ArticlePDF Available

Abstract and Figures

The paper deals with a method and technology for increasing of precision sorting of transparent materials – glass. The recycled glass is a valuable material which is necessary as an ingredient in the process of glass manufacturing. We propose a very fast industrial device based on FPGA (field-programmable gate array) image processing along with a novel algorithm for a robust real-time treatment of acquired data. The system consists from a line-scan camera with fast CameraLink interface, FPGA processing, battery of valves propelled via pneumatics and nozzle block. Very important part of the system is a background configuration software enables precise assessment of required limits as inputs to FPGA processing. The overall performance is discussed from the perspective of laboratory and industrial tests. The proposed FPGA image processing can be utilized for further future enhancements in the field of precise sorting of glass with stuck labels on it. The results give very good performance even in the industrial environment with a lot of dust and dirt and hence the glass does not need to be extensively cleaned.
Content may be subject to copyright.
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
indexed on: http://www.scopus.com
431
Increasing of Precision Technology of Glass Sorting Based on Very Fast Reconfigurable
Image Processing
David Krcmarik, Michal Petru, Ivan Masin
Technical University of Liberec, Studentská 2, 461 17, Liberec 1, Czech Republic. E-mail: michal.petru@tul.cz, da-
vid.krcmarik@tul.cz, ivan.masin@instituti.cz
The paper deals with a method and technology for increasing of precision sorting of transparent materials – glass.
The recycled glass is a valuable material which is necessary as an ingredient in the process of glass manufacturing.
We propose a very fast industrial device based on FPGA (field-programmable gate array) image processing along
with a novel algorithm for a robust real-time treatment of acquired data. The system consists from a line-scan
camera with fast CameraLink interface, FPGA processing, battery of valves propelled via pneumatics and nozzle
block. Very important part of the system is a background configuration software enables precise assessment of
required limits as inputs to FPGA processing. The overall performance is discussed from the perspective of labo-
ratory and industrial tests. The proposed FPGA image processing can be utilized for further future enhancements
in the field of precise sorting of glass with stuck labels on it. The results give very good performance even in the
industrial environment with a lot of dust and dirt and hence the glass does not need to be extensively cleaned.
Keywords: Glass cullet; sorting; optical processing; FPGA; color extraction; pneumatics.
Introduction
Glass is very good recyclable like in comparison with
polymeric materials [25]. Up-to-date goal of glass in-
dustry towards higher level of glass products recycling is
to use as much glass cullet as possible [1]. The use of
cullet avoids CO2 emissions, since cullet requires less
energy to melt, and replaces carbonated raw materials.
Although significant effort has been made by the
upstream European container glass industry, especially
through cullet recycling whose rate has increased from 43
% in 1990 to 73 % in 2015 [2], further improvements are
required to reduce its global environmental footprint,
enhancing not only cullet use but also improving its qua-
lity [3]. Glass cullet can also be used in the construction
industry for production ceramic bricks, tiles and their
glazing, glass-ceramics, foam glass-ceramics, porcelain
[4] or as a light composite material for transport applica-
tions resp. alternative for carbon composites [26,27]. An
essential condition for recycling glass is efficient, fast and
precise sorting of glass cullet. Glass sorting and separa-
ting methods and machines have accordingly been stu-
died and discussed for more than fifty years [5–9]. Gene-
ral glass sorting systems (Fig. 1) typically consist of the
following five sub-systems:
feeding sub-system ensuring the transfer of the
mixture of particles into the inspection zone
inspection sub-system ensuring the measure-
ment or detection of physical properties of par-
ticles
evaluation sub-system ensuring comparison of
actual and required values and transmit signals
for rejecting or accepting of measured particle
sorting sub-systems ensuring by means of actua-
tors physical action on the particles
collecting sub-system ensuring the position of
the sorted particles.
Fig. 1 General glass cullet sorting system
Glass sorting processes are based on the evaluation of
physical characteristics of sorted material or particle. For
this evaluation mechanical, optical, chemical or electrical
effects are usually exploited. Historically and currently
used cullet sorting methods include:
sorting as a function of density [10]
sorting by differential thermal characteristic
[11]
froth flotation [12]
electrostatic separation [12]
electro-optical sorters that recognize the color
of cullet based on their opacity [5]
sorting by transmission of visible light at cer-
tain wavelengths to distinguish between clear,
brown and green glass [8, 13, 14, 15, 16]
sorting using camera with high power multi-
spectral LED light source [17]
spectrophotometric techniques using laser
beam light [18, 19, 20]
spectroscopy using UV absorption [21, 22]
sorting using X-ray fluorescence spectroscopy
[23]
optical sorting usingan array of optical vortices
[24]
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
432
indexed on: http://www.scopus.com
In terms of the steadily increasing volume of recycled
glass and increasing demands for productivity and quality
of sorting process, image processing methods and
spectroscopic methods have the greatest potential from
the above listed methods. The subject of our research was
therefore sorting technology based on image processing.
