills. Proc.
Austr. Inst. Min. Met. 239, 1971.
11. Mori, Y., Jimbo, G. and Yamasaki, M. Residence Time Distribution and
Mixing Characteristics of Powders in Open-Circuit Ball Mill. Kagaku-
Kogaku (English ed.) 2, 1964.
12. Keinenburg, R., von Seebach, H.M. and Strauss, F. Einfluss von Durch-
satz, Mahlkorperfullung und Austragswand aufden Mahlgntfallungsgradund
137
-------
die Mahlgut Verweilzeitverteilung in Rohrmuhlen, Dechema-Mongraphien
69, 1971, Part 2. '
13. Herbst, J.A., Grandy, G.A. and Fuerstenau, D.W. Population Balance
Models for the Design of Continuous Grinding Mills. 10th IMPC, London,
1973.
14. Austin, L.G. A Review Introduction to the Mathematical Description of
Grinding as a Rate Process. Powder Tech. ^, 1971/72.
15. Herbst, J.A. Batch Ball Mill Simulation: A Method for Estimating
Selection and Breakage Functions. M.S. thesis, University of
California, Berkeley, Calif., 1968.
16. Herbst, J.A. and Mika, T.S. Mathematical Simulation of Tumbling Mill
Grinding: An Improved Method. Proc. IX Int. Min. Proc. Congr. Part
III, 1970.
17. Grandy, G.A. and Fuerstenau, D.W. Simulation of Non-Linear Grinding
Systems: Rod-Mill Grinding. Trans. Am. Inst. Min. Engrs., 247, 1970.
18. Herbst, J.A. et al. An Approach to the Estimation of the Parameters of
Lumped Parameter Grinding Models from On-line Measurements.
Zerkleinern, 3rd Europ. Sym. on Comminution, Cannes, 1971. H.C.H.
Rumpf and K. Schonert, eds. Dechema-Monographien 69, 1972, Part I.
19. Austin, L.G. and Klimpel, R.R. Determination of Selection-For-Breakage
Functions in the Batch Grinding Equation by Nonlinear Optimization. I.
and E.G. Fund. £, 1970.
20. Arbiter, N. and Bhrany, U.N. Correlation of Product Size, Capacity and
Power in Tumbling Mills. AIME Trans. 217, 1960.
21. Epstein, B. The Mathematical Description of Certain Breakage Mechanisms
Leading to the Logarithmico-Normal Distribution. J. Frank. Inst. 224,
1947.
22. Shiflett, G.R. A Model for the Swing Hammermill Size Reduction of
Refuse. D. Eng. Dissertation, Univ. of California, Berkeley, Calif.,
1978.
23. Goldfarb, D. and Lapidus, L. Conjugate Gradient Method for Nonlinear
Programming Problems with Linear Constraints. Ind. Eng. Chem. Fund. _7,
1968.
24. PoweTl, M.J.D. An Efficient Method for Finding the Minimum of a
Function Without Computing Derivatives. Computer Journal 2.. 1964.
25. Fiacco, A.V. and McCormic, G.P. Non-Linear Programming — Sequential
Unconstrained Minimization Technique.John Wiley, New York, N.Y., 1968.
26. Radcliffe, C.W. and Suh, C.H. Kinematic and Mechanism Design. John
Wiley, New York, N.Y., 1977.
138
-------
APPENDIX A
COMMINUTION COMPUTER MODEL
LISTING *************************** FUNCTION FFIIM ************************** MONDAY
1...S b 7 ............ ? ................... « ................... 6 .......... 7Z
FUNCTION FFUN(X)
C
C OfiJECUVE FUNCTION FOR GPTIMI/ATION PROGRAM PCOJI
C
DiXENSION FF(3«n »GG<30) iHH(3&) tX(30)
DIMENSION YiYPU5),XPT(lS)iYPTiThETA(15)
QIMtNSION BCJ5il5»«8Bll5«15)«DiSIMP(15)«ZLRO(l5)
LOGICAL ssc
COMHOU / QLOC«T1 / >F»XP.YFtYP,XPTiYPTiRiStTMETA«r.TAlji^,CiXNf)EX.lVT
COMMON / BLOCKS f BiBBiDiFtGtGltH.JC.T.Tt
COrtWM / DLOCK8 / FFiZEKO
SSCs.TRUC.
C
C IF SSg IS TRUE USE THE SUM OF. THR SHUARFO ERRORS AS THE
c OBJECTIVE FUNCTION
c
IF ( .NOT. S5E ) GO TO 20H
CALL HFON(HM.Jf)
DO 10H 1=1 t 7
SUM=SUM+
IDt) CO.MT1MUE
2tlt) CONTINUE
StZMOO=SGRT(l./2.»
S1=X(1)
KAPPA=X(2)
A£.PHA=AOS«X«3))
082l=KAPPft*SI?rtOC**ALPHA/Sl
EftlO
PiLATIOK/.
