Resources, Conservation and Recycling 54 (2010) 571578

Contents lists available at ScienceDirect

Resources, Conservation and Recycling

journal homepage: www.elsevier.com/locate/resconrec

Automatic sorting of lightweight metal scrap by sensing apparent density and

three-dimensional shape
Shigeki Koyanaka , Kenichiro Kobayashi
Research Institute for Environmental Management Technology, National Institute of Advanced Industrial Science and Technology (AIST), 16-1, Onogawa, Tsukuba,
Ibaraki 305-8569, Japan

a r t i c l e

i n f o

Article history:
Received 27 May 2009
Received in revised form 20 October 2009
Accepted 21 October 2009
Keywords:
Automatic sorting
Lightweight metal scrap
Apparent density
Three-dimensional shape
Multivariate analysis
End-of-life vehicle

a b s t r a c t
A new method for the automatic sorting of lightweight metal scrap has been developed to aid the recycling of scrap metal. The sorting system enables separation of relatively large metal pieces according
to the differences in their apparent density and three-dimensional (3D) shape. Shape parameters such
as width, height, volume, and projected area of irregular-shaped metal pieces moving along a conveyer
are measured by means of a 3D imaging camera system consisting of a linear laser and camera with
associated optics. The measured values of the weight and shape parameters are transferred to our own
data-processing software that uses multivariate analysis. Mixed fragments of cast aluminum, wrought
aluminum, and magnesium from an end-of-life vehicles shredder facility were sorted based on the output of the data-processing software. The results show that the developed automatic sorting system is a
highly viable method that could replace conventional dense medium separation and manual sorting.
2009 Elsevier B.V. All rights reserved.

1. Introduction
The motor vehicle industry is paying a great deal of attention to
improving fuel efciency by reducing vehicle weight and thereby
reducing emissions of greenhouse effect gases. As lightweight metals, such as aluminum and magnesium alloys, are suitable materials
to reduce vehicle weight because of their low density and high specic strength, the substitution of light alloy parts for conventional
steel parts is becoming more and more prevalent. With respect to
automotive applications, the adoption of wrought aluminum parts
is now increasing for use in body panels (hoods, roofs, deck lids,
etc.), in addition to the fact that cast aluminum has been widely
used for engine parts (Inaba, 2002; Sakurai, 2007). According to a
strategic target drawn up by the Japan Aluminum Association, the
amount of aluminum parts per automobile in Japan is expected to
reach 250 kg (130 kg of wrought aluminum, 120 kg of cast or forged
aluminum) in 2025 (Okubo, 2005). However, this rapid expansion
of aluminum parts in motor vehicles has caused anxiety in the aluminum recycling business: If scrap is recovered as a mixture of
wrought and cast alloys, the recovered aluminum is not suitable
for use in wrought alloy production, and its usage is limited to cast
alloy, which allows for somewhat contaminant elements of negatively affecting mechanical and chemical properties (Ambrose et al.,
1983). As a consequence, this may lead to a supplydemand imbal-

Corresponding author. Tel.: +81 29 861 8099; fax: +81 29 861 8457.
E-mail address: s-koyanaka@aist.go.jp (S. Koyanaka).
0921-3449/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.resconrec.2009.10.014

ance in the secondary aluminum alloy market, because demand for

cast aluminum is not speculated to increase compared with that for
wrought aluminum in the future. On the other hand, magnesium
alloy parts have been introduced on a smaller scale than aluminum
because of their low density (about 30% lower than aluminum).
The use of magnesium parts is expected to increase steadily against
the background of the serious requirement for weight reduction in
the motor vehicle industry (Gesing et al., 2003). Currently, most of
the magnesium scrap from end-of-life vehicles (ELV) seems to be
recovered along with aluminum scrap, and recycled as an additive
for secondary aluminum alloy. Although this cascade recycling system of magnesium scrap is effective provided consumption is small,
separation from aluminum scrap will become important so as to
recycle it as secondary magnesium alloy when its consumption
grows in the future.
In an ELV shredder facility, shredded lightweight metal fragments are concentrated in a mixture of nonferrous metals after
conventional separation processes such as magnetic separation,
pneumatic separation, and eddy current separation. This metal
mixture usually consists of copper, brass, zinc, lead, nonmagnetic
stainless steel, aluminum and magnesium fragments, and therefore
further separation is necessary to recycle each material. Although
these different metals are currently separated by manual sorting
in most recycling facilities, several automatic separation or sorting
techniques are available for this purpose; color detection can be
applied for recovery of copper and brass, electromagnetic sensing
can be applied for recovery of nonmagnetic stainless steel (Mesina
et al., 2005), and X-ray transmission sensing and dense medium

