IWA Conferences

Influence of air flow rate and backwashing on the hydraulic behaviour of a submerged
biological filter
--Manuscript Draft--

Manuscript Number: IWA-8387R1

Full Title: Influence of air flow rate and backwashing on the hydraulic behaviour of a submerged
biological filter

Article Type: Full Paper

Keywords: Hydraulic behaviour; lava stones; mathematical models, non ideal flow; submerged
biological filter; tracer studies.

Corresponding Author: Yazmín Lucero Cobos-Becerra, M.D.

National Autonomous University of Mexico
Mexico D.F., Coyoacán MEXICO

Corresponding Author Secondary

Information:

Corresponding Author's Institution: National Autonomous University of Mexico

Corresponding Author's Secondary

Institution:

First Author: Yazmín Lucero Cobos-Becerra, M.D.

First Author Secondary Information:

Order of Authors: Yazmín Lucero Cobos-Becerra, M.D.

Simon González-Martínez, phD

Order of Authors Secondary Information:

Abstract: This study evaluated the hydraulic behaviour of a submerged biological filter with and
without aeration, for two particle sizes of filter media, under different hydraulic retention
times using a tracer. The filter media was sieved volcanic stones (with irregular shape)
with average values of 4.7 and 9.5 mm. The effect of backwashing and different air
flow rates over the hydraulic behaviour of the filters were evaluated. The tracer curves
were analyzed with two mathematical models for ideal and non-ideal flow (axial
dispersion and Wolf and Resnick models). Deviations from the ideal behaviour of plug
and completely mixed flow indicate further reduction of the filter effective volume due to
the turbulence caused by air mixing in the case of the aerated filter and the fluid
channelling in the non aerated filter. It was observed that backwashing causes attrition
of media particles, decreasing the void volume of the filter media and, consequently,
the tracer flow is more uniform. The eroded media (surface erosion caused by
backwashing) presented lower dead volume (74 % for aerated filter and 4 % for non
aerated filter) compared with the new media (originally packed, 83 % for aerated filter
and 19 % for non aerated filter). The dead volume fraction decreases when the system
tends to plug flow behaviour.

Powered by Editorial Manager® and Preprint Manager® from Aries Systems Corporation
Manuscript
Click here to download Manuscript: Cobos-Becerra_IWA-8387_WWC2012.doc

Influence of air flow rate and backwashing on the hydraulic

behaviour of a submerged biological filter

Y.L. Cobos-Becerra, S. González-Martínez

Institute of Engineering, National Autonomous University of Mexico (Universidad Nacional

Autónoma de México), Ciudad Universitaria, 04510 Mexico DF, Mexico, (E-mail:
lucerocobos@yahoo.com, sgm@pumas.ii.unam.mx)

Abstract
This study evaluated the hydraulic behaviour of a submerged biological filter with and without
aeration, for two particle sizes of filter media, under different hydraulic retention times using a
tracer. The filter media was sieved volcanic stones (with irregular shape) with average values of 4.7
and 9.5 mm. The effect of backwashing and different air flow rates over the hydraulic behaviour of
the filters were evaluated. The tracer curves were analyzed with two mathematical models for ideal
and non-ideal flow (axial dispersion and Wolf and Resnick models). Deviations from the ideal
behaviour of plug and completely mixed flow indicate further reduction of the filter effective
volume due to the turbulence caused by air mixing in the case of the aerated filter and the fluid
channelling in the non aerated filter. It was observed that backwashing causes attrition of media
particles, decreasing the void volume of the filter media and, consequently, the tracer flow is more
uniform. The eroded media (surface erosion caused by backwashing) presented lower dead volume
(74 % for aerated filter and 4 % for non aerated filter) compared with the new media (originally
packed, 83 % for aerated filter and 19 % for non aerated filter). The dead volume fraction decreases
when the system tends to plug flow behaviour.

Keywords
Hydraulic behaviour; lava stones; mathematical models, non ideal flow; submerged biological
filter; tracer studies.

