Powder Technology, 74 (1993) 179-195 179

Pressure fluctuations in pneumatic conveying systems

Shrikant V. Dhodapkar and George E. Klinzing

Chemical and Petroleum Engineering Department, University of Pittsburgh, Pittsburgh, PA 15261 (USA)

(Received June 24, 1992; in revised form November 3, 1992)

Abstract

The pressure fluctuations generated in pneumatic conveying of solids in a horizontal piping arrangement have
been studied in order to provide a technique for flow pattern identification and control of the pneumatic conveying
operation. Practitioners of pneumatic conveying have used flow noise production in pneumatic conveying as a
technique to diagnose the operation of such systems. The pressure fluctuations measured were placed in power
spectral density function (PSDF) format and related to the flow patterns visually observed. Distinguishing PSDFs
were observed for the various flow pattern regimes. The effect of particle size, system configuration, pressure
transducer type and acquisition parameters were explored. The analysis was carried out on-line so that control
strategies can be employed. The pressure tap spacing did not influence the spectra generated at the various
flow regimes. Static pressure transducers proved more reliable for homogeneous and stratified flows detection
while differential pressure transducers were preferred for the dune/slug flow condition.
-

Introduction figuration, transducer type (differential versus static),

data acquisition parameters, etc., which need to be
carefully addressed.. For vertical pneumatic transport
Solid-gas transport have an inherent ability to produce
Matsumoto and Harakawa [3] examined pressure fluc-
pressure fluctuations. This behavior can be seen clearly
tuations and solid concentration fluctuations using sta-
by observing the pickup and deposition of particles
tistical analysis of these parameters. Qualitative de-
from and to the bottom of the pipe. Pressure fluctuations
scription of the ‘changes in the probability density
can also be caused by the air supply, feeder, line
configuration and solids collector. Experienced prac- function of the pressure fluctuation showed a change
from a unimodal distribution at high gas velocities to
titioners of pneumatic transport have commented re-
peatedly on their ability to diagnose the behavior and a bimodal distribution near lower slug flow conditions.
condition of a transport system by listening to the flow The results of this study are qualitative and specific
noise generated in a piping system. In an effort to to their experimental system.
make this detection ability more widely accessible, the Previous plant experience, the works of Tsuji and
use of pressure transducers to sense pneumatic transport Morikawa and lvlatsumoto and Harakawa suggest the
flow noise has been employed in this. study. The flui- potential for using pressure fluctuations as a means of
dization field has seen many studies addressing pressure flow pattern identification in pneumatic transport. Gen-
fluctuations, however, little work has been seen in erally these studies point out the deficiency of efficient
pneumatic transport. Tsuji and Morikawa [l] have information manipulation in analysis of pressure fluc-
explored pressure fluctuations and flow pattern iden- tuations. An on-line analysis system at a minimum of
tification in pneumatic transport, with Zenz and Othmer cost would advance the ability one has to establish flow
[2] suggesting a relationship between the average pres- pattern identification and provide a means of flow
sure drops and flow patterns. control for greater operability and reduced cost.
Tsuji and Morikawa observed with their analysis of Flow pattern, observations are visual observations
pressure fluctuations and flow patterns in horizontal which strongly depend on the ability of the experi-
flow that the particle properties were important. Each mentalist to translate the visual observation into lin-
of the flow patterns were seen to be associated with guistic expression. A wide range of features can be
the characteristics spectral and statistical estimates of seen in these patterns and a large number of descriptions
the pressure fluctuations. This work, while pioneering would be necessary to account for all of them. In order
the way, was limited in its scope. There are various to reduce the uncertainty or vagueness of these ob-
other parameters such as pipe diameter, system con- servations, it is necessary to classify each of the signals