572

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

separation can be used in separation systems, based on the difference in atomic number or density of materials. However, with
regard to separation within the three types of lightweight metals,
it is difcult to apply these techniques because of similarities in
physical properties (color, electrical conductivity, density, etc.). A
Delft University of Technology group conducted a test using a dual
energy X-ray transmission (DE-XRT) sorting system, one of the latest technologies in the recycling industry, but the result was not
good enough to achieve separation between cast aluminum and
wrought aluminum (Mesina et al., 2007). Other separation techniques based on laser-induced breakdown spectroscopy (Noll et al.,
2008; Stepputat and Noll, 2003) or X-ray uorescence analysis are
potentially applicable, but their high sensitivity to contamination
on the surfaces of metal scrap and low processing speed become
serious problems in an actual ELV shredder facility.
The purpose of this study is to develop a new automatic sorting
technique that can overcome the above-mentioned problems, and
which has high separation efciency and low processing cost. In
this study, we rst demonstrate a sorting system that combines two
cutting-edge instruments: a three-dimensional (3D) imaging camera that enables various application developments by connecting
it to a personal computer (PC); and, a weight meter that measures
the weight of a moving object on a belt conveyer. This sorting technique uses the weight information and the 3D shape parameters
of an inspected piece of scrap, and an identication of the fragment is made by inputting these values into our own calculation
program, which is based on multivariate analysis. In this paper, we
point out the necessity of the separation of wrought aluminum,
cast aluminum and magnesium scraps, and then we introduce
the equipment used in this sorting system, the basic principle for
identication, and the results of the sorting test of three types of
lightweight metal scrap. Finally, we discuss the advantageous features of the newly developed system over conventional separation
techniques, and look at further problems to be examined.

therefore it is difcult to recycle aluminum scrap into wrought aluminum alloy. With regard to wrought aluminum scrap, only the
3000 series used for the bodies of beverage cans and a part of the
5000 series used for the lids of beverage cans are currently recycled as secondary wrought aluminum, all of the others are recycled
as cast aluminum. In particular, ADC12 alloy which allows for a
relatively high contamination of silicon, iron and copper is most
frequently produced as secondary aluminum alloy. As mentioned
in the previous section, a closed recycling system in which wrought
aluminum parts are recycled as wrought aluminum parts is the target. For this purpose, it is necessary to separate cast aluminum and
wrought aluminum scrap during their recycling process, in addition to removing other contaminants. However, the differences in
the physical properties of these aluminum alloys are very small:
The color tones of these alloys are basically the same. The density
and electrical conductivity of these alloys usually spans the range
2.652.85 g/cm3 and 3060% IACS, respectively. For such materials,
accurate separation cannot be expected using conventional color
separation, dense medium separation or electromagnetic sensing
separation.
Currently, the weight of magnesium parts per automobile in
Japan is about 1 kg, which is less than 1/100 that of aluminum parts.
As long as the weight of magnesium parts in an ELV is much smaller
than that of aluminum parts, it is reasonable that magnesium scrap
will be processed as a part of aluminum scrap. The amount of postconsumer magnesium scrap in Japan is speculated to reach around
5000 t in 2015 (NEDO, 2008). Around that year, if magnesium scrap
is not separated from aluminum scrap, there is a possibility that the
average composition of magnesium in the secondary ADC12 alloy
will become larger than the standard value of 0.3%, which would
disrupt the recycling system for aluminum scrap. Meanwhile, the
addition of magnesium scrap to wrought aluminum alloy is not
realistic, because of the contamination problem. Therefore, it is necessary to separate magnesium scrap from aluminum scrap when
consumption of magnesium parts is increased in the future.

