Newtonian flow past a hollow frustum in vertical and inclined plane: An
experimental observation for terminal velocity and drag coefficient

Saroj Kumar Samantaray, Bijay Kumar Rout, Soumya Sanjeeb Mohapa-

tra, Basudeb Munshi

PII: S0032-5910(17)31007-0
DOI: doi:10.1016/j.powtec.2017.12.049
Reference: PTEC 13038

To appear in: Powder Technology

Received date: 4 May 2017

Revised date: 5 December 2017
Accepted date: 14 December 2017

Please cite this article as: Saroj Kumar Samantaray, Bijay Kumar Rout, Soumya Sanjeeb
Mohapatra, Basudeb Munshi, Newtonian flow past a hollow frustum in vertical and
inclined plane: An experimental observation for terminal velocity and drag coefficient,
Powder Technology (2017), doi:10.1016/j.powtec.2017.12.049

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 proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Newtonian Flow past a Hollow Frustum in Vertical and Inclined Plane:

An Experimental Observation for Terminal Velocity and Drag Coefficient

Saroj Kumar Samantaray 1, Bijay Kumar Rout 2, Soumya Sanjeeb Mohapatra 1and

PT
Basudeb Munshi 1*
1
Department of Chemical Engineering, National Institute of Technology Rourkela - 769008,

RI
India.
2
Department of Mechanical Engineering, Birla Institute of Technology and Science Pilani, India

SC
Abstract

NU
This work included the experimental studies of the Newtonian fluid flow over hollow frustum

both in the normal and inclined channel. It encompasses terminal velocity and drag coefficient,
MA
CD data for 0.13 ≤ 𝑅𝑒 ≤ 8.41, 0.19 ≤ 𝑑𝑜 ⁄𝐷 ≤ 0.33 , 0.22 ≤ 𝑑𝑖 ⁄𝑑𝑜 ≤ 0.83 and 40𝑜 ≤ 𝜃 ≤

90𝑜 . The effect of inner diameter to outer diameter ratio, di/do and the outer diameter to the flow
D

channel diameter ratio, do/D on terminal velocity is reported for the specified range of angle of
TE

inclination, θ for several hollow frustums in a series of high viscous Newtonian fluids. The
P

terminal velocity shows an increasing trend with increasing and decreasing do/D and di/do ratios,
CE

respectively. The terminal velocity also increases with increasing the angle of inclination, θ of

the flow channel. Predictive equations are developed for the estimation of CD as a combined
AC

function of Re, di/do, do/D and 𝜃. The statistical sensitivity analysis shows a large variation of

the drag coefficient of the hollow frustum with the Reynolds number than with 𝑑𝑖 ⁄𝑑𝑜 and 𝑑𝑜 ⁄𝐷

ratios. The dependency of CDRe term on the fluid viscosity and θ are expressed in terms of the

developed correlations. Linear variation of CD with 1/Re confirms that flow regime is laminar in

the present study. The experimental drag coefficients are then predicted excellently through

numerical approximation using Reynolds Stress Model (RSM) available in Ansys-15.

Keywords: Hollow frustum, terminal velocity, diameter ratio, drag coefficient, Reynolds number

1
 ACCEPTED MANUSCRIPT

*
Author to whom all the correspondence should be addressed:

Dr. Basudeb Munshi, Department of Chemical Engineering, NIT Rourkela, Odisha, India-
769008, Email id: [email protected], Ph. No.: +91-661-2462259.

PT
1. Introduction

RI
In process industries like fluidized bed reactor, processing of food particle, conveying

SC
solid particle in fluid channel, chemical blending, mineral processing and other solid handling

processes, several kinds of solid and hollow objects are involved. The resistance or drag offered

NU
by a fluid medium, due to relative motion, on the particle is a critical parameter for process
MA
design. The flow over regular shaped particles i.e. sphere, cylinder and the circular disk had

received the maximum attention. Clift et al. [1] and Chhabra [2] summarized details of flow over
D

solid particles in both Newtonian and non-Newtonian media. In numerous processes, the
TE

particles may be sliding or rolling over the inclined planes. For example, the flow of fluid in

inclined packed and fluidized beds, dredging underwater silt and gravel, and storage in large
P
CE

containers, etc. As compared to vertical channel, quite a few works were reported on the flow

over solid objects especially for non-spherical particles in inclined channel. Several kinds of
AC

hollow particles are used in chemical process industries especially as a packing material.

Unfortunately, in literature, no attempt has been communicated to study the fluid dynamic

behavior of flow over hollow object either in vertical or in inclined flow channels. Hence, in this

work, the Newtonian flow over hollow frustum is discussed to provide the supplements to the

existing literature. Before describing the present work, it is useful and instructive to report a brief

record of the background of the present work.

Many researchers, Clift et al.[1], Chhabra [2], Khan and Richardson [3], Uhlherr and

Chhabra [4] studied the affect of the confined channel wall on the drag coefficient. The flow

2
 ACCEPTED MANUSCRIPT

over cylinder was studied by Unnikrishnan and Chhabra [5]. Nitin and Chhabra [6] and Munshi

et al. [7] also reported numerical analyses of flow over circular disk. Chhabra [8] and Sharma and

Chhabra [9] estimated the drag coefficient for non-spherical particles.

PT
The previous studies as discussed above were limited to vertical flow channel. Only few

RI
studies were carried out to estimate the drag coefficient for the inclined flow channel. Carty [10]

SC
concluded that the drag in the inclined plane is larger than the vertical plane due to additional

retardation force from inclined plane. Garde and Sethuraman [11] also conducted the experiment

NU
in both smooth and rough inclined plane. However, the experimental outcomes of Carty [10] did
MA
not match with Garde and Sethuraman [11] , due to the differences in length of flow domain and

ratio of the diameter of sphere (d) to flow channel (D), d/D. [Jan and Chen [12],Chhabra and
D

Ferreira, [13], Chhabra et al. [14], Jalal and Ganji [15] also established set of CD-Re relationships
TE

varying d/D ratio and angle of inclination for the inclined plane filed with incompressible

Newtonian fluid. In all the above literatures, only sperical particle was considerd.
P
CE

In preceding paragraphs, the literatures included the drag coefficient data of solid regular

shaped body in the vertical and inclined flow channels. To the best of the author’s knowledge, no
AC

attempt was reported for the drag coefficient due to flow of Newtonian fluid over any hollow

particle. Looking at the usefulness of the hollow particles in process industries, to fill the void

and to enrich the existing literature, the present work is undertaken to estimate the drag

coefficient of hollow tapered cylindrical body called hollow frustum (Fig. 1A) flowing in

Newtonian fluid in both the vertical and inclined channels. Pure glycerin and its aqueous

solutions are used as the working fluid in the present study.

