� ����� ��� � ��

Could cyclone performance improve with reduced inlet velocity?

P.A. Funk, K. Elsayed, K.M. Yeater, G.A. Holt, D.P. Whitelock

PII: S0032-5910(15)00300-9

DOI: doi: 10.1016/j.powtec.2015.04.026

Reference: PTEC 10936

To appear in: Powder Technology

Received date: 4 November 2014

Revised date: 6 March 2015

Accepted date: 15 April 2015

Please cite this article as: P.A. Funk, K. Elsayed, K.M. Yeater, G.A. Holt, D.P. White-lock, Could cyclone performance improve with reduced inlet velocity?, Powder Technology

(2015), doi: 10.1016/j.powtec.2015.04.026

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

1

Could cyclone performance improve with reduced inlet velocity?

P. A. Funk, K. Elsayed, K. M. Yeater, G. A. Holt, and D. P. Whitelock

The authors are Paul A. Funk, Ph.D., Agricultural Engineer, USDA Agricultural Research Service, Southwestern Cotton Ginning

Research Laboratory, Mesilla Park, NM, Khairy Elsayed, Ph.D., Assistant Professor, Mechanical Power Engineering Department,

Helwan University, Egypt, Kathleen M. Yeater, Ph.D., Statistician, USDA Agricultural Research Service, Plains Area, Fort Collins,

CO., Gregory A. Holt, Ph.D., Research Leader and Agricultural Engineer, USDA Agricultural Research Service, Cotton Production

and Processing Research Unit, Lubbock, TX, and Derek P. Whitelock, Ph.D., Agricultural Engineer, USDA Agricultural Research

Service, Southwestern Cotton Ginning Research Laboratory, Mesilla Park, NM Corresponding author: Paul Funk, PO Box 578,

Mesilla Park, NM 88047 USA; phone: 575-526-6381; paul.funk@ars.usda.gov.

ABSTRACT

Emissions abatement cyclone performance is improved by increasing collection effectiveness or decreasing

energy consumption. The object of this study was to quantify the pressure drop and fine particulate (PM2.5)

collection of 1D3D cyclones (H = 4Dc, h = 1Dc) at inlet velocities from 8 to 18 m s-1

(Stk = 0.7 – 1.5) using

heterogeneous particulate as a test material at inlet concentrations from 3 to 75 g m-3

. Cyclone exhaust was

passed through filters. Laser diffraction particle size distribution analysis was used to estimate PM2.5

emissions. Response surface models showed a strong correlation between cyclone pressure loss (Euler

number) and inlet velocity and predicted a 46% reduction in pressure loss for a 25% reduction in inlet velocity

(Stokes number). The model for PM2.5 emissions was less definitive and, surprisingly, predicted a 31% decrease

in PM2.5 emissions when operating 25% below the design inlet velocity. Operating below the design inlet

velocity (at a lower Stokes number) to reduce pressure losses (Euler number) would reduce both the financial

and the environmental cost of procuring electricity. The unexpected co-benefit suggested by these trials was

that emissions abatement may improve at the same time, though other empirical trials have shown emissions to

be independent of inlet velocity and Stokes number.

Keywords.Keywords.Keywords.Keywords. Cyclones, Emissions, Energy consumption, Fine particulate, PM2.5.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

2

1 INTRODUCTION

1.1 FINE PARTICULATE (PM2.5) IS A CRITERIA POLLUTANT

In the interest of protecting the public health, the United States Clean Air Act required states to develop a

general plan to attain and maintain National Ambient Air Quality Standards (NAAQS) in all areas of the

country and a specific plan to attain the standards for each area designated non-attainment for a NAAQS. These

plans, known as State Implementation Plans (SIPs), are developed by state and local air quality management

agencies and submitted to the U.S. Environmental Protection Agency (EPA) for approval. Particles less than

2.5 microns in aerodynamic equivalent diameter (PM2.5) were first included in the NAAQS in 1997. In 2006

EPA NAAQS for PM2.5 were lowered, to 15 µg m-3

[1], and in 2012 they were lowered again, to 12 µg m-3

[2].

As SIP authors may try to comply with EPA regulations through controlling anthropogenic sources of PM2.5,

more stringent operating permit qualifications for agricultural processing facilities may arise.

1.2 CYCLONES ARE USED FOR PARTICULATE ABATEMENT

Many agricultural processing facility exhaust flows are controlled with cyclones. Cyclones have no moving

parts and a relatively low initial cost. Their operating costs are primarily due to the electrical energy required

for fans to overcome the friction losses (pressure drop, related to Euler number) in the cyclone and connected

ducts. Since generating electricity results in emissions at the power plant, reducing the energy required to

operate cyclones potentially helps the environment [3]. Cyclone performance is a function of both particle

separation and energy consumption. Cyclone performance may be improved by reducing pressure drop, and

thus energy use, as well as by increasing collection efficiency. One option may be to operate cyclones below

their traditional design inlet velocity [4]. Research has shown that operating cyclones that are 41% larger than

normal, so the inlet velocity was 50% of the design inlet velocity, resulted in pressure losses that were only

28% of normal [5]. While numerous modeling studies have examined monodisperse particles, and empirical

studies have reported the impact of inlet velocity on total particulate collection efficiency, this study is the first

to report the impact of cyclone inlet velocity on PM2.5 emissions for heterogeneous material.

1.3 CYCLONES ARE ROBUST TO INLET VELOCITY

The cyclone design called the ‘1D3D’ was first introduced by Texas A&M University [6]. Collection

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

3

efficiency improvements brought about by design modifications were confirmed through USDA ginning lab

tests [7]; [8]. Modified 1D3D cyclones are widely used in agricultural processing, such as by the U.S. cotton

ginning industry [9]. Figure 1 shows dimensions in cm of the modified 1D3D cyclones that were tested, with

the overall height, H, = 4 diameters (D), and the barrel height, h, = 1 D, with the inlet height, a, and width, b,

equal to 0.5 and 0.25 D, respectively. The diameter of the gas outlet, De, = 0.5 D, the vortex finder insertion

depth, S = 0.625 D, and the particle outlet diameter, B, = 0.33 D.