0032-5910/93/$6.00 0 1993 - Elsevier Sequoia. All rights reserved

180

into predetermined categories which can be called flow further flatten out and become one continuous layer.
patterns of flow regimes. The suspension above this layer during unsteady state
Some of the common flow patterns observed in is almost without particles since most of the particles
horizontal pneumatic transport are: accumulate inside the pipe. Finally, the gas velocity on
1. Homogeneous flow the top becomes high enough to ,prevent the particles
2. Stratitied from dropping out. This is characterized by a steady
3. Dense moving strand (at the bottom) pressure drop.
4. Strands on settled layer ‘Plugs or slugs’ refer to a moving packed bed in the
5. Blowing packets or clusters pipe which has the same diameter as the pipe. A
6. Dunes (blowing or moving) degenerate plug flow may refer to a situation when
7. Settled layer these plugs break and reform during their passage
8. Plugs or slugs through the system. Plugs or slugs are mostly observed
‘Homogeneous’ flow refers to uniform dispersion of during low velocity transport of Geldart A, C and
particles. There is no visual opacity gradient in the Geldart D type materials.
flow. It should be pointed out that some of the visually Sketches of the flow patterns for 55 micron glass
homogeneous flows are actually segregated. By splitting beads (powdery, Geldart A) and 450 micron glass beads
the flow into upper and lower halves, the solids flux (granular) further emphasize the differences in flow
were found to be different. The optical homogeneity patterns for fine and coarse materials (Fig. i). Although
will be taken as a sufficient evidence to classify the these sketches represent the possible flow patterns that
flow as ,‘Homogeneous’. may be seen, some of them might not be seen at all
‘Stratified’ flow refers to the segregation of particles in a given system due to component specific limitations.
near the pipe bottom. These particles did not drop out It is also not necessary for all the flow patterns to exist
of the suspension, and they continued to move with a in the flow pattern map of a given material. For example,
visible density gradient near the pipe bottom. A visual Geldart A type of material shows a strong tendency
observation from the bottom was found to be more to form a moving strand but rarely forms a stable slug.
helpful in stratified flow identification.
‘Dense moving strand’ refers to a continuous or
discontinuous moving layer of solids with a dilute sus- Experimental
pension above it. These particles have dropped out of
the suspension and continue to move along the pipe The principal aim of this study was simultaneous
bottom. No significant exchange of solids could be seen acquisition of pressure signals and visual flow pattern
between the layer and the dilute suspension above it. observations. The flow patterns in this pneumatic con-
This flow pattern is mostly observed for powdery ma- veying study were observed through sight-glasses placed
terials. at the feed point and at the steady state conveying
‘Blowing packets or clusters’ refers to a collection section of the pipeline. The feed point behavior is
of particles generally of the size of a dune but much helpful in understanding the feeder characteristics and
loosely bound. At high gas velocity, these clusters look often times the origin of flow patterns in the steady
like dense packets of solids moving quickly across the region. All pipeline connections are carefully aligned
system. At low gas velocity, one can see the cluster with no protrusions into the flow regime. The pipe
and its shape (dune like). was smooth copper tubing. The nature of pressure
The term, ‘dunes’ originates from similar flow patterns fluctuations depends on their origin. Typically in pneu-
observed in deserts. %o types of dunes have been matic conveying systems one or more sources are re-
observed in horizontal systems. The term ‘blowing dunes’ sponsible for a given pressure signal. It would be
refers to dunes which move along the pipe bottom impossible to quantify the individual contributions unless
because of continuous erosion of the front end and each effect can be identified and isolated. Flexibility
deposition of particles at the far end. The term ‘moving and generality of the system components was therefore
dune’ refers to a dune which moves along the pipe stressed.
bottom almost intact (like plugs) due to pressure drop It is necessary to enumerate the possible sources of
across it. pressure fluctuations before such an experimental set
‘Settled layer’ refers to a stationary layer of solids up can be conceived. Pressure fluctuations (both static
at the bottom of the pipe. The formation of steady and differential) can be generated in horizontal systems
layer may take as long as 15 min in some cases. In by:
the beginning, only isolated dunes can be found. Re- 1. Fluctuating solids flow rate
duced cross section at the dunes causes the particles - Due to unsteady flow from the feeder
to re-entrain and then settle out again. These dunes - Accumulation of solids at the feed point
 181

Pattern 1 .,
.“,“’ ”
.::. ..
:. .,
Pattern 1
Homogeneous Flow _. ..,..... : : .‘.‘.’‘. Homogeneous Flow
.

......... ..... ...... .......

...... ......................
................................... ............
............................................... Pattern 2 ............... .......... ........
................... Pattern 2
Dense Moving Strand ....................... ....
.......
..:
.........
..A?
.7 ...
.::. ............
....
i...
.:
..I _ ....
.......
Y,.,.,
;::.;
.:.:.,
“” .:
...............................:z
.. ....
.................
,.:.:
.,.;
“” i. ..........
.....
7..:...;:
.: ....
_ ..........
.*z.:<.:.
.,.I,
2,.%.2:
.........................
..............
......
.....
,~, ,.Z<
.....
..:
.. .........
.:.::.:,;,,,;
...
;..y.( ..:..:
........
. ........
.:::\, .;,
.................
...........
.u:‘!:‘::wc Stratified
..v..s
... .._. .....................
”........................
....‘-.i..~~~....._
.._
.................................. B’,::
::...>:
.....
..:.:.x
......
.6::
...z.
z.:r .,.:....:
“:.;.: ...........................................

Pattern 3 Pattern 3
Strands on Settled Layer Blowing Pat ke ts

Pattern 4 Pattern 4
Blo$&unes Blowing Dunes
<slugs>

Pattern 5 Pattern 5
Settled Liver Settled Layer

Fig. 1. Sketches of the flow patterns for 55 and 450 micron glass beads.

Relative contribution to the total fluctuations depends In order to obtain an air supply which contributes
on the departure of operating conditions from saltation. insignificantly to the overall pressure fluctuations, com-
2. Air supply pressed house air was the obvious choice. However,
- Due to the nature of air supply (e.g. pulsations due certain problems associated with compressed house air
-to rotating lobes in Roots blower) had to be overcome.
- Pulsations due to by-pass valves and relief valves 1. House air exhibits pressure surges which can cause
in the air supply systems non-stationarity of the signal.
3. Material transport 2. The pressure has to be reduced from 80 psig to 5
- Interaction of particle motion with the turbulence psig without causing ‘breathing/chattering’ problems due
structure. to the regulator.
4. Line configuration 3. The gas flow rate had to be steady and independent
- Flow around bends of back pressure.
- Inclined pipes and combination of horizontal and A combination of regulators, surge tank and valves was
vertical pipes found to be the most optimum. Various other config-
5. Collector characteristics urations were tried and were not found to be as suitable.
6. System abnormalities Preliminary investigations indicated that the rotameter
-Such as leakage or broken components did not introduce any pulsations of its own. Replacement
The primary objective was therefore to eliminate these of the glass section with a copper section did not show
effects as much as possible so that pressure fluctuation any significant change in the nature of pulsations for
due to the transport of solids could be studied. single phase air flow. However, existence of sudden
diameter changes in the air supply lines were found
to increase the fluctuations.
Pneumatic conveying system-iayout
The layout of the piping for the 0.0254 m I.D. system Solids delivery system
and the 0.054 m I.D. horizontal systems is shown in Solids were introduced into the system using a screw
Fig. 2. The layout was kept purely horizontal, which feeder (Vibrascrew live bin type) with a helical hollow
means that the center line elevation of the pipe at the screw. The solids delivery could be adjusted by changing
feed and collector is exactly the same. The pipeline the screw RPM. This feeder tank had to be pressurized
was securely fastened on a Unistrut frame to prevent to 5 psig if flow rates in excess of 250 kg h-’ were
any misalignment due to vibrations. The glass section desired.
(viewing section) was connected using flanges. Extreme Feeders introduce their characteristic signature to
care must be taken to ensure that the inside diameter the nature of fluctuations. A rotary feeder or screw
of the copper pipe and glass tube are flush at the feeder can feed pockets of solids, causing pressure
flange. The total length of the system was about 30 fluctuations which are cyclic in nature. The frequency
m. of these cycles depends on the RPM of the feeder and
182