139
-------
LISTING ************************* SUBROUTINE GFIIN ************************** MONDAY
1...5 6 7 ............ 2. .................. t ..... . ............. 6 .......... 72
SUBROUTINE GFUN(GG,X>
C
C INEQUALITY CONSTRAINTS FOR PROGRAM PCON
C
DIMENSION FF(30)*GG(30)iHH(30>«X(J0)
DIMENSION YU5)iXF<15>,XP<15>iYF(l2).YPU5».XPT{15)iYPTU5)
DIMENSION SU5)«fttl5) tTHETAUS)
DIMENSION B<15tl5>,BB(15.15>,0(15.1eJ),JC(l'5il5).T(15,13) ,TI(15ilS)
DIMENSION CUS),SIMP(15) tZ£RO(15)
DIMENSION F(i«5%i5> ,GU5.i5> tRms.is) .hds.is)
REAL JC.KAPPA
COMMON / QLOCKl / XF.XP, YF« YP, XPT , YPT.R t S.THETA . Y ,TAU«N»C , INOtX ,f^T
COMMON / BLOCK2 / B«BB=-X<2)
PILATIOM.
RETURN
ENO
140
-------
LISTING ************************** SUBROUTINE HFUN ************************** MONDAY 1
1...S 6 7 ............ 2 ................... •*..... .............. f, ........ ..72 7
SUBROUTINE HFUN(HHtX)
C
C EQUALITY CONSTRAINTS FOR PROGRAM PCON
C
C
C CALCULATE VALUES FOR THE SELECTION AND BREAKAGE FUNCTIONS OF A
C CONTINUOS GRINDING MILL
C THt VARIABLE NAMES' IN THIS PROGRAM REPRESENT
C X(I)=THE VARIABLES OF THE OPTIMIZATION ROUTINE
C FF(I)sTHE RESIDUALS TO BE SQUARED BY THE OPTIMIZATION ROUTINE
C Y(I)=THE SCREEN SIZES
C XF(I)=THt SIZE DISTRIBUTION OF THC FEED
C YFdlsTHE CUMULATIVE FRACTION PASSING OF THE FFED
C XP(I)sTHE SIZE DISTRIBUTION OF THE PRODUCT
C YP(I)sTHE CUMULATIVE FRACTION PASSING OF THE PRODUCT
C XHTiDsTHE PREDICTED SIZE OISTIBUTION FROM THE MODEL
C YPT(I)=THE PREDICTED CUMULATIVE PASSING
C R«I)=THE RESIDENCE TIME DISTRIBUTION
C THETA(I)sNORMALlZED TIME
C NTsTHE NUMBER OF TIME PERIODS
C S=THE BREAKAGE FUNCTION
C BB(ItJ>=TH£ CUMULATIVE BREAKAGE FUNCTION
C DU.JIsTHE MILL MATRIX
C H(IiJ)=THE CLOSED LOOP CIRCUIT MATRIX
C TAUsTHE MEAN RESIDENCE TIME
C NsTHE NUMBER OF SIZE CLASSES
C YMAXsTHE MAXIMUM SIZE
C iNCEXsCONTROL TO DETERMINE IF OPEN OR CLOSED LOOP MODEL IS USED
C
DIMENSION FF(30).X<30).GG.YPTU3)
DIMENSION SU5).RC15).THETA<15>
DIMENSION OU5»l3»tBB«15tl5)»D(15.15).JCC15.15).T<15.1«>,TI<13.1«)
DIMENSION CU5).SIMP(15>tZERO<15>
DIMENSION F(15.l5)iG(15.l5)iGI(l5tlS)tH(15,13)
REAL JCt KAPPA, MASK
COMMON / BLOCK1 / XF.XP. YF.YP.XPT tYPT.R .S.THETA .Y.TAU.N.C . INDEX , NT
COMMON / BLOCKS / B.BBiDt FiGtGI.Hi u
COMMON / BLOCK3 / FF.ZERO
LOGICAL ONCE
DATA ONCE / .FALSE. /
IF ( ONCE ) GO TO H99
c INITIALIZE ALL CONSTANTS Td ZERO
c
DO 100 1=1.15
TH£TA(I)s0.0
141
-------
LISTING ************************* SUBROUTINE HFUN **************************
1...5 6 7 2 «• 6 72
XF(I)=0.k>
YF
C
C READ IN THE CLASSIFIER FUNCTION IF A CLOSED LOOP MODEL IS DESIRED
C
IF ( INDEX.LT.0 ) READ(5,2)
C
c SET up SIZE CLASSES
c
YdlsYMAX
DO 2BH 1=2,N
200 CONTINUE
C
C CONVERT CUMULATIVE PASSING TO SIZE DISTRIBUTION
C
XF(l)=1.0-YF(l)
XP(1)=1.«-YP(D
DO 300 1=2,N
XF(I)=YF(I-1)-YF(I)
XP(I)aYP(I-l>-YP(I)
300 CONTINUE
Oi»CE=tTRUE.
