Separation of Automobile Shredder Residue by Gravity Separation Using a Gas-Solid Fluidized Bed

The authors investigated the gravity separation of plastic, rubber, and wire harnesses in automobile shredder residue using a gas-solid f luidized bed. Uni-beads (barium silica titanate glass), zircon sand, and glass beads were employed as f luidized particles. Superficial air velocity was changed to see how the plastic, rubber, and wire harnesses f loated and sank in the f luidized bed. Wire harnesses were almost completely separated from the other constituents by using uni-beads and zircon sand. Plastic can be separated from rubber by using glass beads, although separation ef ficiency is relatively low. Precise adjustment of superficial air velocity is essential for attaining high separation efficiency because particle f low and air bubbles in the f luidized bed affect how objects f loat and sink.


Introduction
Waste reuse and recycling are now considered important, and this also holds for automobile shredder residue.Most of the approximately 5 million endof-life vehicles entering the waste stream each year are dismantled, with constituents including engines, tires, steel, and nonferrous metals being reused and recycled, bringing the recycling rate up to a high 75 to 80%.The remaining 20 to 25%, which is shredder residue, amounts to about 1 million tons annually and is nearly all disposed in controlled landfill sites.[1,2] In response to the "End-of-Life Vehicle Recycling Initiative" released by the former Ministry of International Trade and Industry (now the Ministry of Economy, Trade and Industry) in 1997, the Japan Automobile Manufacturers Association (JAMA) un-veiled a voluntary action plan with the numerical goals of raising the recycling rate to at least 95% and reducing landfilled waste to at least one-fifth the 1996 volume, both by 2015.
Automobile shredder residue comprises about 60% organic materials such as plastic, rubber, and urethane foam, and about 40% inorganic materials including glass, steel, wiring called harnesses, and nonferrous metals.JAMA and the automobile industry have developed a variety of technologies to separate these materials, such as sorting, volume reduction, and solidification; dry distillation; and pneumatic separation, and have shown it is quite possible to achieve the above numerical goals.[3][4][5][6] But these technologies have not found general use owing to high processing equipment costs.The authors explored a very simple pneumatic separation method in which air is blown into falling dust from the side to separate the plastic, rubber, urethane foam, wiring, and nonferrous metals in the dust.[7] But while the light urethane foam was separated with high efficiency because of its lightness, it was not easy to separate the other constituents.
In this research we used gravity separation in a gas-solid f luidized bed to test separation of three shredder dust constituents: plastic, rubber, and wire harnesses.Because gas-solid fluidized beds fluidize powders with an air blast from below, therefore resembling liquids, [8,9] objects lighter than their bulk density in a f luidized state float in the bed, while objects that are heavier sink.So far the authors have successfully used this method for coal cleaning [10] and for separating silica stone and pyrophyllite, which have a density difference of about 250 kg m Ҁ3 .[11] We have used uni-beads, zircon sand, and glass beads as f luidized powders to determine bulk densities and dispersion in a f luidized state based on how spheres with various densities float or sink.Densities of the three dust constituents in ascending order were plastic, rubber, and wiring.Only the wiring sank when using uni-beads and zircon sand, while we tried using glass beads to separate plastic and rubber by f loating the former and sinking the latter.Because the constituents have small density differences, their f loating and sinking are greatly affected by the particle f lows and air bubbles in the bed, and hence their separation is likely difficult.To obtain high separation efficiency we therefore conducted a separation experiment that varied superficial air velocity.

Automobile Shredder Dust
We extracted the three constituents of plastic, rubber, and wire harnesses from automobile shredder dust we had obtained from a shredder operator, and used them in our separation experiment.Table 1 gives the total weights and densities of the con-stituents, and their photographs appear in Fig. 1.All constituents weighed about 21 g, and the pieces were 10 to 20 mm in size.Because average density differences were small, at 400 kg m Ҁ3 between plastic and rubber, and 1,400 kg m Ҁ3 between rubber and wiring, and because of the large density dispersion, we anticipated difficulty with gravity separation in a gas-solid f luidized bed.

Fluidized Powder
Our f luidized powders were uni-beads of barium silica titanate glass (Union Co., Ltd.), zircon sand (Lasa Co., Ltd.), and glass beads (Toshiba Barotini Co., Ltd.).For each powder Table 2 shows particle size, true density, bulk density, shape, and the minimum fluidization velocity u mf obtained in the method described below.Uni-beads and zircon sand were used to test the sinking and separation of wiring from the three constituents.Uni-beads are comparatively monodispersed and spherical, and are suited to obtaining basic data showing the tendencies of the three constituents to f loat or sink in the f luidized bed.By contrast, zircon sand has a wide particle diameter range of 90 to 300 µm, but is cheaper than uni-beads, and because the small u mf presents advantages such as smaller blower power, it seems they are suited to practical use.We also tried separating plastic and rubber by using glass beads to float plastic and sink rubber.