I-T- RECYCLE HOPPER

r DATA
ACQUISITION
SYSTEM

SCREW FEEDER

1 COLLECTOR/SEPARATOR
L______l
Fig. 2. Layout of the piping of the 0.0254 m i.d. and 0.054 m I.D. systems.

the ability of the conveying air to pick up the solids alternative was found. The end of the barrel of the
at the feed point. The magnitude of these pulses is so screw feeder was partially covered with duct tape. Partial
large that even for extremely dilute flow, large pressure blockage at the exit causes the solids to compact, and
oscillations can be obtained. The pressure fluctuations thus creating a seal.
due to dilute phase flow are over-shadowed by these
pulses. The magnitude of such oscillations has been
found to decrease with its distance from the feeder, Feed tee configuration
however, limitations on the maximum length of the Various feed tee configurations were probed. It was
system compels us to devise a more reliable non-pulsing found that in the absence of venturi at the feed point,
feeder. the unstepped feed tee affects the system behavior
The following modifications were made to the feed significantly due to solids deposition at the feed point.
point to achieve near perfect steady flow (450 micron), For the sake of uniformity, stepped tee coupled with
an extra length of tube was attached after the screw a modified venturi at the pickup point was used. This
such that a sump of solids dampened the feed fluc- prevented any unwanted build up or accumulation of
tuations. The solids then dropped straight into the feed solids.
tee.
Fine material (55 micron) could be fed with constant
flow rate by pressurizing the feeder and using it as a Collector
blow tank. The screw was shut off. The solids flow rate A cyclone type of collector (0.6 m i.d.) was designed
was controlled by altering the size of the holes in the such that the inlet to the collector is at the same level
screen placed in the vertical section of the feed con- as the horizontal pipe in order not to introduce fluc-
figuration. It is necessary to tap the column of solids tuations in the conveying line. The solids simply drop
occasionally to ensure no compaction takes place. The into the collector and do not cause any additional back
flow remains steady, constant and uniformly distributed pressure.
across the pipe. The conveying line was extended to various distances
Screw feeders have air leakage through the scre% inside the collector to study the effect of exit config-
back into the feed hopper. This backflow’of air causes uration on the pressure fluctuations upstream. The
a reduction in the bulk density of the material, Since actual configuration of the collector did not affect the
the screw feeders are volumetric devices, reduction in fluctuations.
the bulk density results in the reduction of mass flow
rate. Typically in Geldard A type material (with low
U,,), small amount of leaks can cause bubbling in the Piping
screw which results in erratic flow rates. To overcome The piping for two phase section was copper tubing
these problems, either a variable pitch screw (with solid with inside diameters of 0.0254 m and 0.05 m. Copper’s
center) is used or a flapper valve can be put at the good conductivity reduces the electrostatics in the sys-
end of the barrel. Since neither of these solutions could tem. Morris Quikon couplings were used to connect
be implemented due to the physical constraints, another two sections of pipe.
 183

Location of pressure tap pressure transducers and changes in static pressure

From preliminary investigation it was found that for signals as a function of distance.
homogeneous flow the pressure signals obtained from For each tap, a Swagelok male adapter was soldered
a tap located on the top is same as that located at onto the pipe and then a hole (l/32 in dia.) was drilled.
the bottom of the pipe. At or near saltation, the signals Visual inspection confirmed the absence of burrs inside
are significantly different because the bottom taps tend the pipe.
to get clogged. Since clogging can cause indeterminacy
in the measurement of the pressure signal, no further
attempt was made to measure the pressure signals at
the bottom of the pipe. Pneumatic conveying system - measurements
Two types of pressure fluctuations can be recorded,
namely, static pressure and differential pressure. The
static pressure signal may depend on the location of
Pressure measurement
the tap with respect to the feeder and collector and
Omega pressure transducers (silicon diaphragm type)
its radial position. In addition to these factors, differ-
with 8 V DC excitation were used to measure both
ential pressure signals depend on the spacing between
the static and differential pressures. Omega model PX
the taps. No published experimental information is
162 was used for static pressure measurement and
available to help us identify the effects of some of the
Omega model PX164 was used for differential pressure
above mentioned variables while establishing some re-
measurement. The output from the transducers was
lationship between flow patterns and pressure fluctua-
fed into the computer using a 12 line shielded cable.
tions.
Each transducer is connected to the tap by a 150
One of the aims of this research was to study the
cm long l/4 in PVC tube. The length of all the tubes
effects of taps location and relationship between dif-
were kept equal despite the physical proximity of the
ferential and static pressure signals for the same flow.
tap with respect to the transducers. The effect of tube
Two configurations were tried (Fig. 3). Configuration
length on the pressure signal was therefore same for
I has pressure tap P, between two taps of P3 and
all measurements. These taps were purged periodically
pressure tap P, between two taps of P,. P1 and P, are
with high pressure air.
static pressure transducers while P3 and P, are differ-
ential pressure transducers. This will enable us to study
the changes in differential and static pressure fluctua-
tions as a function of distance. Configuration II was Velocity measurement
necessary to study the effect of tap distance in differential The- particle velocity was measured by cross-corre-
lating the electrostatic noise picked up by the coil
probes. The coil probes were mounted on the glass
section using an electrical insulation tape. The coils
were connected to Keithly model 610C electrometers
CONFIGURATION I using coaxial cables (as shown in Fig. 2). The elec-
trometer amplified the signal which could then be read
P
3 by the computer for further processing. Probe sepa-
P4
.._.
I ..?...?
.. rations of 8 in and 12 in were used.
Irn /
FLOW - p1 p2