Considering that the increase in sorting productivity (effi-
ciency) is influenced on the one hand by the image pro-
cessing speed factor (equivalent of production machine
cycle time) and on the other side by the factor of sorting
machine reconfiguration time (equivalent of production
machine changeover time), we are discussing the possibi-
lities of improving the sorting machines in terms of these
factors.
Materials and methods
This paper deals particularly with final stage of cullet
sorting process – glass sorting machine. The sorting ma-
chine consists of two parts hardware and software. First
we will deal with hardware.
2.1 Hardware
We have used a traditional configuration of using
compressed air for blowing out desired/undesired pieces.
The input to the glass sorting machine is glass cullet with
CSP without any organic material. The principal scheme
is depicted in Fig. 2. The vibration feeder needs to have a
gratual circucalar transition into chute for cullets. If this
is not met the cullet are jumping on the chute instead of
sliding. Since the cullet is very abrasive we have found
that optimal material is glass. Tests with stainless steel
showed its impropriety. At the end of glass chute just be-
fore its end there is a high intensity LED light (3 segment
parallel beam Corona II) with a diffuse piece of plastic or
glass in order to have uniformly distributed the light along
the whole width of the chute. This end of glass chute ser-
ves as an area of detection which is sensed by a line ca-
mera (Basler spL 4096-39kc). Few centimeters below is
a set of nozzles (in line holes with 1mm diameter apart
from 6mm) which are propelled by individual valves (BN
16 watts, 80 psi, open delay 3,88ms, close delay 1,84ms).
Each valve has an equal lenght hose connecting it to a
proper nozzle. The valves are initialized by printed circuit
board which is controlled by a supervisor FPGA based
logic. The line camera, area of detection and the light are
all in one line, hence the camera is looking directly into
the light. This approach of sorting is usable only for trans-
parent materials. However if the light is placed on the
same side from area of detection as the camera, the ma-
chine could be used also for non-transparent materials.
A few centimeters below is the decision edge which
has to be properly adjusted. The overall view on the ma-
chine is visible in Fig. 3.
2.2 FPGA
Quite unique is the software which is the heart of the
machine. It consists of two parts FPGA real-time con-
trol and off-line configuration GUI. The block diagram of
the FPGA software is depicted in Fig. 4. The used FPGA
is within a System on the chip (SOC) using a Zynq tech-
nology – MicroZed Zynq 7020.
Fig. 2 Principal scheme of the main part of machine
sorting
Fig. 3 Machine scheme: 1 – vibration feeder, 2 – glass
chute, 3 – area of detection, 4 – lights, 5 – nozzle block,
6 – valve block, 7 – decision edge, 8 – valve control unit,
9 – line camera with camera link interface, 10 – pres-
sured air buffer, 11 – power, 12 – PC with FPGA board
Fig. 4 Block diagram of FPGA data processing (user
inputs are in green boxes)
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
indexed on: http://www.scopus.com
433
Fig. 5 Serialization of data in CameraLink protocol
The line scan camera provides continuously 2048
color pixels (camera uses Bayer encoding). Each pixel
consists from red, green and blue component hence resul-
ting in 24 bits/pixel. The exposure time of line scan ca-
mera was chosen as small as possible 80 µs. The data
from all 2048 color pixels are sent every 82 µs. The pro-
tocol used to send data from camera to FPGA is Camera-
Link Base (4 differential parallel data lines) 3tap, each
color coded to 8 bits and clock speed 40 MHz. The overall
data throughput is quite high 699 MHz. The CameraLink
protocol enables such a high throughput thanks to using 4
parallel data lines and embedded serialization of data – in
other words upon one clock tick on each data line come 7
bits of data – Fig. 5. Data coming to FPGA are processed
in CLGrab block which deserializes data (clock speed of
CameraLink is 40 MHz). As a result CLGrab block gives
for each pixel RGB components and also provides if
currently read data are valid and internal pixel clock. Data
are converted from RGB space into YUV because YUV
space better suits needs for sorting. The conversion is in
RGB2YUV block. The coefficients for conversion (Y =
0.3125R + 0.5625G + 0.125B, U = -0.125R -0.3125G +
0.4375B, V = 0.9375R -0.5G -0.125B) are chosen in such
a way to be fixed point values with 4 decimal places. This
is because of FPGA processing. Next blocks operate with
user inputs which are provided from off-line user inter-
face. First we have to clarify the algorithm strategy used.