Based on the literature gaps, the following objectives are adopted for the present work.

3
 ACCEPTED MANUSCRIPT

 To estimate the effect of the angle of inclination, θ on the terminal velocity and drag

coefficient of the hollow frustum, which is still unreported in open literature.

PT
To estimate the effect of the inner diameter to outer diameter ratio (di/do) of the hollow

frustum and the outer diameter of the hollow frustum to the flow channel diameter ratio

RI
(do/D) on the measured terminal velocity and estimated drag coefficient.

SC
 To develop the correlation for the drag coefficient, CD as function of Re, di/do and do/D

for different angle of inclination, θ and to identify the most dominating parameter among

NU
Re, di/do and do/D to vary the drag coefficient the most.
MA
 To predict the experimental drag coefficients using Ansys Fluent software.

2. Physics around particle

D
TE

The schematic representation of the hollow frustum is given in Fig.1. When a hollow

frustum is moving down in fluid inside an inclined channel, the forces exerting on hollow
P

frustum are drag force (FD), the net body weight of the hollow frustum (𝐹𝐵 sin 𝜃), and solid–
CE

solid resistance (FR). The force balance equation at equilibrium is

AC

𝐹𝐵 sin 𝜃 = 𝐹𝐷 + 𝐹𝑅 (1)

The expression of drag force is

1
𝐹𝐷 = 2 𝐶𝐷 𝜌𝐹 𝑉 2 𝐴 (2)

For smooth plane with 𝐹𝑅 zero, the expression of 𝐶𝐷 becomes

𝜌
𝑚𝑔(1− 𝐹 )
𝜌𝑝
𝐶𝐷 = 2
𝑠𝑖𝑛𝜃 (3)
0.5𝐴𝑉 𝜌𝐹

For vertical channel consideration, Eq. 3 is reduced to

4
 ACCEPTED MANUSCRIPT

𝜌
𝑚𝑔(1− 𝐹 )
𝜌𝑝
𝐶𝐷 = 2
(4)
0.5𝐴𝑉 𝜌𝐹

The diameter ratios (di/do, do/D) and the angle of inclination are the vital parameters to affect

PT
the drag coefficient in inclined plane consideration, hence the total drag coefficient 𝐶𝐷 may be

RI
written as follows

SC
𝐶𝐷 = f (Re, do/D, di/do,𝜃) (5)

3. Experimental

NU
3.1. Materials MA
Hollow frustums (SS 304L) of different dimension (listed in Table 1) and density

7859.56 kg/m3 are machined in wire EDM (Electrical Discharge Machining) and used as hollow
D
TE

obstacles in 30.1×1200 mm circular Perspex tube. The viscosities of working fluid are measured

using Rehometer (RheolabQC, Anton Parr, Germany,) and the densities are estimated by
P

pycnometer. The rheological properties of test fluids are listed in Table 2.

CE

3.2 Experimental setup and Procedure

AC

The inclination angles (900-400) are varied by rotating the moveable end of the tube

around the fixed end. The schematic representation of the experimental setup is shown in

Fig. 1 (B). The test fluids are loaded in the tube 24hrs before starting the experiment to make the

fluid bubble free [12 & 14]. The test location is taken sufficiently away from the inlet to allow

the frustum to reach constant terminal velocity. The hollow frustums are released by forceps and

stopwatch is used to measure settling time. Every experiment is repeated for five times and mean

settling time is taken for the calculation of terminal velocity followed by Reynolds number and

CD.

5
 ACCEPTED MANUSCRIPT

4. Result and Discussion

PT
4.1. Effect of do/D and di/do on measured terminal velocity

Measured terminal velocity is one of the vital parameters for the estimation of drag coefficient

RI
and Reynolds number, and it depends on the ratio of the particle and channel diameters [5 & 9].

SC
In this study, the particle is hollow; hence, it is believed that, the ratio of inner diameter, di to the

NU
outer diameter, do of the hollow frustum may possess some effect on the terminal velocity. Fig. 2

represents the variation of measured terminal velocity of the hollow frustum with do/D using
MA
di/do as the parameter. The angle of inclination is 90o here. The do/D ratio is varied by changing

do keeping D fixed. For a particular di/do, the increase of do makes the particles heavier. Thus,
D

the terminal velocity of the frustum increases with increasing the do/D ratio. The increasing trend
TE

of the velocity with do/D ratio contradicts the declining trend of it for fixed diameter, d solid
P

spherical, cylindrical and conical particle [5 & 9]. In the latter case, the d/D ratio was increased
CE

by reducing the flow channel diameter, D for a particular sphere diameter, d which thus
AC

decreased the terminal velocity due to increasing the wall effect on the free falling sphere. In the

present study, the di/do ratio is also an important parameter. The hollow frustum becomes thinner

and lighter with increasing the di/do ratio for a particular do. Thus, the terminal velocity as shown

in Fig. 2 decreases with increasing di/do.

The effect of angle of inclination (900-400) on the terminal velocity is illustrated in Fig.3

for pure glycerin. It shows an increasing trend of the terminal velocity with do/D ratio for all the

angle of inclinations. The most profound variation of the terminal velocity with the do/D ratio is

observed for vertical flow channel and it becomes less with decreasing the angle of inclining.

6
 ACCEPTED MANUSCRIPT

This occurs due to decreasing the effect of body force on the terminal velocity with decreasing

the angle of inclination of the flow channel. For other operating fluids, the similar trend is

observed.

