You are on page 1of 6

THE EFFECT OF BOTTOM ROUGHNESS ON

THE MINIMUM AGITATOR SPEED REQUIRED


TO JUST FULLY SUSPEND PARTICLES IN A
STIRRED VESSEL

A. Ghionzoli1,2, W. Bujalski2, R. K. Grenville3, A. W. Nienow2, R. W. Sharpe2 and


A. Paglianti4,
1
Laboratory of Process Equipment, Department of Chemical Engineering, Industrial Chemistry and
Materials Science, University of Pisa, Pisa, Italy.
2
Department of Chemical Engineering, The University of Birmingham, Edgbaston, Birmingham, UK.
3
Du Pont Engineeering, Wilmington, DE, USA.
4
Department of Chemical, Mining and Environmental Engineering, University of Bologna, Bologna,
Italy.

Abstract: In this work, the effect of the vessel bottom roughness on the suspension of solid
particles in stirred tank reactors is investigated. The experiments were performed in a baffled
vessel, which was mechanically stirred with a 458 pitched blade turbine. In order to evaluate
the influence of the bottom roughness on particle suspension, four bottoms of different
roughness and 8 different sets of spherical particles were used. The density of the solid particles,
rs, ranged from 2500 kg m23 to 8743 kg m23 and they were characterized by narrow size distri-
butions with a mean diameter, dp, from 128 mm up to 1850 mm. Measurement of the minimum
impeller speed for ‘just complete suspension’, Njs, showed that the roughness of the bottom
had a significant influence. The precise effect depends on the particle size compared to the
size of the roughness elements and to the Kolmogoroff microscale, lK.
Keywords: particle suspension; mixing; bottom roughness; settling velocity, Kolmogoroff
scale, lK.


Correspondence to:
Professor A. Paglianti, INTRODUCTION similar to those of Zwietering (1958). His
Department of Chemical, work implies that the large scale eddies are
Mining and Environmental For solid– liquid systems in stirred tanks, one most important in that different impeller/tank
Engineering, Viale
Risorgimento 2, I-40136
of the most important parameters is the agita- configurations have very different (1̄T )js
Bologna, Italy. tor speed at which all the particles are just values. This difference has been signified by
E-mail: fully suspended, Njs. Understanding the the parameter, S (Nienow, 1997a; Zwietering,
alessandro.paglianti@ suspension mechanism is also important. In 1958) that allows Njs for each impeller/tank
mail.ing.unibo.it
general, two main mechanisms have been configuration to be estimated from
DOI: 10.1205/cherd06152 considered (Nienow, 1997a): either processes
in which suspension depends only on the N js ¼ S½g(rs  rl )=rl 0:45 dp0:2
0263–8762/07/ interaction between the particles and turbu- X 0:13 n0:1 =D0:85 (1)
$30.00 þ 0.00 lent isotropic eddies of the size of the particles
Chemical Engineering
or those in which it depends on the interaction Neither theoretical approach nor the S par-
Research and Design between particles and large-anisotropic ameter considers the impact of the roughness
eddies. In the former case, the process is of the tank bottom to be significant. However,
Trans IChemE, influenced primarily by the mean specific
Part A, May 2007 recently, Thorpe and Stevenson (2003)
energy dissipation rate, (1̄T )js , to some attempted to relate the particle suspension
# 2007 Institution extent independent of impeller type while in phenomena in stirred tanks to that proposed
of Chemical Engineers the latter case, (1̄T )js is strongly dependent for hydraulic conveying in pipelines and also
on the impeller type. compared literature correlations for the two
In the last 50 years, many papers have configurations. These authors concluded
been published on this topic and these that just suspended conditions occur at a
papers have been reviewed in detail certain level of shear stress at the base of
(Nienow, 1997a). This review showed that the vessel on the particle due to turbulence,
most of the published papers gave results relative to the shear stress required to bring

