Powder Technology 301 (2016) 369378

Contents lists available at ScienceDirect

Powder Technology

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

Wall effects on particle separation in air jigs

Weslei Monteiro Ambrs a,, Bogdan Grigore Cazacliu b, Carlos Hoffmann Sampaio a
a
Mineral Processing Laboratory, Federal University of Rio Grande do Sul, 9500 Bento Gonalves Avenue, Zip Code: 91501970, Porto Alegre, Brazil
b
LUNAM, IFSTTAR, Aggregates and Materials Processing Laboratory, Route de Bouaye CS4, 44344 Bouguenais Cedex, Nantes, France

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates the inuence of wall effects on particle separation in an air jigging device. Tests were con-
Received 17 March 2016 ducted with a ternary mixture of concrete, brick and gypsum particles in the size range from 12 to 20 mm for
Received in revised form 13 June 2016 three different jigging times. It is shown that an unusual, remarkable segregation occurred in the direction per-
Accepted 17 June 2016
pendicular to the lateral walls, giving rise to a horizontal segregation pattern. In all cases, heavier particles con-
Available online 19 June 2016
centrated more in the center of the particle bed while lighter particles concentrated along the side walls.
Keywords:
Composition measurements of the stratied bed revealed that wall effects can signicantly inuence the compo-
Wall effect sition of jig products. Linear predictions of jigging evolution suggest that stratication progress linearly after the
Particle separation rst initial stages. Also, time evolution and analysis of local compacities indicate that stratication starts from the
Jigging bottom to the upper layers of particle bed. Based on the experimental results, a mechanism for the pattern for-
Density segregation mation was proposed, which was enhanced by a geometry model of bed rearrangement. Prediction results for
Stratication pattern bed composition correlated well with experimental data and endorsed the role of lateral walls on the segregation
pattern formed. The results obtained also show practical signicance for the use of air jigging in the sorting of
recycled aggregates.
2016 Elsevier B.V. All rights reserved.

1. Introduction
Dry processing technologies such as air jigging have received much
attention over the last years due to its inherent advantage of not using
process water, which offers important economic and environmental
benets. Moreover, modern air jigs have presented a signicant im-
provement in performance compared with that of air jigs during the
20th century [13]. As a result of such development, the use of air jig-
ging is no longer limited to coal and ore beneciation but has also
been investigated as a potential technique for recycling of solid wastes
[46].
Gravity concentration processes are usually performed in the pres-
ence of water. The main reason for using water as means of separation
rather than air is the separation efciency. Taggart [7] proposed an
index that combines the effects of the minerals densities and the density
of the separation medium. This index is known as the concentration
criterion. The concentration criterion related to the concentration of
two minerals in a gravimetric process is the ratio of the density of the
denser mineral minus the density of uid divided by the density of the
less dense mineral minus the density of the uid. This relation results
Corresponding author at: LUNAM, IFSTTAR, Aggregates and Materials Processing
Laboratory, Route de Bouaye CS4, 44344 Bouguenais Cedex, Nantes, France.
E-mail addresses: weslei.ambros@ufrgs.br (W.M. Ambrs), bogdan.cazacliu@ifsttar.fr
(B.G. Cazacliu), sampaio@ufrgs.br (C.H. Sampaio).
in a number which is bigger for processes that use water than for pro-
cesses that use air.
Jigs are versatile gravity concentrators that work efciently with
coarse particles. With the exception of jigs for ne minerals (particular
discharge type of the products), all jigs operate efciently at particle
sizes above 1 to 2 mm [8,9]. The top size of the material to be concen-
trated is only related to operational aspects of the equipment and the
mineral liberation. In coal beneciation, there is a special jig (ROM jig)
that can treat particles up to 350 mm [10].
During the 1950s and 1960s, water jigs were exhaustively studied
and their efciencies were greatly improved [8,9], increasing the differ-
ence in efciency in relation to air jigs. Thus, the use of dry gravity con-
centration processes was restricted to regions with lack of water and to
minerals that can not be wetted. From the new millennium, due to en-
vironmental problems, the cost of water usage increased greatly. There-
by, several studies were directed to the construction of modern air jigs
with reasonable efciencies.
As mentioned, the separation efciency of air jigs is lower than that
of water jigs [8,9]. However, in spite of recent efforts to investigate jig-
ging phenomena by means of numerical simulation techniques [11,
12], water jigs remains the main focus. Little attention has been paid
for peculiar phenomena that take place in air jigs. For instance, Cazacliu
et al. [6] reported that purity of the air jig products could be affected by
disturbing wall effects in such a way that an unequal segregation pat-
tern can occur close to the internal jig walls. This indicates that the ex-
pected vertical stratication can be also accompanied by an unusual