PT
4.2. Drag coefficient and Reynolds number relationship

RI
The drag coefficient and Reynolds number are calculated using the fundamental

SC
properties of the hollow frustum and process fluid and the terminal velocity of the hollow

frustum for different angle of inclination, do/D and di/do ratios. The orientation of the hollow

NU
frustum was normal to the flow direction for the 900, and horizontal for all other acute angles.
MA
Hence, appropriate projected areas and characteristic lengths are considered for the estimation of

CD and Re.
D

According to Eq. 5, drag coefficient is function of Re. The effect of Re on CD is shown in

TE

Fig.4. It shows the expected decreasing trend of CD with Re for all the angles of inclination. The
P

variation of drag coefficient with Reynolds number is matching with the trend obtained by the
CE

previous investigators for regular shaped particle [1-3]. The CD value is obtained maximum for
AC

the vertical channel where the orientation of hollow frustum is normal as shown in Fig. 1(A). For

the other inclined tubes, the moving hollow frustum becomes horizontal shown in Fig.1 (B) and

thus the drag coefficient increases with decreasing the θ. The highest value at 900 is obtained due

to normal orientation which gives lesser projected area than other inclined channels.

For 40o to 80o inclinations, projected areas are same, but as mentioned above the terminal

velocity decreases with decreasing the angle of inclination. According to Eq. 3 and 4 CD is

inversely proportional to V2. Hence, it increases with decreasing the 𝜃 values. The drag

coefficient is also inversely varied with the projected area, A, which is the minimum for the

7
 ACCEPTED MANUSCRIPT

normally oriented frustum. CD is therefore the maximum for the vertical channel in spite of

having the highest terminal velocity.

Empirical expressions of CDRe for sphere are available in Hasan [16] and Duan et al. [17]. Hasan

PT
[16] developed the following Reynolds number and d/D ratio dependent expression for CD for

RI
spherical particles falling in inclined channel.

SC
𝐶𝐷 𝑅𝑒 = 15.717{1 − (𝑑 ⁄𝐷)}−2.5 [16] (6)

NU
He did not explore any effect of the angle of inclination on the drag coefficient expression. The

present work has developed expression of CD given in Eq. 13 and 14 which includes the
MA
dependency of it on all the variables like Re, di/do and do/D ratio. The coefficients of the

correlation in the equation are also function of the inclination, 𝜃. The least-square method is used
D

to find the value of the coefficients i.e. A, B, C, E and F for a particular n, m and k of Eq. 13 and
TE

P, Q, R and S of Eq. 14. For both the correlations i.e. for Eq.13 & 14 the estimated statistical
P

deviations are given in Table 3 and 4. The R2 values closer to 1.0 and other statistical parameter
CE

in the tables assure the viability of both the correlations. The expressions for the statistical
AC

deviations are as follows [18&19]:

Root Mean Square of Logarithmic Deviation (RMSLD)

2
(log 𝐶𝐷 −log 𝐶𝐷 ^ )
√∑𝑁
1 ( ) (7)
𝑁

Sum of Squared of Logarithmic Deviations (SSLD)

2
(log 𝐶𝐷 −log 𝐶𝐷 ^ )
∑𝑁
1 ( ) (8)
𝑁

8
 ACCEPTED MANUSCRIPT

Average Relative Error (ARE)

(𝐶𝐷 −𝐶𝐷 ^ )
[∑𝑁
1 ( ) / 𝑁] × 100 (9)
𝐶𝐷

PT
Sum of Relative Error (SRE)

RI
(𝐶𝐷 −𝐶𝐷 ^ )
[∑𝑁
1 ( )] (10)
𝐶𝐷

SC
Sum of Squared Relative Errors (SSRE)

NU
2
(𝐶𝐷 −𝐶𝐷 ^ )
[∑𝑁
1 ( ) ] (12)
𝐶𝐷
MA
Where 𝐶̂𝐷 is the computed drag coefficient using Eq.13 & 14.
In case of Eq.13, the declining trend of CD with the variations of Reynolds number and di/do,
D

and do/D ratios leads to negative values of n, m and k. The tabulated data (in Table 3) shows that
TE

the magnitude of E and F is much lower than A and B, which shows weak effect of di/do and
P

do/D ratio on the drag coefficient than the Reynolds number. Similar conclusion is drawn for the
CE

empirical model Eq.14, where the magnitude of Q (exponent of Re) is much higher than the R
AC

and S, the exponent of di/do and do/D, respectively (Table 4). These agree with the conclusion

drawn in [14]. The parameters of Eq. 13 i.e. A, B, C, E, F, n, m and k and P, Q, R and S of Eq.

14 are further analyzed and modeled as a function of 𝜃 and given in Table 5 and 6. The

graphical comparison between the drag coefficient calculated by using empirical model

equations and the experimentally obtained drag coefficient is performed in Fig. 4.

𝑑 𝑚 𝑑 𝑘
𝐶𝐷 = 𝐴 + 𝐵𝑅𝑒 −1 + 𝐶𝑅𝑒 𝑛 + 𝐸 (𝑑 𝑖 ) + 𝐹 ( 𝐷𝑜 ) (13)
𝑜

𝑑 𝑅 𝑑 𝑆
𝐶𝐷 = 𝑃 × 𝑅𝑒 𝑄 × (𝑑 𝑖 ) × ( 𝐷𝑂 ) (14)
𝑜

9
 ACCEPTED MANUSCRIPT

According to Eq. 6, the product term 𝐶𝐷 𝑅𝑒 is function of only (d/D) ratio. To examine it

in the present study, the 𝐶𝐷 𝑅𝑒 data of different Newtonian fluids are plotted in Fig. 5 against

(do/D) ratio keeping di/do = 0.5. The figure depicts that 𝐶𝐷 𝑅𝑒 varies with both the fluid viscosity

PT
and angle of inclination along with the Reynolds number. Fig. 5 portrays a decreasing trend of

RI
CDRe with viscosity for both the angle of inclination owing to well-known inverse relation of

SC
terminal velocity with viscosity. The magnitudes of C (stands for viscosity affect in Eq. 13) as

given in Table 3 are lower than B for all the tube inclinations, which confirms the higher effect

NU
of Re on the drag coefficient than the viscosity. The terminal velocity of the hollow frustum

decreases with increasing the fluid viscosity and thus, a lesser variation of 𝐶𝐷 𝑅𝑒 with do/D ratio
MA
is observed in the figure.
D

The variation of the drag coefficient with 1/Re is studied in Fig. 6 where all the present
TE

drag coefficients are plotted against 1/Re. The range of 1/Re in the figure confirms that all the

present experimental data are valid in laminar flow state. The curves in the figure are more linear
P
CE

for lower angles of inclination of the flow channels and at lower Reynolds number. The

deviation from linearity at lower 1/Re i.e. at relatively higher Reynolds number and especially
AC

for the higher 𝜃 is because of the appearance of turbulence in the flow domain.

