PII: S0003-4878(00)00027-2
Ann. occup. Hyg., Vol. 45, No. 2, pp. 115–121, 2001
 2001 British Occupational Hygiene Society
Published by Elsevier Science Ltd. All rights reserved
Printed in Great Britain.
0003–4878/01/$20.00
Changes in the Performances of Filter Media During
Clogging and Cleaning Cycles
S. CALLɆ*, D. BÉMER‡, D. THOMAS†, P. CONTAL† and
D. LECLERC†
†Laboratoire des Sciences du Génie Chimique–CNRS, 1 rue Grandville, 54000 Nancy, France;
‡Institut National de Recherche et Sécurité, Avenue de Bourgogne, 54501 Vandoeuvre les Nancy,
France
The performance of two filter media used in industrial air cleaning were studied both in the
initial state (new filter) and after a number of collection and pulse pressure cleaning cycles.
The main difference between them is that one has anti-clogging properties and the other does
not. The test aerosol is composed of alumina particles with a median volumetric diameter of
2.6 mm (MMAD=4.8 mm) generated at a concentration of 700 mg mⴚ3. Filtration took place at
a velocity of 2 cm sⴚ1. Two parameters, namely pressure drop and efficiency, were monitored
according to the collection and cleaning cycles. The comparison of the filtration efficiency of
the two media and that of the corresponding industrial dust separator at the end of the
cycles showed a close agreement. The separation efficiency calculated with a new medium
(corresponding to initial switch-on of the installation) was low and increased very quickly
during the cycles. Finally, a phenomenological model was developed to represent the increase
in pressure drop of a filter medium after cleaning and was found to be in close agreement
with the experimental values.  2001 British Occupational Hygiene Society. Published by
Elsevier Science Ltd. All rights reserved.
Keywords: filtration; clogging; pulse-jet
INTRODUCTION
Recycling the air of workplaces is a ventilation technique in widespread use mainly to save energy. The
quality of the recirculated air, which is monitored
very closely, is directly linked to the filtration
efficiency for the pollutants present in the air flow.
The quality of the air can be checked by measuring
the pollutant concentration in the recirculated air
regularly and by measuring the separation efficiency
of the filter.
European standards currently classify the filter
media used in general ventilation systems by means
of this efficiency parameter. In the field of industrial
air cleaning, the test methods are based solely on testing new media. The performance measured in this
way is far removed from that of materials in use on
account of the loading and regeneration cycles undergone by the media. The point of this study is to show
Received 8 December 1999; in final form 17 April 2000.
*Author to whom correspondence should be addressed. Tel.:
+33-8340-2000; fax: +33-8350-2097; e-mail: [email protected]
how the performance of a filter medium varies with
the clogging and cleaning cycles, in other words in
real operating conditions (filtration and regeneration).
Two types of medium with different clogging
properties are therefore examined. The paper then
goes on to compare their performance with that of a
dust separator used in wood industry.
CHARACTERISATION OF FILTER MEDIA
Two filter media were selected for this study of
filtration performance in collection and cleaning
cycles. These two media, designated C and D in what
follows, are employed in industrial filters in the form
of a bag and cartridge respectively. The fundamental
difference between these two media is the clogging
related treatment received during manufacture.
Medium C is a polyester needlefelt with no specific
treatment whereas medium D, made up of cellulose
fibres, has a thin layer of very fine fibres on its surface
constituting an anti-clogging coating.
The different physical parameters determined were
the mass per unit area, filter thickness, packing den-
115
116
S. Callé et al.
Nomenclature
af
ag
Cp
dae
⌬P
E
g
m
Q
S
U
W
Z
rf
kck
Ap
rp
filter medium packing density
particle cake packing density
particle concentration (kg.m⫺3)
aerodynamic diameter (m)
pressure drop (Pa)
filtration efficiency
fraction of cleaned surface
air viscosity (Pa s)
volumic flow rate (m3 s⫺1)
filtration surface (m2)
filtration velocity (m s⫺1)
mass of particle per unit area (kg m⫺2)
filter medium thickness (m)
fibre radius (m)
Kozeny and Carman constant
particle specific area (m⫺1)
particle density (kg m⫺3)
sity and fibre size distribution. The mass per unit area
was calculated from the weight and area of a sample.