685 Vol 85 (A5) 685–690


686 GHIONZOLI et al.

it into motion by overcoming the friction between it and the Table 1. Vessel and impeller characteristics.
base. If the latter aspect is important, it follows that, as the Tank diameter, T (mm) 202
roughness of the vessel base changes both shear stresses, Liquid height, H (mm) 202
it should also impact on the process of solid suspension. Volume, V (m3) 0.0065
Therefore, it was decided to choose an impeller/tank con- Impeller type PBT-6b
Clearance, C (mm) 50.5
figuration for which S is available and that is popular in indus- Impeller diameter, D (mm) 67.41
try for solid/liquid systems because it has a low (1̄T )js, and Baffles width, B (mm) 20.2
measure Njs for different levels of base roughness. It was
also hoped to establish a criterion to predict the effect, if
any, of the bottom roughness on the pick-up of solid particles
in stirred tanks. Solid Particles
Eight different types of particles were used covering three
densities, (bronze, lead glass and soda glass particles) and
a range of sizes from 128 mm to 1850 mm. Table 3 lists the
details including the density difference based on the use of
EXPERIMENTAL APPARATUS distilled water at room temperature.
Vessel and Impeller
All the experiments were carried out in a cylindrical vessel, Power Measurements
schematically shown in Figure 1, which was vertically
The power input was measured for the vessel filled with
mounted and open topped. The geometrical characteristics
distilled water. The power absorption of the impeller was
of the vessel and of the impeller used in present work are
obtained by measuring the torque exerted onto the shaft.
shown in Table 1.
The shaft torque was measured using a Visco-Mix DUO,
In all cases, the vessel was filled with distilled water to a
a torque meter produced by Coesfeld Mess-Technik
height, H, equal to the tank diameter, T, and the impeller
(Dortmund, Germany).
was a 458 pitch, six-bladed turbine, diameter D ¼ 1/3T,
placed with a clearance, C ¼ 1/4T off the base. The vessel
was equipped with four equally-spaced baffles of 0.1T. A Impeller Speed Njs: Visual Observation
very narrow space was left between the wall and the baffles
to facilitate the motion of fluid near the walls. The base of the The minimum impeller speed for ‘just complete suspension’,
vessel was flat and made of Perspex. Some of the exper- Njs, was determined visually as was done by Zwietering (1958)
iments were done with the Perspex bottom surface, as in and most other workers (Nienow, 1997a). For the transparent
many of the earlier studies, which can be considered a Perspex bottom, a light was set beneath it and the observer
smooth bottom. The others were performed with abrasive looked from the side. Since the rough bottoms were not trans-
sheets glued on the base. These sheets consisted of grains parent, in that case, the vessel was illuminated and viewed
of aluminum oxide on a paper backing. In Table 2, the main from the side and it was enclosed in a square water trough in
characteristics of the abrasive sheets used are shown. order to prevent distortion. However, observation was rela-
tively easy because the last point of suspension was in the
fillets between the base and the cylindrical wall. According to
Zwietering’s criterion, Njs is the minimum speed when the
particles do not remain at rest on the bottom for longer than
one or two seconds before being lifted up so that the suspen-
sion can be considered just complete.

EXPERIMENTAL MEASUREMENTS AND DATA


ANALYSIS
Power Measurements
The power measurements were carried out in water in
order to check the effect of the bottom roughness on the
power draw of the impeller and Figure 2 shows the power
number, Po calculated from it. For any one base, Po was
essentially constant in the fully turbulent region with the fall
occurring at Reynolds number of about 45 000 due to surface

Table 2. Rough bottom characteristics.

Abrasive sheet Average grit size, e (mm)

P120 127
P60 268
P36 524
Figure 1. Geometrical configuration of the stirred vessel.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A5): 685– 690
THE EFFECT OF BOTTOM ROUGHNESS ON THE MINIMUM AGITATOR SPEED 687

Table 3. Solid particles specifications.

Density difference Average size


Particles Acronym (kg m23) (mm)

Glass powder GP128 1500 128


Lead glass ballotini LGB215 1950 215
Lead glass ballotini LGB300 1950 303
Lead glass ballotini LGB460 1950 460
Blue lead glass BLGB655 1950 655
ballotini
Spherical bronze SBP655 7743 655
powder (89/11)
Lead glass ballotini LGB1090 1950 1090
Lead glass ballotini LGB1850 1950 1850

aeration. Also, Po was the same within experimental error for


all four bases, with a value of 1.7, close to the value reported
earlier (Nienow, 1998). It can be concluded the base rough-
ness has a negligible effect on the turbulent power number.