INTRODUCTION
Uniformity and effective contact between substrate and biomass in a biological filter play an
important role in its performance and behaviour. These parameters are conditioned function of
liquid mixing and have an effect in bioreactor performance. Factors such as water flow, air flow
rate, packed media shape and morphological characteristics and filter porosity also affect the flow
pattern in the system (Smith et al., 1996; Nabizadeh et al., 2000; Escudié et al., 2005; Tembhurkar
and Mhaisalkar, 2006; Fatihah and Donelly, 2008).

Hydraulic characteristics of bioreactors containing fixed biofilms are commonly measured using a
tracer to obtain the residence time distribution (RTD) of the liquid (Wolf and Resnick, 1963; Smith
et al, 1993; 1996; Morgan-Sagastume et al., 1997; de Nardi et al., 1999; Escudié et al., 2005). To
describe the deviations from the ideal completely mixed reactor (CMFR) or ideal plug flow reactor
(PFR) several reactor models have been developed. Deviation from ideal behaviour of PFR or
CMFR can be caused by channelling of the fluid or by the presence of slow moving regions in the
reactor due to recirculation or stagnant pockets of fluid (Levenspiel, 2002; Fatihah and Donelly,
2008).

Although some reactors designed via this ideal approach can be operated with good results under
real conditions, others fail to achieve the expected performance due, in most cases, to substrate
transport limitation. Hydraulic short circuits affect the hydrodynamic behaviour of reactors
originating “dead volumes” (Morgan-Sagastume et al., 1997; Fatihah and Donelly, 2008) which
decreases the reactor effective volume and, as a result, real hydraulic retention time (HRT) and
efficiency decrease proportionally (Morgan-Sagastume et al., 1997). Tracer studies and analysis of
residence time distribution (RTD) are used to quantify these as well as deviations from ideal flow
patterns (Morgan-Sagastume et al., 1999; 1997).

The measurement and analysis of RTD has become an important tool in the study of continuous
flow systems. In order to analyze the RTD using tracer studies three methods may be employed: (1)
the determination of the mean hydraulic retention time and the variance by means of the area under
the curve; (2) the use of mathematical models in order to assess the flow pattern by fitting the RTD
curves shape (such as tanks in series model, axial dispersion model, combined models, etc.
(Levenspiel 2002)); and (3) the calculation of indexes or parameters that have an empirical or semi
empirical support based on the most important characteristics of tracer curves, i.e. its height and
eccentricity (Morgan-Sagastume et al., 1999).

Different types of models may be used to describe non ideal flow within reactors. The most widely
used model for tubular reactors is the axial dispersion model (Smith, 1981; Turan and Ozturk, 1997;
Levenspiel, 2002). This model describes mass transport in the axial direction in terms of an
effective longitudinal dispersion coefficient (DL). By changing the magnitude of this single
parameter, the performance of reactors ranging from an ideal PFR (DL=0) to ideal CMFR (DL=∞)
can be described (Smith, 1981; Smith et al., 1993; Morgan-Sagastume et al., 1997, 1999;
Levenspiel, 2002). According to Morgan-Sagastume et al. (1999) and Smith et al. (1993) there are
models named "combined models" in which the reactor is described according to different flow
types such as plug and completely mixed flow types with dead volume zones in the reactor (Wolf
and Resnick and Chollete and Cloutier models); other models consider interconnected flow regions
in various ways such as bypass, recycle or crossflow. For example Chollete and Cloutier (1959)
combine a mixed tank with a bypass stream and a dead volume zone. Combined models have been
applied in the study of anaerobic sludge reactors, anaerobic filters and aerobic submerged filters
(Morgan-Sagastume et al., 1999).