http://dx.doi.org/10.1016/j.powtec.2016.06.034
0032-5910/ 2016 Elsevier B.V. All rights reserved.
370 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378
horizontal segregation in the zones close to the internal jig walls. Simi-
larly, Windows-Yule and Parker [13] has demonstrated that the extent
of density segregation in granular systems under vibration can be al-
tered merely through the modication of the aspect ratio of the system.
In the framework of air jigging, this implies that the geometry of the jig
chamber can be treated as an additional factor in the separation
efciency.
Also, worthwhile studies can be performed in the scope of the time
evolution of the stratication process in a jigging bed. Although previ-
ous work has demonstrated that the stratication by gravity normally
occurs quickly [14], there remains a need to more fully address the
local changes in composition over jigging time. For instance, Crespo
[12] observed different rates of concentration of the heavy and the
light components at the bottom and top of the bed, respectively. Simi-
larly, Sampaio et al. [15] reported that the bulk of stratication of terna-
ry mixtures of aggregates in an air jig device occurred before that a
threshold of time (about 30 s) was reached, after which little gain in
separation could be noted.
Furthermore, there has been extensive investigation related to seg-
regation in dry granular systems and a variety of properties have been
pointed as capable of undergoing segregation in addition to differences
in particle density, size and shape. These includes frictional forces
[1618], presence of interstitial air [19,20] and differences in the pack-
ing fraction or compacity [21,22]. The synergy between these studies
and particle stratication in air jigs is obvious in such way that further
investigation can give rise to signicant performance improvements.
The purpose of this work is to investigate a distinct stratication pat-
tern that occurs near to the internal walls of the jig device when a gran-
ular mixture is subjected to air jigging. The composition of products
obtained after several sorting tests was observed carefully. On the
basis of the experimental evidences and literature data, a pattern forma-
tion mechanism together with a pattern formation model are proposed
in order to better understand the role of wall effects during the strati-
cation in air jigs. The model denes a pattern geometry of the bed strat-
ication and a level of mixing between layers which depend on the
jigging time.
2. Experimental setup
2.1. Materials
The material used in the tests was a mixture of particles in the size
range of 12 to 20 mm. The components were obtained by crushing
and sieving the following materials: concrete, clay bricks and gypsum.
These materials were chosen in order to create conditions for a strati-
cation based on differences in particle density and packing density (or
compacity), as previously reported by Cazacliu et al. [6]. The authors
showed that, despite the close particle densities of brick and concrete
particles (Table 1), their separation is effective due to the signicant dif-
ferences in the individual compacities. Similarly, it was shown that the
segregation of gypsum from the other two species (brick and concrete
particles) is driven mainly by its relatively low particle density. The
enveloped density was adopted as the particle density and was mea-
sured for each material by means of water displacement after water-
proong. The compacity was determined after three successive
weighing of a known volume of each individual material. In this proce-
dure, one section of the jig was fully lled with particles and, since the
enveloped density was known, the particles volume was calculated
Table 1
Skeletal, enveloped and bulk density of materials.
Material Particle (enveloped) density (g/cm3)
Concrete 2.39
Brick 2.26
Gypsum 1.86
and the compacity was obtained in relation to the section volume. A
narrow size range was chosen to minimize the effects of size
segregation.
2.2. Equipment
The tests were performed in a batch pilot-scale air jig model AllAir
S-500 from AllMineral with a capacity of approximately 50 kg per batch
and a working size range of 1 to 25 mm. This laboratory jig is able to re-
produce the same stratication process experienced in an industrial jig
and can therefore be used to investigate the jigging process on a pilot
scale. The complete jig system consists of feeding unit, separation cham-
ber, control panel and powder ltering units as displayed in Fig. 1. The
jigging cycle can be widely varied by the combined effect of bed expan-
sion and pulse frequency. The former one is controlled by the air ow
provided by a 15 kW blower (Combimac, 49,631/B1Y1), which is ad-
justed in the control panel in function of the percentage of the blower
power (0 to 100%). The blower is able to produce an air ow up to
73 m3/min. The air intake is controlled by a pneumatic utter valve lo-
cated in the air duct between the blower and the separation chamber.
The pulse frequency is controlled in function of the valve rotation in fre-
quencies ranging from 0 to 300 rpm.
During the tests, the mixed particles are placed into the container
and subjected to vibration by two streams of air ow which enter simul-
taneously by the bottom plate of the jigging chamber. The rst air ow is
continuous and is responsible by the pre-expansion of the bed while the
second one is controlled by the utter valve and promotes the bed pul-
sation. The jigging chamber is assembled with different rectangular sec-
tions of Plexiglas (500 500 50 mm) tted one over the other on a
perforated plate ( = 1 mm). The set of rectangular sections made pos-
sible the extraction of the particles bed layer by layer.
2.3. Experimental procedure
A total of three jigging tests were conducted with three different jig-
ging times: 15 s, 45 s and 120 s. Pulse frequency, uidizing air rate and
bed height were xed in 160 rpm, 70% and 150 mm, respectively. This
was the optimal operating point obtained by Sampaio et al. [15] for
the separation of mixtures of concrete, brick and gypsum particles in
an air jigging device. The operating parameters used in the tests gave
a pulse duration of 0.37 s and a bed expansion of about 25 mm. For
each experiment, a ternary mixture with the same bulk volume of ag-
gregates was prepared, which corresponded to a mass fraction of 48%
of concrete, 31% of brick and 21% of gypsum. The total amount of mate-
rials used in each test was of 43,000 g (1%). In order to quantify the
particles stratication, the bed was divided into six sampling zones.
Three vertical layers (top, middle and bottom) were divided
Compacity
0.70 0.00
0.49 0.01
Fig. 1. Air jig model AllAir S-500 from AllMineral. (1) Feeding unit; (2) control panel;
0.40 0.00
(3) separation chamber; (4) powder ltering unit.
 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378