(iuS8 Solids mass flow rate

Mcnon
Average solids flow rate was determined by gain in
weight of the collector. The collector/receiver was placed
on a digital weigh scale with an accuracy of +0.05 kg.
CONFIGURATION II This method is reasonably accurate so long as the
duration of solids feed is not short.
p4
’ 2m ’

, )
Gas volumetric flow rate
FLOW - P, p2 The conveying air was metered differently for different
OIASS air supply sources. A turbine meter (Elster model Ql60)
SEclxu
was used to measure the flow rate when blower was
Fig. 3. Configurations of pressure taps. used for air delivery. A rotameter calibrated at 30 psig
184

with a range of 4-40 SCFM was used with compressed twice the maximum frequency in the signal if aliasing
air. is to be avoided. Aliasing refers to the folding over of
the high frequency data onto the lower spectrum due
Material charactektics insufficient sampling rates.
Based on preliminary experimental studies, the fol-
TABLE 1. Material properties - pneumatic conveying experi- lowing procedure was found to reduce aliasing problems.
ments Step 1: Set the sampling rate equal to the maximum
allowable by the hardware and software combination.
Material d, Particle U, terminal Use the maximum storage allowable on the system.
name (pm) density velocity Step 2: Using the standard procedures, obtain the
(kg m? (m s-l)
power spectrum of the signal. The power in the spectrum
Glass beads 450 2340 3.97 should be negligible at the highest frequencies. If this
Glass beads 55 2380 0.22 is not so, then you must make necessary changes in
Alumina 400 3420 3.40 hardware and/or software to increase the sampling rate.
PVC 137 1180 0.61 Choose the highest frequency in the spectrum (f,,,_J
beyond which the power is negligible.
Step 3: Change the sample rate given by the expression
as follows:
Data acquisition and analysis

Understanding of digital signal processing and tech- where K can take any value between 2 and 3.
niques to analyze times series signals is imperative if The factor of 2 comes from the Nyquist sampling
any useful and reliable information has to be obtained. theorem which states that the maximum frequency
Both acquisition and analysis of pressure signal requires calculable in the spectrum is equal to half the sampling
a detail discussion of the justification of the choice of frequency.
parameters. Various types of analysis can be performed Step 4: Repeat steps 2-4 unitf,,,,,, does not change.
on a time series resulting in different interpretations. Hardware solution to this problem is also available
in the form of frequency splitters and anti-aliasing
Data acquisition filters.
Acquisition and manipulation of experimental data The number of data points depends on the hardware
(such as pressure signals) is performed by digital com- limitations of storage and processing. A large number
puters, Although the transducer output is an analog of data points will also increase the reliability of the
signal which is continuous, discrete representation of estimates. The importance of sample size is greater for
this analog signal is necessary if digital computers are amplitude domain analysis than frequency domain anal-
used. ysis.
According to Bendat and Piersol [4], there are two
main issues involved with digitization of continuous
signal. Flow patterns
1. Sampling One of the first attempts to relate the flow patterns
2. Quantization with gas and solids flow rate was made by Zenz [2].
Sampling refers to the spacing of sample points or Zenz superimposed on a plot of P/L versus U, curve
the sampling rate. Sampling rates are generally software existing flow patterns. Such a representation gives only
controllable. Quantization refers to the accuracy of the a general idea of the flow patterns that can be observed.
A/D converter to represent the signal. If the signal is Various flow patterns and their approximate locations
towards the low end of the input range then higher in the Zenz plot for coarse material can be seen in
gains are advisable to reduce the white noise content. Fig. 4. Similarly, one can locate the flow patterns on
There are two questions to be addressed here: a plot of standard deviation of pressure drop fluctuations
1. How fast should the data be sampled? versus superficial gas velocity (Fig. 5). For finer material,
2. How much of the data is sufficient? the locations and types of flow patterns are different.
Sampling at high frequencies may well represent the Unsteady flow regimes are located in the region of
data if the sampling time has been sufficient. This may high standard deviation. A major drawback of this
result in data redundancy and prove expensive in storage representation is the lack of information that can be
and calculations. The smallest sampling rate of fre- put on a plot without cluttering it. The effect of solids
quency is given by Nyquist’s criterion which states that flow rate and the transition regions also cannot be
the minium frequency of sampling must be equal to represented.
 185

in dilute and homogeneous flow results in flow smooth-

ening. This observation can be corroborated by the
comparison of the power spectrums. It should also be
noted that under these conditions, the fluctuations are
almost independent of solids flow rate. Therefore, the
spectral nature of fluctuations for homogeneous flows
is reasonably invariant.
i
Reduction in gas velocity causes flow stratification
in case of coarse materials while the fine materials
, -- form a dense moving strand at the pipe bottom. Flow
stratification does not cause flow instabilities, and the
r
intensity of fluctuations continues to decrease. The
,r formation of strands, which are often times discontin-
1-t 1
li-

25
uous, is accompanied by increase in the standard de-
viation.
In the case of coarse and hne materials, dune flow
causes a sharp increase in fluctuations. The standard
Fig. 4. Zenz type plot for 450 pm glass beads in 0.0254 m (I.D.
system).
deviation during such conditions is about two orders
of magnitude higher than single phase gas flow.
In the case of fine materials, the height of settled
Oair layer need not be greater than 4 mm before conveying
. l D5 kg/hr velocity is attained in the remaining cross-section. There-
~260 kg/hr I
~327 kg/hr fore, at low solids flow rates one may expect a reduction
in fluctuations. The height of the settled layer is sig-
nificant in case of coarse material and the minimum
100 velocity required to maintain two-phase flow in the

. -
%
d
upper section is also higher, thus resulting in greater
L fluctuations.