Tracer studies can provide information on the residence time distribution to determine the real
residence time and use different mathematical and simulation models that allow the understanding
of the biological system hydraulics and assist its start up, operation and control (Smith et al., 1993;
de Nardi et al., 1999; Escudié et al., 2005). There are few investigations about biological aerated
filters (BAF) hydraulic behaviour in their start up phase. In previous work González-Martínez et al.
(2007, 2011) showed nutrient profiles of a biological aerated filter (BAF) indicating a removal
along the filter depth which leads to the conclusion that a BAF has predominantly plug flow
behaviour but it was not proven using tracer test or mathematical models.

For better understanding of the submerged biological filter hydraulics, tracer studies were carried
out in a pilot-scale submerged filter to evaluate its hydraulic behaviour. The filter was packed with
two different particle sizes of lava stones and evaluated under different hydraulic retention times.
The effects of backwashing and different air flow rates were important to establish possible
deviations from ideal models.

METHOD
Pilot filter
In order to see inside the filter, a pilot filter made of glass sheets; 15 cm square section was built
and packed with lava stones. The effluent was placed 15 cm above the upper stone layers (figure 1)
and a diffuser was placed under the lava stones at the bottom. Table 1 presents the filter main
characteristics.
Table 1. Pilot submerged biological filter characteristics
Characteristics Value
Total height 1.90 m
Section 0.023 m2
Filter bed depth 1.0 m
Total volume 22.5 L
Apparent void volume 15.6 L

Packing bed material

Due to its high porosity, roughness and specific area exposed to the adhesion of microorganisms,
lava stones were used as a filter media (figure 2). Particle sizes of 4.7 mm (between 4.0 and 4.75
mm) and 9.5 mm (between 8.0 and 9.5 mm) were used (table 2). According to Mendoza-Espinoza
and Stephenson (1999) and Moore et al.(2001), the media size has a strong influence on process
performance. Particles with sizes above 6 mm are recommended for COD removal and nitrification
and for tertiary treatment and nutrient removal particle sizes between 2 and 3 mm are
recommended.

Figure 2. Lava stones with average diameter of 9.5 (a) and 4.7 mm (b)

Figure 1. Pilot Filter

Table 2. Lava stones characteristics for both particle sizes

Characteristic Units Size 1 Size 2
Diameter mm 9.5 4.7
Specific area m2/m3 442 1952
Apparent porosity % 55 54
Bulk density kg/m3 664 710
Lava stones density kg/m3 1926 1921

Tracer
Direct Blue 2, diazo dye (=598 nm), was used as tracer. During 2 days before the beginning of
tracer studies, the bed filter was saturated with the tracer to prevent sorption phenomena during the
tests. The excess dye was washed out with tap water until only tap water filled the void volume in
the filter shortly before the test began (time = 0). Tracer solution was continuously pumped from
the bottom of the filter. The dye concentration was analyzed at the effluent with a
spectrophotometer.
Tracer tests
Considering an effective volume of 15.6 L, three different influent flows were adjusted (0.108
L/min, 0.058 L/min and 0.035 L/min) to achieve three different theoretical hydraulic retention times
(HRT) of 2.4, 4.5, and 7.4 h, without aeration. These same conditions were used for both particle
sizes and later with aeration (air flow rate = 10 L/min) to make the comparison between hydraulic
behaviour of an aerated and a non aerated packed filters. In a second stage, tracer test were carried
out using different air flows (0, 2, 4, 6, 10 L/min) for a HRT of 4.5 h, which is common for
biological aerated filters (Mendoza-Espinoza and Stephenson, 1999).

Mathematical models
Response curves were analyzed considering mathematical models for non-ideal flow according to
Wolf and Resnick (1963), Smith (1981), Levenspiel (2002). The Excel Solver tool was used for best
fitting.

Axial Dispersion Model. This model considers the dispersion coefficient as a deviation from plug
flow ideal behaviour due to longitudinal dispersion that may occur along the filter bed. For an ideal
PFR the dispersion coefficient is zero (0) and for an ideal CMFR it is infinite (∞) (Figure 3 and
Equation 1) (Levenspiel, 2002; Smith, 1981).

  t 
 C  1  1 uL 1  t0  
   1  erf   Equation 1
 Co  2   2 DL t 
  t0  

Where:

C0 = Tracer maximal (influent) concentration

C = Tracer in effluent at any given time
t0 = Theoretical residence time
Figure 3. Graphical representation of the t = Time
ideal models for PFR and CMFR (Smith, D /uL = Dispersion coefficient
L
1981).