horizontally into two zones by a rectangular box (400 400 50 mm)
and separately collected after each test. The inner side of the box was
dened as the center region and the outer side of the box, alongside
the edges, was dened as the wall region (Fig. 2). After collecting
the samples, the particles were separated manually (based on color)
and weighed to get the percentage of each material in each zone.
Additionally, a sorting test with a binary mixture of concrete
(40,700 g) and gypsum (8300 g) was carried out in order to verify
how the concrete particles were stacked on the bottom layer after strat-
ication. The conditions used were the same as those of ternary mix-
tures (t = 120 s) and after sorting the gypsum particles at the top
surface were carefully removed by hand.
3. Results and discussion
3.1. Stratication proles
Fig. 3 shows the results obtained after stratication in terms of vol-
ume fraction. In general, the particle bed was arranged in a new cong-
uration, consisting of top and bottom layers lled in a large proportion
with gypsum and concrete, respectively, and a middle layer containing
a mixture composed mainly of brick and concrete. As previously
discussed by Cazacliu et al. [6], although the concentration criterion pro-
vides the greater ease of gypsum separation from the other compo-
nents, it also suggests that separation between brick and concrete is
very difcult to carry out due to the similarity in particles densities. As
this was not conrmed by the separation results and taking into account
the absence of size segregation, it seems probable that differences in in-
dividual compacities engendered by differences in particle shape played
an important role on the bed rearrangement [2123]. As displayed in
Table 1, the compactibility of concrete particles is signicantly higher
than that of brick particles, which suggests that a more compacted
layer is formed when concrete is preferably concentrated at the bottom.
As a general rule, denser particles in all layers were more concentrat-
ed in the center of the jigging chamber while the lighter ones tended to
be more concentrated alongside the walls. As can be seen in Fig. 3, gyp-
sum particles were grouped preferentially close to the walls in all layers,
notably at the top. For all three jigging times, its volume fraction along
the walls in this region remained at around 24% greater than in the cen-
ter, which may indicate that segregation along the horizontal plane rap-
idly achieved a steady distribution in the upper layer. Similarly, the
difference between the concentration of concrete particles in the center
and near the walls in the bottom layer remained approximately con-
stant in 16% over sorting time, but in the opposite direction, i.e. it was
oriented towards the center. Brick particles showed a tendency to con-
centrate near the walls at the middle and bottom layers. Local differ-
ences at middle layer increase over jigging time in 15%, 17% and 26%
for jigging time of 15 s, 45 s and 120 s, respectively. Taking into account
that the brick was the intermediate density material in the bed and
therefore the more difcult to separate from others, it may be supposed
that if the jigging time is large enough, its horizontal distribution over
middle layer also will eventually remain approximately constant.
Fig. 2. Experimental setup and sampling locations of the stratied bed. (A) Vertical layers (top, middle and bottom). (B) Each layer was divided horizontally by a rectangular box in center
region (inner side the box) and wall region (outer side the box).
371
In order to analyze the stratication behavior over jigging time, pre-
dictions of the stratication at 45 s and 120 s were performed by means
of linear extrapolation. In order to estimate the volume fractions at 45 s
it were used experimental data of t = 0 s (mixed state) and 15 s, while
for the estimation for t = 120 s it were used data of t = 15 and 45 s.
Fig. 4 shows predicted data in comparison with the experimental data.
The concentration of particles of the same composition is predicted by
the measured data at 0 and 15 s to be faster than that observed experi-
mentally for t = 45 s. On the other hand, the predictions for t = 120 s
obtained from results at 15 and 45 s are in very good agreement with
the experimental observation. This seems to indicate that particles con-
centration took place rapidly at the beginning of jigging with a decrease
of concentration rate over jigging time after 15 s and became stable up
to 120 s. Thereby, from 15 to 120 s the evolution of bed stratication can
be treated as a linear process in which the effect of jigging time could be
estimated by linear interpolation.
In addition, the standard deviation value related to linear extrapola-
tion, i.e. SD 0.15 was used to estimate the standard deviation of the
measurement of the volume fraction of a granular component in each
sampling location (described in Fig. 2). For this, the measure of the vol-
ume fraction was considered as a random variable with a standard de-
viation of 0.033.
In Fig. 4 (a) it can be also observed that some concentration values
outcomes from the linear extrapolation were lower than 0 and even
N1, as indicated by the values inside the circles. A detailed analysis re-
veals that the bed zones in which these critical values occurs correlate
well with the bed rearrangement as illustrated in Fig. 3. For instance,
the extrapolation predicts a large concentration of concrete at BC (pre-
dicted volume fraction of 1.28) and a negative concentration of gypsum
at the same zone (predicted volume fraction of 0.32). Similar tenden-
cy occurred in the extrapolation from data of t = 15 and 45 s, but to a far
lesser extent (predicted volume fraction of concrete at TW of 0.03).
These results enhance the evidence that vertical segregation occurs
faster during the rst cycles of jigging. Furthermore, it indicates that
the horizontal segregation pattern (i.e. the difference between concen-
tration in the center and in the wall zones) is established at the begin-
ning of jigging action and, despite changes in the segregation extent,
maintains the same conguration over jigging time.
In addition to the volume fraction, the compacity in each zone (cen-
ter and wall of the three layers) was used to examine the bed rearrange-
ment after stratication. At this point we dene the local compacity to
be the local lling fraction in a given region:
mC =dC mB =dB mG =dG
CC 1
VT
Where mC, mB and mG are the mass of concrete, brick and gypsum
collected in one of the six sampling locations, respectively; dC, dB and
dG are the enveloped densities of each material as given in Table 1,
and VT is the total volume of the zone, which either corresponds to the
center or the wall regions of a layer. At the initial state, the mixed parti-
cle bed had a global compacity CC = 0.51 throughout it. Please note
372 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378