H00 CONTINUE
C
C INITIALIZE ALL WORKING ARRAYS TO ZERO
C
00 500 I=ltl3
SI!)=«.B
SIMP(I)=0.0
XpT(I)s0.0
142
-------
LISTING ************************* SUBROUTINE HFliN **************************
1...5 6 7 ............ 2 ................... i* ................... 6 ........ ..72
YpTti)=0.0
00 500 J=lil5
B(I,J)=0.0
BdU«J)=0.0
D(I«J)=0.0
F(I,J)=U.B
G(I.J)=H.0
GI(I.J)=H.0
H(ItJ»=0.H
JC(I«J>=0.0
TKIiJ)=0.0
500 CONTINUE
S11)=X(1)
KAPPA=X(2)
AUPHA=AOS(X(3) )
C
C CALCULATE THE ZERO ORDER PRODUCTION RATE CONSTANTS
C
00 600 I=2iN
S I ZMOO=Y ( I ) /SORT < Y ( 1 ) *T < 2 ) )
ZERO(I)=KAPPA*SIZMOD**ALPHA
600 CONTINUE
C
C CALCULATE BREAKAfiE FUNCTIONS
C
00 700 I=2iN
BB(Iil)=lERO(I)/Sm
7)JB CONTINUE
DO 800 J=2»N
00 800 I=J.N
B8(I.J)=DB(I-J-t-ltl)
800 CONTINUE
DO 900 J=2iN
B(J«l)sBBiJtl) -BE ( J+l » 1 )
DO 900 I=JiM
IK ( I.EQ.J > MASK=0.0
B(I,J>=BG(IiJ)-BB(I+liJ)*MASK
S00 CONTINUE
C
C CALCULATE SELECTION FUNCTION
C
MsN-1
DU 950 I=2»M
S(I)=ZERO(N)/BB(N«I>
950 CONTINUE
C
C CALCULATE THE MATRIX T
C
143
-------
LISTING ************************* SUBROUTINE HFlIN ************************** MONDAY 1
1...5 6 7 2 )/(Sm-S*T(K,j»
1200 CONTINUE
1300 CONTINUE
C
C CALCULATE T INVERSE
C
00 1700 Jrl.N
Do 1700 IsJiN
L=I-1
IF ( L-J ) l«»*JM500tlS00
im*0 CONTINUE
Ti(I.J)=1.0/T(1,1)
GO TO 1700
1500 CONTINUE
DO 1600 KsJiL
TI(ItJ)3Tl-TCXtK)*TI(KiJ)/T(ItI)
1600 CONTINUE
1700 CONTINUE
C
C CALCULATE THE MATRIX JC USING SIMPSONS APPROXIMATION
C
DEL=TriETA<2)-THETA(l>
MsNT-3
00 2000 1=1.N
00 1800 w=l.NT
SrMP(J)=EXPt-SCI>*TAU*THETA
E»/ENsEVEN+SIMP{K*l)
1900 CONTINUE
OOOaOOD+SIMP
-------
LISTING ************************* SUBROUTINE HFUN ************************** MONDAY |
1...5 6 7 ............ 2 ............ . ...... t ................... 6 ........ ..72 7
D(I»J>=D(IiJ)+T(l.K)*JC(K,K)*TI(Kit)
2100 CONTINUE
C
C IF AN OPEN LOOP MODEL IS USED CONTINUE
C IF A CLOSED LOOP VOQEL IS USED 60 TO 2300
C
IF ( INDEX. LT.0 ) 60 TO 2300
C
c MULTIPLY o UY THE FEED SIZE DISTRIBUTION
c
00 2200 IsliN
DO 2200 J=1«N
XPT(I)=XPT(I»+0«I«J)»XF«J)
2200 CONTINUE
60 TO 3300
2300 CONTINUE
C
C FORM THE MATRIX F BY MULTIPLYIN6 I-C BY 0 AND
C FOKM THE MATRIX G BY SUBTRACTING C*0 FROM I
C
00 2600 JsliN
00 2600 IsJiN
FlIiJ) = U.-C(I) >*D(I«J)
Lsl-l
IF ( L-J ) 2<*00t2500i25B0
2<**0 CONTINUE
60 TO 2640
2500 CONTINUE
G*n
2600 CONTINUE
C
C CALCULATE 6 INVERSE
C
00 3000 0=1. N
00 30(10 IsJtN
L=I-1
IF I L-J ) 27a0«2600t28ff0
2700 CONTINUE
GO TO 3000
2800 CONTINUE
00 2930 KaJ.L
6i(I«J>s6I(I*J)-6(ItK)«6I(KtJ)/6II«I)
2900 CONTINUE
3000 CONTINUE
C
C FORM THE CIRCUIT MATRIX H BY MULTIPLYING F BY 61
C
00 3100 JsltN
00 3100 IsJ.N
145
-------
LISTING ************************* SUBROUTINE HFUN ************************** MONDAY /
1...5 6 7 2 U 6 72 ?
00 3100 K=J,I
H(I.J)=H«I.J)+F
IF ( XP(I).NE.0.0 ) FFlI) = (FF(I)/XF( I) )*10B.