5. Sensitivity Analysis of Empirical Models

The empirical model equations (Eq. 13 and 14) contain three different input variables like di/do

and do/D ratios and Reynolds number, Re. The model equations are solved to investigate its

sensitivity to the change of the inputs. It is also very much necessary to isolate the most

dominating parameter(s) to affect the drag coefficient. The present study desires to identify it

using the statistical sensitivity analysis method. During the analysis level values of the inputs

which are given in Table [7] are selected for the application of design of experiments approach.

10
 ACCEPTED MANUSCRIPT

The statistical analysis of the first empirical model (Eq. 13) is discussed here. The

experimental combinations for the sensitivity study are generated by selecting 23 full factorial

designs. These combinations allow comparison of the effects of each factor at two level values.

PT
The generated design combinations are provided in Table 8 with the computed CD values at

different angle of inclinations, θ of the flow channel. The simulated CD values are consolidated

RI
and various statistical tests are applied on the data to perform the statistical sensitivity analysis.

SC
In this regard, the mean effects of each factor are estimated and Analysis of Variance (ANOVA)

NU
technique required for the analysis of full factorial (23) designs is performed. Representative

results of the analysis for θ value equal to 900 are presented. Fig. 7 shows that the change in level
MA
values of Re has significant effect on the CD values. Remaining factors (di/do and do/D) do not

affect much the CD values. The associated ANOVA table for the simulated results is given in
D
TE

Table 9. The ANOVA table indicates that the level change of Re has significant influence on the

first empirical model i.e. on the CD values computed by using Eq. 13 as compared to the effect
P

of the level change of the di/do or di /D factors. The computations details of getting ANOVA
CE

table is available in [20]. Here the intent is to distribute the variability present in the simulated
AC

data using the empirical model among the identified factors. The column titles Adj. SS stands for

Adjusted Sum of squares and Adj. MS stands for adjusted Mean Square obtained from the Table

9. The observed statistical sensitivity results for other channel inclinations are similar to the

above.

In the similar fashion the statistical sensitivity analysis of the predictive empirical second model

(Eq. 14) is carried out. The associated computed drag coefficient values for the full factorial (23)

designs are given in Table 10 for different tube inclinations. Here also the simulated values of

11
 ACCEPTED MANUSCRIPT

CD are consolidated and statistical sensitivity tests are conducted. The mean effects of each factor

are obtained and ANOVA technique is applied. A representative result of the analysis for θ equal

to 900 is also presented here. Fig. 8 shows that the change in level values of Re has significant

PT
effect on the Cd values. Remaining factors (di/do and do/D) do not have any effect on the CD

RI
values. The associated ANOVA table for the simulated result is given Table 11. The ANOVA

table (Table 11) too indicates that the level change of Re has significant influence on the

SC
empirical model i.e. Eq. 14 with the high change of CD values due to the change of Re as

NU
compared to the level change of the di/do or di/D factors.

Overall, it can therefore be concluded that the change of Reynolds number on the drag
MA
coefficient has large effect as compared to negligible effect due to change of the di/do and do/D

ratios.
D
TE

6. Prediction of Drag Coefficient through Numerical Estimation

P
CE

Experimental drag coefficients are predicted numerically, by employing the three dimensional –

pressure based –steady state CFD simulation using the commercial Fluent module of Ansys-15.
AC

The following steady state model equations are solved under consideration in the present study.

∇. (𝜌𝑢) = 0 (15)

∇. (𝜌𝑢𝑢) = −∇𝑃 + ∇𝜏 + 𝜌𝑔 (16)

Where 𝜌 is the fluid density, 𝑢 is the velocity of the fluid, 𝑃 is the pressure, 𝜏 is the stress tensor

and, 𝑔 is the acceleration due to gravity.

Hollow frustum is placed vertically at the center of the flow domain in the vertical channel.

However, for the inclined channel, the horizontal frustum is placed off-centered in the mid height

12
 ACCEPTED MANUSCRIPT

of channel. The exact off-centered location of the hollow frustum is determined by adjusting its

position to match the experimental data. In the present study, for all the channel inclinations

(<90o) and diameter rations, the hollow frustum is kept 0.87 mm away from the wall. The top

PT
view of the generated mesh is shown in Fig. 9 for both the vertical and inclined channels. At the

RI
inlet of the flow channel, experimentally obtained terminal velocity is assigned. The surface of

the hollow frustum and the wall of the flow domain are assigned as wall with no-slip condition.

SC
The boundary, opposite to inlet is specified as pressure outlet. The acceleration due to gravity is

NU
used as gsinθ for all the tube inclinations.

6.1. Solution Strategy

MA
According to the range of the experimental Reynolds number flow domain falls in laminar zone,
D

hence, the laminar model is tried first, however, it is observed that after Re = 2.82, gradually the
TE

difference between the experimental and simulated drag coefficient increases. In order to reduce
P

the error various turbulent models available in Ansys Fluent are tested, but a consistent minimum
CE

error is found with the Reynolds Stress Model (RSM) for all Re. Hence, the results obtained with

RSM model are compared with the experimental data here. The SIMPLE (Semi-Implicit Method
AC

for Pressure Linked Equations) algorithm is used to couple the pressure and momentum terms.

All transport equations are discretized using second order upwind difference scheme. The

convergence criterion for each transport equation is set to be 10-5.

6.2. Mesh Independency Test

Meshing is very important part for the solution purpose and the sensitivity on the mesh quality

and quantity affect largely the results. The triangular meshes with fine relevance center and high

smoothing are chosen. After 85-relevance center, no further improvements in the results are

13
 ACCEPTED MANUSCRIPT

noticed. The number of nodes and elements are varied along with the diameter ratio. The mesh

independency tests are carried. In Table 12 the changes of CD values with number grid is given

for Re = 0.359475 with do/D= 0.33 and di/do=0.7. This schema is also tested for other Reynolds

PT
numbers, di/do and do/D.