Fig. 3. Volume fraction in each sampling location for the three jigging tests. Sampling locations were abbreviated as TC (top center), MC (middle center), BC (bottom center), TW (top wall),
MW (middle wall) and BW (bottom wall). The percentage displayed above each column represents the mass fraction of the location in relation to the total mass into the jig.

that, for the purpose of comparison, the system height changes after jig-
ging were considered negligible.
As expected, after successive jigging cycles a very compacted layer
was formed on the bottom and the compacity distribution differed sig-
nicantly over the partially segregated bed, as illustrated in Fig. 5. It was
apparent beforehand that a reduction in local compacity could occur for
particles in contact with the sidewalls due to packing effects [24,25]. In
agreement with this behavior, the local compacity near the walls
remained lower than in the center on every occasion. At the bottom,
the difference in compacity between center and wall zones remained
constant in approximately 0.13 over the three jigging times. On the
other hand, differences in local compacity in the horizontal plane varied
signicantly in the other two layers, increasing from 0.100 to 0.173 in
the middle and decreasing from 0.168 to 0.075 in the top, which clearly
dene an opposite tendency.
In the center, it was observed that the rate at which compacity
changed at the bottom layer was faster than the compacity variation
in other layers and stabilized at CC 0.68 for t = 45 s. This suggests
that the heavy component was concentrated at the bottom layer faster
than the lighter components concentrated at the upper layers. In addi-
tion, this also indicates that stratication continues to evolve even
after the occurrence of high compaction at the bottom layer. This nding

Fig. 4. Correlation between experimental results and estimative of the volume fractions by means of linear extrapolation. (A) Prediction of stratication at 45 s from data of t = 0 and 15 s.
(B) Prediction of stratication at 120 s from data of t = 15 and 45 s. The points represent the mass fraction of each component (concrete, brick or gypsum) in each one of the six sampling
locations. Inside the circles are indicated the number of predicted values that exceed the minimum and maximum limits of concentration.
 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378
Fig. 5. Local compacity calculated for each sampling location along the three jigging times.
agrees qualitatively with results reported by Crespo [12]. Interestingly,
the compacity variation over time both at the top and at the middle
layers, although in opposite ways, showed a very similar rate of change.
In particular, for t = 15, 45 and 120 s the increase in compacity at the
middle was 0.02, 0.06 and 0.12, whereas the decrease in compacity at
the top was 0.02, 0.09 and 0.12, respectively. This conrms that the var-
iation of the total compacity over jigging time could be neglected at this
stage.
These previous results apparently indicate that a compacted layer
was rapidly formed at the bottom and the interchange of particles was
particularly high between the top and middle layers. In short, the evolu-
tion of local compacity in the center zone suggests that the mixed bed
starts to stratify from the bottom to the top, which is in agreement
with the behavior observed by Crespo [12].
Compared to the signicant variation over time in the central region,
the local compacity near the walls remained almost entirely unchanged
(with differences inferior to 0.03 in all cases). This low compacity vari-
ation is not necessarily contradicted by the results displayed in Fig. 3,
which shows, for instance, an increasing concentration of gypsum
near the walls at the top layer. This may suggest that segregation close
to the walls evolves during jigging time, but at constant compacity,
which is not the case in the center zone. This can also indicate that pack-
ing effects near the walls can be signicant enough to obliterate varia-
tions in compacity related to variation in local composition.
3.2. Comparison between vertical and horizontal segregation
The overall results were evaluated by comparing the segregation ex-
tent in both vertical and horizontal planes. Fig. 6 summarizes the results
Fig. 6. Comparison between vertical and horizontal segregation for jigging tests. The
values displayed above of each column represent the ratio between the horizontal and
vertical segregation extent.
373
obtained and displays the ratios between horizontal and vertical segre-
gation. In the present case the segregation in the vertical direction was
assessed in function of the absolute difference between the volume con-
centration of the major material in a given layer (gypsum at the top,
brick at the middle and concrete at the bottom) and the average volume
concentration of the same material in the others vertical layers of the
bed. Similarly, the segregation in the horizontal direction was evaluated
by the absolute difference between the volume concentration of each
material in the center and near the wall zones.
It is rst notable from Fig. 6 that horizontal segregation was decided-
ly secondary compared to vertical segregation. However, it is noticed
that it plays an important role on the bed rearrangement. Segregation
kinetics was signicantly different for the two directions. In the hori-
zontal plane, average differences in composition between center and
side wall regions over time remained in 0.090 (0.014) for gypsum,
0.091 ( 0.010) for brick and 0.180 ( 0.014) for concrete. On the
other hand, using the method previously described, average differences
in composition for the vertical axis over time increased signicantly
mostly in a linear form over time, ranging from 0.490 (0.190) for gyp-
sum, 0.122 (0.064) for brick and 0.312 (0.092) for concrete. It is no-
table from this analysis that, in contrast to vertical segregation, a
horizontal stratication prole rises rapidly at the beginning of strati-
cation process and changes very little over jigging time, as can be noted
by observation of the standard deviations values. In addition, average
differences in composition suggest that heaviest particles were more
strongly affected by wall effects than the others. This evidence also sug-
gests that the bed thickness should be a parameter of signicant inu-
ence on the extent of horizontal stratication although the analysis of
its relevance is beyond the scope of the current work. What is more,
the analysis of the average changes in the composition also provides
support for the hypothesis discussed in Section 3.1 related to compacity
changes, which point out that the clustering of concrete particles at the
bottom is faster than the concentration of gypsum particles at the top.
It can be also observed from Fig. 6 that in all cases the segregation ra-
tios values obtained were lower for gypsum than for brick and concrete.
This evidence can be related to the relatively low particle density of gyp-
sum, which favors its upward motion and so on its accumulation at the
top layer. In this case, it may be presumed that when signicant particle
density differences are present, the density-driven mechanism en-
hances vertical segregation in such a way that can far exceed horizontal
segregation.
On the other hand, high values of segregation ratio were obtained for
brick and concrete, including a segregation ratio exceptionally N1 for
brick at the jigging time of 15 s, which indicates a predominance of hor-
izontal over vertical segregation. As mentioned before, it should be
noted that the close particle density of brick and concrete (2.26 and
2.39, respectively) makes them difcult to separate only due differences
in density, although other mechanisms may engender their segregation
as the difference in the maximal packing fraction [6,21], which in turn
374 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378