CONTINUE
DO 3500 1=4,N
J=I-3
HH(J)=XPT
CONTINUE
RETURN
1 FORMAT(IS)
2 FORMAT<15F5.3)
3 FUKMAT(6F10.3)
plLATION,
146
-------
LISTING ************************* SUBROUTINE MORPRNT *********************** MONDAY 1
1...5 6 7 ............ ? ................... '»... ................ 6 ....... ,..72 7
SUBROUTINE MORPRNT iXPU5).YF<15),YP(15)tXPTI15),YPT(15)
DIMENSION S(15)tR«15),THETA(15).ERRORO5)
DIMENSION B(15»l5).BB(15»lS)tD(lS,15)tJC(15»15),TU5,15).TI<15,15)
DIMENSION C< 15), SIMP (15). ZERO (15)
DIMENSION F(15,l5) ,G(15,15).GI(1«,15),H(15,15)
REAL JC.KAPPA
COMMON / 8LOCK1 / XF.XP . YF«YP»XPT . YPT.R .S.THETA.Y , TAU«N , C » INDEX, M
COMMON / BLOCK2 / B.BB.O.F.G.GI .H.vC.T.TI
COMMON / BLOCKS / FF.ZERO
C
C INITIALIZE THE SIZE DISTRIBUTION ERROR TO ZERO
C
DO 100 1=1, N
100 CONTINUE
C
C CALCULATE THE SIZE DISTRIBUTION ERROR
C
DO 2'31 1 = 1, M
E*«OR(I)=XPT(I)-XPm
2k) a CONTINUE
MKITE(gtl)
WRITE (6. 2) «I,Y»XF«YF«I=liN)
WRITE(6t3> TAU
WRITE ( 61 <») ((I,R(I),THETA(I)>,I=1,NT)
WRITE(6,5)
WRITE (6,6) ((I,Y«XPIsltNI
WHITE(6,7)
WKITE(6«6>
WRITE (6 ,9)
WHITE(6<10)
WRITE(6,11)
WRITE (6, 10) ((B8'J=1«N)»I=1.N)
WRITE(6,12)
WRITE (6, 18) ((T(I«J>,J=1»N),I=1,N)
WRITE(6,13)
WHITE (6, 10) ((TKItJ)iJsltN)»Isl«N)
«JC(ItJ)tJsl.N)«Isl«N)
W, tlsliN)
IF I INDEX. GT.0 ) GO TO 30fl
WHITE(6,16)
WRITE(6«fl) «I,C,I=1«N)
WMlTE(6fi7)
WRITE (6, 13) ((F«I.J),J=ltN),I=l,N)
WRITE(6,18)
147
-------
LISTING ************************* SUBROUTINF MQRPRNT *********************** MONDAY 1
1...S 6 7 ............ 2 ................... «» ................... 6 .......... 72 7
«G(I«d) tJ=l«N).I = l.N)
300
2
3
8
9
10
12
13
It
15
16
17
18
19
20
DILATION.
«GKI«J)iU=l.N).I = l.N
FORMAT<10X,I2,1X.F10.5,3H X.F10.S.IX.F10.5,IX.F10.3,IX,F10.5•
lX«F10.5«lX«F10.5«lXtF10.5l
FORMAT!lhl./////15X»27H THE SELECTION FUNCTION IS ////
10X,3H I ,<»X,6H S(I) /)
FUKMAT<10XtI2,lX.F10.5)
FORMAT
FORMAT!////19X.28H THE CLASSIFIER FUNCTION IS ////
10X,3H I ,HX,6H C!I) /)
FORMAT{1H1/////1SX,22H THE MATRIX FlltJ) IS ////)
FORMAT!////1SX«22H THE MATRIX G(I IS ////)
FORMAT{////lBXi37H THE CLOSED CIRCUIT MATRIX H(I,J) IS ////)
END
148
-------
LISTING **************************** PROGRAK GRIND ************************ TUESDAY I
1...5 b 7 2 4 6 72 7
PHOGRAM GRINDCINPUT,OUTPUT,TAPE5=INPUT,TAPE6=OUTPUT)
C
C IF VALUES FOR FLOWRAlEt MOISTURE. SI, KAPPA, AND ALPHA ARC
C SUF-PLlEO FROI* A KNOWLEDGE OF THE OPERATING CONDITIONS OF THE SHREDDER
c THAT THE. USER WISHES TO EXAMINE THE PROGRAM WILL CALCULATE A
C PREDICTED PKCDUCT SI2E DISTRIBUTION ANC COMPARE IT TO AN
C EXPERIMENTALLY DETERMINED ONE FOR THE SAME OPERATING CONDITIONS
C ANL ESTIMATE THE POWEK NECESSARY TO OPERATE THE SHREDDER
C A bRUENDLER M8-4 SHREDDER WITH 2.5 Cf GRATES IS ASSUMED
C
REAL KAPHA.MC
READ<5,1) 0
READ(5,1) MC
READ(S.l) SI
READ(5,1) KAPPA
READ(5,1) ALPHA
TlMs3.6*13.61*Q**.£64
P=18.61*fi**1.107*U.-MC)**.358
MC=MC*10B.
WRITE<6<2) O.^C.P
CALL SHREDJTIW,si,KAPPA,ALPHAI
STCP
1 FCJRMATlFm.B)
2 FuRKAT«lHl,5X,18H THE FLOW RATE IS tF«f.l,17H TONNES PER HOUR ///
S 5X.25H THE ""OISTURE CONTENT IS ,F5.1,9H PFRCENT ///
$ 5X.33H THE ESTIMATED POWER REQUIRED IS .Ffi.l.llH KILOWATTS )
piLATION.
149
-------
LISTING ************************* SUBROUTINE SHRED *************•*********.»* TUESDAY 1
1...5 6 7 2 u 6 72 7
SUBROUTINE SHRED(TIMtSi.KAPPAtALPHA)
C
C CALCULATE VALUES FOR THE SELECTION AND BREAKAGE FUNCTIONS CF A,
C COIuTlNUOS GRINDING MILL
C THE VARIABLE NAMES IN THIS PROGRAM REPRESENT
c x(i)=THE VARIABLES OF THE OPTIMIZATION ROUTINE
C FF(1)=THE RESIDUALS TO BE SQUARED BY THE OPTIMIZATION ROUTINE
C Y(I»=THE SCREEN-SIZES
C XF(1)=THE SIZE DISTRIBUTION OF THE FEED
C YF(I»=THE CUMULATIVE FRACTION PASSING OF THE FEED
C XP(I)sTHE SIZE DISTRIBUTION OF THE PRODUCT
C YPCIJsTHE CUMULATIVE FRACTION PASSING OF THE PRODUCT
C XPT(I)=THE PREDICTED SIZE DISTlQtTICN FROM THE MODEL
C YPT(I>=THE PREDICTED CUMULATIVE PASSING
C RU)=THE RESIDENCE TIME DISTRIBUTICN
C THETA(II=NORMALIZED TIME
C NT=THE NLMBER OF TIME PERIODS
C S(I)=THE SELECTION FUNCTION
C CtX(30)tGG(3(M,HH(3e>
DIMENSION Y(15)iXFC15),XP(15).YF(15).YP<15>.XPTC15>«YPT(15»
DIMENSION S(15)tR(15).THETA(15)
DIMENSION B(15tlSt«BB(15tl5)iO«T(15.15>«TI(l$tlS>
DIMCNSION CdSJ.SIMPUSI.ZEROUS)
DIMENSION F(15.15) tG(15%15)«GH15,15) tHdSil1?)