Dispersion is the general concept used to describe the process which can take place when a liquid
flows through a tubular reactor. The effects of dispersion are particularly important in packed bed
reactors. Neglecting the fluid dispersion may result in inefficient transport of substrate through the
reactor and an inaccurate estimation of effective volume of the system (Turan and Ozturk, 1997).

Wolf and Resnick Model. This model considers that a real system (non-ideal) may behave as a
combination of plug flow (P) and completely mixed flow (M) containing a dead volume (m) which,
combined, affect the hydraulic behaviour (Wolf and Resnick, 1963). The corresponding fractions of
these flow types and the dead volume can be estimated using this model (Equation 2).
C    1  t 
   1  EXP       p 1  m    Equation 2
 Co    1  p 1  m    t0  
Here the term (1-p) represents the fraction of non-plug flow and m the dead volume. Modifying
Equation 2 results in Equation 3:
 C  log e  t 
Log 1        p 1  m   Equation 3
 C0   1  p 1  m    t0 

Figure 4. Graphical representation of
Wolf and Resnick Model
Complete mixed flow
Plug flow P= p(1-m)
Figure 4 shows the graphical representation of
Equation 3 where (1-C/C0) represents the fraction of
tracer that remains in the system as a function of the
relative time (t/t0). The dotted line intersects the x-axis
at a value of 1 and is tangent to the observed flow (log
e/(1-p)(1-m)). Using the slope value of the previously
described line and equations 4 and 5, the plug flow (P),
completely mixed flow (M) fractions and the dead
volume (m) can be calculated.
M= (1-p)(1-m) Equation 4
Equation 5

RESULTS AND DISCUSSION

Tracer test results
The objective of tracer studies was to estimate dead volume formation in a SUBMERGED FILTER
under different HRT and air flow rates. The principle underlying the RTD is based on the fact that
factors such as interstitial superficial upflow velocity, mixing and backwashing could significantly
affect the reactor performance.

In the first experimental phase, tracer tests were carried out in clean lava stone beds (no biomass
were present) with particle sizes of 9.5 and 4.7 mm with and without aeration. Considering the
effective (apparent) void volume in the filter (15.6 L) and three different water flows, the
theoretically HRT during the experiments were 2.4 h (114 min), 4.5 h (269 min) and 7.4 h (446
min).

Figures 5a and 5b present the tracer curves for both filter beds with and without aeration: 1) for tests
with aeration (dotted lines), the curves indicate that the tracer begins to appear in the effluent within
a few minutes after the test began and it replaces all the effective volume filter in less time than
expected. These curves have an almost linear behaviour, most of the time indicating that the tracer
concentration in the effluent increases steadily with time. By decreasing the water flow rate (higher
HRT) the slope of these "curves" decreases. 2) During the tests without aeration (solid lines), the
tracer has a radically different behaviour compared with the test with aeration. The tracer starts to
appear in the effluent after several minutes while its concentration increases slowly. The maximum
tracer concentration appeared in the effluent earlier than predicted based on the theoretical HRT.

During the tests without aeration, formation of "channels" was observed. These are deviations from
the ideal behaviour where, in specific places, the tracer flows faster than in most of the packing bed
(figure 6). In the bed with smaller particles (4.7 mm) fewer channels were observed than in the bed
filter with the larger stones (9.5 mm). Although both filter beds have the same apparent porosity (55
%), for the three tested flow rates (HRT) the tracer concentration in the effluent appeared faster in
the bed filter with the larger particles than in the other one. The smaller particles allow a more
uniform flow through the filter bed as the particles present smaller spaces between them (figure 6).
To avoid the formation of the same channels, backwashing of the filter was performed between tests
to redistribute the particles inside the filter bed.
Figure 5. Tracer test for 9.5 (a) and 4.7 mm (b) lava stones, filter with aeration (dotted lines) and
filter without aeration (solid lines).