has its base differences in particle shape [26]. However, the mere differ-
ences in compacity of brick and concrete are not able to explain the for-
mation of a horizontal stratication prole, as well as the reported
evolution of stratication at constant compacity at the bottom layer
(Fig. 5). One way to explain these unusual results is to assume that
the horizontal stratication does not take place directly due to differ-
ences in local compacities, but in function of the period in which the
bed is in motion. Based on this assumption, a brief discussion of the pos-
sible mechanisms behind the horizontal segregation phenomenon is
presented in the next section.
3.3. Wall-induced transverse settling
As previously reported, it is evident from results that the wall dis-
tance of particles plays an important role on the nal stratication pat-
tern. Visual observation of the external lateral walls and the bed surface
revealed that a distinct segregation pattern took place along the jig
walls (Figs. 7 and 8). In an attempt to explain the observed phenomena,
a stratication mechanism is proposed based on experimental and liter-
ature data.
During jigging, when the particle bed is expanded, a loose packing is
produced and heavy particles move downwards through the bed,
tending to minimize the potential energy of the system. In practice,
this means that heavy particles displace the light ones upwards during
the stratication process, as outlined by Mayer [27]. These displace-
ments can be explained by a mechanism based on void penetration, as
proposed in previous studies [17,22,28,29]. In the framework of the
present study, the spatial distribution of voids taking into account the
connement imposed by the lateral walls could be a useful mean to ex-
plain the empirical results. Thus, it can be assumed that stratication is
controlled by the local porosity of the bed, as also reported by Crespo
[12].
Given the constraints imposed by the side walls on the particle bed,
it is reasonable to suppose that the vertical velocity of particles
decreases rapidly as the wall is approached, which give rise to a dual ef-
fect on the jigging cycle. On the one hand, the particle bed would be di-
lated with a slightly concave prole, i.e. the expansion near the center
would be greater than near the walls. On the other hand, a gap would
be formed beneath the center of particle bed when it loses contact
with the jig base. Therefore, it would exists a difference among local po-
rosities (void fraction) for the fraction of time during which the particle
bed is dilated. As a result, heavy particles of concrete nearby walls
would be more likely to transversely settle into the gap in the center. Si-
multaneously, light grains of gypsum and brick which were previously
carried up by the air ow would tend to ll the voids left by concrete
particles near the walls. The cumulative effect of several jigging cycles
would lead to a clustering of concrete particles on the center of the bot-
tom layer and in the lower part of the middle layer while the lighter par-
ticles would be concentrated near the side walls. This mechanism is in
line with data displayed in Figs. 3 and 5 which suggests a fast clustering
of concrete at the bottom and a gradual accumulation of lighter particles
close to the walls. Furthermore, experimental data also suggests that
similar phenomena could have occurred for brick and gypsum at the
top layer. As a result, the stratied bed outcomes from the jigging pro-
cess would not be rearranged in at strata but in a softly concave prole,
where a mound of heavy particles would be surrounded by the light
ones (Fig. 9).
The mechanism suggested is in line with that proposed by Xue et al.
[29], who estimated that a particle in a bed under vibration is ve times
more prone to migrate into a void further away from the lateral wall
than into one closer to the wall. In fact, within the proposed mechanism
the transverse settling of heavy particles towards the center of the bed,
enhanced by differences in particle density typically found in jigging
beds, is the key step for segregation pattern formation. The assumption
that the concentration of light particles along the walls occurs as an ef-
fect of this settling behavior also provides an explanation for the greater
overall extent of horizontal segregation of heavy particles compared to
the lighter ones. Furthermore, although the validation of the proposed

