Professional Documents
Culture Documents
net/publication/340025681
CITATION READS
1 543
4 authors, including:
Wang Fei
Tongji University
23 PUBLICATIONS 36 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Variable exhaust air volume system for workshop with multiple intermittent-releasing pollutant sources View project
All content following this page was uploaded by Ruiyan Zhang on 12 April 2020.
ABSTRACT
Adsorption filters in air cleaners are usually produced as honeycomb structure filled with
granular adsorbent, which can be seen as numerous parallel arranged fixed beds. In order to
investigate if the friction factor correlations for packed bed can be used in adsorption filter,
experimental study was conducted using adsorption filters filled with spherical, columnar and
crushed activated carbon from 2 to 20 mesh. The range of void fraction of adsorption filters is
0.29 to 0.92, and the range of particle Reynolds number is 8 to 1072. Comparing the
experimental data with the calculation results using 12 published friction factor correlations,
results show that large differences exist between the experimental data and the calculated
values. Considering such large differences, friction factor correlations for packed bed are
likely not suitable for prediction of pressure drop in adsorption filter.
INTRODUCTION
Adsorption is currently considered to be one of the most effective method for removing the
gaseous contaminations such as volatile organic compounds (VOCs) from air in building
environment and protecting the persons indoor (Zhang et al., 2011). As a kind of technical
application of adsorption, adsorption filters are commonly used in portable air cleaners and air
handling units in air conditioning system.
Not like the industrial waste gas treatment equipment designed to remove as much waste gas
as possible before air is released, a considerable portion of adsorption filters for indoor air
cleaning are designed to remove gaseous contaminants slightly but continuously from the
circulating air. Therefore, they have significantly lower resistance and lower removing
efficiency compared with the former (Chen et al., 2005). For selection of an adsorption filter,
its resistance and efficiency are important information, which obviously correlate with its
friction factor (Pei and Zhang, 2010; Mugge et al., 2005) and are fundamentally determined
by its structure (such as the size and shape of the filled granular adsorbent and the degree of
fullness).
The correlations of friction factor for fixed bed are usually adopted for calculating the
pressure drop in adsorption equipment with the structure similar with a fixed bed, namely a
flow path filled with granular matter. The pressure drop in packed bed can be correlated in
terms with the dimensionless friction factor, as
(1)
where is pressure drop, is dimensionless particle friction factor, is fluid density, is
superficial velocity and is particle diameter.
Most of published equations for calculating pressure drop in fixed bed can be easily converted
the form of correlations of friction factor with Reynolds number and several structural
parameters, as shown in Table 1.
Table 1. friction factor correlations for the main range of in adsorption filters
Source Equation
Chilton and Colburn (1931) (2)
(3)
Carman (1937) (4)
Rose (1945) (5)
Rose and Rizk (1949) (6)
Ergun (1952) (7)
Hicks (1970) (8)
Brauer (1971) (9)
Tallmadge (1970)
(10)
(16)
Montillet et al. (2007)
(17)
for
for
*
surface elevation changes are about two orders of magnitude smaller than the local channel diameter
is void fraction of adsorption filter, is diameter of a unit in adsorption filter, and the
Reynold number is defined as
(18)
where is fluid viscosity.
Although extensive research has been carried out on pressure drop calculation of fixed bed,
the structural characteristics of most adsorption filters make them more extreme cases
compared with the applicable range of these correlations. Firstly, the diameter ratio between
the flow channel units and the adsorbent particles are relatively small, making the low
resistance area near the wall, namely the wall effect, could not be neglected. Secondly, the
granules filled in adsorption filters are usually irregular in shape and size, producing greater
randomness in the way of stacking. Thirdly, the Reynolds number of adsorption filters at their
normal working flow rate is relatively small while the void ratio can be really large in some
cases. So, there is no clear published evidence that which equation of the fixed bed friction
factor is the best for adsorption filters.
In order to solve the above-mentioned problem, in this paper, the resistance curves of the
adsorption filters with typical skeletal structure were experimentally measured in the range of
their common working flow rates, with emphasis on changing thickness, particle shape and
size distribution, to cover as many common adsorbents as possible. The results of above
experiment were compared with the predicted results using various friction factor equations,
and the overall deviation by those equations was investigated, to find if there is one best
suited for adsorption filters.
Since the roughness differenced by material is not included in all of the above friction factor
formulas, activated carbon was selected as the representative adsorbent in this experiment
because of its pervasiveness, and it has the largest variety of shapes, containing most of the
common shapes of other adsorbents. The physical properties of activated carbon used are
shown in Table 2.
The approximation of diameter using sphericity and screen size was introduced in the
experiment, because it is difficult to accurately measure the precise geometry of mass of
activated carbon particles. Particles of different screen size are obtained by sorting them
using sieves in Tyler standard. Then the particle diameter can be converted by the
following formulas (Levenspiel, 2014)
(19)
(20)
The void fraction of the adsorption filter is calculated by the following formula,
(21)
where is the total volume of the filled activated carbon particles, is the volume
of filter.