Figure 6. Tracer tests without aeration, bottom of the filter

For the second experimental phase, the influence of air flow rate and backwashing in the filter
hydraulic behaviour was tested. For these tests, the 4.7 mm filter media was selected because 1) the
tracer takes longer to flow through the filter bed with and without aeration, 2) visual assessments
indicate that the smaller size has fewer channelling than the large size (figure 6) and 3) Moore et al.
(2001) and Mendoza-Espinosa and Stephenson (1999) recommend smaller particle sizes that
present higher surface area exposed for biofilm development.

Backwashing procedures expose the filter media to continuous attrition. Abrasion of the media
particles causes removal of the corners, edges and surface layers by grains rubbing against each
other, the particles resulting smaller than the parent particles (Humby and Fitzpatrick, 1996). Sixty
backwashes were performed in the 4.7 mm bed filter; after these procedures the particles had more
regular shape and presented a better distribution in the packing bed. Figure 7 shows the tracer
curves for an HRT of 269 min under different air flow rates (0, 2, 4, 6 and 10 L/min).

The mixing caused by different air flow rates

(2 to 10 L/min) did not present a significant
influence in the flow patters observed on the
4.7 mm filter bed subjected to repeated
backwashing. Comparing the tracer curves
for new filter media (originally packed,
figure 5b) with the media after repeated
backwashing (figure 7), it can be concluded
that erosion of media filter, caused by
repeated backwashing, influences the flow
patterns: The tracer remains longer in the
eroded bed than in the newly packed bed for Figure 7. Tracer test under different air flows for
both conditions, with or without aeration. 4.7 mm media filter with backwashing.

Mathematical modelling
Ideal models. These models suggest that, when the dispersion coefficient tends to zero (DL/uL= 0),
the axial dispersion is negligible and the system presents a plug flow behaviour. On the other hand,
when the dispersion coefficient tends to infinity (DL/uL = ∞), the dispersion is very large and the
system approaches the completely mixed hydraulic pattern (Smith, 1981; Smith et al., 1993;
Morgan-Sagastume et al., 1997; Levenspiel, 2002; Tembhurkar and Mhaisalkar, 2006).

Figure 8 compares the tracer curves for 4.7

mm new and eroded media, with and
without aeration, using ideal models. Under
ideal plug flow conditions, no tracer should
appear in the effluent until t equals t0
(theoretical HRT, empty bed volume divided
by influent flow rate) and the tracer
concentration C in the effluent should be
equal to C0 (DL/uL=0). In the case of
complete mix, shortly after the tracer begins
to flow in the filter, the tracer begins to
appear in the effluent, increasing its Figure 8. Comparison of tracer curves for new and
concentration with time tending to an eroded media (4.7 mm) with CMFR and PFR
asymptote; when t equals t 0 the tracer models.
concentration in the effluent reaches
approximately 63 % of C0 (DL/uL=).
Between these opposite conditions intermediate degrees of dispersion in the reactor can be
attributed to effective volume reduction of the reactor caused by fluid channelling, intense mixing
caused by the air bubbles and the fluid volume displaced by the air bubbles. For the new media
(solid lines) and eroded media (dotted lines) tracer curves do not fit any ideal model, either CMFR
or PFR (figure 8). The non-ideal flow models according to Levenspiel (2002), Smith (1981) and
Wolf and Resnick (1963) to describe the hydraulic behaviour were selected.

Axial dispersion model. To evaluate the extent of short circuiting within the packed bed, this model
assesses the dispersion coefficient and actual HRT.

New media. The dispersion coefficient and effective volume for new filter media (9.5 and 4.7 mm)
and eroded media (4.7 mm) were calculated. In the new filter media, the hydraulic behaviour
between both beds (9.5 and 4.7 mm) was similar. The dispersion coefficients and effective volumes
calculated for each HRT, for the filter with and without aeration, are presented in table 3.