2. Demand for separation of lightweight metal scrap

3. Materials and methods
Representative alloys of aluminum and magnesium and their
chemical compositions are listed in Table 1. Production of each
alloy in Japan and their major uses are also shown (JAA, 2005).
As seen in this Table, the tolerance of wrought aluminum alloy to
impure elements is smaller than that of cast aluminum alloy, and

3.1. Sample materials

The prerequisite in this research is that lightweight metals have
been separated from other nonferrous metals in advance using

Table 1
Representative alloys of aluminum and magnesium and their chemical compositions.
Series

Wrought aluminum alloy

1000 (Al)
2000 (AlCu)
3000 (AlMn)
4000 (AlSi)
5000 (AlMg)
6000 (AlMgSi)
7000 (AlZnMg)
Cast aluminum alloy
Sand casting permanent mold
casting (AlSiMg)
Die casting (AlSiCu)
Magnesium alloy
Die casting thixomolding

Production
(103 t)

406
23
449
28
386
883
39

418
1038

11

Major alloy

1100
2117, 2036
3003, 3004, 3005
4004, 4104
5005, 5052, 5082,
5182
6061, 6063

Major use

Foil, Fin
Can body
Can lid, body
panel
Sash, body
panel

7N01, 7003, 7016,

7046

Chemical composition (wt%)

Si

Fe

Cu

Mn

Zn

Fe+Si<1.0
<0.8
<0.6
9.010.5
<0.20

<0.7
<0.7
<0.8
<0.40

0.050.20
2.23.0
0.050.20
<0.25
<0.10

<0.05
<0.20
1.01.5
<0.10
0.200.50

<0.10
<0.25
<0.10
<0.20
<0.25

1.02.0
4.05.0

0.20.6

<0.35

<0.10

<0.10

<0.10

0.450.90

<0.30

<0.35

<0.20

<0.25

AC2A, AC4A, AC4C,

AC5A, AC8A, AC8B
ADC10, ADC12

Engine, wheel

6.57.5

<0.55

Engine,
transmission

9.612.0

<1.3

AZ91D
AM60B, AM50A

Case
Steering wheel,
seat frame

<0.10
<0.05

<0.004
<0.004

1.53.5

<0.03
<0.008

0.200.7

4.05.0

Mg

0.200.50

1.02.0

<0.35

<0.35

0.200.45

<0.5

<1.0

<0.3

0.351.0
<0.20

Al 8.39.7
Al 5.66.4

0.130.50
0.260.5

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

573

Fig. 1. Examples of lightweight scrap metal fragments used in this study.

Fig. 2. Schematic diagram of the developed sorting system.

conventional separation techniques such as color detection, electromagnetic sensing, dense medium separation, or manual sorting.
In fact, aluminum scrap that contains only the three types of
lightweight metals stated above is currently recovered successfully
in most ELV shredder facilities.
Sample materials for the sorting experiments were st-sized
fragments of cast aluminum, wrought aluminum, and magnesium
alloy scraps generated at an ELV shredder facility in Japan. These
fragments, sized between 10 and 200 mm and well separated from
other materials, were taken at random from the aluminum scrap
and manually sorted into three groups: cast aluminum (104 pieces),
wrought aluminum (192 pieces), and magnesium (246 pieces).
Fig. 1 shows photographs of some of these pieces of scrap. From
their weight and shape, it is believed that cast aluminum and
wrought aluminum mainly originated from the engine and body
parts, respectively, and that all of the magnesium fragments were
die-cast parts, such as steering columns and seat frames. In addition, it was observed that the surfaces of the magnesium pieces
were corroded, whereas those of the aluminum pieces were not.
Although some fragments were covered with mud, oil, or paint,
the experiments were conducted without removing the contamination.

and an optical CCD, and the height of each fragment is determined

by laser line triangulation. As shown in Fig. 3, the laser draws a
line on the surface of a moving fragment when the attached optical sensor detects it, and a digital 3D image of the entire fragment
is constructed in the 3D imaging camera from the movement of
the reected ray. The developed system can process a fragment
having a size of 2250 mm in the horizontal direction (width and
length), and a size of 260 mm in the vertical direction (maximum
height) with a resolution of 0.04 mm. Several dimensions or 3D
shape parameters of the fragment, i.e. volume, vertically projected

3.2. Sorting system

Fig. 2 shows a schematic diagram of the developed sorting system, which consists of a belt conveyor, weight meter (Anritsu
Industrial Solutions Co. Ltd., KW6205), 3D imaging camera (Sick Co.
Ltd., IVC-3D), air compressor, air cylinders, electromagnetic valves,
and a PC. The 3D imaging camera is equipped with a linear laser

Fig. 3. Measurement of 3D shape parameter of a fragment moving on a belt conveyor.