The thickness of each medium was determined by
observation with an optical microscope. The packing
density was calculated from these two parameters and
the density of the fibres. Finally, image analysis carried out with a scanning electron microscope gave
access to the fibre size distribution. One observation
of each of the filters C and D with the electron microscope is given in Fig. 1.
The arithmetic mean diameter of the fibres as well
as the spread of the distribution were calculated. In
the case of filter D, the distribution of the two types
of fibres were considered separately. The results
obtained are given in Table 1.
The characterisation of the new filters was followed by measurement of the initial fractional filtration efficiency according to the standard Pr EN 779
(1998). A polydisperse test aerosol was generated
upstream of the filter and samples taken alternately
upstream and downstream of the filter medium using
an optical counter. The separation efficiency for a
given diameter (corresponding to one channel of the
counter) was calculated from the number of particles
of this diameter counted upstream and downstream of
the filter. The efficiency value uncertainty is determined from the Student law. The aerosol selected was
composed of KCl particles generated by means of
nebulisation from a 1% in volume solution using a
De Vilbis 40 nebuliser. This choice stemmed from
the fact that the information given by the counter is
only valid for particles whose refractive index is close
to that of the latex particles used to calibrate the
apparatus. The error on the number of particles
sampled when using a KCl aerosol according to the
calculations is about 5% (Bémer, 1990).
As shown in what follows, this initial efficiency
information is particularly important in highlighting
Fig. 1. Microscopic observations of new medium C and medium D.
Clogging and cleaning cycles of filter media
117
Table 1. Physical parameters of the two filter media tested
Filter media
C
D
Mass per unit area
Thickness Z (µm) Packing density af
(g m⫺2)
475
135
1400
400
the difference in the performance of a filter medium
between initial state and after a certain number of
clogging and cleaning cycles.
METHODS
The test rig employed for the various experiments
is shown in Fig. 2. Its central section comprised a
filter holder which presents a filtration surface of 130
mm diameter. At one end of the rig, a rotary brush
generator (Palas RBG 1000) supplied with filtered
and dried air generated alumina particles (Abralis
OPTI 3S). The generation mass flow rate was about
10 mg min⫺1. The particle size distribution of the
aerosol at the outlet was determined by sampling on
an isopore membrane and analysis using a Coulter
counter (Multisizer). The characteristics obtained
were the following: volume median diameter 2.6 µm
(MMAD=4.8 µm) and geometric standard deviation
1.5. The total air flow rate in the installation was held
0.246
0.225
Arithmetic mean
diameter (µm)
Standard deviation
of the fibre size
distribution
14.9
13.9/0.25
0.4
10/0.2
constant by means of a control and by-pass system
to obtain a filtration velocity of 2 cm s⫺1 (66.3 Nm3
h⫺1 m⫺2). Two sampling lines upstream and downstream of the filter medium comprised a profiled
probe coupled to an elbow (⭋ 16 mm), a valve and
a duct linked to an optical counter (PMS Lasair 1001
or 310). The technique to determine the efficiency as
a function of the particle diameter was therefore the
same as that used to characterise the initial state of
the filtration media. Both sections of the filter were
fitted with pressure transducers that allowed continuous monitoring of the pressure drop of the filter.
The cleaning technique employed was a reverse
flow pulse pressure. To achieve this a cross-flow compressed air inlet was installed downstream of the filter. A time delay circuit was employed to set the
valve opening time at the fixed value of 0.4 s. The
overpressure recorded on the filter was 3080 Pa. Cleaning was triggered at a maximum pressure drop,
designated ⌬Pmax. Different tests were conducted: test
Fig. 2. Representation of the test rig to study the clogging and cleaning cycles.