Table 3. Effective volumes and dispersion coefficients for new media (9.5 and 4.7 mm). The
theoretical effective volume is 15.6 L.
Theoret- Aerated filter Non aerated filter
ical Calculated Calculated Calculated Calculated
Dispersion Dispersion
HRT HRT effective HRT effective
coefficient coefficient
(min) (min) volume (L) (min) volume (L)
144 116 12.2 0.005 25 2.7 0.27
269 230 13.0 0.003 46 2.6 0.38
446 360 12.6 0.002 88 3.1 0.50

For the aerated filter the calculated effective volumes at different HRT are lower than for the filter
without aeration. It can be concluded that the aerated filter has higher dead volume caused mainly
by the turbulence and the fluid displaced volume caused by air bubbles. For the new filter media,
the dead volume fraction of the aerated filter is 83 % and 22 % for the filter without aeration.

Based on the dispersion coefficient, for the non aerated filter, the hydraulic behaviour reflects more
closely a plug flow reactor with 22 % of dead volume and a certain degree of dispersion. The
observed dispersion coefficients are in the range between 0.002 and 0.005, which indicate an
intermediate dispersion in the reactor according to the values reported by Tembhurkar and
Mhaisalkar (2006) and Tay and Show (1998). The dispersion coefficient increases with decreasing
HRT in the reactor. For the filter with aeration, the calculated dispersion coefficients show that
higher HRT produce higher longitudinal dispersion due to turbulence caused by air.

Eroded media. The channels formed in the filter without aeration and eroded media (4.7 mm
particles) were not as significant as those observed in the new media. The hydraulic behaviour of
eroded and new media adjusted to the axial dispersion model is presented in figure 9. The eroded
filter media presented less dead volumes (4 %) than the new media (17 %). The dead volume
fractions were estimated using the same dispersion coefficient of 0.003.

For the filter with aeration, the effective volume for eroded media is larger (4.0 L) than for new
media (2.6 L) estimated using the same dispersion coefficient of 0.38; the tracer curves fitted to the
axial dispersion model are shown in figure 10. The eroded media presented less dead volume (74
%) than the new media (83 %).

Flow patterns of eroded and new media are different because the backwashing 1) causes attrition of
the particle surface reducing the edges to a more regular shape, 2) decreases the void volume of the
media, 3) improves the particle size distribution (Fitzpatrick, 1998) and, consequently, a more
uniform flow of the water.
Figure 9. Axial dispersion model fitted to tracer Figure 10. Axial dispersion model fitted to
curves for the filter without aeration before and tracer curves for the filter with aeration before
after backwashing. and after backwashing.

Wolf and Resnick model. This model is a "combined model" where the reactor is described in terms
of different flow types such as plug and completely mixed flow and dead volume zones.

New media. Using the model developed by Wolf and Resnick (1963), the flow pattern and the dead
volume fraction in the filter can be estimated. Table 4 shows the estimated flow patterns, dead
volume fractions and the equation that describes the hydraulic behaviour of the filter with and
without air when the bed filter has not yet been exposed to attrition caused by backwashing (new
media).

Table 4. Hydraulic behaviour of the filter without and with aeration for 4.7 and 9.5 mm particles
using new media; Wolf and Resnick model.
Plug Completely Dead
Filter Particle Expression for the hydraulic
flow mixed flow volume
type size behaviour
(%) (%) (%)
With 9.5 mm 3 12 84 C C0  1  EXP  8.03  t t0  0.034  
aeration 4.7 mm 3 14 83 C C0  1  EXP  7.17  t t0  0.026  

Without 9.5 mm 72 8 20 C C0  1  EXP  13  t t0  0.722  

aeration 4.7 mm 74 7 19 C C0  1  EXP  13.35  t t0  0.743 

Figures 11a and 12a show that the tracer curves for the new filter media (4.7 mm) have the same
slope (indicated by the line) for the three HRT tested and for both conditions, with and without
aeration. Same hydraulic behaviour was observed for the 9.5 mm new media. Figures 11b and 12b
show the Wolf and Resnick model fitted to the experimental data.