REAL JC«KAPPA,MASK
COMMON / BLOCKl / XFtXP.YF.YP.XPT,tPT.RtS,THETAiYiTAUiM.CiINDEXtlU
COMMON / BLOCK2 / B.BBtD.FtGt GI ,H t«.C t T.TI
COMMON / BLOCK3 / FFtZERO
LOGICAL ONCE
DATA ONCE / .FALSE. /
IF ( ONCE ) GO TO 400
C
C INITIALIZE ALL CONSTANTS TO ZERO
C
00 100 1=1.15
C(I)=U.0
K(I)s0.0
THETA(I)=0.0
T(I)=».U
XF(I)=a.0
YF(I»s0.0
150
-------
LISTING ************************* SUBROUTINE SHRED ************************ TUESDAY I
1...5 6 7 ............ 2 ................... t 1=1
C
c SET UP SIZE CLASSES
C
Y(1»=YMAX
DO 200 I=2=0.0
XPT(I)=0.0
YPT(I)=U.0
ZtHO(I)=0.0
DO 500 0=1.13
151
-------
LISTING ************************* SUBROUTINE SHRED **•**•««*«******»****«** TUESDAY 1
1...5 b 7 ............ 2 ................... «t ..... ........ ...... 6 .......... 72 7
Bs0.0
F(I,J>=0.0
G(IiJ>=0.0
GJ(I»J)=0.0
H(IiJ>=0.0
JC(I.J)=0.0
T)
ZERO«I)=KAPPA*SI2MOD**ALPHA
600 CONTINUE
C
C CALCULATE BREAKAGE! FLNCTIONS
C
00 70B 1=2. N
BrU I • 1 ) s2ERO ( I ) /S II)
700 CONTINUE
DO 800 J=2.N
DO 600 I=J.N
BBlIt J)=EB(I-J+1«1)
800 CONTINUE
DO 900 J=2,N
DO 900 I=J.N
MASK=1.0
IF ( I.EC.J ) MASK=0.0
B < I » J ) sBE ( I . J ) -BB ( 1 + 1 , J > *MASK
900 CONTINUE
C
C CALCULATE SELECTION FUNCTION
C
MSN-1
DO 990 1=2, M
S(I)s2ERO(N)/BB(N.I)
950 COWTH4UE
C
C CALCULATE THE MATRIX T
C
DO 1300 J=1.N
DO 1300 I=J.N
Lsl-l
IF ( L-J I 1000,1100,1100
1000 CONTINUE
152
-------
LISTING ************************* SUBROUTINF SHRED ************************ TUESDAY
1...5 6 7 ............ 2 ................... it ................... 6 .......... 72
T(ItJ)sl.0
GO TO 1300
1100 CONTINUE
OJ 1200 K=JiL
T=T
1200 CONTINUE
1300 CONTINUE
C
C CALCULATE T INVERSE
C
DO 1720 d = l»N
DO 1700 I=J»N
Lsl-l
IF I L-J ) I«»00tl500tl500
CONTINUE
GO TO 1700
1500 CONTINUE
DO 1600 K=J,L
TKItJ)=7I(ItJ)-T«I»K)*Tl(K,J)/T(l«I>
CONTINUE
CONTINUE
c
C CALCULATE THE MATRIX JC USING SIMPSONS APPROXIMATION
C
DELaTHETA|2)-THETA(l>
H=NT-3
DO 20*0 1=1. M
00 1^01 UsltNT
SIMP(J)=EXP(-S(1)"TAU*THETA(J»*R(J)
1800 CONTINUE
000=0.0
DO 1900 K=liM,2
ODDsOOO+SlMPfK)
1900 CONTINUE
OUD=000+SIMP(NT-1)
JC(I.I) = «OEL/3.)*U.*ODD+2.*EVEN+SIMP(NT)>
200fl CONTINUE
C FORM THE MILL MATRIX 0 8Y MULTIPLYING T 8Y JC B* TI
C
00 2100 J=1«N
00 2100 IsJiN
On 2100 KrJ.I
D(I.J)=D(I.J)+T(ItK)*JC(K,K)*Tl(K,J)
2100 CONTINUE
C
C IF AN OPEN LOOP MODEL IS USED CONTINUE
C IF A CLOSED LOOP MODEL IS USED 60 TO 2300
153
-------
LISTING ************************* SUBROUTINE SHRED ************************ TUESDAY f
1...5 b 7 ............ 2 ................... 4 ................... 6 .......... 72 1
C
IF ( INOCX.LT.0 ) GO TO 2300
C
C MULTIPLY 0 BY THE FEED SIZE DISTRIBUTION
C
DO 2200 IsltN
DO 2200 JsltN
XPT < I ) =XPT < I ) *0 { 1 1 J ) *XF < J )
22 HU CONTINUE
GO TO 3300
2300 CONTINUE
C
C FOKM THE MATRIX F BY MULTIPLYING I-C 8Y 0 AND
C FORM THE .HATH IX G BY SUBTRACTING C*0 FROM I
C