574

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

Table 2
Parameters for data analysis.
X0 : material
X1 : weight/volume
X2 : volume
X3 : vertically projected area
X4 : length

X5 : width
X6 : maximum height
X7 : height of the center of
gravity
X8 : X2 /X3
X9 : X2 /(X4 X5 )

X10 : X3 /(X4 X5 )
Xn : X7 /X6
X12 : X2 /X3 /X7
X13 : X4 /X5
X14 : X5 /X4

area, length, width, maximum height and height of the center of the
gravity, are measured by calculation using the digital 3D image. The
sphericity of the fragment, which is usually taken into account in
conventional image analysis, cannot be measured using the current
developed system. These data are immediately transferred to the
PC and analyzed by our own data-processing software using multivariate analysis. After identication, the fragment is sorted by a
burst of compressed air in accordance with an output signal from
the PC.
3.3. Algorithm for identication
The sorting method we developed is based on the concept of
multivariate analysis: That is, an unknown fragment is identied
by an algorithm using the discriminant functions determined from
a database of weight and 3D shapes of sampled fragments. The
mixture to be separated should rst be separated or sorted by
other means when applying this method. However, that is not
critical problem if the samples are restricted to lightweight metals, because most of these fragments can be identied by manual
sorting, although this requires considerable time.
As the parameters for data analysis, sample fragment materials
(X0 ) and 14 variables (X1 X14 ) listed in Table 2 were recorded in
the database by repeated measurement. The data considered here
are a combination of X0 X14 parameters obtained by a one-time
measurement (henceforth referred to as one case). The case data
for the same fragment were separately recorded if the orientation

of the fragment on the belt conveyor was different. Especially for

a fragment having a complex shape, a lot of case data are needed
because the identication of this method is a process in which a
measured case is checked against preliminarily recorded case data
in the database. After sufcient data were recorded in the database,
all case data were sorted and categorized into several sections
according to the size of the apparent density X1 . Then a multitude
of discriminant analyses were carried out in each apparent density section with the explained variable of X0 and the explanatory
variable of X1 X14 . The reason of this procedure is discussed later.
Fig. 4 shows a schematic illustration of the algorithm used
to identify whether a fragment is cast aluminum (Alc), wrought
aluminum (Alw), or magnesium (Mg). The cases of X1 > 3.6 g/cm3
and X1 < 0.6 g/cm3 were immediately identied as cast aluminum
and wrought aluminum, respectively. Discriminant analyses were
applied to the cases of the other apparent density sections in which
the diversity of the material was high. For example, the cases that
belong to the section of 1.9 < X1 < 2.1 g/cm3 were identied over
two to four discriminant analyses. The number of case data at the
third discriminant analysis was limited to a maximum of around
50, because accuracy of the analysis deteriorated when the number of case data was more than that. This is the reason that the
ow of this algorithm was fragmented after the second analysis.
On each discriminant analysis, the discriminant function Zn given
by Eq. (1) was calculated and its constants an on were decided. Here
the subscript n means the step of discriminant analyses
Zn = an + bn X1 + cn X2 + + on X14

(1)

The threshold values for distinction n n were set to maximize

the result of distinction. Similar calculation procedures were conducted in the other apparent density sections, and the constantand threshold-values of the discriminant functions were decided.
All of these values and ow of the analyses were saved into the PC
beforehand. In the sorting experiment, the material of each sample
fragment was inspected individually by a calculation substituting
measured values of X1 X14 into the data-processing software. It

Fig. 4. Algorithm for identication of three types of light metals.