Fig. 7. Snapshots of the particle bed before and after jigging for t = 120 s. (A) Mixed particles before the sorting test. (B) Stratied bed, where an arch-shaped pattern under the gypsum
strata could be clearly noted. (C) Top view of the top layer totally covered by gypsum after stratication. (D) Top view of the middle layer, in which gypsum particles were noticeably more
concentrated close to the walls. (E) Top view of the bottom layer, where the visual difference between center and wall regions was discrete.
 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378
Fig. 8. Shape of the cluster formed by concrete particles after jigging of a binary mixture, gypsum-concrete particles. (A) The occurrence of a mound resembling a frustum of a square
pyramid could be clearly observed along the surface area separating gypsum from concrete particles after jigging. (B) A considerable amount of gypsum particles were concentrated
alongside the walls.
mechanism is beyond the scope of this study, it seems also able to ex-
plain the fast clustering of concrete particles at the central part of the
bottom as well as the low compacity changes in the zone near the walls.
3.4. Stratication model
Based on the foregoing analysis, a simplied stratication pattern for
prediction of the density segregation after jigging was developed, based
on the hypothetical stratication pattern displayed in Fig. 9(c). It was
assumed that after stratication the particles are distributed on the jig
container as illustrated in Fig. 10.
In the present case, it was determined the mass fraction composition
in each predetermined zone (TC, TW, MC, MW, BC and BW) for the con-
guration of the test setup considering the above-mentioned stratica-
tion pattern model. For this purpose, it was adopted the following
assumptions:
Wall effects. They were represented by the segregation pattern
displayed in Fig. 10. The shape of each pyramidal strata was dened
by considering that the ratio B-bi is proportional to the ratio H-hi,
where bi and hi are the top base length and the height of the strata i,
respectively. The distance w from where starts the inuence of the
walls on stratication was given by the intersection between the con-
tainer base and the line that crosses the vertex of each stratum, which
denes an acute angle 90. Thereby, when 90 then w 0 and
wall effects are negligible;
Packing effects. The packing perturbation induced near the walls was
assumed as postulated by Ben-Am [24]. In this approach, it is admit-
ted the existence of a perturbed volume of the packed bed Vw, in
which the distance between each point and the wall is less than d/2,
the half of particle diameter. In this volume the compacity is Kw.CC,
with Kw b 1, while it remains equal to CC in the rest of the container.
By means of several experiments, De Larrard [30] estimated that for
crushed grains Kw = 0.73, which was the value adopted in the present
Fig. 9. Illustration of the ideal stratication pattern of a homogeneous ternary mixture (A): in the absence of wall effects (B); and when wall effects are signicantly enough to give rise to a
stratication pattern as proposed by the mechanism reported (C).
375
model;
Mixing. It was assumed that a given amount of brick particles
remained mixed with concrete and gypsum after jigging in a propor-
tion specied by the mixing index Imix. So a proportion Imix of concrete
or gypsum is replaced by brick particles in concrete or gypsum zones
while a proportion Imix of brick is replaced by concrete or gypsum par-
ticles so that the total mass of the system is maintained constant. As
previously reported in section 3.1, changes in the bed height were
considered negligible, which means that it was assumed the conser-
vation of the total volume of the packed bed. The value of Imix is an
input parameter.
Given the input parameters, the rst step is to calculate the shape of
each stratum dened by the height hi and the top base length bi.
Given the following relation:
3H h
tan  i  3
w Bbi
w
2
After some simple algebra, the following equation can be found:
3Hbi B 2w
hi 4
2w
The solution of Eq. (4) together with the equation of volume for the
pyramidal shape provide the values of bi and hi for each stratum. In the
present case, the respective values for gypsum strata were obtained by
difference, once Vgypsum and H were known.
After determining the dimension of strata, the total volume VZ of
each zone in each region can be calculated by simple geometric relation-
ship. Then, the mass of each component in each zone can be determined
directly from the individual compacity and density taking into account
376 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378

Fig. 10. Bed rearrangement scheme as assumed by the proposed model. (A) After jigging,
each one of the n strata takes the form resembling a frustum of a square pyramid where
Fig. 12. Evolution of the mixing index Imix and R2 over jigging time. The mixing index
heavier components are covered by the lighter ones. The top base length of each
decreased linearly along time while R2 is getting close to one as the number of jigging
stratum increases progressively over bed height H until becomes equal to the container
cycles (jigging time) increased.
length B at the top. (B) Detail of the geometric conguration at the wall region, dened
by the distance w from the container edges.

extent can be indirectly estimated in terms of Imix, since mixing rate is

packing effects on the walls:
mi K w :CC i :V Z :i
When the distance from the walls is less than d/2 = 8 mm, Kw = 0.73
and is otherwise one. The mass fraction in each zone is then given by:
mi mj
xi 1I mix Imix
mi mi
Where i, j = concrete, brick or gypsum.
For each jigging time, the best t values of parameters Imix and
were founded by the minimization of the difference between experi-
mental and numerical values (sum of squared residuals).
Fig. 11 shows model prediction results for the three jigging times
considering the best parameters values. As can be seen, the model was
sufciently accurate to predict the mass composition for t = 45 s and
t = 120 s (R2 = 0.908 and 0.924, respectively). However, as anticipated,
it did not align precisely with the shorter jigging time (R2 = 0.803 for
t = 15 s). Despite this limitation, the decreasing of the mixing index
Imix over jigging time provides important insight into the stratication
degree during the jigging process. Although the role of jigging time
have not been considered in the model, its effect on the segregation
expected to decrease over jigging time.
The decreasing of Imix was virtually linear in relation to the jigging
5 time, as illustrated in Fig. 12. As previously discussed in section 3.1,
the stratication rate can be considered a linear process after a certain
jigging time so that the use of a mixing index as Imix seems a plausible
way to represent the stratication extent. Furthermore, extrapolating
the linear regression equation for t = 0 (mixed state) results in Imix =
0.352 instead of Imix = 0.45, which corresponds to the original mixed
6
state. This seems to indicate that a faster segregation (non-linear over
jigging time) occurs at the beginning of jigging process followed by a
sudden change of the stratication evolution (linear over jigging time)
at some stage when t b 15 s.
Unlike the mixing index, the value of for the three jigging times
remained very close to each other, indicating that the stratication pat-
tern as previously dened varied slightly over the jigging process. In ad-
dition, the evolution of R2 over jigging time seems to indicate that the
segregation pattern takes place fastly (before 15 s) but it is not fully
established in short jigging times (15 and 45 s). In the main, it indicates
that the differences in mass concentration between the center and side
wall zones varied very little over jigging time, which is supported by ex-
perimental data with the exception of brick particles at the middle layer.
In order to verify the relevance of each individual assumption on the
model, simulations were performed after one of the assumptions had

Fig. 11. Correlation between simulations and experimental results.