00 2600 wIsl.N
00 2600 I=JtN
L=I-1
IF ( L-J ) 2
2900 CONTINUE
3000 CONTINUE
c
C FORM THE CIRCUIT MATRIX H BY MULTIPLYING F BY Gl
C
00 3100 JsltN
00 3100 IsJtN
00 3100 Ksjtl
H(ItJ)sH*GI(KiJ)
3100 CONTINUE
c
C MULTIPLY H BY THE FEED SIZE OISTRIBUTICN
154
-------
LISTING ************************** SUBROUTINE SHRED ************************ TUESDAY 1
1...5 6 7 2..... H ft ...72 7
C
00 32rffl IsltN
DO 3200 Jsl.N
XPT(II=XPT(I)*H
3200 CONTINUE
338« CONTINUE
c
C CONVENT THE PREDICTED PRODUCT SIZE DISTRIBUTION TO CUMULATIVE PASSING
c AND CALCULATE PERCENT ERROR OBJECTIVE FUNCTIONS
c
FF(l>sABSjXPm-XPT
DO 3400 Is2iN
YPT(I)=YPT(I-1)-XPT
-------
LISTING ************************* SUBROUTINE MORPRNT ********************** TUESDAY I
1...5 6 7.... 2 U 6 72 7
Mt*ITE(6,10)
300
1
2
3
8
9
Ifl
11
12
13
14
15
16
17
18
19
20
PILATION.
WMITE(6,10)
W»UTE(6,20)
WHITE(6,1H)
CONTINUE
RtTURN
FuRMAT(lHl/////10X,llH PEED DATA
15X.32H SIZES ARE GIVEN IN CENTIMETERS /
15X,43H XF(I)a£XPERlMENTAL FEED SIZE DISTRIBUTION /
15X.44H YF(I>sEXPERIM£NTAL FEED CUMULATIVE PASSING ////
10X.3H I ,6X.1£H SIZE CLASS .7X.7H XF(I) ,<*X,7H YF(I) /)
FORMAr(10X,I2,F10.5,3H X,F10.S,IX.F10.5,IX.FIB.5)
FORMAT{////13X,34H R(IJsRESlDENCE TIME DISTRIBUTION /
15X.36H THETA(I>=tJORMALIZED RESIDENCE TIME /
1SX.26H TAtsMEAN RESIDENCE TIMEs .F10.5.SH SECONDS ////
10X.3H I ,2X,6H H(I) ,5X«10H THETA «4X,aH YPT >3x,
JH u'^CUNT /
SI+X.7H ERROR /)
FORMAT(10X*l2,lX«F10,5i3H X«Fl0.5,lX,Fl0.5,lX.Fl0.5,lX,Fl0.!i,
lXtF10.5,lX,F10.5«lX>F10.S)
FORMAT<1H1,/////15X,27H THE SELECTION FUNCTION IS ////
1UX.3H I ,4X,6H S(I> /)
FOHMAT(1MX,I2,1X,F10.5>
FURMAT(////15X,33H THE BREAKAGE FUNCTION BUiJ) IS ////)
FORMAT(9X,IX,£10.3,IX,£10.3,IX,£13.3.IX,£10.3,IX,£10.3,IX,£10.3.
IX,£10.3,IX,E10.3,IX,£10.3,IX,£10.3)
FORMATUH1/////15X,
45H THE CUMULATIVE BREAKAGE FUNCTION BB(T.J) IS ////)
FOKMAT(////1SX,22H THE MATRIX Ttl.vl IS ////)
FOHMAT(1H1/////15X,34H THE INVERSE OF TtI,J)sJI(I,J) IS ////)
FORMAT(////15X,23H THE MATHIX JC IS ////)
FORMAT<1H1/////15X,27H THE MILL KATRIX 0(1,J) TS ////)
FORMAT(////lSXi28H THE CLASSIFIER FUNCTION IS ////
10X,3H I ,<4X,6H C(I) /)
FORMAT(1H1/////15X,22H THE MATRIX F(I«J) TS ////>
FOHMAT(////1SX,22H THE MATRIX G(I,o) IS ////)
FORMAT(1H1/////15X,34H THE INVERSE OF G IS ////)
FORMAT(V///15X.37H THE CLOSED CIRCLIT MATRIX H(1.J> IS ////)
EMO
156
-------
THE FLOto RATE IS 7.4 TONNES PEH HOUR
THE MOISTURE CONTENT JS 30.0 PERCENT
THE ESTIMATED POWER REQUIRED IS lgtf.2 KILOWATTS
157
-------
FEED DATA
SIZES ARE GIVEN IN CENTIMETERS
XF(I)=LXHERIMENTAL FEED SIZE DISTRIBUTION
YF(I)=EXPERIMENTAL FEED CUMULATIVE PASSING
SIZE CLASS
XF(I)
YFU)
1
2
3
4
5
6
7
8
9
10
40.00000
20,00000
10.0*000
5.000100
2.50000
1.25000
.62500
.31250
.15625
.07813
X
X
X
X
X
X
X
X
X
X
20.00000
10.fl (10(10
5.0000U
2.50000
1.25000
.62500
.31250
.15625
.07813
t).