RI
SC
6.3. Discussion on the Predicted Results

NU
It is always beneficial to validate the accuracy of the model approximation with the experimental

results. The obtained drag coefficient values in each simulation are compared with the
MA
experimental data in Fig. 10. The figure shows an excellent prediction of CD by RSM model. The

statistical error analysis is also carried in line of Eq. (7-12) and presented in Table 13. The
D

minimum deviation and R2 values close to 1.0 are obtained in all the cases .According to the
TE

range of the Reynolds number, it is expected not to have any separation of flow near top the
P

hollow frustum. In agreement to this, a fore and aft symmetry is observed in Fig. 11 for both the
CE

vertical and inclined channels. It confirms the laminar flow behavior of fluid flow over the
AC

hollow frustum and it accords the experimental laminar flow regime according to the Reynolds

number. In view of these observations, it is thus safe to claim the applicability and viability of

the RSM models of Ansys Fluent to predict the drag coefficient of hollow frustum.

7. Conclusions

In this work, the experimental drag coefficient of the hollow frustum particles of different

dimensions settling in both the vertical and inclined plane of constant diameter was reported to

show the effect of the channel inclination, ratio of the physical dimensions and Reynolds number

on the variation of the drag coefficient of the particle in Newtonian fluids. The experimental drag

14
 ACCEPTED MANUSCRIPT

coefficient of the hollow frustum are valid for 0.13 ≤ 𝑅𝑒 ≤ 8.41, 0.19 ≤ 𝑑𝑜 ⁄𝐷 ≤ 0.33

, 0.22 ≤ 𝑑𝑖 ⁄𝑑𝑜 ≤ 0.83 and 40𝑜 ≤ 𝜃 ≤ 90𝑜 . The terminal velocity was increased monotonically

with the frustum outer diameter to channel diameter ratio (do/D) at a particular di/do. The

PT
increasing trend was just opposite of the decreasing trend of the drag coefficient of the regular

RI
shaped solid particles. The terminal velocity of the hollow frustum at any do/D ratio was

SC
increased with enhancing the di/do ratio. It was also increased due to increase of the downward

body force with the angle of inclination. CD value was enhanced with decreasing the angle of

NU
inclination of the flow channel. Appropriate predictive correlations for CD as a function of Re,

di/do, do/D and θ were developed using the least square method. Good prediction capabilities of
MA
the predictive empirical models were observed. The Reynolds stress model (RSM) of Ansys

Fluent was used successfully to predict the experimental drag coefficients for the complete range
D
TE

of experimental Re. The angle of inclination and fluid viscosity sensitive 𝐶𝐷 𝑅𝑒 term showed a

declining trend with do/D ratio. A linear variation of CD with 1/Re was observed at lower angles
P

of inclinations; it occurred due to the dominations of the viscous force over gravitational force.
CE

The statistical sensitivity analysis showed a large dependency of the drag coefficient of the
AC

hollow frustum on the Reynolds number as compared to its insignificant variations due to change

of 𝑑𝑖 ⁄𝑑𝑜 and 𝑑𝑜 ⁄𝐷 ratios.

Nomenclatures

Symbols

𝐴 Projected area (m2)

𝐶𝐷 Drag coefficient

15
 ACCEPTED MANUSCRIPT

𝐶̂𝐷 Computed drag coefficient using empirical equations

𝐷 Diameter of flow channel(mm)

PT
𝑑𝑜 Outer diameter of hollow frustum (mm)

di Inner diameter of hollow frustum (mm)

RI
𝐹𝐷 Drag force(N)

SC
𝐹𝑅 Solid –Solid resistance (N)

NU
𝐹𝐵 Body force (N)

G Percentage of Glycerin in aqueous solution

MA
g Acceleration due to gravity (m/s2 )

𝑅 Radius of sphere(mm)
D

𝑉 Measured terminal velocity (m/s)

TE

Greek Letters
P

𝜃 Angle of inclination (Degree)

CE

𝜌𝑆 Density of solid (kg/m3)

𝜌𝐹 Density of fluid (kg/m3)

AC

𝜇 Kinetic viscosity(Pa. s)

References

[1] R. Clift, J.R. Grace, M.E. Weber, Bubbles, Drops and particles, Academic Press,
New York, 1978.
[2] R. P. Chhabra, Bubbles, Drops and Particles in Non-Newtonian Fluids, CRC
Press, Boca Raton, FL, 1993.
[3] A.R. Khan, J.F. Richardson, The resistance to motion of a solid sphere in a fluid,
Chem. Eng. Commun.62 (1987) 135–150.

16
 ACCEPTED MANUSCRIPT

[4] P. H. T. Uhlherr, R.P. Chhabra, Wall effect for the fall of spheres in cylindrical
tubes at high Reynolds number, Can. J. Chem. Eng.73 (1995) 918-923.
[5] A. Unnikrishnan , R.P. Chhabra, An Experimental Study of Motion of Cylinders
in Newtonian Fluids: Wall Effects and Drag Coefficient, Can. J. Chem. Eng. 69
(1991) 729-735.

PT
[6] S. Nitin and R. P. Chhabra, Wall Effects in Two-Dimensional Axisymmetric
Flow Over a Circular Disk Oriented Normal to Flow in a Cylindrical Tube, Can.