Experimental Apparatus
Our experimental apparatus, shown in Fig. 2, was the same as that in a previous report.[11] A f luidized bed was created in an acrylic cylinder 48 cm high with a 20.5-cm inside diameter and wall thickness of 0.5 cm, and an antistatic agent was applied inside to reduce the inf luence of static electricity.At the bed bottom was an air distributor made with cloth between two porous stainless steel plates having holes with 0.3-cm diameters, a pitch of 0.6 cm, and an opening ratio of 22.67%.Powder was prepared so that it would be at a height of 15 cm in the fluidized bed.Air was supplied by a blower to fluidize the powder, and superficial air velocity was fine-adjusted by opening and closing the motor valve.Superficial air velocity u 0 and pressure drop ∆P were calculated by first using pressure transducers to read, as voltages, the orifice f low meter pressure and the pressure difference between the bottom of the fluidized bed and the atmosphere, then using previously obtained relational equations for voltage/superficial air velocity and voltage/pressure drop.∆P was measured while gradually decreasing u 0 , and the value of u 0 at which ∆P begins falling from a constant value was used for the minimum f luidization velocity u mf .

Measuring Bulk Density and Its Dispersion in a Fluidized State
The bulk density ρ fb of each powder in a fluidized state and its dispersion ∆ρ fb were determined by investigating the floating/sinking in a fluidized bed of 3.75-cm diameter spheres (table tennis balls containing a certain amount of steel shot) whose densities ρ sp were adjusted for each 10 kg m Ҁ3 .We f luidized powders at prescribed superficial air velocity u 0 /u mf , put a sphere into the center of the top stratum, stopped the air after 5 min, and measured the distance from the bed bottom to the sphere's center of gravity, h sp .When ρ sp ρ fb the sphere f loated, when ρ sp Ϸρ fb the sphere stayed in the middle of the bed without either rising or falling, and when ρ sp ρ fb the sphere sank.We determined h sp for each sphere's ρ sp .ρ sp1 was used for the largest density sphere that f loated, and ρ sp2 for the smallest density sphere that sank.These values were used in Eqs. 1 and 2 below to calculate ρ fb and ∆ρ fb , respectively.ρ fb is the average density of spheres that do not rise or fall, and ∆ρ fb is their density range.h sp was measured three times at prescribed superficial air velocities in each powder.Averages and standard deviations of ρ fb and ∆ρ fb were determined using the three measured values.(2)

Assessment of Residue Floating-Sinking and Separation
We f luidized powders at prescribed superficial air velocities u 0 /u mf .When using uni-beads we separately added plastic, rubber, and wiring; with zircon sand we added the three constituents together; and with glass beads we added plastic and rubber together.When using uni-beads we added the three constituents together at the u 0 /u mf at which plastic and rubber best f loated, and wire harnesses best sank.We stopped the air one minute after adding the residue constituents, measured the distance from the bed bottom to each constituent's center of gravity i, and determined the weight percentage w i of each constituent at each 3-cm stratum from the bed bottom.
To assess separation, we determined the purity x p and recovery rate x r , defined below.In uni-beads and zircon sand, plastic and rubber f loated while wiring sank.Hence the values of x p were the weight percentage of plastic and rubber in the dust that floated in the top stratum at i҃12 to 15, and the weight percentage of wiring in the dust that sank to the bottom stratum at i҃0 to 3. For x r values we used the weight percentage of the portion of total plastic and rubber that f loated in the top stratum at i҃12 to 15, and the weight percentage of the portion of total wiring that sank to the bottom stratum at i҃0 to 3. In glass beads plastic f loated and rubber sank.Thus for their x p values we used the weight percentage of dust plastic that f loated in the top stratum at i҃12 to 15, and the weight percentage of dust rubber that sank to the bottom stratum at i҃0 to 3. For x r values we used the weight percentage of the portion of total plastic that f loated in the top stratum at i҃12 to 15, and the weight percentage of the portion of total rubber that sank to the bottom stratum at i҃0 to 3. The experiment was performed three times under each set of conditions, and the average and standard deviation of each value were determined.