The obtained image from camera when visualized can be
typically as shown in Fig. 6. In fact camera gives at cer-
tain time only one row (e.g. ln01, ln02). So the shown
image is a collection of several lines (namely lines within
rows A to K) over some time. The lines can even overlap
depending on the speed of the falling glass (typical values
are around 2.2 m/s) and on the shutter of the line scan
camera. In either case the functionality does not change.
The cullet moves from up to down. We group always a
certain number of consecutive lines into a single row
for example lines ln01 up to ln20 are grouped into row A,
then line count starts again and other 20 lines are grouped
into row B. Similarly we group the pixels into columns.
So for example the camera is put into the sorting machine
in such a way that the area of detection starts at pixel 10
and ends with pixel 569. First 35 pixels (e.g. px10 to px
44) are then grouped into column 1 then next 35 pixels
are grouped into column 2 etc. It is necessary to point out
that our rows and columns are only virtual. Rows are
guided by the time and columns are guided by the posi-
tion of the pixels within the sensing chip. The pixel count
which forms a single column was chosen in such a way
that a certain nozzle propelled with a particular valve is
blowing air into this pixel area. The line count which form
s a single rows was chosen accordingly to pixel count in
such a way that a particular cell defined by column and
row forms approximately a square although this is not ne-
cessary for correct operation. The pixel count, line count
and start pixel of area of interest are all user configurable.
Fig. 6 Part of typical image from camera
Fig. 7 Controlling PCB for valves
So each pixel belongs to a certain cell. Each cell
(bank) is treated separately. The user has to choose seve-
ral other parameters within GUI. Parameter called li-
mit_obj (see Fig. 4) defines a maximal Y component of a
pixel which describes a potential pixel of an object (glass
or CSP) – it is usually around 245 (span of Y is from 0 to
255 – 255 means white color which means there is no ob-
stacle between light source and camera). The user has to
define also up to 10 sets of senary numbers – upper limit
of Y, lower limit of Y, upper limit of U, lower limit of U,
upper limit of V and lower limit of V. If all particular pi-
xel components fall within such limits (or another 9 sets
of limits), the pixel is said to fulfill the condition for the
sorted color. Hence every pixel is tested for an „object“
and for a „color“. If „object“ (OBJECT) is met the par-
ticular bank to which pixel belongs is incremented and
similarly if „color“ (COLOR) is met the particular bank
is also incremented. After each 20 lines forming one vir-
tual row we evaluate the counters for OBJECT and
COLOR for each bank (cell) and if the conditions (1) and
(2) are both met a particular valve is labeled as Enabled.
OB_Bxx > obj_pres (1)
CO_Bxx > col_pres * OB_Bxx (2)
Otherwise a particular valve is labeled as Disabled. In
equations (1) and (2) OB stands for OBJECT, CO stands
for COLOR, Bxx marks a particular bank (eg. B20 is
Bank20) and two left parameters obj_pres and col_pres
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
434
indexed on: http://www.scopus.com
are user defined. Parameter obj_pres describes the per-
centage of cell which has to be filled with any part of
falling object (e.g. for pixel count 35 and line count 20 we
have maximum of 700 potential pixels, if obj_pres is set
to 350 it says that if at least 50 % of pixels in a bank
belong to potential object we have to consider that there
is a object of interest). Parameter col_pres is a fixed point
decimal number with 5 decimal points describing what
percentage of object pixels within a bank has to meet the
any of the 10 color specification set for a particular color
to sort in order to really enable the valve. So the core of
the real-time evaluation is within the blocks LimObj,
LimCol, Banks and Eval (see Fig. 4). Valves have speci-
fication which defines the maximum allowed time when
enabled. This time for our valves is around 0,5s. This is
due to fact that there is a high current flowing into the
valves when on and they need to cool down. Normally the
valves are on only a small amount a time. But just as a
protection a special counter which control the amount of
continuous operating time for each valve is implemented.
If such a counter is exceeded the particular valve is forced
for a certain time to be off. Next block is a delay block.
As shown in Fig. 2 there is a certain distance (in our case
cca 10 cm) between the area of detection and actual air
pressure nozzles. So the 80bit-length RESULT is stored
for a certain amount of cycles (cycles is a virtual row time
duration) and after that is fed into MSPI (proprietary mo-
dified SPI) interface to the printed circuit board (PCB)
which controls each valve – Fig. 7.