Solid Suspension Measurements---Effect of the


Bottom Roughness
The minimum impeller speed for just complete suspension,
Njs, was determined in 32 different combinations where either
the vessel bottom or the particle types were varied. The
amount of solid particles stirred in the vessel, X, was also
varied. The lower limit was the minimum that allowed an
accurate assessment of Njs and depended on both the par-
ticle type and the bottom used while the upper limit was
held to 2% by weight at most. Generally, about 10 concen-
trations were used for each bottom/impeller combination.
All the experiments were repeated twice by the same obser-
ver and by comparing the measurements taken under the
same conditions, it was found that the reproducibility of the
results was 1.5% on average and the maximum difference
was less than 5%.
Figure 3 shows four examples of Njs as a function of
concentration X covering the smallest to the largest particles
used. Inspection of these figures shows that with the smallest
particles [Figure 3(a)], the smooth base required the highest
speed whilst with the biggest particles [Figure 3(d)], the
roughest base required the highest. Indeed, in general, the
order of the speeds for the different bases varied with particle
size. However, in each case in Figure 3 and for the other par-
ticles (data not shown), the effect of concentration was very
similar and was very close to that given in equation (1).

Figure 3. Some examples of Njs versus concentration for different


particles with base roughness as a parameter: (a) GP 128; (b) LGB
300; (c) BLGB 655; (d) LGB 1850.

In order to see the effect of bottom roughness on solid


suspension, S was first calculated for the Perspex base tank
using equation (1) to give SS for a smooth base. Figure 4
shows the values of SS obtained as function of particle size.
Size does appear to be having a small effect but the average
SS value of 4.4 + 10% is in good agreement with that reported
earlier for this geometry (Ibrahim and Nienow, 1996).
Figure 2. Effect of bottom roughness on power number. Subsequently, the parameter, DS/SS ¼ (SR 2 SS)/SS was

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A5): 685–690
688 GHIONZOLI et al.

calculated where SR referred to a certain rough base with SR


and SS each referring to the same particle size. This
approach allowed the effect of the bottom roughness only
to be developed while neglecting the small influence of par-
ticle size that was revealed by the data of Figure 4.
Figure 5 shows the effect of the particle size on DS/SS as a
function of particle size. The data allows the identification of
two different regimes, dependent on the sign of DS/SS. All the
data for the three different bottom roughnesses follow a similar
trend. Thus, for the smallest particle sizes (128 mm), DS/SS has
negative values, i.e., the soda glass powder, the smallest par-
ticles used, seems to be suspended much more easily from
rough bottoms, with the rougher bottoms being more effective.
Figure 5. Effect of the bottom roughness on DS/SS.
With increasing particles size, DS/SS increases to take positive
values and reach a maximum; i.e., particle suspension is more
and more hindered. Eventually, DS/SS decreases and
approaches the zero reference line. Thus, with the biggest it does not give any physical insight into the phenomenon.
particles, the influence of the rough surface is least. In order to achieve a better understanding of the results, it
The effect of the different levels of roughness as indicated in is useful to consider several studies which have been carried
Figure 5 can be considered in more detail. Thus, the least out both on the effect of eddy size on solid suspension in a
rough bottom (P120) gives the highest increase in DS/SS for stirred tank and on the effect of roughness on turbulence
particles with a mean diameter of 303 mm, the middle rough- structure in a channel.
ness (P60) shows a maximum in DS/SS at 460 mm and the Firstly, the interaction between the small eddies in the iner-
roughest one (P36) has a maximum at 1090 mm. Moreover, tial sub-range of turbulence and the particles plays an import-
P120 does not have significant effect on suspension of par- ant role in the pick-up mechanism, as pointed out by Baldi
ticles bigger than 460 mm, while P36 approaches a zero et al. (1978). According to those authors, only eddies with a
value of DS/SS much more slowly and its influence is still size (defined by the local 1T ) of the order of the particles
notable for particles of 1850 mm. Finally, the maximum value are involved in causing solid suspension. If the eddies are
of DS/SS for the roughest bottom (P36) is much higher than smaller than dp, they do not contain sufficient energy to lift
that for the least rough (P120) and for the middle roughness the particles, while if the eddies are larger, they approach
(P60) an intermediate value is found. Thus, with decreasing the base too seldom to lift them completely.
bottom roughness, it would seem that, as might be expected, However, the eddy sizes depend not only on local 1T from
the influence on solid suspension becomes less strong. Kolmogoroff’s theory. Close to the bottom, where the particles
Based on the analysis of Figure 5 linking the maxima in are settled, eddy sizes are also influenced by the wall rough-
DS/SS to the different degrees of roughness and the particle ness. Cui et al. (2003), who studied the flow in a channel with
size, it was decided to combine the latter two parameters into spaced ribs to create roughness showed, using large-eddy
a dimensionless roughness, dp/e, where e is the mean grit simulation, that surface roughness elements have a great
size of the bottom (see Table 2). The resulting data have influence on the flow structure near the wall; in particular,
been plotted in this way in Figure 6. It can be seen that this the flow structures depend on the roughness height. Thus,
method of handling the data has substantially brought it all the results by Cui et al. (2003) suggest a possible reason
together and that for all the bottoms, the ratio DS/SS reaches for the influence of the roughness of the bottom on the
a peak at a value of dp/e of about 2. suspension of particles; the eddy sizes close to the bottom
surface have a length scale of the order of the roughness
DISCUSSION elements and, depending on their size, they affect, in different
ways, the suspension of the particles.
All the analysis of the data so far shows that the bottom However, in addition, there is also evidence that the level
roughness has a significant effect on solid suspension, but of turbulence in a stirred vessel, expressed as 1̄T , affects