For the new media (particle sizes of 9.5 and 4.7 mm) the dead volume fraction is higher in the filter
with aeration (84 %) compared to the filter without aeration (20 %); the last one presents
predominantly plug flow compared with the filter with aeration. Dead volumes are the result of
turbulent mixing and liquid displacement caused by air bubbles that reduce the effective volume in
the aerated filter. In the filter without aeration, channelling and stagnant zones are the cause for
dead volumes.

Eroded media. Under different air flow rates the tracer curves present the same slope (figure 13).
Table 4 shows the estimated flow types, dead volume fractions and the mathematical expressions
that describe the hydraulic behaviour of the filter with and without air when the bed filter has been
exposed to attrition. The filter with aeration presents a higher dead volume fraction (79 %) than the
filter without aeration (8 %). Backwashing causes attrition and redistribution of media particles and,
consequently, a decrease of void volume.

Figure 11. Non aerated filter with 4.7 mm new media. Wolf and Resnick model (a) and tracer
curves fitting (b).

Figure 12. Aerated filter with 4.7 mm new media. Wolf and Resnick model (a) and tracer curves
fitting (b).

Table 5. Hydraulic behaviour of the filters without and with aeration for 4.7 mm eroded media
using the Wolf and Resnick model.
Plug Completely Dead
Particle Filter Expression for the hydraulic
flow mixed flow volume
size type behaviour
(%) (%) (%)
With C C0  1  EXP  8.03  t t0  0.034  
6 15 79
aeration
4.7 mm
Without C C0  1  EXP  7.17  t t0  0.026  
83 9 8
aeration

The dispersion model allows a better fit of the experimental data than the Wolf and Resnick model.
Flow types and fractions of dead volumes can be estimated using the Wolf and Resnick model
while the effective volume and, therefore, the dead volumes can be better estimated using the axial
dispersion model. The dead volumes presented no significant differences when calculated with
either model.

CONCLUSIONS
Tracer curves in a submerged filter were analyzed with two mathematical models for ideal and non-
ideal flow (axial dispersion and Wolf and Resnick models). The dispersion model allows a better fit
of the experimental data than the Wolf and Resnick model. Flow types and fractions of dead
volumes can be estimated using the Wolf and Resnick model while the effective volume and,
therefore, the dead volumes can be better estimated using the axial dispersion model.

Figure 13. 4.7 mm eroded media. Wolf and Resnick model (a) and tracer curves fitting (b).

Deviations from the ideal behaviour of plug and completely mixed flows indicate losses of the filter
effective volume due to the turbulence and water volume displacement caused by the air bubbles in
the filter with aeration and fluid channelling in the filter without aeration.

Based on dispersion coefficients calculated with the axial dispersion model, the filter without
aeration shows that the hydraulic behaviour approaches a plug flow reactor; the dispersion
coefficients present values between 0.002 and 0.005, indicating intermediate dispersion in the
reactor: The dispersion coefficients increase with decreasing HRT. In the filter with aeration, the
calculated dispersion coefficients show that higher HRT produce higher longitudinal dispersion due
to turbulence caused by air.

The eroded media presented lower dead volumes (79 % for the filter with aeration and 8 % for the
filter without aeration) compared with the new media (83 % for the filter with aeration and 19 % for
the filter without aeration). The flow patterns of eroded and new media were different because the
more regular shape of the particles decreases the void volume of the filter media.

The dead volume fraction decreases when the system tends to plug flow. The filter with aeration
presented completely mixed flow fractions of 14 % for new media and 15 % for eroded media.

ACKNOWLEDGMENT
This project was possible thanks to the support of The National Council for Science and
Technology (CONACYT), SEP-CONACYT project 103922.