The design inlet velocity for the 1D3D cyclone is 16.26 ± 2.03 m s-1

[10], [11], corresponding to a Stokes

number of 1.4 for PM2.5 in the conditions tested. But as publications presenting empirical cyclone collection

efficiencies at various inlet velocities have shown [12]; [13]; [14]; [15], cyclones of this design, and other

designs, work well over a range of inlet velocities.

Differences in cyclone design, cyclone diameter, particulate inlet concentration and particle size distribution

prevent direct comparisons of published results. However, a plot of the collective results (Figure 2) illustrates

that gravimetric total suspended particulate (TSP) collection efficiency does not have a clear relationship to

inlet velocity (or Stokes number) over the range of inlet velocities reported. This experiment was designed to

test the relationship between PM2.5 emissions and inlet velocity for the 1D3D cyclone.

1.4 CYCLONE PERFORMANCE INDICIES

Gas cyclones performance characteristics are estimated using the static pressure drop from the inlet to the

outlet (∆�) and the grade efficiency curve. The static pressure drop (∆�) between the inlet and gas outlet of a

cyclone is proportional to the square of the flow rate (Q), with a proportionality resistance coefficient (��)

defined on the basis of the inlet velocity (Vi = Q/a b), thus:

∆� = �� ����

�

� (1)

Alternatively, Dewil et al. [16] recommended avoiding the inlet velocity as a characteristic velocity and

replacing it with the average velocity in the cyclone body (�� =�

����/�

). Thus the Euler number (Eu) is related

to �� and the pressure drop as:

� = ∆

������

� (2)

In this study, (following the recommendation of Svarovsky [17]), Eu will be used as a dimensionless

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

4

performance parameter [18].

The separation (grade) efficiency of cyclones represents the variation of the separation efficiency with the

particle size. The particle size recovered at an efficiency of 50% is called the cut-size or d50 [16]. For

estimating the cut-size, two different approaches are found in the literature [17]. In the first approach, the cut

size is reported in ��. The second approach uses the Stokes number. In this study, based on the above-

mentioned recommendation, the Stokes number based on the cut-size is:

�� =��� ���

����� (3)

Where �� is the particle diameter, �� is the particle density, and � is the gas viscosity. In this study, a

similar Stokes number will be used for the Stokes number when the particle diameter equals 2.5 �, as:

���.� =(�.�� ��)����

����� (4)

1.5 TEST OBJECTIVE

An agricultural processing facility typically has several pneumatic conveying systems that use cyclones to

control emissions. Each system has a fixed airflow requirement based on pipe size and minimum carrying

velocity. Cyclone inlet concentration varies with processing rate. Cyclone inlet velocity could be modified by

substitution of a different sized cyclone. The object of this study was to evaluate the performance (pressure

drop and PM2.5 particulate collection) of 1D3D cyclones over a range of inlet velocities using typical

agricultural processing facility particulate matter over a range of inlet concentrations.

2 MATERIALS AND METHODS

2.1 TEST CYCLONES

Two modified 1D3D cyclones, 30.5 cm in diameter, were fabricated for this study (Figure 1). The 1D3D

cyclones had a barrel that was one diameter in height (1D) and a cone that was three diameters in height (3D).

All dimensions in Figure 1 were determined based on published ratios to barrel diameter [9]. The percentage

of processing facilities using the modified 1D3D design has continued increasing as older abatement devices

and older cyclone designs have been replaced through repairs or new construction. This design is called

“modified” because it has a 0.5 D x 0.25 D inlet and a D/3 particle outlet, both of which were not part of the

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

5

original design [6]. There also was an expansion chamber (product receiver, or dust bin) at the bottom; this

modification is not as widely used as the first two, but it is still common. These test cyclones were fabricated

following current industry practices and using 24 gage galvanized lock forming-quality steel sheet metal. All

seams were lapped away from the direction of airflow, spot welded and polished until smooth. All joints

between connected sections (i.e., cones and barrels) were sealed to prevent leaks. A sealed container 28 cm in

diameter by 30.5 cm high was affixed to the bottom of the expansion chamber to hold collected particulate and

to prevent air flow (though industry applications are not always able to prevent air flow at the dust outlet.)

2.2 TEST APPARATUS

Testing was conducted using an apparatus specifically built for this research (Figure 3). A variable speed

feed conveyor belt was used to meter test material into the blades of a variable speed trash fan (MF7, Murray

Co., Dallas, TX) with a 1.49 kW DC motor (M-200-A, T.B. Wood’s Sons Co., Chambersburg, PA). The trash

fan conveyed the test material in air through a Venturi tube, through a 10.2 cm diameter pipe, and into the test

cyclone. All cyclone exhaust air passed through a filter aided by a filter fan (2206 Alum, New York Blower,

Willowbrook, IL) with a 5.59 kW AC motor (M3769T, Baldor, Ft. Smith, AK). Prior to each test a Pitot-static

tube was used to measure the air velocity in the duct before the cyclone in order to set the cyclone inlet velocity

(by adjusting the speed of the trash fan) according to the velocity called for by the experiment design. During

each test the differential pressure through the Venturi tube was monitored, providing feedback control for a

variable frequency drive (VFD) (VLT 8000 AQUA, Danfoss, Nordborg, Denmark) that controlled the filter

fan’s motor. The filter fan produced additional fan pressure to maintain constant air flow as flow through the

filter became more restricted with particle accumulation.