Superficial Air Velocity and Pressure Drop
For each powder we measured pressure drop ∆P while gradually decreasing superficial air velocity u 0 .Results appear in Fig. 3.When u 0 is big, ∆P is constant, showing that all powders are f luidized, but under minimum f luidizing velocity u mf , ∆P decreases linearly, showing that the powders are fixed.Uni-beads and zircon sand have about the same bulk density, and they have about the same ∆P values when in a f luidized state, but because zircon sand has many particles smaller than those in uni-beads, it has a smaller u mf value than uni-beads, as shown in Table 2.

Bulk Density and Its Dispersion in a
Fluidized State We varied the superficial air velocity u 0 /u mf in each powder and checked the positions in the f luidized bed h sp of spheres with various densities ρ sp , then determined the bulk density ρ fb and its dispersion ∆ρ fd in a f luidized state.Figs. 4 and 5 show the h sp at ρ sp for various u 0 /u mf , and the ρ fb and its ∆ρ fb for various u 0 /u mf when using uni-beads.At all u 0 /u mf there were spheres which f loated at low density, but which, when their densities were larger, stayed in the middle stratum of the fluidized bed without either rising or falling, and which completely sank at still greater densities.The average density ρ fb of spheres that stayed in the bed's midregion remained almost constant even if u 0 /u mf was changed, but its dispersion ∆ρ fb increased with u 0 /u mf .This is likely because powder f luidization becomes more intense as superficial air velocity increases, which destabilizes the f loating and sinking of spheres.Figs. 6 and 7 show the results for zircon sand.∆ρ fb increased with u 0 /u mf just as with uni-beads, but ρ fb tended to decline instead of remaining constant.Because the bulk volume of a fluidized powder generally increases as superficial air velocity increases, ρ fb decreases.Using zircon sand we performed sphere f loat/sink experiments over a broader u 0 /u mf range than with uni-beads, and it seems this tendency appeared more salient.It is also possible Density of sphere, ρ sp (kg m Ҁ3 ) Density of sphere, ρ sp (kg m Ҁ3 ) Superficial air velocity, u 0 /u mf that the wide range of zircon sand particle sizes and the non-spherical particle shape engendered porosity changes, and this is conceivably the reason for the difference observed in change tendencies between unibeads and zircon sand.Figs. 8 and 9 show the results for glass beads.Unlike uni-beads and zircon sand, ρ fb and ∆ρ fb remain nearly constant even if u 0 /u mf is changed.A possible explanation is that glass beads have a lower density than uni-beads and zircon sand, but the reason is unclear.
These results show that uni-beads and zircon sand have bulk densities of ρ fb Ϸ2300 to 2400 kg m Ҁ3 when f luidized, which are between the average densities of rubber and wire harnesses.Glass beads, by contrast, have a bulk density of ρ fb Ϸ1400 kg m Ҁ3 when f luidized, about the same as rubber's average density.