.
It was also observed that formation of stationary

=A:
50 layer is favored at low solids how rates while at high
flow rates one may observe dunes or slugs.

A Pressure Jluctuations - single phase gas flow

0.
0 5 lb 1; zo 25
Extreme care must be taken to identify and isolate
Ug. m/s
the sources of pressure fluctuations. Some of these
fluctuations might not be phenomenological but related
Fig. 5. Effect of gas velocity on standard deviation and their
relationship with the flow patterns (450 w glass beads in 0.0254
only to the transducers. In actual practice, the pressure
m (I.D.) system). signal converted into an analog voltage signal by the
transducer is recorded by the computer in a digital
form. Due to the inherent instability of A/D board
Figure 4 is typical in horizontal conveying conditions. ( f 1 bit) and the noise due to power supply, it is not
The average pressure drop in the conveying line in- possible to obtain a zero amplitude signal for zero flow.
creases with solids mass flow rate. The pressure drop Analysis of transducer output at zero flow provides
curve shows a minimum. At gas velocities less than a reference for further measurement and also helps to
saltation, which corresponds to the pressure minimum, identify any spurious noise. From the power spectral
a non-moving layer can be observed. The pressure drop analysis, it can be seen that the noise is mostly wide
for such conditions is greater because of higher actual band white noise. No familiar components (such as 60
fluid velocity due to reduced cross-section. It can also Hz) were found. The maximum power at the peak
be seen that the saltation velocity increases with solids frequency is less than 0.5 Pa2. The noise was truly
flow rate (Fig. 4). The fluctuations in pressure are random and had a very low power, indicating that
reflected in the standard deviation plots (Fig. 5) since results will not be affected by the transducer or A/D
variance represents the sum total of powers of all board noise.
frequencies. As can be seen, for gas velocities greater Air movers are essential components of pneumatic
than about 10 m s-‘, the standard deviation for air conveying systems. To study the fundamental nature
alone is higher. This implies that the addition of solids of the pressure fluctuations introduced by the flow of
186

solids, one must use a steady and non-pulsing air supply. of components near the far end of the spectrum confirms
Although external pulsations can be minimized by the sufficiency of 1000 Hz as the sampling rate and
proper design of the air supply, pulsations due to this suggests that a transducer with response time of
turbulence are characteristic to the system. It has been 1 millisecond is adequate for these measurements.
found that the air supply must use a steady and non- The power spectrum analysis for the same static
pulsing air supply. Although external pulsations can be transducer when the blower was used as the air supply
minimized by proper design of the air supply, pulsations shows most of fluctuations have frequencies lower than
due to turbulence are characteristic to the system. It 100 Hz but they are at least two orders of magnitude
has been found that the air supply and the conveying higher. Due to the gas by-pass configuration of the
line has its own characteristic pressure signature. Data blower, higher fluctuations are observed at lower su-
on static pressure was acquired from two transducers perficial velocities. At low flow rates, the by-pass valve
located 0.5 m apart and data on differential pressure is almost completely open, thus causing maximum in-
signal was acquired for tap spacings of 1 m and 2 m. terference.
Comparison of static pressure signals at two locations
Frequency domain ana&si.s on the system indicates that there is no significant
Figure 6 shows the power spectral density functions difference between the variance (total power) of the
for static wall pressure signal at different gas velocities two signals. Although the spectrums look alike, there
when compressed air was used as the air supply. A is a small shift in the frequencies of various components.
wide range of frequencies up to 400 Hz were found Composition of differential pressure signal depends
in the spectrum. The higher frequency components tend on the tap spacing. It was found that more higher
to disappear with reduction in the gas velocity. Absence frequency components are obtained at reduced tap

18.0
16.0
14.0
12.0
10.0
8.0
6.0

4.0
2.0
0.0
0.0 100.0 266.0 380.0 406.6 see.0

0.0
8.8 1ee.e

Frequency _ Hz

1 .e
8.8 U,=4.6 m/s
0.6

0.4

8.2

0.6
0.0

Frequencu, Hz
Fig.’ 6. Power spectral density function for air only flow.
 187

3.0

2.5

2.0

1.5

1.0 W,=O kg/hr

0.5

0.8
8.0 1ee.e 208 .e 300.0 488.0 see.0

4.5

4.8

3.5

3.8
Us=95 kg/hr
2.5

2.8

1.5

1.0

8.5

0.e
.e 28.8 40.0 60.0 80.8 lee.8

Frequency, n2

lee.0

8B.B

60.0
W,=354 kg/hr

40 .‘0

20.0

8.0 l""""'l
2.0 4.8 6.0 8.8 ie.e 12.0 14.0

Frequency, nI:
Fig. 7. Effect of solids flow rate on PSDF of static pressure fluctuations for 450 /*m glass beads at lJ,=9.3 m s-‘.

spacing. Unlike static pressure signals, the higher fre- In this study, the effect of the following parameters
quency components do not disappear with reduction on the nature of fluctuations were studied:
in gas velocity. Each of the components loses some 1. Effect of solids flow rate
power as the gas velocity is reduced, thus retaining the 2. Effect of gas flow rate
general appearance of the spectrum. 3. Effect of particle size
4. Effect of particle density
5. Effect of feed configuration and air supply
Pressure fluctuations - gas-solid flow 6. Effect of tap location
The nature of pressure fluctuations due to gas flow
in pipe at various gas velocities has been studied so
far. It can be concluded that the nature of fluctuations Frequency domain analysis of static pressure
depend on the gas velocity, the type of air supply and fluctuations
the tap spacing (in case of differential transducers).
Location of the transducers did not make significant Effect of solids flow rate
difference for single phase (gas) flow. When solids are The effect of solids flow rate on the changes in the
added to the air stream, not only do they modify the power spectrum cannot be viewed in isolation. Particle
fluctuations but the resultant fluctuations also depend size, solids loading and gas velocity are found to affect
on the mode of feeding the solids. The nature and the spectral composition of the pressure fluctuations.
extent of modification of the fluctuations due to the In fact, there is a direct relationship’ between flow
addition of solids depends on the characteristics of the patterns and nature of pressure fluctuations (and its
solids and solids loading. power spectrum). Flow pattern maps generated for
 14.0