Ambient and cyclone exhaust air temperatures (type K thermocouples), and differential pressures across the

Venturi tube, the cyclone, the filter and the Pitot-static tube downstream of the filter fan were also measured

and recorded. The recorded differential pressure from the Pitot-static tube downstream of the filter fan

provided an independent value of instantaneous air velocity throughout each test. The 4-20 mA pressure

transmitter signals (Model 614, Dwyer Instruments, Inc., Michigan City, IN) were converted to voltages by a

custom-made signal-conditioning card. Voltage signals were recorded continuously at 1 s intervals using a data

logger (Model 34970A, with 34908A switch units, Agilent Technologies, Santa Clara, CA) and averaged over

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

6

the duration of the test to determine their respective values.

2.3 FILTERS

Particulate matter escaping the dust cyclone was collected on 56 cm x 61 cm hydrophobic filter media (Type

FP2063, Hi-Q Environmental Products Co., Inc., San Diego, CA) made of borosilicate glass microfibers with a

minimal amount of an acrylic resin binder to maintain filter integrity. Filters had 97% collection efficiency for

0.3 micron dioctyl phthalate [19]. Filters were conditioned in a controlled atmosphere environmental chamber

(21 C and 35% relative humidity) for 72 hours or more before weighing, following guidelines published by the

EPA [20]. Each filter was carefully handled with gloves, folded in quarters, and stored and weighed in a

dedicated 4 mil anti-static polybag (Part No. 49115, Protektive Pak, Chino, CA) to prevent the incidental loss

of glass fiber or particulate. The bagged and conditioned filters were weighed before and after testing in an

environmental chamber on an electronic balance (PG1003-S, Mettler-Toledo Inc., Columbus, OH) after passing

them through an anti-static device. The balance was leveled on a brass slab and housed inside an acrylic box to

minimize the effects of air currents and vibrations. Each filter was weighed three times. If the standard

deviation of the weights exceeded 10 mg, the filter was re-weighed.

2.4 EXPERIMENTAL DESIGN

This response surface experiment used a central composite design. There were nine combinations of two

continuous variables: inlet concentration and cyclone inlet velocity. Five replicates at the center point

estimated variability (Figure 4). Four corner and four axial points were equidistant from the center to attain a

rotatable design with uniform precision of estimation in all directions. Commensurately, the resulting five

levels of inlet velocity and five levels of particulate inlet concentration provided an estimate of curvature for

potential non-linear responses. The experiment design called for inlet velocities of from 8 to 18 m s-1

(corresponding to a PM2.5 Stokes number between 0.7 and 1.5). This range was selected based on earlier work

[5] and included the design inlet velocity for 1D3D cyclones (16.26 m s-1

). The experiment design called for

inlet concentrations from 3 to 75 g m-3

, the range of inlet concentrations from published research [21].

A comparison was made between two identical cyclones (A and B) across the full range of inlet velocity and

inlet concentration. Thirteen runs (four corner points, four axial points, and five replicated center points) were

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

7

conducted with each cyclone for a total of 26 experimental runs. The runs were conducted in a completely

randomized design. Uncontrolled variables such as air temperature were measured throughout the six weeks of

testing to control for potential covariance.

2.5 PARTICLE SIZE DISTRIBUTION ANALYSIS

All filters, cyclone catch and feed material samples were sent to the USDA-ARS Air Quality Laboratory in

Lubbock, TX for Particle Size Distribution (PSD) analysis. These were conditioned in an environmental

chamber for at least 48 hours at 21 C and 35% relative humidity. Bagged cyclone catch and feed samples from

each test were weighed on an electronic balance (A & D HP-20K, Data Weighing Systems, Elk Grove, IL).

After removing sample material, the bags were reweighed. Cyclone catch from each test were sieved using

standard procedures [22] and equipment (Ro-tap, W.S. Tyler, Mentor, OH) to obtain coarse gradation above

106 µm. Particles larger than 100 µm aerodynamic equivalent diameter (AED) have a high probability of being

captured by the cyclone. The fine portion (below 106 µm) was sub-sampled three times for PSD analysis.

Particle size analyses were conducted on a laser diffraction (LD) system (Beckman Coulter LS230,

Beckman Coulter Inc., Miami, FL) with software version 3.29 to calculate the PSD. The instrument measured

particles over the range of 0.375 to 2000 µm that were suspended in a 5% lithium chloride/methanol electrolyte

solution. The LD particle-size analyses consisted of the following procedures: The electrolyte was pre-filtered

using a filtration system that removed all particles larger than 0.2 µm. A background count of the filtered

electrolyte was made with the LD system to ensure that there was minimal particulate contamination of the

electrolyte. Background counts of less than 300 particles per cm3 of electrolyte were considered acceptable.

For PSD analysis of particulate on filters, the particulate must be extracted from the filter media and suspended

in the electrolyte solution. Particulate was extracted from the loaded filters by cutting three 2-cm diameter

samples from a heavily loaded area of the filter, placing the samples in a 100-ml beaker with 50 ml of

electrolyte, and processing the sample in an ultrasonic bath for 5 min. The particulate/electrolyte solution was

then introduced into the fluid module of the LD system to run a PSD analysis.

Optical model parameters used with the LD system software to determine the sample PSDs were 1.56 for the

real part of the sample refractive index, 0.01 for the imaginary part, and 1.326 for the suspension fluid

refractive index. [These values are used in the Lorenz-Mie solution to Maxwell’s equations. These equations

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

8

describer light scattering by spherical particles in a transparent medium.] LD system results in terms of

equivalent spherical diameter were corrected to aerodynamic equivalent diameter (AED) by multiplying

equivalent spherical diameter by the square root of the product of the density (2.65 g cm-3

) divided by the shape

factor (1.40). This PSD analysis was used to estimate PM2.5 emissions (particulate 2.5 µm AED and smaller).

2.6 TEST MATERIAL

Since this research was intended to assist agricultural processing facilities with energy conservation as well

as regulatory compliance, agricultural particulate from the cotton ginning industry was used. The range of

particulate inlet concentration values chosen for this study, from 3 to 75 g m-3

, was influenced by that reported

by earlier empirical studies: 2-16 g m-3

[5]; 7-15 g m-3

[23]; 8-10 g m-3

[24]; 8-100 g m-3

[25]; and 75 g m-3

[7].