RI
J. Chem. Eng.83 (2005) 450-457.
[7] B. Munshi, R. P. Chhabra and P. S. Ghoshdastidar, A numerical study of steady

SC
incompressible Newtonian fluid flow over a disk at moderate Reynolds numbers,
Can. J. Chem. Eng.77 (1999) 113–118.
[8] R.P. Chhabra, Wall effects on free-settling velocity of non-spherical particles in

NU
viscous media in cylindrical tubes, Powder Technol.85 (1995) 83-90.
[9] M. K. Sharma and R. P. Chhabra, An experimental study of free fall of cones in
Newtonian and non-Newtonian media: drag coefficient and wall effects, Chem.
MA
Eng. Process.30 (1991) 61-67.
[10] J. J. Carty, Resistance Coefficients for Sphere on Plane Boundary, B. S. Thesis,
Massachusetts Institute of Technology, 1957.
D

[11] R. J. Garde, S .Sethuraman, Variation of the drag coefficient of a sphere rolling

along a boundary, La Houille Blanche 7 (1967) 727-732.
TE

[12] C.D. Jan, J.C. Chen, Movements of a sphere rolling down an inclined plane, J. of.
Hydraulic Res. 35 (5) (1997) 689–706.
P

[13] R.P. Chhabra, J.M. Ferreira, An analytical study of the motion of a sphere rolling
CE

down a smooth inclined plane in an incompressible Newtonian fluid, Powder

Technol. 104 (1999) 130–138.
[14] R.P. Chhabra, Mukesh Kumar, Rishitosh Prasad, Drag on spheres in rolling
AC

motion in inclined smooth tubes filled with incompressible liquids, Powder

Technol. 113 (2000) 114–118 .
[15] M. Jalaal , D.D. Ganji, An analytical study on motion of a sphere rolling down an
inclined plane submerged in a Newtonian fluid, Powder Technol. 198 (2010) 82–
92.
[16] M.A. Hasan, Wall effects on the motion of a rolling sphere in a closely fitting
tube, Chem. Eng. Journal, 33 (1986) 97 – 101.
[17] Z. Duan, B. He, Y. Duan, Sphere Drag and Heat Transfer, Sci. Rep., 5 (2015)
12304.
[18] R. Barati, S. A. A. S. Neyshabouri, G. Ahmadi, Development of empirical models
with high accuracy for estimation of drag coefficient of flow around a smooth
sphere: An evolutionary approach, Powder Technol. 257 (2014) 11–19.
[19] R. Barati, S. A. A. S. Neyshabouri, G. Ahmadi, Sphere drag revisited using
Shuffled Complex Evolution algorithm, River Flow, (2014) 345-353.

17
 ACCEPTED MANUSCRIPT

[20] D. C. Montgomery, Design and Analysis of Experiments, John Wiley & Sons,
Inc., New York 5th Ed, 2001.

PT
RI
SC
NU
MA
D
P TE
CE
AC

18
 ACCEPTED MANUSCRIPT

List of Figures
Fig.1. Schematic representation of flow over hollow frustum in (A) vertical channel and (B) in
inclined channel.
Fig.2. Variations of terminal velocity with d0/D for different di/do in 100% glycerin and 900 angle

PT
of inclination
Fig.3. Variation of terminal velocity with do/D and θ for di/do = 0.5 in 100% Glycerin

RI
Fig.4. Drag coefficient against Reynolds number plot for angle of inclination 900- 400

SC
Fig.5. Dependency of CDRe on do/D and the viscosity of fluid for (A) 900 and (B) 500 angle of
inclination at di/do=0.5

NU
Fig.6. Variation of CD with 1/Re for different angle of inclination
Fig. 7 Mean Effect Plot at θ = 900 for Eq.13
MA
Fig. 8. Mean Effect Plot at θ = 900 for Eq.14

Fig.9. Top view of generated mesh for (A) vertical and (B) inclined plane
D

Fig.10. Comparison of Experimental and simulation results for different tube inclination
TE

Fig. 11. Velocity distribution around hollow frustum in (A) vertical and (B) inclined channel
P

List of Tables
CE

Table 1. Physical dimensions of hollow frustum

Table 2. Physical properties of test fluids
AC

Table 3. Obtained coefficients and statistical errors of Eq.13 for different angle of inclination, θ
Table 4. Obtained coefficients and statistical errors of Eq.14 for different angle of inclination, θ
Table 5. Relationship among the coefficients of Eq. 13 and angle of inclination, θ
Table 6. Relationship among the coefficients of Eq. 14 and angle of inclination, θ
Table 7. Identified factors and their level values

Table 8. Treatment combinations of factors and the simulated responses values using first
empirical model (Eq. 13) at different θ values

Table 9. Analysis of Variance for CD = 900 (Eq. 13)

Table 10. Treatment combinations of factors and the simulated responses using second empirical
model (Eq. 14) at different θ values

19
 ACCEPTED MANUSCRIPT

Table 11. Analysis of Variance for CD = 900 (Eq. 14)

Table 12. Mesh independence test for Re= 0.359475, do/D=0.33 and di/do=0.7. Experimental
value =276.18
Table 13. Statistical deviation and goodness of CFD simulation

PT
RI
SC
NU
MA
D
P TE
CE
AC

20
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
(A) (B)
D

Fig.1. Schematic representation of flow over hollow frustum in (A) vertical channel and (B) in inclined
channel
P TE
CE
AC

21
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D
TE

Fig.2. Variations of terminal velocity with d0/D for different di/do in 100% glycerin and 900 angle
P

of inclination
CE
AC

22
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D
P TE
CE

Fig.3. Variation of terminal velocity with do/D and θ for di/do = 0.5 in 100% Glycerin
AC

23
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D
TE

Fig.4. Drag coefficient against Reynolds number plot for angle of inclination 900- 400
P
CE
AC

24
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
A B
D
TE

Fig.5. Dependency of CDRe on do/D and the viscosity of fluid for (A) 900 and (B) 500 angle of inclination
at di/do=0.5
P
CE
AC

25
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D
TE

Fig.6. Variation of CD with 1/Re for different angle of inclination

P
CE
AC

26
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D

Fig. 7 Mean Effect Plot at θ = 900 for Eq.13

P TE
CE
AC

27
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
Fig. 8. Mean Effect Plot at θ = 900 for Eq.14
D
P TE
CE
AC

28
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
(A) (B)
MA
Fig.9. Top view of generated mesh for (A) vertical and (B) inclined plane
D
P TE
CE
AC

29
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D
TE

Fig.10. Comparison of Experimental and simulation results for different tube inclination
P
CE
AC

30
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
D
TE

(A) (B)
P

Fig. 11. Velocity distribution around hollow frustum in (A) vertical and (B) inclined channel
CE
AC