2.3 Graphical user interface
The off-line graphical user interface is very important
since there are many parameters which have to be confi-
gured into FPGA. The interface greatly simplifies the pro-
cess. It is written in C# and interfaces the FPGA via ARM
processor. The communication link between the configu-
ration PC and MicroZed is via Ethernet so it is even pos-
sible to configure the FPGA remotely. In ARM processor
there is set of functions which call drivers directly
communicating with appropriate parts of FPGA. More
details about the software is behind the scope of this ar-
ticle. User interface provides several options – image
grabbing, assesment of user parameters (parameter in
green from Fig. 4) based on the color which is wanted to
be sorted, serial interface for sending selected parameters
to FPGA and testing of individual valves. The image
grabbing option is implemented as a circular buffer of ob-
tained lines from line scan camera which are saved into
DDR3 which is part of MicroZed. User has several opti-
ons of start/stop grabbing. Either using a GUI buttons, or
use buttons on MicroZed board or the grabbing can be
started by an internal event. The last option is mostly
used. The user simply takes fistful of material to be sorted
and cast it into the sorting machine. The program automa-
tically start grabbing the image after first detection of ob-
ject. After that user can download the image and start to
process it. Found parameters are saved under a certain
profile (e.g. profile Green glass, Brown glass, CSP, …)
and using a built in serial interface send to the FPGA.
FPGA stores the data into an SD card so after restart is
immediately starts to do its jobs according to the last con-
figuration made. This serial communication is easily fe-
asible since CameraLink standart has two differential li-
nes for sending and receiving data (SerTC, SerTFG). Tes-
ting is needed because the valves from time to time stop
working properly. Hence user can active arbitrary valve
and check for proper function.
Results and discussion
One of the keys which are crucial for proper function
is to choose before mentioned parameters appropriately.
Along with the unique described algorithm this is a very
important thing. In Fig. 8 there is a GUI with obtained
image from the line scan camera.
Fig. 8 Initial GUI
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
indexed on: http://www.scopus.com
435
Fig. 9 Ignoring black content (Y<20)
Fig. 10 Found Y limits with required portion of histo-
gram 80%, in U and V histograms are visible two
graphs as a result of pixel chosen with the condition
95<=Y<=169 (such pixels shown in red above)
Let’s consider a case that a user wants to create a new
profile which would blow out all green glass and rest will
stay. First of all he has to define a value for parameter
limit_obj (see Fig. 4). This gives the information which
pixels are considered as object pixels and which are only
background (background can be formed by a dust which
could by deposited for some time on the chute until it is
brushed away by cullet). First he has to select with a click
a particular piece of glass of interest. Then using the but-
tons in the upper right corner he starts to process the piece
of glass. He defines the limit of black which he wants to
ignore (typically Y<20). The glass on the edges has
nearly always artifacts due to the refraction of light which
gives his edges very dark color. We want to avoid such
color not to come into our process. The result is on Fig.
9. Next he use three buttons “Histogram Y”, “Histogram
U” and “Histogram V” which computes appropriate qu-
antities. Using text boxes defining wanted portion of se-
lected histograms, the user can set limits on Y, U and V
which satisfactorily satisfie his need. It is advisable to use
such portions of histograms so that the resulting Y, U, V
limits are as much as possible restrictive. If they are too
wide the pixels which satisfy the conditions can occur
also within glass pieces which are not wanted to sort
Fig. 10. When chosen Y limits, the U histogram is slightly
changes according to the restrictions of Y and similarly
the same applies for V limits with respect to U (Fig. 11).
Also the amount of pixels which satisfy the conditions for
Y and U is less then only for Y condion (similarly for
conditions Y, U and V with restpect to Y and U) – see red
dots in a piece of glass in Fig. 11.
Fig. 11 Subsequent finding limits for U and V with vi-
sible smaller amount of pixels which satisfy (in red)
The found limits are hence (note that range for Y is 0
to 255, range for U is -128 to 127 and range for V is -128
to 127):
Ylim00Do = 95
Ylim00Up = 169
Ulim00Do = -61
Ulim00Up = -49
Vlim00Do = -57
Vlim00Up = -50.
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
436
indexed on: http://www.scopus.com
When the limits are applied and with the assumption
that the user chooses other important parameters
obj_pres=50% and col_pres=50% the hit result is shown
in Fig. 12. It is important to note that only a portion of the
piece of glass is hit. It is also apparent from Fig. 13 that
only one piece of glass is hit and the other pieces of glass
are intact. This is the reason why there are up to 10 sets
of such limits which are all tested. Hence we can save
current limits and continue with different piece of glass
of desired green color – Fig. 14.
Fig. 12 Left - pixels which satisfy the Y, U, V limits,
right - hit results (valves which are enabled – red marks
the places of the piece of glass where the air pressure is
applied due to enabled valves)
Fig. 13 Overall view of hits due to previously found con-
ditions
Although we were finding optimal limits for only one
piece of glass (Fig. 14 left), the pixels which satisfy the
new conditions are also in other pieces (Fig. 14 center)
and the resulting hits of pressured air is shown in Fig. 14
right. When so far 2 set of conditions are combined toge-
ther the result is shown in Fig. 15 left. After adding 2
more conditions the result is in Fig. 15 right.