Figure 4. Effect of particle size on the Zwietering constant SS for a


smooth base. Figure 6. Effect of the dimensionless roughness, dp/e, on DS/SS.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A5): 685– 690
THE EFFECT OF BOTTOM ROUGHNESS ON THE MINIMUM AGITATOR SPEED 689

the interactions between particles and fluid, as it provokes a


change in the flow field around the particle compared to the
case without turbulence. This change is indicated by the
different settling velocity, us, of particles in a turbulent fluid
compared to the terminal settling velocity, ut, in a still one.
This difference has been explained by Magelli et al. (1990)
and Brucato et al. (1998). They examined the suspension
of millimetre and submillimetre particles and found that the
ratio between the particle settling velocity in a turbulent fluid
and that in the fluid at rest varied in the range 0.2 –1 and
0.15–1 respectively. They were then able to attribute this
difference to changes in the particle drag coefficient, CD,
which determines us and ut in the two flow fields. In particular,
they expressed the effect of turbulence on CD in terms of lK Figure 7. Effect of ratio lK/dp on DS/SS.
and showed that when the turbulence scale, lK, is small
enough in comparison with the particle size, i.e., lK/
dp , 0.1, the turbulence causes an increase in the particle and being suspended from it by the flow near the bottom.
drag coefficient. On the contrary, when the ratio lK/dp is Here the turbulence interacts with the particles, enhancing
greater than 0.1, i.e., when the particle diameter is less the drag coefficient and decreasing their settling velocity rela-
than 10 lK, the eddies have no significant influence on the tive to a quiescent system. With increasing lK/dp, and there-
particles motion, i.e., the drag coefficient is not affected by fore increasing lK for a fixed particle size, the settling velocity
turbulence. increases until it reaches us, i.e., when it is equal to the
Therefore, because of the similarity between particle settling velocity in the quiescent fluid. However, as previously
suspension and settling velocity, it is interesting to analyse pointed out, close to the bottom of the vessel, the roughness
how the parameter DS/SS is related to the ratio between elements increase locally the size of the eddies (Cui et al.,
the Kolmogoroff microscale, lK, and the particle diameter, 2003), even though in this work, 1̄T remains essentially the
dp. The Kolmogoroff length scale is given by same. Therefore, according to this hypothesis, for a fixed par-
ticle size, an increase of the bottom roughness causes an
lK ¼ (n3 =1T )1=4 (2) increase of the size of the dissipating eddies close to the
bottom of the vessel. As a result, the settling velocity rises
where n is the liquid kinematic viscosity and 1T is the local and therefore the value DS/SS rises too. Support for this
specific energy dissipation rate. It would seem appropriate analysis comes from the fact that for the least rough bottom
to use a value of 1T for the base of the vessel but this is (P120), the mean increment of DS/SS, with respect to the
not known. However, local 1T values are proportional to the smooth bottom, is about 1.3%, for the next roughness
mean value, 1̄T and that parameter was used previously (P60), it is 13% and for the roughest (P36), it is about 20%.
(Magelli et al., 1990; Brucato et al., 1998). 1̄T is equivalent Such arguments, though more empirically-based, have
to the specific power dissipation throughout the vessel been used in the terminal velocity-slip velocity theory of
given by solid –liquid mass transfer (Nienow, 1997b).
If lK/dp . 0.1 so that the particle size is relatively small
1̄T ¼ P=rV ¼ PoN 3 D5 =V (3) relative to lK, its settling velocity is the same as its terminal
velocity in the still fluid. This is equivalent to saying that the
Because of the low solid concentration, the power number for particles move with the large anisotropic eddies, again an
the solid –liquid mixture has been considered equal to the argument that has been used for estimating mass transfer
power number of the single-phase (Figure 2). coefficients in solid–liquid systems (Nienow, 1997b).
Following the approach of Magelli et al. (1990) and Brucato
et al. (1998), in Figure 7, the dimensionless parameter, DS/
SS, has been plotted against the dimensionless parameter,
lK/dp. The experimental results show that at lK/dp  0.1 –
0.2, DS/SS is approximately constant for each and depends
significantly on the roughness of the base whilst for greater
values, DS/SS for all the bases converge and follow the
same downward trend. The observation of a change in beha-
vior at values of lK/dp of this order is in good agreement with
earlier work (Magelli et al., 1990; Brucato et al., 1998), an
example of which is shown in Figure 8 for the convenience
of the readers.
Building further on the above analysis, it suggests that the
influence of the turbulence on solid suspension varies
according to whether the value of lK/dp is . or . 0.1. If
lK/dp , 0.1, the suspension of solid particles depends on
the interactions between turbulent eddies and the solid par-
ticles. The latter are in dynamic equilibrium, falling from the Figure 8. Effect of ratio lK/dp on the ratio between settling velocity,
bulk of the tank towards the bottom under the action of gravity us, and terminal velocity, ut, (from Magelli et al., 1990).