 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378
been intentionally omitted (Fig. 13.b, c and d). When packing effects on
the walls were omitted, the model prediction remained virtually un-
changed. If mixing effects are not considered (Imix = 0), the model
loses in prediction capacity. Finally, in the case of the absence of wall ef-
fects ( = 90), the degree of correlation between experimental and
numerical results is signicantly lower than the opposite case. Further-
more, a balance can be noted between mixing effects (controlled by Imix)
and wall effects (given by value) so that when one of them is sup-
pressed the other one tends to be enlarged in order to best t to the
data. These evidences suggest that the bed rearrangement as assumed
by the model (Fig. 10) is able to effectively represent the stratication
prole experimentally observed whereas appreciable improvements
in the prediction can be achieved by taking into account mixing effects.
On the whole, the simulation results obtained were in good agree-
ment with experimental data. The model could predict well the volume
fractions over different zones if the jigging time, i.e. the number o jig-
ging cycles, is enough to achieve a signicant stratication. Evidently,
it is also necessary that the bed constituents have appreciable differ-
ences in density in order to have a mixing index Imix as low as possible,
Fig. 13. Correlation between experimental (t = 120 s) and numerical results. (A) Complete model; (B) Elimination of packing effects (Kw = 1); (C) Removal of mixing effects (Imix = 0);
and (D) Ignoring wall effects on stratication pattern ( = 90).
377
keeping the greater signicance of wall effects on the bed rearrange-
ment. Although the model does not take into account variations in the
jigging cycle (i.e. in the pulsation frequency and expansion amplitude),
it provides a good way to better understand the effects of the lateral
walls on the segregation prole as well as the density stratication
mechanism in air jigs.
4. Conclusions
This study indicates the role of wall effects during stratication in air
jigs. The bed composition after jigging tests was carefully examined at
the central zone and alongside the jig walls. In all cases, there was a no-
table difference in composition between the two zones in such a way
that heavy particles were more concentrated in the center zone of the
stratied bed, giving rise to an unexpected horizontal stratication pro-
le. Experimental results revealed that segregation in the horizontal
plane took place rapidly at the beginning of jigging process and
remained approximately constant as vertical segregation evolved.
From a practical standpoint, this means that wall effects can
378 W.M. Ambrs et al. / Powder Technology 301 (2016) 369378
signicantly inuence the composition of jig products particularly if
short residence times are used.
Similarly, it was found that the stratication rate is fast during the
very few moments after the beginning of jigging process. After a certain
threshold, particle stratication can be approximated as a linear process
over jigging time for the remaining process duration (between one and
two minutes). Changes in local compacity also suggest that the structur-
al rearrangement of particle bed starts from the bottom to the top. The
time evolution of segregation indicated that stratication process con-
tinues to occur even at constant compacity in some bed zones.
Based on the aforementioned observations, a hypothetical mecha-
nism of pattern formation and a prediction model were proposed. The
inuence of side walls on the segregation pattern formation could be
described in the light of the variations in the void fraction in the direc-
tion perpendicular to the lateral walls. The model predictions were in
good agreement with experimental data, especially for long jigging
times. Individual analysis of each model assumption provided consider-
able insight into the role of each phenomenon during the bed rear-
rangement. Comparisons with the experimental results have
enhanced the inuence of wall effects and mixing effects on segregation
pattern formation while packing effects along the walls seems to be
negligible.
Most notably, this is the rst study that investigates the role of wall
effects during stratication in air jigs. It should be noted that the results
were based on a pilot scale batch air jig. Complementary studies are
needed to evaluate the role played by wall effects in industrial
equipments.
Acknowledgements
The authors acknowledge the nancial support of the Brazilian Fed-
eral Agency for Support and Evaluation of Graduate Education (CAPES,
Brazil) under Science without Borders Program and Capes-Cofecub Re-
Maud project.
References
[1] R. Weinstein, R. Snoby, Advances in dry jigging improves coal quality, Min. Eng. 1
(2007) 2934.
[2] C.H. Sampaio, W. Aliaga, E.T. Pacheco, E. Petter, H. Wotruba, Coal beneciation of