118
S. Callé et al.
on filter C with cleaning at ⌬Pmax=600 Pa then 350
Pa, and test on filter D with cleaning at ⌬Pmax=350
Pa. The particle detachment produced by the reverse
flow pulse pressure led to a fall in pressure drop to
a value termed residual. The maximum pressure drop
for each medium was chosen partially on account of
their respective resistances.
For each of the two media tested, the protocol was
the following.
The pressure drop was monitored during the cycles
until stabilisation of the residual pressure drop
occurred.
The efficiency was determined during the cycles
after the first and the fourtieth cleaning and compared
to the previously measured initial fractional
efficiency.
RESULTS
Observation of each medium during the cycles
revealed layered detachment of the cake particles formed on the surface of the filter. Koch et al. (1996)
also referred to this type of observation following
pressure-pulse cleaning. It is called patchy cleaning.
The higher the quantity of particles collected, the
larger the cleaned surface area. In contrast, as the
cycles progressed it appeared that the average surface
area of the cleaned zones diminished.
Change in pressure drop with the cycles
Figures 3 and 4 show the changes in pressure drop
of media C and D respectively with the cycles during
cleaning at a maximum pressure drop of 350 Pa
(40 cycles).
The first collection cycle of each of the new filters
followed three stages. The particles were firstly collected inside the filter, which led to a small increase
in pressure drop. A transition zone was then observed
during which the particles began to accumulate on the
surface of the filter medium until the formation of a
layer that then increased linearly in the same way as
the pressure drop. These collection cycle mechanisms
have been studied in details by Lee and Liu (1982),
Fig. 3. Change in the pressure drop of medium C as the cycles
progressed.
Fig. 4. Change in the pressure drop of medium D as the cycles
progressed.
Brown (1993), Walsh et al. (1996), Japuntich et al.
(1997) and Thomas et al. (1999).
The change in pressure drop had a specific pattern
during the collection periods following one cleaning
cycle. In contrast to the first collection stage of the
new filter, the pressure drop firstly increased rapidily
then rose more or less linearly. This phenomena,
observed by Humphries (1981) on woven cloths and
needlefelted fabrics clogged by an aluminium oxide
powder, is a characteristic of patchy cleaning. Observations of the filter medium after cleaning represented
in Fig. 5 tends to confirm this hypothesis.
The flow of particles through the cleaned gaps has
a much higher velocity than around the remaining
layers. The pressure drop therefore increases very
quickly.
From this hypothesis, a model of the change in
pressure drop with time after one cleaning cycle was
constructed. The aim is to give a way of characterising each experimental cleaning cycle in term of pressure drop according to the hypothesis of patchy cleaning. The key parameter of the calculation is the
cleaned fraction, g, defined as the ratio of the cleaned
surface area to the total filtration surface area. For a
given value of g and by dividing the cleaned filter
into two systems, one formed by the cleaned parts of
Fig. 5. Patchy cleaning illustration.
Clogging and cleaning cycles of filter media
119
the filter the second by the non cleaned zones, the
pressure drop and filtration flow rate were determined. The pressure drop calculation models are
respectively that of Darcy (1) for the new filter and
that of Kozeny–Carman (2) for the dust cake.
3
⌬Pf = m·U·Z·
16af 2·(1 + 56·a3f)
⌬Pg = m·kck·U·A2p
r2f
ag W
·
(1⫺ag)3 rp
(1)
(2)
From the data relative to filter medium C, curves
of the change in pressure drop after cleaning for different values of the cleaned fraction could be
deduced. They are shown in Fig. 6. The considerable
influence of the cleaned fraction was observed.
For both types of medium, the residual pressure
drop after each cleaning cycle increases as the cycles
progress until reaching a more or less constant value.