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A5): 685–690
690 GHIONZOLI et al.

Therefore, in that range of lK/dp, the speed of just characteristic for large particles. It also showed that the wall
suspension, Njs, should not be influenced by the impact of collision frequency has a great influence on the horizontal
the roughness on the turbulent eddy sizes. However, as component of the particle velocity resulting overall in, for
suggested by Thorpe and Stevenson (2003), the protruding small particles, an increase in this component near the wall
roughness elements can provide particles with an upward and for large particles, a decrease.
impulse, thus making the pick up easier. This could explain Clearly, as proposed by Thorpe and Stevenson (2003),
why in this range of lK/dp, for the smaller particles, the particle-base interactions are important for the understanding
rough bottoms help solid suspension sufficiently to give of solid suspension in stirred vessels and this study is the first
DS/SS values that are negative. to specifically address that issue by looking at base rough-
Finally, it is worth noting, again as proposed by Thorpe and ness coupled with a number of different types of particle.
Stevenson (2003), that there is also probably some impact of However, though this corroboration of their hypothesis
the base roughness on the ability of particles to move across suggests that the Zwietering analysis has a fundamental
it and be lifted due to particle-base frictional stresses. Analys- weakness as it is lacking any consideration of this mechan-
ing such effects is difficult but the maximum in DS/SS at ism, the differences between the rough and the smooth
dp/e  2 might be linked also to that mechanism. Particles bases is not so marked that the Zwietering approach can
of just that approximate size might fit best in the interstices no longer be recommended. It is still probably the best
and be most difficult to move whilst much bigger ones than method available for determining this vital operating par-
that level of roughness would ‘sit’ on top of the protruding ameter, Njs, for solid –liquid agitated systems. On the other
elements. Smaller particles would not be so restricted hand, this study certainly suggests that more work might
and for very tiny ones, the enhancement noted above could usefully be undertaken involving the use of more extreme
dominate leading to negative values of DS/SS. shapes such as plates or needles where the change in the
A recent paper by Uby (2006) has also considered the stress balance might be much greater.
implications of earlier work on slurry transport in pipes and
channels for solid suspension on stirred vessels. His analysis
suggested that for very tiny particles, a rough surface created REFERENCES
some turbulence in the otherwise laminar boundary layer and Baldi, G., Conti, R. and Alaria, E., 1978, Complete suspension of par-
this led to lower velocities being required to transport such ticles in mechanically agitated vessels, Chem Eng Sci, 33: 21–25.
particles in rough pipes compared to smooth. Thus, he con- Brucato, A., Grisafi, F. and Montante, G., 1998, Particle drag coeffi-
cients in turbulent fluids, Chem Eng Sci, 53: 3295– 3314.
cluded, as shown here, that Njs would be higher in smooth
Cui, J., Patel, V.C. and Lin, C.L., 2003, Large-eddy simulation of
based vessels than in rough for small particles, i.e., negative turbulent flow in a channel with rib roughness, Int J Heat Fluid