12.0

10.0
rr,=lLZ 1/s
6.0

6.8

4.0

2.0

0.0 rm -r-T-J
8.0 5.8 15.0 28.0 3e.n

Frequency, Hz

48.8
as.0
a0.8
25.8

28.6
ug=9.3 m/s
15.8

lb.0

5.0

0.0
0.0 1e.a 15.0 20.0 25.8 38.6

Frequency, Hz

7.Clk

6.tIk

5.0k

4.fBk

3.Bk
Ug=6.5 m/s
2.8k

1 .Bk

8.0k I
0.B 2.e 4.0 6.0 0.8 1t3.e 12.0 14.0

Frequency, Ha

Fig. 8. Effect of superficial gas velocity on power spectral density function for 55 Mm glass beads in 0.0254 m I.D. system.

different materials are unique with regard to the shape can be made in the case of 55 micron glass beads.
and location of the regions of similar flow patterns, Therefore, to study the effect of solids flow rate or gas
and also in the extent and nature of transition patterns. flow rate on the nature of pressure fluctuations, it is
From the flow pattern map for 450 pm glass beads, essential to view it in relation to the existing flow
one may observe that at low solids flow rates the changes pattern.
of flow patterns due to reduction in gas velocity is Comparison of power spectrums for various solids
different than at high solids flow rate. At lower solids flow rates at a given gas velocity for 450 micron glass
flow rates and for conditions below saltation velocity, beads (Fig. 7) indicates the following:
the formation of a stationary layer is preferred over 1. The small flow rates of solids (95 kg h- ’ in our
the formation of dunes or slugs. Similar observations case) at gas velocities approaching saltation velocity
 189

Pattern 1
Homoaeneous Flow

1.883L

:
1.854k

5. 0aSk

1.796k

I
1.767k ;
I I I I
0.0 8.5 1.0 1.5 2.0

tine sac

F
: 6.0

4.0

b
Z a.0
:
a.0
0.0 50.8 100.0 130.0 200.0 250.0

Fig. 9. Typical flow pattern, static wall pressure time trace and power spectrum for homogeneous flow of 450 pm glass beads.

has a flow smoothening effect. The power spectrum spectrum but merely modifies it. One may conclude
due to single phase flow of gas has significant components that pressure fluctuations due to single phase turbulence
up to 100 Hz. After the addition of solids, all these dominate under such conditions. With further increase
components are absorbed with the addition of only a in solids flow rate, blowing packets are observed. These
singular very low frequency (less than 2 Hz) component. packets are characterized by their distinct peaks in the
The reduction in overall power can be confirmed by power spectrum and absence of fluctuations higher than
the reduction in the standard deviation of pressure 5 Hz.
fluctuations. 4. The conveying characteristic of Geldart A solids
2. At higher solids flow rates (same gas velocity), are different from Geldart B solids. The range of flow
addition of solids tends to increase the fluctuations. patterns observed were also found to be different. By
Although the high frequency fluctuations are absorbed, classifying our results for Geldart A and Geldart B
the low frequency fluctuations have a magnitude ten materials separately, it is possible to point out some
times higher. The formation of light and puffy dunes of the differences in their characteristics and associated
may be observed under these flow conditions. flow patterns.
3. The changes in power spectrum due to changes Unlike 450 micron glass beads which absorb all the
in solids flow rates at high gas velocities (three times high frequency fluctuations, 55 micron glass beads create
the saltation velocity) are quite different. The power a distinct band near 5 Hz. The width of this band is
spectrum due to single phase gas flow has components inversely proportional to the solids flow rate. Although
which are five times those seen when solids are present. the addition of solids does lead to reduction in the
The addition of solids does not drastically change the total power of the spectrum, certain low frequency
190

Pattern 2
Stratified

255.6

242.4 1 I I
I I I
0.0 8.5 1.2 1.5 2.0

Frequency, Hz

Fig. 10. Typical flow pattern, differential pressure time trace and power spectrum for power spectrum for blowing packets of 450
wrn glass beads.

components (O-10 Hz) are not completely removed. The density of the particles is generally 2-3 orders
This probably due to the small size of particles which of magnitude higher than the gas. Regardless of the
are of the same order of magnitude as the turbulent size of the particles, the mass of the particles per unit
eddies. The appearance of components at low fre- volume of the suspension is much higher than the pure
quencies (up to 0.5 Hz) is associated with strand gas. The particles tend to absorb all fluctuations higher
formation. The number and transition of flow patterns than 10 Hz. This can be compared to the pure gas
in Geldart A material is much more complex than in spectrum where significant power at 300 Hz is not
Geldart B material. A simple explanation of the effect uncommon.
of solids flow rate will be misleading since it depends Figure 8 shows the effect of gas velocity on the static
on the transitions in the flow ‘patterns. pressure fluctuations for fine particles. For homogeneous
flow, uniform distribution of power about the dominant
frequency is seen (Fig. 8). Appearance of a strand at
Effect of gas flow rate the bottom of the pipe is reflected by a low frequency
The energy content of the flow is proportional to (but high magnitude) component. Since the pressure
the gas velocity. From the power spectrums of the pure tap is located on the top of the pipe where the flow
gas flow, it is evident that at higher gas velocity greater is homogeneous, the spectrum obtained is a combination
power is available (especially at higher frequencies). of spectrum for homogeneous flow and a low frequency
The kinetic energy of gas is transmitted to the particles component due to the strand (see Fig. 8). With further
as- they accelerate in the conveying ‘line. reduction in gas velocity, the strand that is moving
 191

i.0Sk 1 I I I I I
1.. 0 2’.e 3..e 4’.0 5’.e

tine se0

188.8

:
68.0

P 80.0
10.8 15.8 20.0 25.0 30.8

Frequency. Hz

Fig. 11. Apical flow pattern, differential pressure time trace and power spectrum for power spectrum for blowing packets of 450
pm glass beads.