Different hypotheses can be put forward to account
for this pattern. This increase in residual pressure
drop could be explained by two coexisting phenomena. The quantity of particles inside the filter medium
increases as the cycles progress while remaining a
non regenerable fraction. The pressure drop after each
cycle therefore increases up to a threshold corresponding to maximum filling of the filter. In addition,
the deposit on the surface of the filter medium is more
and more cohesive and thefore better resists the cleaning pulse pressure.
Figure 7 shows the changes in the residual pressure
drop of media C and D as a function of generated
mass per unit area, cleaned at a maximum pressure
drop of 350 Pa, as well as the trend curves fitted to
the experimental points. It would clearly appear that
the anti-clogging treatment of medium D facilitates
detachment of the particles. This therefore has a
residual pressure drop that increases much more
slowly than that of filter C. After cleaning, medium
D recovers, in terms of pressure drop, to properties
closer to its initial state than medium C.
Fig. 6. Change of pressure drop after cleaning as a function of
the fraction cleaned. The different curves represent different
values of g : 0.1 (top), 0.3, 0.5, 0.7, 0.9 (bottom).
Fig. 7. Comparison of the change in residual pressure drop for
medium C (䊐) and medium D (왖).
Comparative tests on medium C cleaned for two
maximum pressure drops (350 and 600 Pa) gave the
results shown in Fig. 8. The residual pressure drop
following cleaning at 600 Pa was always higher than
that following cleaning at 350 Pa. The explanations
put forward earlier also apply in this case. The later
the cleaning is carried out, the greater the quantity of
particles collected on the surface and inside the
medium and the greater the tendency towards particle
cohesion. The particle cake remaining on the surface
after cleaning may also have been more compressed.
These arguments support the hypothesis that the
quantity of dust particles that can potentially be cleaned varies inversely with the maximum pressure
drop.
Changes in filtration efficiency with the cycles
The filtration efficiency as a function of the particle
diameter was monitored at different stages of the first
collection cycle and just after the first cleaning cycle
of filter medium C. The pressure drop of the filter at
each of these points was respectively 18 (new state),
150, 300, 600 (maximum pressure drop) and 39 Pa
Fig. 8. Comparison of the changes in the residual pressure drop
of filter C cleaned at two different maximum pressure drops:
⌬Pmax=350 Pa (䊏) and ⌬Pmax=600 Pa (䊊).
120
S. Callé et al.
Fig. 9. Change in the filtration efficiency of medium C in new
state at different steps of the first clogging for 0.25 µm particles
(t=0; 20; 33; 55 min) and just after the first cleaning (t=55
min).
(cleaned state). The results are shown in Fig. 9 for a
diameter of 0.25 µm, close to the minimum filtration
efficiency. The filtration efficiency increases very
quickly at the start of the collection cycle, reaching
values close to 100% when the pressure drop of the
filter reaches its maximum value. Then, just after cleaning, the efficiency falls to a value slightly above its
initial value. The efficiency therefore tends to
increase to a limit value during the cycles and with
the increase in the quantity of particles remaining.
The fractional filtration efficiency at the end of 40
cycles (when the residual pressure drop had reached
a more or less constant value) was compared to the
initial efficiency for each of the two media. The
curves are shown in Fig. 10 for filter medium D cleaned at 350 Pa. The first conclusion is that the separation efficiency of the filter medium increases until
reaching, for particles with an average diameter
between 0.15 and 10 µm, values greater than 90%.
At the end of 40 cycles, the fractional efficiency
showed much less variation. The efficiency was more
than 90% at all sizes, but it must be remembered that
the penetration (100% minus efficiency) would still
vary several-fold with particle size.
Fig. 10. Filtration efficiency of medium D as a function of
particle size in initial state (䊊) and in cleaned state after 40
cycles (왖), ⌬Pmax=350 Pa.
Comparison with the separation efficiency of an industrial dust separator
A comparison between the efficiency as a function
of the particle diameter of a medium tested on the
test rig shown in Fig. 2 at the end of the cleaning
cycles and that of an industrial collector for wood
dust operating at a filtration flow rate of 12 000 m3
h⫺1 was carried out. The medium C studied forms
part of the industrial filter whose efficiency was measured on site.