REFERENCES
Chollete, A., Cloutier, L. (1959). Mixing efficiency determinations for continous flow systems. The
Canadian Journal of Chemical Engineering. 37(3), 105-112.
de Nardi, I. R., Zaiat. M., Foresti, E. (1999). Influence of the tracer characteristics on hydrodynamic
models of packed bed bioreactors. Bioprocess Engineering. 21(5), 469-476.
Escudié, R., Conte, T., Steyer, J. P., Delgenès, J. P. (2005). Hydrodynamic and biokinetic models of
an anaerobic fixed bed reactor. Process Biochemistry. 40(7), 2311-2323.
Fatihah, S., Donelli, T. (2008). Effect of media volume on mixing of biological aerated filters.
Jounal of Environmental Engineering Science. 7(10), 229-236.
Fitzpatrick, C.S.B. (1998). Media properties and their effect on filter performance and backwashing.
Water Science and Technology. 38(6): 105-111
González-Martínez, S., Millán, T., González-Barceló, O. (2007). Biological aerated filtration of
municipal wastewater using a low-cost filtration media. Water Science and Technology.
55(7), 255-262.
González-Martínez, S., González-Barceló, O., Flores-Torres, C. A. (2011). Wastewater treatment in
an anaerobic filter using small lava stones as filter media without temperature control. Water
Science and Technology. 63(6), 1188-1195.
Humby, M.S., Fitzpatrick, C.B.S. (1996). Attrition of granular filter media during backwashing
with combined air and water. Water Research. 30(2): 291-294.
Levenspiel, O. (2002). Chemical Reaction Engineering. 2nd edition. Reverté. Spain.
Moore, R., Quarmby, J., Stephenson, T. (2001). The effects of media size on the performance of
biological aerated filters. Water Research. 35(10), 2514-2522.
Mendoza-Espinoza, L., Stephenston, T. (1999). A review of biological aerated filters (BAFs) for
wastewater treatment. Environmental Engineering Science. 16(3), 201-216.
Morgan-Sagastume, J. M., Jiménez, B., Noyola, A. (1999). Alternative index for interpretarion of
RTD curves. Journal of Environmental Engineering. 125(3), 290-294.
Morgan-Sagastume, J. M., Jiménez, B., Noyola A. (1997). Tracer studies in a laboratory and pilot
scale UASB reactor. Environmental Technology. 18(8), 817-825.
Nadizadeh, R., Mesdaghinia, A.R., Nasseri, S., Mahvi, A.H., Shariat, M. (2000). Analyzing flow
characteristics and influence of biological growth on dispersion in aerate submerged fixed-
film reactors (ASFFR). Biofilm - Online Journal, 5(1).
https://tspace.library.utoronto.ca/bitstream/1807/79/2/bf00001.pdf (accessed 15 may 2012).
Smith, L.C., Elliot, D.J., James, A. (1996). Mixing in upflow anaerobic filters and its influence on
performance and scale-up. Water Research 30(12), 3061-3073.
Smith, L.C., Elliot, D.J., James, A. (1993). Characterisation of mixing patterns in an anaerobic
digester by means of tracer curve analysis. Ecological Modelling. 69(3-4), 267-285.
Smith, J.M. (1981). Chemical Engineering Kinetics. 3rd edition, McGraw Hill, New York.
Tay, J.H., Show, K.Y. (1998). Media induced hydraulic behavior and performance of upflow
biofilters. Journal of Environmental Engineering. 124(8), 720-729.
Tembhurkar, A.R., Mhaisalkar, V.A. (2006). Study of hydrodynamic behavior of a laboratory scale
upflow anaerobic fixed film bed reactor. Journal of Environmental Science and Engineering
48(2), 75-80.
Turan, M., Ozturk, I. (1997). Comparative evaluation of longitudinal dispersion of liquid in non-
biological and anaerobic fixed film reactors. Environmental Technology. 18(1), 45-53.
Wolf, D., Resnick, W. (1963). Residence time distribution in real systems. I&EC Fundamentals,
2(4), 287-293.