31
 ACCEPTED MANUSCRIPT

Table 1. Physical dimensions of hollow frustum

Sl. No. do di dc de H
1 10 7
2 10 5
3 10 3

PT
4 9 7
5 9 5 8 2 12

RI
6 9 2
7 6 5
8 6 3

SC
9 6 2

NU
MA
D
P TE
CE
AC

32
 ACCEPTED MANUSCRIPT

Table 2. Physical properties of test fluids

Soln. No. Fluid Temperature Density Viscosity
(K) (kg/m-3) (Pa.s)
1 100% Glycerin 298 1257.09 0.7533
2 98% Glycerin 298 1251.906 0.5378

PT
3 96% Glycerin 298 1246.723 0.4025
4 94% Glycerin 298 1241.539 0.3054
5 92% Glycerin 298 1236.356 0.2606

RI
6 90% Glycerin 298 1231.172 0.2094

SC
NU
MA
D
P TE
CE
AC

33
 ACCEPTED MANUSCRIPT

Table 3. Obtained coefficients and statistical errors of Eq.13 for different angle of inclination, θ

θ A B C E F n m k R2
90 0.00641 -100.6782 94.045 1.0002 1.002218 -0.1 -0.03 -0.6 0.97

PT
80 0.1137 -35.9284 9.088 0.012 0.1541 -0.01 -0.1 -0.4 0.978
70 0.1089 -39.2276 11.1669 0.0011 0.5167 -0.1 -0.1 -0.5 0.9932
60 0.1261 -47.7332 12.1739 0.11 0.1185 -0.1 -0.3 -0.6 0.989

RI
50 0.1124 -43.3212 19.9854 0.2246 0.3914 -0.1 -0.4 -0.7 0.993

SC
45 0.1681 -50.8861 34.2011 0.0611 0.0967 -0.1 -0.5 -0.4 0.979
40 0.0061 -47.3412 42.1485 0.0111 0.5527 -0.01 -0.1 -0.2 0.979

NU
θ RMSLD SSLD ARE SRE SSRE
90 0.04527 0.11067 1.6308 0.88063 0.65994
80 0.040162 0.087101 1.9308 1.04267 0.5829
MA
70 0.02058 0.02288 0.89091 0.48109 0.12282
60 0.11154 0.07183 2.55944 1.3821 0.31066
50 0.02207 0.0263 0.1574 0.08503 0.14795
D

45 0.2683 0.03888 1.0352 0.559 0.22438

TE

40 0.0349 0.0658 4.1189 2.2242 0.3159

P
CE
AC

34
 ACCEPTED MANUSCRIPT

Table 4.

Obtained coefficients and statistical errors of Eq.14 for different angle of inclination, θ

PT
θ P Q R S R2 RMSLD SSLD ARE SRE SSRE
90 224.259 -0.598 -0.0152 -0.0241 0.97 0.0348 0.0654 0.16 0.087 0.3516
80 42 -0.583 0.0101 -0.035 0.96 0.0384 0.079625 5.48576 2.96231 0.5137

RI
70 46.3 -0.714 0.00047 -0.00142 0.98 0.0371 0.0741 4.7126 2.5448 0.3402
60 61.259 -0.723 0.0427 -0.015 0.97 0.88401 0.421995 1.512012 0.08164 0.3041

SC
50 63.8 -0.682 0.0124 -0.03451 0.99 0.02461 0.03112 3.51282 1.89692 0.1535
45 86.22 -0.543 0.00017 0.002117 0.98 0.01687 0.01538 0.6051 0.3267 0.0852

NU
40 94.99 -0.415 0.00147 -0.00575 0.97 0.017 0.0156 0.047 0.625 0.0839

MA
D
P TE
CE
AC

35
 ACCEPTED MANUSCRIPT

Table 5. Relationship among the coefficients of Eq. 13 and angle of inclination, θ

Coefficient Fitting line equation

2 2
A
0.5869𝑒 (−((𝜃−47.21)/1.836) ) + 0.1344𝑒 (−((𝜃+67.5)/18.36) )

PT
2 2
B
−49.11𝑒 (−((𝜃−51.11)/33) ) − 7.32 × 1016 𝑒 (−((𝜃−504.2)/70.69) )

RI
SC
2 2
C
1.544 × 1042 𝑒 (−((𝜃−750.3)/68.61) ) − 5.976 × 1026 𝑒 (−((𝜃+1954)/262) )

2 2
E (−((𝜃−94.22)/6.47) ) (−((𝜃+53.37)/6.773) )

NU
1.532𝑒 − 0.2072𝑒

F 2.539 × 106 − 3.366 × 106 cos(0.001679𝜃) − 3.614

× 105 sin(0.001679𝜃) + 8.273 × 105 cos(0.001679𝜃) + 1.797
MA
× 105 sin(0.001679𝜃)

n −7.243 × 107 + 9.656 × 107 cos(−0.0003535𝜃) − 1.729

× 106 sin(−0.0003535𝜃) − 2.413 × 107 cos(−0.0003535𝜃)
D

+ 8.646 × 105 sin(−0.0003535𝜃)

TE

m 3.827 × 109 − 5.102 × 109 cos(0.0002293𝜃) − 8.04

× 107 sin(0.0002293𝜃) + 1.275 × 109 cos(0.0002293𝜃)
P

+ 4.019 × 107 sin(0.0002293𝜃)

CE

k 1.146𝑠𝑖𝑛(0.04752 × 𝜃 + 0.7786) + 0.7248𝑠𝑖𝑛(0.08552 × 𝜃 + 1.001)

AC

36
 ACCEPTED MANUSCRIPT

Table 6. Relationship among the coefficients of Eq. 14 and angle of inclination, θ

Coefficient Fitting line equation

P 1.171𝑒 + 14 − 1.562𝑒 + 14 × cos(𝜃 × 6.623𝑒 − 05 ) − 6.154𝑒 + 11 × sin(𝜃 × 6.623𝑒 −
05 ) + 3.905𝑒 + 13 × cos(2𝜃 × 6.623𝑒 − 05 ) + 3.077𝑒 + 11 × sin(2𝜃 × 6.623𝑒 − 05 )

PT
Q −1.48𝑒 + 09 + 1.974𝑒 + 09 × cos(𝜃 × 0.000184 ) + 1.885𝑒 + 07 × sin(𝜃 × 0.000184) −