Candiota mine by dry jigging, Fuel Process. Technol. 89 (2008) 198202.
[3] F. Boylu, E. Tali, T. etinel, M.S. elik, Effect of uidizing characteristics on upgrading
of lignitic coals in gravity based air jig, Int. J. Miner. Process. 129 (2014) 2735.
[4] T.P.R. Jong, L. Fabrizi, Dry separation of mixed construction and demolition waste,
Recycl. Int. (2004) 2427.
[5] A. Coelho, J. de Brito, Preparation of Concrete Aggregates from Construction and De-
molition Waste (CDW), in: F. Pacheco-Torgal, V.W.Y. Tam, J.A. Labrincha, Y. Ding, J.
de Brito (Eds.), Handbook of Recycled Concrete and Demolition Waste, Woodhead
Publishing Limited, Philadelphia, USA 2013, pp. 210245.
[6] B. Cazacliu, C.H. Sampaio, G. Miltzarek, C. Petter, L. Le Guen, R. Paranhos, F. Huchet,
A.P. Kirchheim, The potential of using air jigging to sort recycled aggregates, J. Clean.
Prod. 66 (2014) 4653.
[7] A.F. Taggart, Handbook of Mineral Dressing, Ores and Industrial Minerals, John
Wiley & Sons Inc, 1945.
[8] R.O. Burt, Gravity Concentration Technology, Elsevier, Amsterdam, 1984.
[9] C.H. Sampaio, L.M.M. Tavares, Beneciamento gravimtrico: Uma introduo Aos
Processos de concentrao Mineral e Reciclagem de Materiais Por Densidade,
Editora UFRGS, 2005.
[10] M.R. Riazi, R. Gupta, Coal Production and Processing Technology, CRC Press, Taylor &
Francis, Florida, U.S.A., 2016
[11] S.M. Viduka, Y.Q. Feng, K. Hapgood, M.P. Schwarz, Discrete particle simulation of
solid separation in a jigging device, Int. J. Miner. Process. 123 (2013) 108119.
[12] E.F. Crespo, Modeling segregation and dispersion in jigging beds in terms of the bed
porosity distribution, Miner. Eng. 85 (2016) 3848.
[13] C.R.K. Windows-Yule, D.J. Parker, Self-diffusion, local clustering and global segrega-
tion in binary granular systems: the role of system geometry, Powder Technol. 261
(2014) 133142.
[14] L.M. Tavares, Monte Carlo simulations on the potential energy theory of jigging, Coal
Prep. 20 (1999) 7183.
[15] C.H. Sampaio, B.G. Cazacliu, G.L. Miltzarek, F. Huchet, L. Le Guen, C.O. Petter, R.
Paranhos, W.M. Ambrs, M.L.S. Oliveira, Stratication in air jigs of concrete/brick/
gypsum particles, Constr. Build. Mater. 109 (2016) 6372.
[16] M. Majid, P. Walzel, Convection and segregation in vertically vibrated granular beds,
Powder Technol. 192 (2009) 311317.
[17] M.J. Metzger, B. Remy, B.J. Glasser, All the Brazil are not on top: vibration induced
granular size segregation of binary, ternary and multi-sized mixtures, Powder
Technol. 205 (2011) 4251.
[18] N. Khola, C. Wassgren, Correlations for shear-induced percolation segregation in
granular shear ows, Powder Technol. 288 (2016) 441452.
[19] X. Yan, Q. Shi, M. Hou, K. Lu, Effects of air on the segregation of particles in a shaken
granular bed, Phys. Rev. Lett. 91 (2003) 014302.
[20] M. Klein, L.L. Tsai, M.S. Rosen, T. Pavlin, D. Candela, R.L. Walsworth, Interstitial gas
and density segregation of vertically vibrated granular media, Phys. Rev. E 74
(2006) (010301(R)).
[21] C.R.A. Abreu, F.W. Tavares, M. Castier, Inuence of particle shape on the packing and
on the segregation of spherocylinders via Monte Carlo simulations, Powder Technol.
134 (2003) 167180.
[22] M.A. Aguirre, A. Calvo, Compacity inuence in reorganization of two dimensional
granular packings: numerical model, Powder Technol. 201 (2010) 109113.
[23] R. Escudi, N. Epstein, J.R. Grace, H.T. Bi, Effect of particle shape on liquid-uidized
beds of binary (and ternary) solids mixtures: segregation vs mixing, Chem. Eng.
Sci. 61 (2006) 15281539.
[24] R. Ben-am, Study of the Texture of Packings of Grains. Applications to the Determi-
nation of the Permeability of Binary Mixtures in the Molecular, Intermediate, Lami-
nar Regime(State Thesis) University of Nancy, 1970.
[25] J.D. Mcwhirter, M.E. Crawford, D.E. Klein, Wall region porosity distributions for
packed beds of uniform spheres with modied and unmodied walls, Transp. Po-
rous Media 27 (1997) 99118.
[26] C.R.K. Windows-Yule, G.J.M. Douglas, D.J. Parker, Competition between geometrical-
ly induced and density-driven segregation mechanisms in vibrouidized granular
systems, Phys. Rev. E 91 (2015) 032205.
[27] F.W. Mayer, Fundamentals of Potential Theory of the Jigging Process, VII Internation-
al Mineral Processing Congress, New York, Part 2, 1964 750786.
[28] A. Jain, M.J. Metzger, B.J. Glasser, Effect of particle size distribution on segregation in
vibrated systems, Powder Technol. 237 (2013) 543553.
[29] K. Xue, Y. Zheng, B. Fan, F. Li, C. Bai, The origin of granular convection in vertically
vibrated particle beds: the differential shear ow eld, Eur. Phys. J. E 36 (2013).
[30] F. de Larrard, T. Sedran, D. Angot, Prvision de la compacit Des Melanges
Granulaires Par le modle de Suspension Solide. II: Validations et Cas Des mlanges
conns, Bulletin de Liaison Des Laboratoires Des Ponts et Chausses, Vol. 1941994
7186.