Fig. 14 Left – chosen piece of glass after black removal
and application of 80% Y, U, V limits, center – overall
pixel satisfaction for new set of limits, right – overall hit
results
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
indexed on: http://www.scopus.com
437
Fig. 15 Left - combined condition sets 1 and 2, right –
combined condition sets 1 through 4
The proposed method of finding the required limits
depends only on the initial set of material for configura-
tion. This is always available. The method convergence
for finding the whole set of required Y, U, V limits is very
fast usually within 6 sets. The overall system is very
robust. Let’s suppose that during sorting the user finds out
that some kind of green glass is not blown away. The only
thing he has to do is to pick several pieces of such glass
and add the configuration for that material to the current
sets. User can also choose sorting several kinds of mate-
rial at once (e.g. he can set the machine to blow away all
green and brown glass). The system is also prepared and
tested for reverse sorting – the demanded material is not
blown out and the rest is blown out. The proposed FPGA
structure enables such modifications very easily.
Conclusion
The proposed system is a complex sorting machine
with all necessary parts: hardware, FPGA real-time
communication, GUI for easy configuration and proposed
and tested algorithms and techniques. Our goal was to
enhance the availability of sorting universal sorting tech-
nology, which we believe is a way for solving part of
current environment challenges. We have tested many ti-
mes the efficiency of proposed device and algorithms. For
CSP sorting the percentual efficiency is around 98%. For
color sorting the results are stated in Table 1.
Tab. 1 Efficiency of sorting
Color component Efficiency of sorting
Green 68%
Brown 74%
Olive 73%
Transparent 78%
A video of test when sorting is attached. It shows real
environment during tests – conveyor belt, dirty cullet with
a lot of CSP. The process of finding the Y, U, V limits is
very important. So far we achieve this with manual
setting of limits according to required portion of histo-
grams. We would like to continue with full automatic
configuration. The user would cast a fistful of glass to the
sorting machine and mark the required pieces to be sorted
out and the system would find alone all the necessary cha-
racteristics.
Acknowledgement
The result was obtained through the financial support
of the Ministry of Education, Youth and Sports of the
Czech Republic and the European Union (European
Structural and Investment Funds – Operational Pro-
gramme Research, Development and Education) in the
frames of the project Modular platform for autono-
mous chassis of specialized electric vehicles for freight
and equipment transportation”, Reg. No.
CZ.02.1.01/0.0/0.0/16_025/0007293.
References
MINKO, N.; BOLOTIN, V.; ZHERNOVAYA, N.
(1999). Technological, energy, and environmental
aspects of collecting and recycling of cullet (A Re-
view). Glass and Ceramics, Vol. 56, N. 6, pp. 131-
133.
Glass recycling hits 73% in the EU. [Internet].
Available from: http://feve.org/glass-recycling-
hits-73-eu/ [Accessed: 16 July 2018]
TESTA, M.; MALANDRINO, O.; SESSA, M.R.;
SUPINO S. and SICA D. (2017). Long-Term
Sustainability from the Perspective of Cullet
Recycling in the Container Glass Industry: Evi-
dence from Italy. Sustainability, Vol. 9, pp. 1752 -
1771.
SILVA, R.V.; de BRITO, J.; LYE, C.Q.; DHIR,
R.K. (2017). The role of glass waste in the produ-
ction of ceramic-based products and other applica-
tions: A review. Journal of Cleaner Production,
Vol. 167, N. 20, pp. 346-364.
PYE, L.D.; STEVENS H.J.; La COURSE W.C.
(eds) (1970). Introduction to Glass Science: Pro-
ceedings of a Tutorial Symposium. State Uni-
versity of New York.
LIMBACHIYA, M. C.; LIMBACHIYA, J. R.,
editors (2004). Sustainable Waste Management
and Recycling: Glass Waste - Vol. 1: Challenges
and Opportunities; Kingston University London.
STESSEL, R.I. (1996). Recycling and Resource
Recovery Engineering: Principles of Waste Pro-
cessing. Springer.
DUBANOWITZ, A. J. (2000). Design of a Mate-
rials Recovery Facility (MRF) For Processing the
Recyclable Materials of New York City’s Munici-
pal Solid Waste. M.S. thesis, Columbia University.
BARNUM R. A. (2008). The Influence of Batch
Segregation and Bulk Flow on Glass Quality. In:
66th Conference on Glass Problems (ed. Kriven,
W.M). John Wiley & Sons, pp: 91-103.