The test material consisted of soil particles, fibers, leaf particles, and stems. The composition of this test

material represents the type of material that might be handled by dust cyclones at an agricultural processing

facility. Each test run used 0.8 kg of test material so that the filter would have close to its maximum load of

dust (pre-trial runs with more than 0.8 kg resulted in filter damage). The largest possible quantity was selected

to minimize experimental error by maximizing test run time and filter weight, and hence signal-to-noise ratio.

Eight sub-samples each weighing 100 g were evenly distributed on eight contiguous sections 30.5 cm long on

the variable speed feed conveyor. The conveyor speed was set before each test so that the test material metered

into the airstream of the test apparatus at the dust inlet concentration required by the experimental design.

Lower inlet concentration trials took more time; time was recorded to independently verify inlet concentration.

2.7 TEST MATERIAL CHARACTERIZATION

Nine samples of test material were collected throughout the test period for particle size distribution analysis.

The size distribution of the test material was somewhat bi-modal. The major constituents were leaf particles

(in the size range between 250 and 850 µm) and soil particles (greater than 10 µm and less than 106 µm), as

presented in Table 1. Note that for this test material, considered representative of that handled by cotton gin

cyclones, only a small part (6%) was particulate less than 10 µm and considerably less than that (0.9% of the

total) was fine particulate less than 2.5 µm.

The sieved portion of each feed sample, the 40.8% that was below 106 µm, was sub-sampled three times for

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

9

particle size distribution analysis. The sieved test material had an aerodynamic equivalent mass median

diameter (AED) of 41.56 µm and a geometric standard deviation of 3.07. The percentage of sieved material ≤

2.5 µm was 2.266%, with a standard deviation of 0.389 percentage points (17%); the percentage of sieved

material ≤ 10 µm was 14.742%, with a standard deviation of 2.893 percentage points (20%). Figure 5 presents

the size distribution of the sieved portion of the test material. Test material variability had the potential to

contribute to uncertainty in the results.

2.8 REPORTED UNITS – PM2.5 PER KG FED

Publications analyzing cyclone performance historically have reported collection efficiency as the collected

percent of mass entering, or they have reported the cut point – that particle size for which the collection

efficiency is 50 percent [26]; [27]. The fraction escaping is quantified by isokinetically sampling a portion of

the exhaust as per EPA Method 201A [28], or by passing all of the cyclone exhaust through a filter weighed

before and after the test under standard conditions (as with this study). The mass lost is compared to the sum

of the mass collected and lost, or to the weight metered into the air stream (as in these tests), or the exhaust

concentration is compared to results from isokinetically sampling the inflow. In all cases, whether sampling a

portion of the gas flow or filtering all of it, whether reporting overall collection efficiency or cut point, results

have also varied with the inlet concentration and particle size distribution of material entering the cyclone [29].

Reporting emissions concentration would have been confusing in a test where inlet concentration varied

over a wide range. Since the rate of material escaping a cyclone has nearly a direct relationship to the rate of

material entering it, emissions concentration would have been confounded by inlet concentration. Therefore,

emitted PM2.5 was normalized to 1000 g of feed material. Emissions values are presented as grams PM2.5

collected on the filter per kg of all material entering the cyclone. The PM2.5 mass on each filter was multiplied

by 1 kg and divided by the mass of material entering the cyclones for each test run (nominally 800 g). The

normalized mass of PM2.5 escaping the cyclone and collected on the filter was expected to give an independent

indication of performance that would reveal the influence of inlet concentration and inlet velocity. Since the

same mass of feed material was used in each test run, reporting fine particulate mass escaping the cyclone per

run is similar to reporting an emissions factor, as is done with AP 42 process rate tables [30]. This relative

performance metric is specific to these trials; it is not an indication of absolute collection efficiency, nor is it a

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

10

universal predictor of emissions concentration.

2.9 DATA ANALYSIS

The theoretical collection efficiency of a cyclone is largely determined by particle velocity [21] and particle

velocity is closely related to air velocity. For this reason, actual values of local air density (based on

temperature, barometric pressure and relative humidity) at the time of each test were calculated and used to

determine the Pitot-static tube velocity pressure corresponding to the desired inlet velocity and inlet

concentration. Uncontrolled variables (run number, date, barometric pressure, temperature, relative humidity,

air density and cyclone) were tested for significance (proc mixed, SAS 9.2, SAS Inc., Research Triangle Park,

NC) at the 5% level to verify that they did not significantly impact the response variables (PM2.5 emissions and

pressure drop). Insignificant uncontrolled variables thus could be omitted from the analysis.

Response surface methodology was performed using experimental design and analysis software (Design-

Expert 8.0.7.1 (DX8), Stat-Ease, Inc., Minneapolis, MN). Actual values for inlet velocity and inlet

concentration (based on local air density) were used in constructing the response surface models. Two

responses, PM2.5 emissions (emitted grams per kg of material entering the cyclone), and pressure drop (Pa),

were modeled. Numerous models were tested in the process of arriving at the “best” one; the criteria for

choosing one model over another, in decreasing order of importance, were: 1) its terms made sense from the

standpoint of the laws of physics; 2) it was simpler; 3) it had a better fit (ANOVA) than a lower-order model, or

adding the higher order terms significantly increased the sum of squares; 4) its lack of fit was still insignificant;

5) it resulted in significantly higher coefficients of determination, or adjusted R2 values.

3 RESULTS

3.1 UNCONTROLLED VARIABLES

Table 2 lists the F value and the probability of exceeding it for each input variable using the proc mixed

output type 3 tests of fixed effects (SAS 9.2, SAS Inc., Research Triangle Park, NC) by response variable; there

were 16 degrees of freedom. For both responses, each one of the uncontrolled variables were not significant –

justification for omitting uncontrolled variables from the response surface models. Inlet velocity was

significant for both PM2.5 and pressure drop. Inlet concentration appeared to have a significant influence on

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

11

PM2.5 emissions, but not on pressure drop, based on this data. There was not a significant interaction between

inlet velocity and inlet concentration.