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

575

4. Results and discussion

Fig. 5. Weight distributions of fragments.

should be noted that most of the sorting results were obtained

against unrecorded case data, because it was very difcult to place
irregular-shaped fragments on the conveyor in the exact same orientation (i.e. 0.04 mm accuracy) as its former measurement.
When we consider mass processing, number of variables for
data analysis, number of case data to register in the database (nc ),
number of apparent density sections (ns ), number of discriminant
functions (nd ) and setting of the threshold values for distinction are very important factors. Actually, these numbers must be
decided according to the property of the group of fragments that are
sorted. In this study, these numbers were set at nc = 5453 (cast aluminum), 7865 (wrought aluminum), 11056 (magnesium), ns = 21
and nd = 641.

Distributions of fragment weight and several 3D shape parameters that were claried during the process of making the database
are shown in Figs. 5 and 6. As shown in Fig. 5, the average weight
increases in the order of magnesium, wrought aluminum, and cast
aluminum, but it is obvious that these materials cannot be identied by the weight of each individual fragment. Fig. 6(a)(d) were
obtained by 15 measurements per fragment by changing its posture
on the belt conveyor. Except for the fact that the 3D shape parameters of the magnesium fragments tend to be smaller than that of the
aluminum fragments, we cannot nd any clear difference between
these fragments, and therefore they cannot be identied from these
individual parameters. In Fig. 6(a), the distributions of volume X2
of both aluminum fragments are almost the same, however, this
result contains the meaningful error as mentioned below.
Fig. 7 shows the distribution of the apparent density X1 calculated from the data of Figs. 5 and 6(a). This gure clearly shows
that the apparent density of each material measured by this system
tends to be lower than its actual density (aluminum 2.7, magnesium
1.7 g/cm3 ). When the laser beam scans the surface of a fragment,
some blind spots are inevitably generated at the back or inside
the fragment. The 3D imaging camera counted these spots as part
of the fragment volume and thus overestimated the volume. An
important point to emphasize is the fact that the degree of this
measurement error is related to the fragment material. Namely,
wrought aluminum scrap characterized by fragments of twisted
thin and plate-like parts tends to carry a larger error and have
lower apparent density, compared to cast aluminum scrap characterized by fragments of broken thick and bulky parts (see Fig. 8).
It is clear that the trend in the degree of measurement error shown
in Fig. 8 is irrelevant to the scale of the fragments, and therefore
the separation among aluminum fragments based on a difference
in apparent density is likely to be achieved regardless of their size,

Fig. 6. Distributions of several 3D shape parameters of the fragments.

576

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

were separated relatively well in the region of 1.0 < X1 < 1.8 (see
Fig. 9(d)). These results suggest that when we draw attention to
a narrow region of the apparent density, there are some statistical differences among the 3D shapes of each materials fragment
that belong to the same region of apparent density. That is the
reason all case data in the database were sorted and separated
into several sections according to the size of the apparent density when we set the algorithm for the identication. The statistical
differences in the 3D shape can be extracted using multivariate
analysis.
The result of the sorting experiment of a mixture of the three
types of lightweight metals is listed in Table 3, which is one of
the typical results obtained by three tests for each fragment by
changing its orientation on the belt conveyor. As this Table shows,
the developed system can perform a sorting with an accuracy
of approximately 90% by count basis, although the separation
between wrought aluminum and magnesium is rather difcult
because of the similarity of apparent density. If the accuracy is
calculated using weight basis, the following results are obtained:
Fig. 7. Distribution of the apparent density calculated from the data contained in
Figs. 4 and 5(a).