June 2019, Vol. 19, No. 3
M
ANUFACTURING
T
ECHNOLOGY
ISSN 1213–
2489
438
indexed on: http://www.scopus.com
KIMMEL, K. S.; HAWK, N. A.; KELLER, M.A.
WHITMORE, F. (2002). Patent US6464082.
Cullet sorting using density variations.
KIMMEL, K. S.; HAWK, N. A. (2000). Patent
US6112903. Cullet sorting by differential thermal
characteristics.
STIRLING, H. (1977). The recovery of waste
glass cullet for recycling purposes by means of
electro-optical sorters. Conservation & Recycling,
Vol.1, N. 2, pp. 209-219.
AFSARI, F.; AFSARI, B.; FLYNN, P. C.;
KOPELIOFF, D.; SHOOK, F. S.; VENDRELL,
L. P. (2002). Patent US8436268B1. Method of
and apparatus for type and color sorting of cullet.
STELTE, N. (1994). Patent US5333739. Method
and apparatus for sorting bulk material.
REICHERT, A.; HOBERG, H. (1987). Photomet-
ric Sorting of cullet. In: Environmental Techno-
logy: Proceedings of the Second European Confe-
rence on Environmental Technology (eds. de
Waal, K.J.A.; van den Brink, K.J.A.), Springer,
pp. 457 – 465.
BONIFAZI, G.; SERANTI, S. (2006). Imaging
spectroscopy based strategies for ceramic glass
contaminants removal in glass recycling. Waste
Management. Vol. 26, N.6, pp. 627-639.
Redwave CX / CXF. [Internet]. Available from:
http://www.redwave.com/en/products/redwave-
cx-cxf/ [Accessed: 16 July 2018]
BONIFAZI, G. (2004). Classical Imaging and Di-
gital Imaging Spectrophotometric Techniques in
Cullets (Glass Fragments) Sorting. In: Intelligent
Robots and Computer Vision XXII: Algorithms,
Techniques, and Active Vision (eds. Casasent, D.
P.; Hall E. L.; Röning, J.), SPIE, pp. 264-277.
CRAPARO, J. C.; WEISBERG, A.; de SARO, R.
(2008). Measurements of Batch and Cullet Using
Laser Induced Breakdown Spectroscopy. In: 66th
Conference on Glass Problems (ed. Kriven,
W.M.), John Wiley & Sons, pp. 105-118.
SUMINA, K.; AKAI, T., YAMASHITA, M;
YAZAWA T. (2004). Patent JP2004219125. Me-
thod for sorting glass cullet and sorting apparatus
therefor.
HUBER, R.; PANSINGERM, C. (2007). Patent
application AU2006203360A1. A method for de-
tecting and sorting glass.
HUBER, R. (2013). Waste glass sorting based on
UV absorption and fluorescence. In: 16th Spectro-
Net Collaboration Forum, Karlsruhe.
Redwave XRF-G. [Internet]. Available from:
http://www.redwave.com/en/recycling/glass/sen-
sor-based/redwave-xrf-g/ [Accessed: 16 July
2018]
GUO, C.S.; YU, Y.N.; HONG, Z. (2010). Optical
sorting using an array of optical vortices with
fractional topological charge. Optics Communica-
tions, Vol. 283, N. 9, pp. 1889–1893.
MAJERNÍK, J.; KMEC, J.; KARKOVA, M.;
PODAŘIL, M. (2017). Possibilities for change of
thermoplastic tensile properties using admixture
of recyclable material. Manufacturing Techno-
logy, Vol. 17, N. 5, pp. 778–782.
PETRŮ, M.; MARTINEC, T.; MLÝNEK, J.
(2016). Numerical model description of fibres
winding process for new technology of winding
fibres on the frames. Manufacturing Technology,
Vol. 16, N. 4, pp. 778–785.
PETRŮ, M.; MLÝNEK, J.; MARTINEC, T.
(2018). Numerical modelling for optimization of
fibres winding process of manufacturing techno-
logy for the non-circular aerospaces frames. Ma-
nufacturing Technology, Vol. 18, N. 1, pp. 90–98.
10.21062/ujep/309.2019/a/1213-2489/MT/19/3/431
Copyright © 2019. Published by Manufacturing Technology. All rights reserved.
... The n = 8 investigations on color classification in our dataset, predominantly applied VIS-RGB sensors (6 of 8 investigations, 75.0%), but also VIS-HSI have been applied (2 of 8 investigations, 25.0%). Among the n = 8 color classification investigations, n = 5 investigations (62.5%) conducted classifications at the particle level to reduce the influence of, e.g., labels, colored bottle caps, or contaminates (Ata et al., 2005;Krcmarik et al., 2019;Tachwali et al., 2007;Wang et al., 2019b;Zhou et al., 2021). Fig. 5e shows the application of optical sensors for different material classes. ...