3.2 RESPONSE SURFACE MODEL

Table 3 lists the selected response surface models. A linear model was selected for PM2.5 emitted per kg

entering the cyclone. Higher-order models improved the adjusted R2 slightly, but were difficult to justify since

the higher-order terms were not significant. The significance of inlet velocity and inlet concentration in the

DX8 model was confirmed by the SAS analysis (Table 2). The adjusted R2 of the selected model for PM2.5 was

0.6599. PM2.5 emitted per kg entering the cyclone was plotted as a response surface over inlet velocity and

inlet concentration (Figure 6).

The selected pressure drop model included terms for inlet velocity to the first and second power. The

adjusted R2 of the selected model for pressure drop was 0.9396. Considering the range of predicted and

measured values for pressure drop in antecedent publications, which vary by a factor of 2.8 for this particular

design [31], and the fact that this model had only inlet velocity as a predictor, it fit this data fairly well. The

influence of inlet concentration on pressure drop has long been documented [32]. Pressure drop decreases as

inlet concentration increases up to about 200 g m-3

[33]. Inlet concentration levels were below that range for

this experiment, but this study lacked the necessary power to detect statistical significance for that variable. In

these trials the Stokes number for PM2.5 was directly proportional to inlet velocity (Stk = 0.0863 Vi), and the

Euler number was a linear function of pressure drop (Eu = 21.72 + 0.074∆P).

When substituting an inlet velocity of 12.19 m s-1

for 16.26 m s-1

in the selected response surface models (a

25% reduction, equivalent to reducing the Stokes number from 1.403 to 1.052) they predicted a 31% decrease

in PM2.5 emissions: from 0.473 to 0.324 g kg-1

; a 46% reduction in pressure loss: from 514 to 280 Pa, and a

29% reduction in Euler number, from 59.8 to 42.4, at an inlet concentration of 45 g m-3

(Table 4). Table 4 also

presents a 95% confidence interval range based on the response variables models (assumed: normal

distribution, model standard deviation and number of observations). The lack of overlap for the 95%

confidence interval ranges on PM2.5 emissions, pressure drop, and Euler number indicate a statistically

significant difference between the two inlet velocities; they predict that an agricultural processing facility may

both save energy and reduce emissions simply by operating its 1D3D cyclones at a lower inlet velocity.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

12

3.3 DISCUSSION

The significance of these results comes from the nexus of emissions and energy. The energy cost of

operating cyclone abatement devices is directly proportional to pressure drop, and producing electricity to

power abatement devices results in air pollution, favoring lower inlet velocities. At the same time, fine

particulate emissions did not increase as expected, but appeared to decrease slightly at lower inlet velocities.

For this particular configuration, set of operating conditions, and inlet material PSD there was a strong

correlation between inlet velocity and pressure drop (model adjusted R2 = 0.94), and a weaker association

between inlet velocity and emissions (model adjusted R2 = 0.66). The second order pressure drop relationship

was expected as it agreed with published results from both modeling and empirical trials. The relationship

between emissions and inlet velocity contradicted many classical numerical models [34]; [35]; [36] and some

empirical research [37] that predict continuously increasing collection efficiency with increasing inlet velocity.

This is best illustrated by Figure 7, a plot of PM2.5 emissions over Stokes number for each run in this trial. A

lower Stokes number predicts that a particle will better follow a streamline, implying an increase in emissions

as particles would be more likely to be carried out of the cyclone by the air. However, this data set showed

emissions decreasing at lower Stokes numbers. Correlation coefficients were between 0.5 and 0.6, partly due

to varying inlet concentration. Other empirical research indicates little correlation between emissions and inlet

velocity, especially with heterogeneous particulate (Figure 2), or at least an upper limit to inlet velocity beyond

which emissions increase. This conflict has been discussed elsewhere [11]. Some authors hypothesize that

turbulence or interior surface roughness induces particle bounce, causing particle re-entrainment [38]; [39], a

phenomenon that is thought to increase with gas and particle velocity. This trial suggested that inlet velocity,

and Stokes number, in 30.5 cm diameter 1D3D cyclones handling heterogeneous particulate, may have little

impact on emissions. Perhaps particle agglomeration, a mechanism for fine particulate capture, is favored by

lower turbulence.

Considering the potential importance of these findings in terms of energy savings and reduced power plant

emissions, the challenge of balancing energy savings to the increased capital cost of larger cyclones, and the

potential difficulty in terms of changing existing environmental regulations, it is important that these

experiments be replicated under field conditions. If confirmation of these experiments were to result in more

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

13

relaxed regulation, it may lead to reductions in the energy consumption of pneumatic conveying systems at

agricultural processing facilities and perhaps elsewhere. This work has a high priority for two reasons. First,

energy costs are rising relative to other costs of operation, and pneumatic conveyance fans account for more

than half of the energy used by some agricultural processing facilities [40]. Second, energy production is a

significant source of air pollution, so reducing energy requirements of abatement devices would benefit the

environment [3].

4 CONCLUSION

Cyclones were tested to better understand the effects of inlet velocity and inlet concentration on pressure

drop and emissions. Modified 1D3D cyclones were operated at inlet velocities from about half to slightly more

than design, from 8 to 18 m s-1

, over a range of inlet concentrations from 3 to 75 g m-3

. Cyclone pressure drop,

∆P, was measured directly. Emitted total mass collected on filters and particle size distribution analysis was

used to estimate PM2.5 emissions, presented in this paper as mass of PM2.5 for each kg of total material fed to

the cyclone. Cyclone performance models were developed using response surface methodology. Based on the

obtained results, the following conclusions were drawn:

• Response surface models showed a strong correlation between cyclone pressure loss and inlet velocity

and predicted a 46% reduction in pressure loss for a 25% reduction in inlet velocity, as well as a 29%

reduction in Euler number for a 25% reduction in Stokes number.