.H5400
.32800
.50100
,05500
.01800
.01400
.01000
.00600
.00(400
.01000
.94600
.61800
.11700
.H6200
. ait nee
.H3000
.02000
.01400
,«1000
0.
R(I)=HES1DENCE TIME DISTRIBUTION
THETA(I)sNORMALIZED RESIDENCE TIME
TAUsMEAN RESIDENCE TIMEs 55,69183 SECONDS
I
1
2
3
4
5
6
7
8
R(I)
.25700
m.8030B
13,36400
7. 7610k)
5.1f040
3,34100
1.74800
,51400
THETA(I)
.01000
.36000
.71000
1.06000
1.41100
1.76100
2.11100
2.46100
158
-------
PRODUCT DATA
SIZES ARE GIVLN IN CENTIMETERS
XPtI)=EXHERIMENTAL PRODUCT SIZE DISTRIBUTION
YPtI)=EXPERlMENTAL PRODUCT CUMULATIVE PASSING
XPT(I»=PHEDICTED PRODUCT SIZE DISTRIBUTION
YPT(I)sPHEDlCTED PRODUCT CUMULATIVE PASSING
vo
I
1
2
3
H
5
6
7
a
9
IB
SIZE
40.00000
20.00000
10.00800
5.0000*
2.50000
1.25000
.62500
.31250
.15625
.07813
CLASS
X
X
X
X
X
X
X
X
X
X
20.00000
10.000*0
5.00000
2.50000
1.25000
.62500
.31250
.15625
.07613
0.
XP(I)
0.
0.
0.
.03700
.03100
.13500
.20600
.22400
.16100
.20600
XPT(I)
«.
H.
0.
-.031H3
.04048
.15281
.20074
.17749
.13716
.3043b
ERROR
0.
0.
0.
-..06843
.UH948
.01784
..00526
-.04651
-.82384
.09836
YP(1)
1.00000
1.00000
1.00000
.96300
.93200
.79700
.59100
.36700
.20600
0.
YPT(I)
l.BHBkta
1.0400B
1. 014040
1.0314£
.99095
.83611
.63737
.45586
.32272
.01836
PERCENT
ERROU
rf.
a.
u.
184.93'ie6
30.57773
13.21530
2.55"i45
20.76469
l«*,ad^64
47.74^87
-------
THE SELECTION FUNCTION IS
1
2
3
7
6
9
IB
.291HS
.16695
.09563
.41797
.01030
.00590
.00338
0
o>
o
THE BREAKAGE FUNCTION B
-------
THE CUMULATIVE: BREAKAGE FUNCTION BBCI.J) is
0.
.235E+0*
.135E+00
.772E-B1
.2S3E-01
Ifl31E-02
0.
.717E+00 0.
0.
0.
.717E+00 0*
.HllC+00 .
.135E+00
,253t>01
.145E-01
.235E+HD
.772E-JJ1
.253E-01
0.
0.
0.
0.
fl.
0.
0.
0.
.717E+H* B.
.717E+00
.135E+0H
.142E-01
.235E+00
.13SE+00
.772E-01
0.
0.
0.
0.
0.
0.
.717E+00
.02 -,
-.H11E-02 -,73lE-«2 -.122E-01 -.2<(7E*01 -.
,iHflr>ai
-------
THt INVERSE OF T(IiJ)=TI(I.J) IS
.lk.BE+01 0.
.717E+00
.5B1E+00
.377C+00
.330E+00
.289E+00
.100F401 0.
,7l7t+00 .100E+01
.b88E+00 .717E+B0
.%01£+B0 .5B8E+0B
.433E+00 .501F.+00
.256E+B0
.330E+B0
.289E+B0
.29l£+0f)
.377E+0B
.33BE+00
0.
0.
.588E+0H
.501E+0«
e.
0.
.100E+B1 0,
.717E+0H
.501E+00
.433E+00
0.
0.
0.
0.
0.
.717F+0D
.501E+0H
B.
B,
a,
B,
0,
,717E+B3
,5fiflE+80
.5-5ME + 80
B.
0.
B.
0.
0.
0.
a.
a.
a.
3.
.130E+01
.717E+00
0.
0.
0.
0.
0.
B.
0.
.100F+01
ro
THE KATR1X JC(I.J) IS
.112E+B0
0.
0.
0.
0.
0.
0.
0.
0.
0.
M.
.239t+B0
0.
0.
0.
0.
H.
u.
£.
0.
0.
0.
.771E+00
0.
0*
0.
0.
to*
0.
0.
0.
0.
0.
.209F.+ 01
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
.431E+01
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
• 7 B*4 1 H
B.
0.
0.
0.
0.
0.
0.
B.
0.
K01 B.
* "oC t H
H.
B.
B.
0.
0.
0.
0.
B .
B.
t01 0.
,llflE<
0.
0.
1. 0'.
1. 0'.
9. 0.
1. B.
<*. 0.
T. B.
3 » 0 «
Kkt2 0. 0.
.133E*02 B.
<».
.157F+02
-------
THE HILL MATRIX DU.J) IS
.112E+0B
.913E-B1
.322E+BB
.706E+00
.107E+01
.123E+01
.112E+81
.859E+00
.590E40B
.943E+00
0.
.239t+00
.382E+00
.816E+U8
.123E+B1
. J39e+ui
.126E+B1
.972t-m0
.666L+00
,186E+al
0.
0.