RI
4.933𝑒 + 08 × cos(2𝜃 × 0.000184) − 9.424𝑒 + 06 × sin(2𝜃 × 0.000184)

SC
R 0.01726 × sin( 0.2447 × 𝜃 − 6.432 ) + 0.02639 × sin(0.07218 × 𝜃 − 2.644 )

NU
S −0.01785 − 0.005434 × cos(𝜃 × 0.2127 ) − 0.01283 × sin(𝜃 × 0.2127 ) + 0.0207 ×
cos(2𝜃 × 0.2127 ) − 0.02424 × sin(2𝜃 × 0.2127 )
MA
D
P TE
CE
AC

37
 ACCEPTED MANUSCRIPT

Table 7. Identified factors and their level values

Factor Low Level High level
di/do 0.2222 0.8333
do/D 0.1993 0.3322
Re 0.1397 8.4164

PT
RI
SC
NU
MA
D
P TE
CE
AC

38
 ACCEPTED MANUSCRIPT

Table 8. Treatment combinations of factors and the simulated responses values using first
empirical model (Eq. 13) at different θ values

di/do do/D Re CD-900 CD-800 CD -700 CD-600 CD-500 CD-450 CD-400

0.2222 0.1993 0.1397 838.8678 266.8726 295.6625 357.1164 336.1673 406.3757 382.646

PT
0.2222 0.1993 8.4164 91.65448 13.58673 14.95287 16.12041 23.0311 34.16745 47.66714
0.2222 0.3322 0.1397 838.1713 266.8183 295.4016 357.0341 335.8034 406.3416 382.5719
0.2222 0.3322 8.4164 90.95802 13.53243 14.69194 16.03807 22.66713 34.13338 47.59302

RI
0.8333 0.1993 0.1397 838.8271 266.8709 295.6623 357.0599 335.999 406.313 382.6444
0.8333 0.1993 8.4164 91.6138 13.585 14.95271 16.06387 22.86276 34.10476 47.66555

SC
0.8333 0.3322 0.1397 838.1306 266.8166 295.4014 356.9776 335.635 406.2789 382.5703
0.8333 0.3322 8.4164 90.91733 13.5307 14.69178 15.98152 22.49879 34.07069 47.59142

NU
MA
D
P TE
CE
AC

39
 ACCEPTED MANUSCRIPT

Table 9. Analysis of Variance for CD = 900 (Eq. 13)

Source DF Adj SS Adj MS F-Value P-Value

di/do 1 0 0 *
do/D 1 1 1 *

PT
Re 1 1116655 1116655 *
Error 4 0 0
Total 7 1116656

RI
SC
NU
MA
D
P TE
CE
AC

40
 ACCEPTED MANUSCRIPT

Table 10. Treatment combinations of factors and the simulated responses using second empirical
model (Eq. 14) at different θ values

di/do do/D Re CD-900 CD-800 CD-700 CD-600 CD-500 CD-450 CD-400

0.2222 0.1993 0.1397 773.9872 137.8857 189.0595 244.2341 253.4372 250.1338 216.5155

PT
0.2222 0.1993 8.4164 66.73224 12.64212 10.13278 12.61586 15.48667 27.01907 39.5201
0.2222 0.3322 0.1397 764.5153 135.4419 188.9224 242.3695 249.0077 250.4045 215.8803
0.2222 0.3322 8.4164 65.91558 12.41806 10.12543 12.51954 15.216 27.04831 39.40416

RI
0.8333 0.1993 0.1397 758.5918 139.7388 189.177 258.4155 257.6254 250.19 216.9366
0.8333 0.1993 8.4164 65.40486 12.81203 10.13907 13.3484 15.7426 27.02514 39.59696

SC
0.8333 0.3322 0.1397 749.3083 137.2622 189.0398 256.4426 253.1227 250.4608 216.3002
0.8333 0.3322 8.4164 64.60445 12.58496 10.13172 13.24649 15.46746 27.05439 39.4808

NU
MA
D
P TE
CE
AC

41
 ACCEPTED MANUSCRIPT

Table 11. Analysis of Variance for CD = 900 (Eq. 14)

Source DF Adj SS Adj MS F-Value P-Value

di/do 1 138 138 4.11 0.113
do/D 1 52 52 1.54 0.282

PT
Re 1 968655 968655 28813.46 0.000
Error 4 134 34
Total 7 968979

RI
SC
NU
MA
D
P TE
CE
AC

42
 ACCEPTED MANUSCRIPT

Table 12. Mesh independence test for Re= 0.359475, do/D=0.33 and di/do=0.7. Experimental
value =276.18
No. of Elements No. of Nodes CD
19521 91470 241.24

PT
26343 126083 248.11
39069 192224 257.14

RI
43978 218656 266.88

SC
48847 245460 275.14

51792 261788 275.54

NU
MA
D
P TE
CE
AC

43
 ACCEPTED MANUSCRIPT

Table 13. Statistical deviation and goodness of CFD simulation

θ RMSLD SSLD ARE SRE SSRE R2

90 0.025568 0.0353 0.164468 0.088807 0.185898 0.9832

PT
80 0.01785 0.017206 0.010148 0.00548 0.090282 0.9849
70 0.012884 0.008964 0.285175 0.153995 0.047712 0.9941
60 0.012226 0.008072 0.05428 0.02931 0.043319 0.9966

RI
50 0.008322 0.003739 0.06821 0.03683 0.01997 0.9974

SC
45 0.009093 0.004465 0.11005 0.05943 0.023812 0.9943
40 0.0088 0.0042 0.1143 0.0617 0.0221 0.9923

NU
MA
D
P TE
CE
AC

44
 ACCEPTED MANUSCRIPT

Graphical abstract

PT
RI
SC
NU
MA
D
P TE
CE
AC

45
 ACCEPTED MANUSCRIPT

Highlights
 Terminal velocity, V and drag coefficient, CD of hollow frustum are reported.
 Wall effects on both CD & V are reported by varying diameter ratios di/do and do/D.
 V increases and CD decreases with diameter ratio, do/D (a reverse trend).


PT
Angle dependent correlations for CD are developed as function of Re, do/D, di/do.
 Product of CDRe is affected largely by the fluid viscosity and inclination angle.

RI
SC
NU
MA
D
P TE
CE
AC

46