values of DS/SS. Flow, 24: 372–388.
Ibrahim, S. and Nienow, A.W., 1996, Particle suspension in the
turbulent regime: the effect of impeller type and impeller/vessel
configuration, Trans IChemE, Part A, 74: 679– 688.
CONCLUSIONS Magelli, F., Fajner, D., Nocentini, M. and Pasquali, G., 1990, Solids
distribution in vessel stirred with multiple impellers, Chem Eng
Experimental results showing the influence of bottom Sci, 45: 615–625.
roughness, e, on the just-completely suspended speed, Njs, Nienow, A.W., 1997a, The suspension of solid particles, in Harnby,
of different particles have been presented. They indicate N., Edwards, M.F. and Nienow, A.W. (eds). Mixing in the Process
Industries (2nd edition) (paperback revision) 364– 393 (Butterworth
that suspension of the small particles, with diameter less Heinemann, London, UK).
than 10 lK is helped by the roughness, i.e., Njs is less than Nienow, A.W., 1997b, The mixer as a reactor—liquid/solid systems,
with a smooth bottom. With increasing size, dp, rough bot- in Harnby, N., Edwards, M.F. and Nienow, A.W. (eds). Mixing in the
toms begin to hinder suspension and this trend is followed Process Industries (2nd edition) (paperback revision), 394– 411.
(Butterworth Heinemann, London, UK).
until dp/e  2 for all particle sizes and bottom roughnesses.
Nienow, A.W., 1998, Hydrodynamics of stirred bioreactors, in Pohor-
Eventually, with further increases in dp/e, the influence of ecki, R. (ed.). Fluid Mechanics Problems in Biotechnology App
the rough surface became weaker and weaker. It was also Mech Rev, 51: 3 –32.
found that with decreasing bottom roughness, the effect on Sommerfeld, M., 2003, Analysis of collision effects for turbulent gas--
solid suspension becomes less strong. particle flow in a horizontal channel: Part I Particle transport
Int J Multiphase Flow, 29: 675– 699.
Interaction between the roughness elements, the particles Thorpe, R.B. and Stevenson, P., 2003, Suspension of particles from
and the turbulent field is proposed as a means to explain this the bottom of pipes and stirred tanks by gassed and ungassed
behaviour. Studies on the settling velocity of particles in stir- flows, Can J Chem Eng, 81: 351–359.
red vessels and on the effects of rib roughness in pipe flow Uby, L., 2006, New regimes and correlations for just suspended
speed, Proc 12th Eur Conf on Mixing, Bologna, 27– 30 June,
are useful for giving a physical interpretation of the mechan-
AIDIC, Milan, 249– 255.
isms involved in particle pick-up from a rough bottom. The Zwietering, Th.N., 1958, Suspending of solid particles in liquid by
importance of the Kolmogoroff length scale, lK, compared agitators, Chem Eng Sci, 8: 244 –253.
to dp has also been established with a critical value of
lK/dp  0.1 as in other aspects of solid –liquid mixing.
However, this behaviour also presents some similarities ACKNOWLEDGEMENT
with that predicted by the numerical calculations of Sommer- Financial support from Du Pont Engineering, Wilmington,
feld (2003) on the effect of wall roughness on particle trans- Delaware, USA is gratefully acknowledged.
port. His predictions showed that for small particles, wall
roughness induces a reduction of the particle-wall collision The manuscript was received 15 September 2006 and accepted
frequency, whereas it leads to a drastic increase of this for publication after revision 8 February 2007.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A5): 685– 690