• The model for PM2.5 emissions was less definitive and, surprisingly, predicted a 31% decrease in PM2.5

emissions when operating 25% below the design inlet velocity (at a 25% lower Stokes number).

• Operating below the design inlet velocity to reduce pressure losses would require larger, costlier

cyclones, but may reduce both the financial and the environmental cost of procuring electricity.

VITAE

Dr. Paul Funk is an Agricultural Engineer working on diverse projects including non-chemical harvest

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

14

preparation, renewable energy production, specialty crop mechanization, trace analytical chemical detection of

contaminants, and air quality. He earned an M.S. from the University of Minnesota and a Ph.D. from the

University of Arizona, both in Agricultural Engineering. He was a professor of Mechanical Engineering before

becoming a Research Scientist with the USDA Agricultural Research Service in Mesilla Park, New Mexico.

Dr. Khairy Elsayed is a Mechanical Engineer working on diverse research projects including Cyclone

separators, CFD simulations, surrogate based optimization and shape optimization. He earned an M.S. from

Helwan University in Egypt and a Ph.D. from Vrije Universiteit Brussel (VUB), both in Mechanical

Engineering. He is a Professor/Senior Research Scientist at VUB and Assistant Professor at Helwan

University.

Dr. Kathleen M. Yeater is an Area Statistician for the United States Department of Agriculture, Agricultural

Research Service, where she provides consultative support to research scientists on the development and

implementation of scientific studies, sampling methodology, and data analysis. She is actively involved in

working with researchers in developing and presenting inferences from statistical results as well as consulting

and training a diverse group of research personnel. Kathy received a Ph.D. in Biometry and Statistics from the

Department of Crop Sciences at the University of Illinois.

Dr. Greg Holt is the Research Leader at the Cotton Production and Processing Research Unit in Lubbock,

Texas. He has conducted research related to development of more efficient pollution abatement devices for

cotton gins, optimization of cotton harvesting and ginning machinery, sensor development for microwave

imaging of cotton moisture, and value-added processing of cotton byproducts and agricultural fibers. He earned

B.S. and M.S. degrees from Texas A&M University in Agricultural Engineering and a Ph.D. from Texas Tech

University in Industrial Engineering.

Dr. Derek Whitelock is an Agricultural Engineer conducting cotton ginning and agricultural air quality

research at the USDA-ARS Southwestern Cotton Ginning Research Lab in Mesilla Park, New Mexico. He

earned B.S. and M.S. degrees from Texas A&M University and a Ph.D. from Oklahoma State University in

Agricultural Engineering. His research, mainly in the areas of air quality and seed-cotton and lint cleaning,

covers a broad spectrum of issues in both saw and roller ginning.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

Paul Funk Khairy Elsaye

ACKNOWLEDGEMENTS

The authors wish to recognize the ou

and Branyan Sanxter at the Southweste

Quality Laboratory.

REFERENCES

[1] CFR, "National ambient air qua

in Code of Federal Regulations

[2] CFR, "National Ambient Air Q

[3] P. Funk, "The environmental co

Manag. Assoc., vol. 60, no. 6, p

[4] M. Gillum, S. Hughs and B. Ar

Trans. ASABE, vol. 25, no. 1, p

[5] K. Baker, P. Funk and S. Hugh

Appl. Eng. Agric., vol. 20, no. 4

[6] C. Parnell and D. Davis, "Predi

processing particulate emission

[7] R. Baker and S. Hughs, "Evalu

no. 1, pp. 11-16, 1998.

[8] R. Baker and S. Hughs, "Influe

cyclone performance," Trans. A

[9] D. Whitelock, C. Armijo, M. B

Appl. Eng. Agric., vol. 25, no. 4

[10] C. Parnell, "Design of cyclone

18, 1980.

[11] C. Parnell, "Cyclone design for

in Beltwide Cotton Conf., Cord

[12] R. Baker and V. Stedronski, "C

Oil Mill Press, pp. 7-8, 17 June

[13] R. Wesley, W. Mayfield and O

gins. Technical Bulletin No. 14

ayed Kathleen Yeater Greg Holt

outstanding contributions of their technical staff, inclu

stern Cotton Ginning Research Laboratory and Bud W

quality standards for particultae matter; final rule.

ons, Washington, DC, U.S. Government Printing O

r Quality Standards for Particulate Matter; Final R

l cost of reducing agricultural PM2.5 emissions," J

6, pp. 681-687, 2010.

Armijo, "Use of secondary cyclones for reducing

, pp. 210-213, 1982.

ghs, "Over-sized cyclones for low pressure cotton

o. 4, pp. 413-415, 2004.

edicted effects of the use of the new cyclone desig

ions," in ASAE Southwest Regional Meeting, St Jo

aluation of two long-cone cyclone designs," Trans.

luence of air inlet and outlet design and trash exit s

s. ASABE, vol. 42, no. 1, pp. 17-21, 1999.

. Buser and S. Hughs, "Using cyclones effetively a

o. 4, pp. 563-576, 2009.

ne collectors to minimize dust emissions," Oil Mill

for air pollution abatement assiciated with agricult

ordova, Tenn., 1996.

"Collection efficiency of small diameter cyclones,

une 1967.

O. McCaskill, "An evaluation of the cyclone colle

1439," USDA Agricultural Research Service, Belt

15

Derek Whitelock

cluding Juan Gomez

Welch at the Air

le. 40 CFR, Part 50.,"

g Office, 2006.

l Rule," 2013.