.771E+00
.9H7E-f0fl
«m0E+ai
.158E+01
.1H3E+81
.1B9E+01
•749E+00
.119E+81
0.
0.
0.
.209E+01
.159E+H1
.177E+81
.159E+H1
.122C+H1
.832E+BH
.133E+B1
0.
B.
0.
0.
.<*31E+01
.196E+01
.175F+01
, 13tE + Bl
.911E+00
.l^SE+ai
0.
0.
0.
8.
0.
,
•
.
.
•
.IS9E+01
0,
H,
H,
0!
0.
0.
0.
0,
0,
0,
0.
It,
nec+02
&,
it.
k),
I. x 0.
i. a.
I. 9,
.133E+0? 0.
.155E+B1
.163E+B1
THE CLASSIFIER FUNCTION IS
1
2
3
4
S
6
7
8
9
10
cm
1.00000
1.00000
1.000BB
.99000
0.
0.
0.
0*
0.
0.
-------
THE MATRIX FlltJ) IS
•
•
•
.706E-B2
,1*7E+01
.123C+01
.112E+01
.859E+8B
.59BE+0M
.9H3E+0B
B.
8.
0.
.816E-02
.123E+01
.139E-HJ1
.126E+01
.972C+00
.666E+B8
.1B6E+01
B.
0.
B.
.947E-82
.140E+01
.15BE+01
«1<»3E+01
.109E+01
,7"»9E+B0
.119E+01
0.
0.
0.
.209E-81
.159E+01
.177E+B1
.159E+01
.122E+H1
.832E+ca
.133E+H1
B.
B.
0.
B.
.i»31E + 81
.196E+01
.175E+01
.13'tE + Bl
.yllE+00
.!<4bE4Bl
0.
0.
0.
0.
0.
.704E+B1
.189E+01
.mtt-fBl
.979E*0H
.155E+B1
B.
U.
».
«.
M.
H.
.968E+B1
.151E+D1
.103L+M1
.163E+B1
B.
0.
k>.
0.
0.
0.
0.
.iiar+02
.1B7L+H1
,169t+Hl
H.
*.
B.
0.
«.
«.
t.
c.
.133F+0?
.17?L>Ml
0.
B.
0.
B.
t).
0.
0.
0.
0.
.157F+02
THE MATRIX SJItJ) IS
0.
913E-01 .
322C-I-00 -,
699E+00 -.
U.
8.
0.
0.
0.
0.
0.
0*
0.
0.
761L+00
382E+00
b08e+0fl -.937E+0B -.1B7E+B1
0. 0.
B.
0.
0.
0.
0.
0.
0.
B.
0.
.1BBE+B1
B.
0.
B.
0.
B.
0.
0.
0.
0.
.100E+01
0.
0.
0.
0.
0.
B!
.100E+B1
B.
B.
f.
B.
B.
B.
0.
0.
0.
0.
.100E+01
0.
0.
B.
B.
*.
0.
0.
9..
«.
C.
.100F+01
0.
a.
0.
0.
8.
0.
0.
0.
0.
.100F+01
-------
THE INVERSE OF 6
CJI
THE CLOSED CIRCUIT MATRIX H(I,J) IS
0*
0.
B.
-,2«»5E-01
.632E-01
.135E+00
.152E+00
.130E+00
.938E-01
.157E+00
0.
0.
0.
-.294E-01
.b9"*£-Bl
.134E+00
.153E+B0
.131E+B0
.9HBK-B1
•159&400
0.
0.
0.
-.387E-01
.
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-80-115
3. RECIPIENT'S ACCESSIOWNO.
4. TITLEAND SUBTITLE
SIGNIFICANCE OF SIZE REDUCTION
SOLID WASTE MANAGEMENT
Volume 2
IN
5. REPORT DATE
August 1980 (Issuing Date)
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
George M. Savage
George J. Trezek
9. PERFORMING ORGANIZATION NAME AND ADDRESS
University of California
Berkeley, California 94607
10. PROGRAM ELEMENT NO.
C73D1C
11. CONTRACT/GRANT NO.
R805414
12. SPONSORING AGENCY NAME AND ADDRESS
Municipal Environmental Research Laboratory—-Gin.,OH
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, Ohio 45268
13. TYPE OF REPORT AND PERIOD COVERED
Final 7/P5/7P. - 7/95/7Q
14. SPONSOrflNG'AGENCY'c'obE'
EPA-600/14
15. SUPPLEMENTARY NOTES
See also Volume 1, EPA-600/2-77-131, NTIS PB 272096, and Volume 3 (in preparation)
Project Officer: Carl ton C. Wiles 513/684-7871
16. ABSTRACT
This report presents results of shredder tests using raw municipal solid waste,
air-classified light fraction, and screened light fraction. The tests simulated
single- and multiple-stage size reduction, using a 10-ton per hour swing hammermill
and a small, high-speed fixed hammer shredder. The tests are generalized so that
the characteristic particle size and energy consumption can be predicted. Various
hardfacing materials and their ability to perform with different solid waste mate-
rials were also tested. A computerized comminution model of a size-discrete, time-
continuous nature has been developed to predict size distribution and energy
consumption.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
comminution
waste treatment
refuse
resource recovery
size reduction
solid waste
shredder design
13B
18. DISTRIBUTION STATEMENT
Release to public
19. SECURITY CLASS (ThisReport)
unclassified
21. NO. OF PAGES
180
20. SECURITY CLASS (Thispage)
unclassified
22. PRICE
EPA Form 2220-1 (9-73)
166
------- |