Chemical Engineering Journal 358 (2019) 225–235

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

Analysis of the turbulent flow and trailing vortices induced by new design T
grooved blade impellers in a baffled tank
Sergio A. Martínez-Delgadilloa, , Alejandro Alonzo-Garciab, Víctor X. Mendoza-Escamillac,
⁎

Israel González-Neriad, J. Antonio Yáñez-Varelad

a
Depto. Ciencias Básicas, Universidad Autónoma Metropolitana Azcapotzalco, Av. San Pablo 180, Azcapotzalco, CP 02200 México, D.F., Mexico
b
CONACyT-Centro de Ingeniería y Desarrollo Industrial, Av. Playa pie de la cuesta 702, Desarrollo San Pablo, 76125 Querétaro, Querétaro, Mexico
c
Depto. Electrónica, Universidad Autónoma Metropolitana Azcapotzalco, Av. San Pablo 180, Azcapotzalco, CP 02200 México, D.F., Mexico
d
Posgrado en Ing. de Procesos, Universidad Autónoma Metropolitana Azcapotzalco, Av. San Pablo 180, Azcapotzalco, CP 02200 México, D.F., Mexico

HIGHLIGHTS

• The performance of two new design turbines with grooved blades 45° was evaluated.
• Dissipated power was reduced until 6% with new designs, compared with the 4PBT.
• The trailing vortices induced by new impellers were modeled with the DES approach.
• Validation of the modeling was carried out by PIV experimental tests.

ARTICLE INFO ABSTRACT

Keywords: U or V shaped grooves have shown reduction in energy and power consumption in blunt bodies immersed in turbulent
Grooved blade impeller streams. In this work, the performance of two new design turbines with four blades rotated 45°, one with U and the
PIV other with V grooved shape blades, were evaluated in a baffled stirred tank reactor and they were compared with the
DES one of the regular 4PBT. The power consumption, mean velocities, turbulent kinetic energy production (TKE), energy
Energy dissipation rate
dissipation rate (ε) and trailing vortices induced by new design impellers with grooved walls and 4PBT were evaluated
Agitated tank
experimentally and modeled by detached eddy simulation (DES) approach, which has not been tested in blade walls
with small perturbations. Validation of the results was carried out by PIV (Particle image velocimetry) experimental
tests. The evolution of new and distinctive trailing vortices located in the upper and lower faces of the grooved new
design impellers is presented. They were measured by angled resolved PIV technique, and visualized in three di-
mensions by the DES method. The V-grooves shaped impeller generated a larger region of influence of the turbulent
kinetic energy along the discharge region than the other impellers. Moreover, presented higher mean dissipation rate
levels for the most of the measured angles and induced higher number of trailing vortices in the blades vicinity than the
other turbines. It was found that in a wide range of the Reynolds number (Re) (40 × 103 < Re < 125 × 103) the
dissipated power measured experimentally from shaft, presented reductions of the order of 6% and 4% for the U and V
grooved new models, respectively, compared with that of the regular 4PBT impeller.

1. Introduction describe the mean flow and turbulent characteristics of the flow inside ATRs
as function of variations in parameters as impeller diameter, clearance, ec-
Agitated tank reactors (ATRs) are devices commonly used in chemical, centricity, number of baffles, blade angle and flow regime [1–4]. It is known
mineral process engineering, food, and wastewater treatment industries with that the impeller blades are responsible for the conversion of mechanical
the aim of increasing the reaction rate and chemical-physical homogeniza- energy to momentum, defining hence the overall flow patterns and mixing
tion. While ATRs are easy to build and operate, the flow patterns induced by characteristics present in the vessel. Thus, an analysis of its influence is a
impellers are composed of a wide spectrum of temporal and spatial scales. For critical issue for designing purposes. In that sense, the mixing quality in an
turbines with standard designs, a variety of work have been devoted to ATR relies greatly on the nature of the vortices created by the blades, its

⁎
Corresponding author.
E-mail address: samd@correo.azc.uam.mx (S.A. Martínez-Delgadillo).

https://doi.org/10.1016/j.cej.2018.10.015
Received 22 May 2018; Received in revised form 26 July 2018; Accepted 2 October 2018
Available online 04 October 2018
1385-8947/ © 2018 Elsevier B.V. All rights reserved.
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