J. Air and Waste

g gin emissions,"

on gin exhausts,"

signs on agricultural

Joseph, MI, 1979.

ns. ASABE, vol. 42,

it size on 1D3D

y at cotton gins,"

ill Gazetteer, pp. 16-

ultural operations,"

es," Cotton Gin and

ollector for cotton

eltsville, MD, 1972.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

16

[14] M. Gillum and S. Hughs, "Velocity effects on operating parameters of series cyclones," Trans.

ASABE, vol. 26, no. 2, pp. 606-609, 1983.

[15] W. Faulkner and B. Shaw, "Efficiency and pressure drop of cyclones across a range of inlet

velocities," Appl. Eng. Agric., vol. 22, no. 1, pp. 155-161, 2006.

[16] R. Dewil, J. Baeyens and B. Caerts, "CFB cyclones at high temperature: operational results and

design assessment.," Particuology, vol. 6, no. 3, pp. 149-156, 2008.

[17] L. Svarovsky, "Solid-Gas Separation," in Gas fluidization technology, New York, John Wiley

and Sons, Ltd., 1986, p. Chapter 8.

[18] K. Elsayed, "Design of a novel gas cyclone vortex finder using the adjoint method," Separation

and Purification Technology, vol. 142, pp. 274-286, 2015.

[19] ASTM, "Standard D2986-95a. Standard practice for evaluation of air assay media by the

monodisperse DOP (Dioctyl Phthalate) smoke test (Withdrawn 2004)," ASTM International,

West Conshohocken, PA, 1999.

[20] Environmental Protection Agency (EPA), "Appendix L: Reference method for the determination

of fine particulate matter as PM2.5 in the atmosphere," in 40 CFR 50, Washington, DC, U.S.

Government Printing Office, 1997.

[21] P. Funk and K. Baker, "Dust cyclone technology for gins - a literature review," J. Cotton Sci.,

vol. 17, no. 1, pp. 40-51, 2013.

[22] ASTM, "Standard C136. Standard test method for sieve analysis of fine and coarse aggregates,"

ASTM International, West Conshohocken, 2006.

[23] S. Hughs and R. Baker, "Effectiveness of model cyclone designs in collecting gin trash

particulate emissions," Appl. Eng. Agrc., vol. 14, no. 3, pp. 317-322, 1998.

[24] P. Funk, S. Hughs and G. Holt, "Entrance velocity optimization for modified dust cyclones," J.

Cotton Sci., vol. 4, pp. 178-182, 2000.

[25] P. Funk, S. Hughs and G. Holt, "Dust cyclone design," Appl. Eng. Agric., vol. 17, no. 4, pp. 441-

444, 2001.

[26] W. Faulkner, M. Buser, D. Whitelock and B. Shaw, "Effects of cyclone diameter on performance

of 1D3D cyclones: collection efficiency," Trans. ASABE, vol. 50, no. 3, pp. 1053-1059, 2007.

[27] W. Faulkner, M. Buser, D. Whitelock and B. Shaw, "Effects of cyclone diameter on performance

of 1D3D cyclones: cutpoint and slope," Trans. ASABE, vol. 51, no. 1, pp. 287-292, 2008.

[28] CFR, "Method 201A - Determination of PM10 and PM2.5 emissions from stationary sources

(constant sampling rate procedure)," in 40 CFR 51 Appendix M, 2010.

[29] A. Hoffman, A. v. Santen, R. Allen and R. Clift, "Effects of geometry and solid loading on the

performance of gas cyclones," Powder Tech., vol. 70, pp. 83-91, 1992.

[30] EPA, "AP 42 Fifth Edition Volume I; Chapter 9: Food and Agricultural Industries; 9.7: Cotton

Ginning," 1996. [Online]. Available: http://www.epa.gov/ttn/chief/ap42/ch09/final/c9s07.pdf.

[31] L. Wang, C. Parnell, B. Shaw and R. Lacey, "A theoretical approach for predicting number of

turns and cyclone pressure drop," Trans. ASABE, vol. 49, no. 2, pp. 491-503, 2006.

[32] C. Shepherd and C. Lapple, "Flow pattern and pressure drop in cyclone dust collectors,"

Industrial & Eng. Chem., vol. 31, no. 8, pp. 972-984, 1939.

[33] C. Cortes and A. Gil, "Modeling the gas and particle flow inside cyclone separators," Progress in

Energy and Combustion Sci., vol. 33, pp. 409-452, 2007.

[34] C. Lapple, "Processes use many collector types," Chem. Eng., vol. 58, pp. 144-151, 1951.

[35] W. Barth, "Design and layout of the cyclone separator on the basis of new investigations.,"

Brennst. Warme Kraft, vol. 8, pp. 1-9, 1956.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

17

[36] D. Leith and W. Licht, "The collection efficiency of cyclone type particle collectors: a new

theoretical approach.," in AIChE Symposium Series, 1972.

[37] Y. Zhu and K. W. Lee, "Experimental study on small cyclones operating at high flowrates," J.

Aerosol Sci., vol. 30, no. 10, pp. 1303-1315, 1999.

[38] H. Mothes and F. Löffler, "Prediction of particle removal in cyclone separators," International

Chem. Eng., vol. 28, pp. 231-240, 1988.

[39] E. Li and Y. Wang, "A new collection theory of cyclone separators," AIChE J., vol. 35, pp. 666-

669, 1989.

[40] P. Funk and R. Hardin IV, "Cotton gin electrical energy use trends and 2009 audit results," Appl.

Eng. Agric., vol. 28, no. 4, pp. 503-510, 2012.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

18

CAPTIONS FOR FIGURES AND TABLES

Figure 1. Modified 1D3D cyclones used in this study were 30.5 cm (12 in) in diameter. This design, both with

and without the expansion chamber, is widely used in agricultural processing. Note the protrusion of the

vortex finder below the tangential inlet and the cone below the top of the expansion chamber. Two cyclones

were built to the same specifications for this experiment. Dimensions are in cm.