Residue Floating and Sinking
We performed f loat/sink experiments with plastic, rubber, and wire harnesses in several powder types while varying the superficial air velocity u 0 /u mf .Fig. 10 shows the f loat/sink results when putting the three constituents into uni-beads separately.The vertical axis shows heights i in the f luidized bed at 3-cm strata, and the horizontal axis shows the weight percentage w i of the three constituents in each stratum.When u 0 /u mf ҃1.05, plastic and rubber were completely af loat, and wiring sank with difficulty.This is probably because superficial velocity was low and the powder was not very f luid.As u 0 /u mf increased, a growing percentage of the wiring sank, and when u 0 /u mf ҃1.30 the wiring almost completely sank, but plastic and rubber did not completely float.This could be because when superficial velocity is high, powder f luidization is intense, and especially because f luidized bed surface vibration is great.Fig. 11 shows the purity x p and recovery rate x r obtained in the experiment for each constituent in relation to u 0 /u mf .The x p of plastic and rubber increased as u 0 /u mf increased, and it was nearly 100% when u 0 /u mf ͧ1.15.Meanwhile, the wiring x p was 100% when u 0 /u mf ͨ1.15, but it decreased when u 0 /u mf 1.15.Change in x r was the opposite from that of x p because the x r of plastic and rubber was 100% when u 0 /u mf ͨ1.15, but it decreased when u 0 /u mf 190 KONA No.21 (2003) 1.15, while the x r of wiring was nearly 100% when u 0 /u mf ͧ1.15, but was sharply reduced when u 0 /u mf 1.15.As noted above, these factors are clear from the connection to low powder fluidity and f luidization intensity.We performed a f loat/sink experiment by putting the three constituents together into the fluidized bed at u 0 /u mf ҃1.15, which exhibited the best plastic and rubber f loating and wire sinking.Results for plastic and rubber were x p ҃99.1Ȁ0.9 and x r ҃ 100.0Ȁ0.0, and for wiring x p ҃98.8Ȁ1.7 and x r ҃92.2Ȁ 4.1.These results were about the same as when putting the constituents in separately, and showed that it is possible to separate wiring from the three constituents with high efficiency.
Figs. 12 and 13 show, respectively, the f loat/sink results and the x p and x r according to u 0 /u mf when the three constituents were put in zircon sand together.The f loat/sink tendencies of the constituents were about the same as those for uni-beads.While the change trends of x p and x r too were about the same, the changes were more distinct in zircon sand.As observed in the previous section, conceivable reasons include the facts that zircon sand has a wider particle size distribution than uni-beads and is nonspherical, and that the u 0 /u mf is larger.When u 0 /u mf ҃1.6 the purity and recovery rate values of all constituents were about 100%, showing that zircon sand, which appears to be more suited to practical use than unibeads, can also separate out wiring with high efficiency.
Figs. 14 and 15 show, respectively, the f loat/sink results and the x p and x r according to u 0 /u mf when plastic and rubber were put in glass beads together.As u 0 /u mf increased, plastic f loated with greater diffi- culty and rubber sank more readily, and despite the use of different dust constituents, float/sink tendencies were about the same as in uni-beads and zircon sand.The change tendencies of x p and x r also coincide qualitatively.The small difference between the average densities of plastic and rubber meant that items of about the same density were mixed, and thus the two were not completely separated in different strata, but when u 0 /u mf ҃1.20 we obtained better-thanexpected separation with a high x p value of about 90 and x r of 70 to 90.When using uni-beads and zircon sand the ρ fb falls between the average densities of rubber and wiring, suggesting that separating them is easy, but even though the ρ fb of glass beads is about the same as rubber's average density, rubber almost completely sank especially when u 0 /u mf was large.Hence the above results likely do not represent f loating and sinking in line with density differences.As described in another paper, [11] the probable reason is that because objects put into the f luidized bed are affected by the particle f lows and air bubbles in the bed, whether they f loat or sink is not determined solely by density difference.Effects of the flow will be different depending on an object's size and shape, and therefore one expects that even among objects with the same density, some will sink and some will not.The ρ fb values obtained in this research are based on the f loating and sinking of spheres 3.75 cm in diameter.Because wire harnesses differ totally in size and shape from these spheres, and are accordingly affected differently by bed f lows, it seems to follow that there are f loat/sink differences between spheres and wiring with the same density.

Conclusion
We used gravity separation in a gas-solid fluidized bed to test the separation of wire harnesses from plas-tic and rubber in automobile shredder residue.We investigated the separation of wiring from the three constituents in uni-beads and zircon sand, and the separation of plastic and rubber in glass beads.Results are as follows.1) In uni-beads and zircon sand, plastic and rubber f loated in the bed's top stratum, and wiring sank to the bottom stratum.It was thus possible to separate wiring from the three constituents with high purity and a high recovery rate.The difference between the average densities of plastic and rubber is not great, and we anticipated difficulty in separating these mixed items with similar densities, but by using glass beads we obtained better-thanexpected separation with purity of about 90% and a recovery rate of 70 to 90%. 2) Floating and sinking of objects in the gas-solid fluidized bed depends not only on density differences, but also on the state of powder fluidization, the effect of bed f lows on objects, and other factors, but we found that controlling superficial air velocity makes it possible to make objects float or sink according to density differences, and to attain high-efficiency separation.

Fig. 8 )Fig. 9
Fig.8 Dependence of sphere position on density at various superficial velocities.Powder is glass beads.

Fig. 10 Fig. 11
Fig.10 Weight percentages of plastic, rubber, and wire harnesses at each height in the f luidized bed at various air velocities

Fig. 12
Fig. 12Weight percentages of plastic, rubber, and wire harnesses at each height in the f luidized bed at various air velocities

Fig. 14
Fig. 14 Weight percentages of plastic and rubber at each height in the f luidized bed at various air velocities

Table 2
1) ρ sp2 ѿρ sp1 2 Powder characteristics Note: Averaged values and standard deviations of u mf were obtained from three measurements.
e y Fig. 2 Schematic drawing of the experimental apparatus ∆ρ fb ҃ ρ sp2 Ҁρ sp1 Dependence of purities and recovery rates of plastic and rubber on superficial air velocity.Powder is glass beads.