Full-text available
Article
Digital technologies hold enormous potential for improving the performance of future-generation sorting and processing plants; however, this potential remains largely untapped. Improved sensor-based material flow characterization (SBMC) methods could enable new sensor applications such as adaptive plant control, improved sensor-based sorting (SBS), and more far-reaching data utilizations along the value chain. This review aims to expedite research on SBMC by (i) providing a comprehensive overview of existing SBMC publications, (ii) summarizing existing SBMC methods, and (iii) identifying future research potentials in SBMC. By conducting a systematic literature search covering the period 2000-2021, we identified 198 peer-reviewed journal articles on SBMC applications based on optical sensors and machine learning algorithms for dry-mechanical recycling of non-hazardous waste. The review shows that SBMC has received increasing attention in recent years, with more than half of the reviewed publications published between 2019 and 2021. While applications were initially focused solely on SBS, the last decade has seen a trend toward new applications, including sensor-based material flow monitoring, quality control, and process monitoring/control. However, SBMC at the material flow and process level remains largely unexplored, and significant potential exists in upscaling investigations from laboratory to plant scale. Future research will benefit from a broader application of deep learning methods, increased use of low-cost sensors and new sensor technologies, and the use of data streams from existing SBS equipment. These advancements could significantly improve the performance of future-generation sorting and processing plants, keep more materials in closed loops, and help paving the way towards circular economy.
Full-text available
Article
This article deals with the issue of mathematical calculating the trajectory of the end-effector of an industrial robot in the manufacture of aerospace composites. Robots are used to define the winding orientation of the fibre strands on a non-bearing 3D core. The 3D core is attached to the robot-end-effector and is led through a fibre-processing head according to a suitably defined robot trajectory during winding of the fibre on the core. The quality of the composite depends greatly on the correct winding angles of the fibres on the frame and on the homogeneity of the individual winding layers. The implementation of these two conditions is related to determining the correct trajectory of the industrial robot, which is part of the composite production technology. The numerical modelling of a passage of the on-bearing 3D core through a fibre-processing head is described in the article. Differential evolution algorithm and matrix calculus are applied to the numerical calculation of optimized robot-end-effector trajectory to achieve optimal angles of windings of fibres on the frame. The numerical calculations of the trajectory of the robot-end-effector were used for verified for the calculated trajectory of the robot-end-effector in the real conditions of robotic laboratory of department of machinery construction.
Full-text available
Article
Glass manufacturing is a high-volume process, during which large substance quantities are transformed into commercial products, and significant amounts of non-renewable resources and energy (i.e., thermal fuels and electrical power) are consumed. The main purpose of this study is to give a critical explanation of the performance of the Italian container glass industry from the perspective of cullet being recycled, to outline the opportunities for transition towards circular business models that stimulate innovation in new sectors based on reverse-cycle activities for recycling. In 2015, disparate performances have been achieved as regards the container glass recycling rate in northern, central, and southern Italy, accounting for around 73%, 64%, and 55%, respectively. In fact, only northern Italy is in line with European targets, as by 2025 it will only need to increase its current performance by two percentage points, unlike central and southern Italy that will have to increase performance by, respectively, 11% and 20%. This shows a need to improve the efficiency of municipal waste collection systems in central and southern Italy, where undifferentiated waste still holds appreciable amounts of glass. Consequently, we propose several improvement channels, from the revision of waste legislation to the re-engineering of waste management supply chains.
Full-text available
Article
Currently, traditional materials are very often replaced by composite materials in many industrial areas. The advantages of these materials consist mainly in their lightweight, high strength and flexibility, corrosion resistance and a long lifespan. The use of composites reaches its large development in the field of aerospace. This article discusses quality of the manufacturing process technology of a specially shaped composite frame in 3D space. The used technology is based on a winding of carbon, glass, organic filament rovings on a polyurethane core. Polyurethane core which is a geometry of frame with and without a circular cross section. Quality production of said type of composite frame depends primarily on the correct winding of fibers on a polyurethane core. It is especially needed to ensure the correct angles of the fibers winding on the polyurethane core and the homogeneity of individual winding layers. The quality of fibers winding also depends on the material properties of polyurethane core and fibers. The article describes mathematical model for use an industrial robot in filament winding and how to calculate the trajectory of the robot. When winding fibers on the polyurethane core which is fastened to the robot-end-effector so that during the winding process goes through a fibre-processing head on the basis of the suitably determined robot-end-effector trajectory. We use for description numerical model and matrix calculus to enumerate the trajectory of the robot-end-effector to determine the desired passage of the frame through the fibre-processing head. The calculation of the trajectory was programmed in the Delphi development environment. Equations and relations of the numerical model are important for use a real solving of the passage of a polyurethane core through fibre-processing head.