The technique used to determine the separation
efficiency of the industrial filter was based on
employing a fluorescent tracer aerosol generated
upstream of the filter (Bémer et al., 1998). The concentrations upstream and downstream of the separator
were measured in particle size ranges using an Andersen 1 CFM cascade impactor and allowed determination of its filtration efficiency.
The separation efficiency of filter C, cleaned at 350
Pa, after 40 cycles and that of the industrial filter are
shown in Fig. 11. Both curves are very close. These
results, although obtained in non identical conditions
(flow rate, concentration, nature of the particles, etc.),
show the extent to which the collection and cleaning
cycles must be taken into account to determine correctly the real separation efficiency of the filter
medium used in industrial separators. Finally, they
show that the tests carried out on new filter medium
employed in industrial filtration units, in other words
without taking collection and cleaning cycles into
account, are meaningless.
CONCLUSION
This work forms part of a more general occupational risk prevention study encompassing the
monitoring of the performance of industrial filtration units.
Monitoring the pressure drop as the cycles progress
has shown that from a certain number of cycles
onwards, the pressure drop of the medium changes in
a more or less constant manner. The separation
efficiency as a function of the particle diameter
Fig. 11. Filtration efficiency of medium C cleaned at 350 Pa
after 40 cycles (왖) (measured on the test rig) and of an industrial filter (䊐) (measured on site) as a function of particle size.
Clogging and cleaning cycles of filter media
measurements taken at different stages of the cycles
(new state, clogged then cleaned state) have demonstrated the high filtration performance of the medium.
This work has finally highlighted the hypothesis of
patchy cleaning for high dust concentration conditions on both filter media tested.
Later other filter media with variable clogging
properties will be tested. The study will be followed
by a parametric experimental study primarily
focussed on particle size, filtration velocity and cleaning characteristics (intensity, direction of the nozzle).
Modelling of the pattern of pressure drop with the
cycles is also currently in progress to further examine
the hypothesis of patchy cleaning. Finally the last aim
is to carry on with the in situ measurements to
improve the test method developped for filter media.
REFERENCES
Bémer D. Exploitation des propriétés physiques d’un aérosol
pour la mesure de la taille des particules et de leur concentration—applications. Thesis, Université de Paris XII, 1990.
Bémer D, Regnier R, Calle S. Measurement of the fractional
121
efficiency of an industrial dust separator using a fluorescent
aerosol. Journal of Aerosol Science 1998;29(S1):S1077–8.
Brown RC. Air filtration—an integrated approach to the theory
and applications of fibrous filters. H.S.E. Oxford: Pergamon
Press, 1993.
Humphries W. Influence of cloth structure on dust dislodgement from fabric filters. Powder Technology 1981;28:189–
94.
Japuntich DA, Stenhouse JIT, Liu BYH. Effective pore diameter and monodisperse particle collection of fibrous filters.
Journal of Aerosol Science 1997;28(1):147–58.
Koch D, Seville J, Clift R. Dust cake detachment from gas
filters. Powder Technology 1996;86:21–9.
Lee KW, Liu BYH. Experimental study of aerosol filtration by
fibrous filters. Aerosol Science and Technology
1982;1:35–46.
Norme européenne Pr EN 779 Revision. Particulate air filters
for general ventilation—determination of the filtration performance, 1998.
Thomas D, Contal P, Renaudin V, Penicot P, Leclerc D, Vendel J. Modelling pressure drop in HEPA filters during
dynamic filtration. Journal of Aerosol Science
1999;30(2):235–46.
Walsh DC, Stenhouse JIT, Scurrah KL, Graef A. The effect of
solid and liquid particles loading on fibrous filter material
perfromances.
Journal
of
Aerosol
Science
1996;27(S1):S617–8.