The tested adsorption filters have two different thickness, 1.5 cm and 0.8 cm, which are
within the usual thickness range. The volume of filter is obtained by filling 200-300 mesh
sand in the adsorption filter and measuring the volume of filled sand. The total volume of the
filled activated carbon particles was measured by volume displacement technique using water.
In order to eliminate the error caused by water adsorption, the activated carbon was pre-
soaked to saturated, and then the excess water on the surface was removed by tissue paper
before measuring.
The experimental apparatus is shown in Figure 1. The adsorption filters were vertically fixed
in the air duct at the direction perpendicular to the wind speed. The fans connected with
frequency converter and the orifice plate flowmeter worked together to obtain a certain
volume flow of air, making wind speed for each of the tested adsorption filter ranges from 0.1
to 2 meters per second. The pressure drop across the adsorbent filter was measured by a
micromanometer.
Since the measured resistance values ranges from 1 Pa to 491 Pa, which is largely differenced
with each other, the measured and calculated values are compared with relative errors but not
the absolute errors. The measured values are regarded as true values.
In Figure 2, the overall accuracy of these friction factor correlations is compared. The relative
errors for all cases of each formula are divided into five levels: 1. The relative errors are
greater than one order of magnitude (the calculated value differs from the measured value by
more than ten times), and the calculated values are greater than the measured values, 2. the
relative errors are greater than 30%, but less than one order of magnitude, and the calculated
values are greater than the measured values, 3. the relative errors are less than 30%, 4. the
relative errors are greater than 30%, but less than one order of magnitude, and the predicted
values are less than the measured value, 5. the relative errors are greater than one order of
magnitude, and the calculated values are greater than the measured values. The proportion of
the relative errors of each level in all prediction cases are marked in dark red, dark yellow,
green, light yellow and pink in Figure 2, respectively.
Figure 2. The overall comparison of relative errors of different friction factor correlations
Results show that big difference exists between the experimental data and the calculated
values. Friction factor correlations of Chilton and Colburn, Carman, Ergun, Hicks, Brauer,
Tallmadge, Macdonald and Eisfeld and Schnitzlein can predict the order of magnitude of
pressure drop in adsorption filter in more than 90% of all 558 cases. However, the ratio when
the relative errors are less than 30% is very small, only about 10%, or even less. And only the
correlation of Eisfeld and Schnitzlein has achieved the order of magnitude accuracy in all
cases.
The difference between predicted values is compared in Figure 3. For each case, the largest
predicted value is represented by a solid yellow line, the smallest predicted value is
represented by a yellow dashed line, and the corresponding measured value is represented as a
black line. The data sets are sorted by the measured values.
Figure 3. Measured pressure loss and the range of predicted pressure loss
It can be seen in Figure 3, large differences exist among the calculated values of different
friction factor correlations, which can be 2 orders of magnitude in the area of small pressure
loss. However, in most cases where the measured pressure loss is lower than 100 Pa, the
measured values lie in the middle of the band of the predicted values. When the pressure loss
is greater than 100Pa, the measured values tend to be smaller than the predicted values. This
phenomenon may be due to the enhancement of the wall effect at high air velocity,
considering the diameter ratio in the adsorption filter is usually smaller than the fixed bed,
while the wall effect is ignored or underestimated in friction factor correlations.
Figure 4. Pearson correlation coefficients between the relative errors and and .
The Pearson correlation coefficients of Reynolds numbers and relative errors are all negative,
and the absolute value are less than 0.4, which means for all correlations the larger the
Reynolds number is, the lower is the prediction accuracy, but the relationship between the
two is not significant.
On the contrary, the Pearson correlation coefficients of diameter ratios and relative errors are
positive, and the absolute values are relatively lager except for correlations of Chilton and
Colbum, Rose and Rose and Rizk, which means the larger the diameter ratio is, the higher is
the prediction accuracy, and for most correlations it has more significant influence than of
Reynolds number.
In a similar way, the accuracy of friction factor correlations of Chilton and Colburn, Rose and
Rose and Rizk increase when the void fraction increases, while the accuracy of correlations of
Carman, Ergun, Hicks, Brauer, Tallmadge, Kuerten, Macdonald, Eisfeld and Schnitzlein and
Montillet increase when the void fraction decreases. However, considering the prediction
accuracy of the first three correlations are relatively low among all correlations, the
improvement of the prediction accuracy seems to be limited when using them by large void
fraction.
CONCLUSIONS
This work experimentally investigates the pressure drop in different adsorption filters and
evaluates the applicability of 12 friction factor correlations in prediction of the pressure drop.
And the scope of application of each correlation is briefly discussed.
For all tests, the differences between the calculated values and the measured values are mostly
in an order of magnitude. But only in less than 10% cases the relative errors are less than 30%.
Such accuracy is not sufficient for engineering application. And the potential improvement of
accuracy is small, even if the friction factor correlations are selected according to the
operating conditions.
Therefore, in general, the friction coefficient correlations of the packed bed are likely to be
unsuitable for predicting the pressure drop in adsorption filters. It is necessary to establish a
friction factor correlation and corresponding resistance prediction formula more suitable for
adsorption filters.
REFERENCES