Article
Polymeric materials are thanks its processing and utility properties materials in demand of common and special use. They are also largely replacing conventional materials. As the popularity of polymeric materials grows, also the amount of its waste increases. For this reason, there is introduced the term recycling as a method of processing, re-use of the waste, into technologies of polymeric processing. So, this paper deals with the possibilities of introduction of recycled material. The main part of this paper is created by an experiment that explores the changes of tensile properties of test specimen according to the selected percentage of additives in the volume of the basic granulate. The test specimen was produced by mixing pure granules with the addition of recycled and re-granulated materials. The conclusion of this work presents a comparison of the results of each tensile test that provide an overview of the behaviour and properties of the materials tested.
Article
This paper presents a literature review relating to the potential waste glass collection and processing as glass cullet for its use as raw material in secondary markets. Emphasis is given to the application of glass cullet in the construction industry, other than as construction aggregate, especially in ceramic-based products, including ceramic bricks, tiles and their glazing, glass-ceramics, foam glass-ceramics, and porcelain. These applications also include the use of glass cullet as a filtration medium, constituent in epoxy resins, in the production of glass fibres, elastomeric roof coatings, aesthetic finishing materials, abrasive material for surface cleaning, and paint filler. The analysis and evaluation of the vast amount of experimental research showed that glass cullet is a potentially valuable resource for the manufacture of ceramic-based products, where it can be used as substitute for expensive natural resources, improving the products’ physical, mechanical and environmental performance.
Chapter
The municipal solid waste produced per annum in the Federal Republic of Germany contains approximately 2.6 mio tons of waste glass. In the past 4 years the recycling rate of waste glass has increased annually on average by 11%. It amounted to 890.000 tons in 1985, meaning a share of approx. 34% of the total mass.
Chapter
ERCo has developed a laser-based technology for rapid compositional measurements of batch, real-time sorting of cullet, and in-situ measurements of molten glass. This technology, termed LIBS (Laser induced Breakdown Spectroscopy) can determine whether or not the batch was formulated accurately in order to control glass quality. It can also be used to determine if individual batch ingredients are within specifications. In the case of cullet feedstocks, the sensor can serve as part of a system to sort cullet by color and ensure that it is free of contaminants. In-situ compositional measurements of molten glass are achieved through immersing a LIBS probe directly into the melt in a glass furnace. This technology has been successfully demonstrated in ERCo's LIBS laboratory for batch analysis, cullet sorting, and glass melt measurements. A commercial batch analyzer has been operating in a PPG fiberglass plant since August 2004. LIBS utilizes a highly concentrated laser pulse to rapidly vaporize and ionize nanograms of the material being studied. As this vapor cools, it radiates light at specific wavelengths corresponding to the elemental constituents (e.g. silicon, aluminum, iron) of the material. The strengths of the emissions correlate to the concentrations of each of the elemental constituents. By collecting the radiated light with a spectrometer capable of resolving and measuring these wavelengths, the elemental composition of the sample is found.
Article
The current situation with respect to collecting, preparation, and recycling of cullet abroad and in the CIS countries is discussed. Information concerning cullet recycling received from Russian and CIS glass factories is summarized.
Article
Cullets optical sorting represents one of the oldest selection procedure applied to the field of solid waste recycling. From the original sorting strategies, mainly addressed to separate non-transparent elements (ceramics, stones, metal particles, etc.) from transparent ones (glass fragments), the attention was addressed to define procedures and actions able to separate the cullets according to their color characteristics and, more recently, to recognize transparent ceramic glass from glass. Cullets sorting is currently realized adopting, as detecting architecture, laser beam technology based devices. The sorting logic is mainly analogical. An "on-off" logic is applied. Detection is, in fact, based on the evaluation of the "characteristics" of the energy (transparent or non-transparent fragment) and the spectra (fragment color attributes) received by a detector after that cullets were crossed by a suitable laser beam light. Such an approach presents some limits related with the technology utilized and the material characteristics. The technological limits are linked to the physical dimension and the mechanical arrangement of the optics carrying out and in the signals, and with the pneumatic architectures enabling the modification of cullets trajectory to realize sorting, according to their characteristics (color and transmittance). Furthermore such devices are practically "blind" in the recognition of ceramic glasses, whose presence in the final selected material to melt, damage the full recycled glass fusion compromising the quality of the final product. In the following it will be described the work developed, and the results achieved, in order to design a full integrated classical digital imaging and spectrophotometric based approach addressed to develop suitable sorting strategies able to perform, at industrial recycling scale, the distinction of cullets both in terms of color and material typologies, that is "real glass" from "ceramic glass" fragments.