as far as they belong in the measurement range of this sorting system. In the case of magnesium scrap, such a trend was not seen,
as all the magnesium fragments used here were die-cast parts.
Wrought magnesium parts are not widely used currently because
of the difculty in mechanical processing of the material. Although
the difference in the distribution curve according to the material
becomes clearer in Fig. 7, a region of apparent density where the
distributions overlap exists still. Therefore, it is difcult to identify a
fragment with sufcient accuracy, even if we measure its apparent
density.
Fig. 9(a)(d) shows the relation between the apparent density X1 and several 3D shape parameters of X2 , X3 , X8 , and X11 .
In these gures, sectional average values of the 3D shape parameters at 0.2 g/cm3 intervals of the apparent density are plotted.
Here we realize that the fragment material is characterized by the
relation between the apparent density and a 3D shape parameter: Wrought aluminum fragments have an apparent density of
about 0.5 < X1 < 1.5, and their X11 value in this region tend to be
smaller than that of cast aluminum and magnesium fragments
(see Fig. 9(a)). Magnesium fragments have an apparent density
of about 0.7 < X1 < 1.8, and their X3 value in this region tend to
be smaller than that of cast and wrought aluminum fragments
(see Fig. 9(b)). Cast aluminum fragments have an apparent density about 1.0 < X1 < 3.0, and their X8 value in this region tend to
be larger than that of wrought aluminum and magnesium fragments (see Fig. 9(c)). Moreover, the X2 values of these fragments

88% of wrought aluminum recovered at 91% purity.

97% of cast aluminum recovered at 92% purity.
85% of magnesium recovered at 88% purity.
This result surpasses the recently reported sorting performance
for the three types of lightweight metals using the dual energy
X-ray transmission (DE-XRT) sorting system developed by Delft
University of Technology (Mesina et al., 2007). In particular, our
sorting system is superior in the separation of wrought aluminum
and cast aluminum. If the sample fragments were limited to these
two types of aluminum, and the fragments identied as magnesium
were counted as wrought aluminum, the accuracy of the sorting
would be as calculated at Table 4. With respect to the separation
of wrought aluminum and cast aluminum, as far as we know, no
automatic sorting system equals our system in accuracy.
The other important features of the developed sorting system
are summarized as follows: First, sorting of dirty scrap is possible. The line of reected laser light on a moving fragment was
detected without any trouble, even if it was covered with oil, paint,
or mud. When the fragments surface is thinly contaminated, the
shape parameters are not changed to that large an extent and our
sorting algorithm using multivariate analysis can predict the material directly. In addition, we can deal with a heavily contaminated
fragment, although one was not found in this research, by calculating the discriminant function after recording its case data into the
database. Second, relatively high-speed processing is possible; we
conrmed that sorting accuracy did not deteriorate even at a belt
speed of 2.0 m/s. The developed system can process a fragment having a weight of 2600 g as per the specication of the weight meter.

Fig. 8. Typical example of the difference in degree of measurement error of fragment volume.

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

577

Fig. 9. Relation between the apparent density and several 3D shape parameters.

Table 3
Result of the sorting experiment of a mixture of three types of lightweight metals.

Number of tested fragments

Number of correctly identied fragments
Correctly identied rate

Wrought
aluminum

Cast aluminum

Magnesium

Total

576
518
0.90

312
287
0.92

738
627
0.85

1626
1432
0.88

The processing speed of this system is currently about 1 piece per

second. Thus, the capacity of this sorting system is estimated at
216 kg/h, if we consider the average weight of a fragment as being
60 g. It is expected that the processing speed will increase as the
calculation speed of the PC used increases (the current CPU is an
Intel Pentium 4 3.6 GHz). When we put it to practical use, the capacity can be further increased by running several systems in parallel.
Third, a low-cost system conguration is possible; the apparatus
shown in Fig. 2 can be produced for less than $40,000. As for the
running costs, the power consumption of the system, including 3D
image camera, weight meter, belt conveyor, air compressor, personal computer and display, totals less than 1 kW. Even if several
processing lines were installed in parallel to increase processing
efciency, the total cost would be much less expensive compared
with an X-ray sorting system, which is the best and most popular
nonferrous metal sorting equipment currently available.

On the other hand, the developed sorting system has a problem

that needs to be addressed at some point in the future: This system gives both correct and incorrect results for the same fragment
according to its conguration on the belt conveyor if the database
and algorithm for identication are not perfect. This means the
accuracy of the sorting system varies for each sorting test. The
result in Table 4 is the most typical example among many tests, and
the count basis accuracy for the three types of lightweight metals
changed actually in the range of 8595%. Moreover, we must draw
attention to the representativeness of the tested fragments. When
some fragments having totally different shapes from ones already
recorded in the database are added to the sorting samples, the accuracy would deteriorate unless the database and algorithm were
updated. Although we believe that a database necessary for adequate sorting performance can be realized by a limited number of
case data provided the sorting samples are restricted to lightweight

Table 4
Calculated sorting accuracy between wrought aluminum and cast aluminum.