along the pipe bottom becomes denser and occasional which manifests itself by decrease in total power. De-
dunes are observed. In Fig. 8 the frequencies around creasing the gas flow rate below the saltation velocity
2 Hz increase at least by an order of magnitude. No leads to flow instabilities. The particles no longer flow
single frequency dominates the spectrum because a as discrete particles but they flow in aggregates or
regular dune flow does not exist. Further decrease in clusters. In general, it may be concluded that decrease
gas velocity causes the commencement of regular dune in gas velocity decreases the dominant frequency. The
flow. The magnitude of low frequency pulsations are power at this frequency will depend on the actual flow
3 orders of magnitude higher compared to the homo- pattern in the pipe.
geneous flow (Fig. 8(d)). A single dominant peak can
now be observed. The 55 micron glass beads belong Effect of tap location
to Geldart A classification which implies that they are One of the most important problems that needs to
aeratable. Due to their state of high aeration, the dunes be addressed is to study the effect of tap location on
(and slugs) of these type of materials move like a wave. the static pressure fluctuations. The magnitude of static
Unlike 450 pm particles, where the front end of the .pressure decreases from feed point to the collector.
dune was being eroded constantly, the dunes of 55 pm The variance of the static pressure also changes as a
particles moved mostly enmasse. function of the distance from feeder. Static pressure
Similar comparison can be made for coarse particles at any point in the system is affected by both the
(450 pm glass beads). The reduction in gas velocity downstream conditions and upstream conditions. The
causes reduction in energy input into the gas-solid flow effects of the upstream conditions and downstream
192

‘,’ _.’ _‘.”

: _ ;:,: i ,~.,. z Pattern 4
,.‘;.,
. Blowing Dunes
;;;;:
.;,,.;
:.,;__,;y::
” .:-
:;.!:! v..;.:;. : ,, :‘~.‘:..‘,.
.., ,,:,,,,, <slugs>

1
1.862k

1.773k

pl
1.684k
E
I.596k
:
1.587k

1.419k-),
0.e 5.0 le.e 15.0 20.0

0.0 2.0 4.8 6.0 6.0

Fig. 12. Typical flow pattern, static wall pressure time trace and power spectrum for power spectrum for moving dunes (450 pm
glass beads).

conditions can be quite opposite. For example, a flow The feed section plays an important role in the
blockage upstream can cause a reduction in the static development of flow patterns. The nature of pressure
pressure while a flow blockage downstream will cause fluctuations and smoothness of flow are strongly de-
an increase in the pressure. This cumulative effect on pendent on the design of feed section. In order to
the static pressure makes the interpretation more dif- study the nature of pressure of fluctuations due to flow
ficult. The interpretation may be further complicated patterns alone, the feed tee was modified to reduce
by effect that tend to cancel each other. low frequency pulsations due to the feeder. However,
it is of practical interest to find out the effect of feed
From the experimental results on pure gas flow, it
tee configuration on the nature of pressure fluctuations.
can be concluded that the location of taps affects the
From the analysis of the experimental data for a
measurement of static pressure signal to a small extent.
wide range of conditions, it can be concluded that the
Comparison of spectrums from two static pressure effect of feed tee configuration is significant at high
transducers (located 3 m apart) for homogeneous flow gas velocities. The entrainment of solids by the gas
shows no significant difference in the nature of the (pickup) at the feed point is aided by the stepped
spectrum. Such is not the case when the flow is mostly configuration. Therefore, a stepped configuration of the
dune like. The flow pattern dynamics for dune flows feed tee may be utilized to smoothen the flow of gas-
are so complex that the coherence between two widely solid suspensions. However, at lower gas velocity the
spaced pressure transducers is not expected to be high, solids tend to deposit at the feed point regardless
however, the pressure fluctuations from closely placed of the existence of stepped configuration. The
taps (0,5 m) were found to be almost identical. relative contribution due to the lower frequency
 193

Pattern 4
Blowing Dunes
<slugs>

2.33k

f
e
c 1.83k
:

1.33k,
I I I I
8.8 5.8 10.8 15.0 20.0

Frequency. Hz

Fig. 13. Typical flow pattern, static wall pressure time trace and power spectrum for eroding dunes (450 pm glass beads).

components in an unstepped tee configuration is much and density) and solids loading are accounted for, it
higher. may be very difficult to account for the system con-
Differential pressure transducers are less susceptible figuration and component related effects.
to upstream and downstream conditions than the static
pressure transducer. Thus differential pressure trans- Coarse material - 450 micron glass beads
ducers are preferred for dune/slug flow conditions. 1. Spectrum for homogeneous dilute flow (Fig. 9)
Homogeneous flow could be due to
(i) low solids flow rate at moderate gas flow rate -
Representative spectrums of static/differential characterized by absorption of most of the higher
pressure fluctuations at various flow patterns frequency gas components
(ii) very high gas flow rate - characterized by more
It will be instructive to compare the typicaZspectrums, gas like spectrum
corresponding signals and flow patterns. Nevertheless The pressure-time trace for a smooth and homo-
it is much more insightful to look at this combination geneous flow is characterized by high frequency and
of the time signal and spectrums than isolated spectrums low amplitude fluctuations. The intensity of pressure
or simple pressure-time traces. fluctuations is lower than that due to pure gas flow
However, generalization of power spectrums for each (especially at high gas velocities). The spectral density
type of flow patterns observed in horizontal pneumatic function for homogeneous flow is merely a modification,
conveying may not be totally possible. Even if the to the gas flow spectrum. Addition of a low frequency
differences in the particle characteristics (due to size component (less than 5 Hz) can also be seen.
194

‘. Pattern 5
.‘.‘. Settled Layer
t..\.,(.,,;,.\
._ :,..*..
.,.;.. ‘.~,.‘.~~
..,.., I.,\‘,,.,,_
‘.(WC,:!”
!.‘.!:.K,?:,.