Figure 2. Plot showing total suspended particulate gravimetric collection efficiency (percent) for different inlet

velocities (meters per second) tabulated by four research teams over five decades. Cyclones were 30 to 76 cm

(12 to 30 inches) in diameter. These results illustrate that dust cyclone performance is relatively insensitive to

cyclone inlet velocity. Sources: Baker and Stedronski (1967); Wesley, Mayfield and McCaskill (1972);

Gilum and Hughs (1983); Faulkner and Shaw (2006) where “2006a” = alumina and “2006c” = cornstarch as

the test particulate. The design inlet velocity of a 1D3D cyclone, 16.26 m s-1

(3200 ft min-1

) is indicated by the

vertical line.

Figure 3. Experiment apparatus. A: variable speed test material feed conveyor; B: variable speed trash fan; C:

Venturi tube flow meter; D: removable Pitot-static tube; E: test cyclone with sealed collection bucket; F: filter

to collect cyclone losses; G: variable speed filter fan; H: fixed Pitot-static tube; T: temperature sensing

(thermocouple) locations. Pressure sensing locations: Venturi tube (1-2); cyclone pressure loss (3-4); and filter

pressure loss (5-6). Not to scale.

Figure 4. Experiment design showing the combination of the two controlled variables, inlet velocity and inlet

concentration. Each circle represents the target values for test conditions for one test of each cyclone. Five

replicates were run at the center (39 g m-3

and 13 m s-1

).

Figure 5. Sieved test material aerodynamic equivalent diameter size distribution (particles < 106 µm).

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

19

Figure 6. PM2.5 emissions (in g kg-1

test material entering cyclone) over the tested range of inlet concentration

and velocity. The response surface is bounded by actual data points; gradations are uniformly distributed

between them. Emissions per kg of material entering the cyclones increased significantly with increasing inlet

velocity and emissions increased with decreasing inlet concentration.

Figure 7. PM2.5 emissions (in g kg-1

test material entering cyclone) as a function of Stokes number for PM2.5

separated by the cyclone. Normally, emissions would be greater at lower Stokes numbers, where drag forces

are greater than inertial forces. This result, for heterogeneous particulate, indicates that other factors may be

present that the Stokes number does not account for. Perhaps particle agglomeration, a mechanism for fine

particulate capture, is favored by lower turbulence.

Table 1. Size distribution of test material (gin trash) used in the experiment.

Table 2. Test statistics indicating the influence of uncontrolled and controlled variables for each response

variable (bold indicates significance).

Table 3. Selected response surface models for cyclone performance (PM2.5 emissions per kg fed and pressure

drop) as functions of inlet concentration (g m-3

) and actual inlet velocity (m s-1

). Model statistics are from

Design Expert 8.0.7.1 (2010, Stat Ease, Minneapolis, Minnesota).

Table 4. Response surface model results and 95% confidence intervals for two velocities at 45 g m-3

inlet

concentration.

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

20

Figure 1

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

Figure 2

21

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

22

Figure 3

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

Figure 4

23

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

24

Figure 5

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

Figure 6

25

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

26

Figure 7

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

27

Table 1

Particle Size

< 2.5

µm

< 10

µm

< 106

µm

< 150

µm

< 250

µm

< 850

µm

> 850

µm

Category

Percent 0.90% 5.10% 34.80% 6.10% 7.80% 31.20% 14.00%

Cumulative

Total 0.90% 6.00% 40.80% 46.90% 54.80% 86.00% 100.00%

Table 2

PM2.5 (g/kg fed)

Pressure Drop (Pa)

and Euler Number

(Eu)

Statistic F Value Pr > F F Value Pr > F

Uncontrolled

Variables

Run Number 0.46 0.5051 0.02 0.8818

Date 0.72 0.4083 0.21 0.6545

Barometric Pressure 0.21 0.6552 0.84 0.374

Temperature 0.25 0.6251 1.1 0.3099

Relative Humidity 0.03 0.8656 2.59 0.127

Air Density 0.25 0.6268 0.9 0.358

Cyclone A or B 0.11 0.7474 0.14 0.7095

Controlled

Variables

Inlet Velocity 15.92 0.0011 237.86 <.0001

Inlet Concentration 8.16 0.0114 1.29 0.2734

Velocity*Concentration 2.73 0.119 2.66 0.1238

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

28

Table 3

Response [a]

Model [b]

Model adj.

R2

Std.

Dev.

PM2.5 (g/kg) = + 0.0051617 + 0.036549*V – 0.0028133*C 0.6599 0.095

∆P (Pa) = + 80.679 – 14.504*V + 2.5314*V2 0.9396 40.87

PM2.5 (g/kg) = + 0.0051617 + 0.42351*Stk – 0.0028133*C 0.6599 0.095

Eu = + 21.72 + 0.074*(80.679 - 168.065*Stk +

339.89*Stk2)

0.9534 3.024

[a] Eu = Euler Number

[b] V = Inlet Velocity (m s

-1), C = Inlet Concentration (g m

-3), Stk = Stokes Number

Table 4

Response Inlet Velocity = 16.26 m s

-1 (3200 fpm)

(Stokes Number, Stk = 1.403)

Inlet Velocity = 12.19 m s-1

(2400 fpm)

(Stokes Number, Stk = 1.052)

Estimate 95% Confidence

Interval Estimate

95% Confidence

Interval

PM2.5

(g/kg) 0.473 0.438 - 0.511 0.324 0.289 - 0.362

∆P (Pa) 514 498 - 530 280 264 - 296

Eu 59.8 58.6 - 60.9 42.4 41.3 - 43.6

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

Graphical Abstract

29

AC

CEPTED

MAN

USC

RIP

T

ACCEPTED MANUSCRIPT

30

Highlights

• The significance of these results comes from the nexus of emissions and energy.

• Producing electricity to power abatement devices results in air pollution.

• Reducing inlet velocity reduced pressure drop and hence energy required.

• Empirical PM2.5 emissions also appeared to decrease at lower inlet velocities.