Number of tested fragments

Number of correctly identied fragments
Correctly identied rate

Wrought aluminum

Cast aluminum

Total

312
287
0.92

576
572
0.99

888
859
0.97

578

S. Koyanaka, K. Kobayashi / Resources, Conservation and Recycling 54 (2010) 571578

metals, this has not been conrmed. Therefore, the practicality of

the developed sorting system to other ELV shredder factories has
not been claried at present. Because the shape of shredded nonferrous metal fragments is affected by the characteristics of the
shredding-machine and processed scrap, further study using fragments sampled from different ELV shredder facilities is expected to
evaluate the usefulness of this sorting technique.

Acknowledgements

5. Conclusions

References

A novel automatic sorting technique using a 3D imaging camera and a weight meter was demonstrated for the recycling of
lightweight metal scrap. This sorting technique was capable of
separation among several hundred pieces of wrought aluminum,
cast aluminum, and magnesium fragments obtained from an ELV
shredder facility, with approximately 90% accuracy according to
the difference in the apparent density and 3D shape parameters.
Sorting accuracy was not affected by contamination on the surfaces of the fragments. Although this sorting technique is currently
at the research stage, the cost of this sorting system seems to be
much lower than X-ray sorting system, and to be a highly probable
method that could replace conventional dense medium separation
or manual sorting. The adaptability for mass processing remains
a matter to be discussed further because this sorting technique is
based on the data accumulation of past measurements. This means
that the database of fragment properties and the algorithm for
the identication become complex when dealing with mass processing. The database and algorithm developed in this study have
not been optimized yet from a viewpoint of simplicity. An effective fragment-sampling scheme to downsize the database scale,
and a simple, easy calculation technique for data analysis will be
investigated later.

Ambrose F, Brown Jr RD, Montagna D, Makar HV. Hot-Crush technique for separation of cast- and wrought-aluminum alloy scrap. Conserv Recycl 1983;6:
639.
Gesing A, Torek P, Dalton R, Wolanski R. Assuring recyclability of automotive magnesium alloys: chemical composition-based sorting of magnesium shredded
scrap. In: Proceedings of symposium on aluminum held at the TMS 2003 annual
meeting; 2003. p. 1523.
Inaba T. Aluminum alloy sheet for automotive body panel. R&D Kobe Steel Engineering Reports (in Japanese, with English abstract) 2002;52:7982.
JAA
(Japan
Aluminum
Association).
Aluminum
statistics
homepage
(http://www.aluminum.or.jp/).
Mesina MB, De dong TPR, Dalmijn WL. Scrap stainless steel detection using a pulsed
electromagnetic eld. Int J Miner Process 2005;76:2131.
Mesina MB, De dong TPR, Dalmijn WL. Automatic sorting of scrap metals with a combined electromagnetic and dual energy X-ray transmission sensor. Int J Miner
Process 2007;82:22232.
NEDO (New Energy and Industrial Technology Development Organization of Japan).
Ledger of Forged Magnesium Parts Technological Development Project; 2008.
p. III-64 [in Japanese].
Noll R, Sturm V, Aydin , Eilers D, Gehlen C, Hhne M, et al. Laser-induced breakdown
spectroscopyfrom research to industry, new frontiers for process control.
Spectrochim Acta B 2008;63:115966.
Okubo M. Recycling of aluminum material in automotive applications. Alutopia
2005;35:1522 [in Japanese].
Sakurai T. Aluminum alloy sheet trends for automotive body panels. R&D Kobe Steel
Engineering Reports 2007;57:4550 [in Japanese with English Abstract].
Stepputat M, Noll R. On-line detection of heavy metals and brominated ame retardants in technical polymers with laser-induced breakdown spectrometry. Appl
Opt 2003;42:621020.

This study was conducted as a part of Forged Magnesium Parts

Technological Development Project of New Energy and Industrial
Technology Development Organization (NEDO) of Japan. All metal
scrap fragments used in this study were provided from the Tochigi
plant of YAMANAKA Co. Ltd., Japan.