2.59k

1.94k
1

51.29k-
*
L

:
L

0.0 20.8 48.0 60.0 80.0

Frequency, Hz

Fig. 14. Differential pressure time trace and power spectrum for settled layer (150 pm glass beads D,s (450 pm glass beads).
D, = 0.0254 m.

2. Stratified flow (Fig. 10) 4. Dunes

The stratification of the flow can be seen at reduced Two types of dunes were observed
gas velocity and lower solids flow rates. Segregation of (i) Moving dunes - these dunes are small clusters of
particles in the lower half of the pipe and further solids moving along the pipe bottom. The whole dune
absorption of already reduced gas fluctuations result is seen to move as one unit. They are relatively light
in a.very low amplitude signal. The near zero component and puffy.
in the spectrum is indicative of the fluctuating nature The pressure-time trace (Fig. 12) for such flow pat-
of stratified layer.
terns can be compared with similar ‘observations in
3. Blowing packets (Fig. 11)
slugging beds. The static pressure oscillates in a very
These packets are like very fast moving clusters or
regular manner with sharp peaks. The size and velocity
aggregates of solids. They are generated at the feed
point when the flow rate is high. The pressure-time of these dunes is not uniform, therefore, a range of
trace is characterized by small peaks occurring almost components in the power spectrum are seen. The relative
at regular intervals. The total duration of these peaks contribution due to high frequency fluctuations (greater
is about 0.3 s. Correspondingly, the spectrum analysis than 4 Hz) is small.
shows the appearance of unique peaks. The intensity (ii) Eroding dunes - dunes which move along the
of these peaks is an order of magnitude higher than bottom of the pipe by the action of erosion on the
those observed for homogeneous conditions. front end and deposition on the other side. These dunes
 195

move due to the eroding action. Their frequency is depending on the amount of increase of solids flow.
much smaller and they move over slowly. See Fig. 13. The pressure fluctuations go through the sequence of
The pressure-time trace for eroding dunes shows increasing the decreasing with the associated flow pat-
isolated and widely spaced peaks. The variations are terns. Geldard A powders are found to interact more
not as regular as seen for moving dunes. The spectral intimately with the gas pressure fluctuations while Gel-
density function has a characteristic low frequency dart B powders do not have this degree of intimacy
component (0.2-0.4 Hz) which is almost two orders of with the gas pressure fluctuations.
magnitude higher than those observed for moving dunes. At a fixed solids flow rate at high gas velocity, the
5. Stationary layer (Fig. 14) pressure fluctuations are dominantly those of the gas
The formation of stationary layer is an unsteady flow alone. Lowering the gas velocities the solids absorb
process which can take as long as 15 min to reach a the energy to give pressure fluctuations below 10 Hz
steady state. During the unsteady process of the for- with associated larger amplitude pressure fluctuations.
mation of the stationary layer, the static pressure shows The feed tee configurations influences the smoothness
a continual rise. Often times, this results in a non- of the flow only for high gas velocity conditions.
stationary signal which is difficult to analyze. The flow pattern behavior can be surmnarized for
The pressure-time trace on a 80 s window shows each of the patterns observed.
that the pressure is almost invariant. The spectrum is Homogeneous flow - high frequency components
devoid of any fluctuations except some very low fre- present at low amplitudes.
quency fluctuations (0.05 Hz) which are due to surface Stratified flow - significant fluctuating components
waves on this stationary layer. near 0 Hz at generally lower amplitudes.
Blowing packets - regular pressure fluctuations arise,
unique peaks in the spectrum. The amplitude is an
order of magnitude greater than homogeneous flow.
Conclusions
Dunes
(1) Moving dunes - irregular fluctuations less than 4
This study of pressure fluctuations analysis proved Hz frequency.
to be a successful technique to identity flow patterns (2) Eroding dunes - smaller frequency than the moving
in horizontal pneumatic transport by employing the dunes.
power spectral density function of the pressure fluc- Stationary - little pressure fluctuations with very
tuation. One can identify with the PSDF the flow pattern low frequency ~0.05 Hz.
present in the pipeline and thus provide one with the
ability to modify the operational parameters such as
feed rates to control the flow pattern and thus the References
overall energy losses and operability of the system. On-
line measurements can be made and sampling frequency 1 Y. Tsuji and Y. Morikawa, Inr. J. Muftiphuse Flow, 8 (1982)
329.
and data storage is realistic for such operations. Static 2 F. Zenz and D. A. Othmer, Fluidization and Fluid Particle
and differential pressure transducer signals were found Systems, Reinhold, New York, 1960, p. 314.
function best for homogeneous and stratified flow and 3 S. Matsumoto and H. Harakawa, Int. J. Multiphase Flow, 13
dune/slug flow respectively. The pressure tap spacing (1987) 123.
changes the magnitude of the PSDF while not affecting 4 J. A. Bendat and A. G. Piersol, Measurement and Analysis
if Random Data, Wiley, New York, 1966.
its shape. 5 G. E. P. Box and G. M. Jenkins, Time SeriesAnalysis: Forecasting
The increase in solids flow rates can change flow and Contml, Holden-Day, San Francisco, CA, 1967, pp. 1-166.
patterns from homogeneous conditions through strat- 6 G. M. Jenkins and D. G. Watts, Spectral AnalysysLF and its
ified flow, blowing packets, dunes and settled layers Applications, Holden-Day, San Francisco, CA, 1968.