position, the turbulence kinetic energy (k) and dissipation rate (ε) and its four equally spaced baffles (B = 0.1 T) and an impellers (D = T/3,
energy consumption. Works as those of Ge et al. [5] and Kumaresan and w = 14.3 mm and C = T/3). This is depicted in Fig. 2a. The working
Joshi [6] have reported valuable calculations of parameters as power number fluid used was water (ρ = 998.2 kg/m3 and µ = 1.003 mPa·s). As a
(NP), pumping number (NQ), k and ε present in hydrofoil, Rushton or axial consequence of the refractive index effects the SRT was placed inside a
turbines with plane or slightly twisted blades. Nevertheless, in such works, cubic tank filled with the same level of water (H = T). The angular
the effect of small scale flow perturbations imposed on blades, which have velocity was set to N = 500 rpm, at such conditions the flow was fully
been proven to be beneficial from an energetic point of view has not been turbulent (Re = ρND2/µ ≈ 52,000).
addressed. Vasconselos et al. (1999) studied the turbulent flow induced under The velocity fields measured were acquired from a 2D PIV system (TSI
gasified or un-gassed conditions using a Rushton turbine with streamlined or Incorporated), integrated for a double-pulse Nd:YAG laser, that generates a
perforated blades [7]. Depending among the design, the NP is capable to beam with a wavelength of 532 nm and energy of 75 mJ per pulse. An
decrease at expense of an increase in mixing time, at the same rotational optical array converts the round laser beam into a laser sheet of approxi-
speed. Also, diverse correlations between the inter-stage exchange number, mately 1 mm thick. 400 pairs of photos with a capture rate of 14 frames/s
NP and mixing time were stablished. Ankamma Rao and Sivashanmugam were taken with a high-resolution cross-correlation camera (2360 × 1776
(2010) reported power consumption reductions in a Rushton turbine when pixels) with a time difference of 100 µs. The test section covered an area of
small triangular or rectangular slices are made in the blade tips, for a range ≈0.101 m × 0.075 m situated half way between two baffles, achieving a
Re from 50 × 103 to 200 × 103 [8]. According to the authors, such reduc- spatial resolution of ≈43 µm/px being the vector displacement of 0.69 mm.
tions are related to decreases in the shear forces that occur in the blade tips. A shaft encoder was adapted to provide the trigger signal to the camera. In
More recently, Steiros et al. [9] and Başbuğ et al. [10], implemented direct order to assess the angled resolved measurements, 18 planes located at
numerical simulation (DNS) to analyze the effect of squared fractal shaped different blade angles were measured (Δθ = 5°). The blade angle θ = 0°
blades in a Rushton turbine. Authors obtained reductions in NP, which in correspond to the mid-plane of a blade, Fig. 2b. TSI Insight 4G software was
turn, were explained as function of changes in the recirculation wake formed applied to analyze images via the Nyquist recursive grid algorithm, with an
behind the blades. So, it is recognized that with small scale wall modifications interrogation window of 32 by 32 pixels and 50% overlap between them.
NP can be reduced. Also, most of works have focused on wheel turbines. Torque measurements were acquired by a Futek FSH1980 rotary torque
Another alternative not explored yet in order to reduce the drag forces and sensor with a maximum torque capacity of 1.41 N·m at a rate of 50 Hz,
hence NP in bodies influenced by an adverse gradient pressure is based on precision of 0.5%, with temperature control. The value used for calculations
grooved surfaces. From an hydrodynamic point of view, grooved surfaces has was the mean of 900 measurements.
proven to be capable of reducing shear forces and promote delay in the se-
paration point on blunt bodies. For example, Lim and Lee [11], obtained a 2.3. Turbulent kinetic energy and dissipation rate calculations
drag reduction of 19% in the subcritical flow over a circular cylinder, emu-
lating the effect of the small scale riblets with sharp edges, which reduced In order to obtain turbulent data free from the periodic waves induced
drag in flat plates [12]. Thus, as the aerodynamic features of a circular cy- by blade motions, the triple decomposition procedure shown in the work of
linder (blunt body) found a large similitude with those of a rotating blade, Sharp and Adrian [14] is applied by using Eq. (1). Here, u″ represents a
and considering that the U and V grooves have been successful used as fluctuating velocity component, U the instantaneous velocity being u |
passive devices to obtain drag benefits [12,13], in this work, the perfor- the average velocity obtained at a fixed blade position.
mances of two new design axial impellers namely, U-grooved and V-grooved
are evaluated. Both designs composed of four blades rotated 45° with u =U u| (1)
grooves; one with U-shaped grooves and the other with V-shaped grooves. To Because of the 2D PIV is not capable to measure the third velocity
gain insight about the velocity components, turbulent parameters and trailing component and according with the pseudo-isotropic assumption of the
vortices, produced by these new impellers, the angle resolved PIV technique turbulence, the k is calculated as function of the two measured fluc-
was applied. In addition, NP estimations based on torque measurements over tuating velocity components as shown in Eq. (2). Khan et al. (2006)
the range (40 × 103 < Re < 125 × 103) were made. The results were verified negligible deviations between 2D and 3D PIV maps, for both,
compared with those of the regular 4PBT impeller. To understand the in- magnitudes and locations [15].
fluence of grooved blades on the trailing vortices, the detached eddy simu-
lation technique (DES) was applied. The main objective of this work is based 3 ''2 2
k= (u + v'' )
upon the actual necessity of designing new type of turbines of simple man- 4 (2)
ufacture capable of optimize mixing and decrease power consumption in long
The turbulence dissipation rate ε is composed of 12 terms, however
time operation mixing devices. Also, the angle resolved maps of velocities,
from the 2 velocity components measured, just 4 terms could be ob-
turbulent variables, vorticity and trailing vortices could be used to validate
tained. In order to estimate the contribution of the non-measured ve-
the application of other turbulence models applied to walls with small scale
locity the turbulence is considered statistically isotropic. An interesting
perturbations.
approach to calculate ε from 2D-PIV with spatial resolutions (Δ) larger
than the Kolmogorov length scale K = (v 3/ T )1/4 is the large eddy ap-
2. Materials and methods
proach [16]. In this method, the contribution of the scales smaller than
the interrogation window of PIV is modeled by a sub-grid scale model
2.1. Impeller models
which compensates ε values. In this work, the Eq. (3) as defined by
Khan (2005) was implemented [17].
Two new different blade designs were tested and it performance was
compared with a regular 4PBT impeller with plane blades. The new designs 2 2
3/2
2 2
have groove blades, one with curved blades, resembling U-grooves, and the u'' v'' u'' v''
= (Cs ) 2 4 +4 +2 +2
other with triangular V-grooves as shown in Fig. 1. For simplicity, during x y y x
(3)
the rest of the document, the U and V grooved impellers will be named as U
and V model respectively. The new design impellers dimensions are defined The Smagorinsky constant Cs was adjusted for each impeller de-
based on the blades width w = 14.3 mm. pending on the radio Δ/λk, according to Meyers and Sagaut [18]. From
the experiments Cs ≈ 0.16 for the three impeller models, so the 0.16
2.2. Experimental arrangement and operational conditions value was used as a constant. In the discussion of the results, the maps
will be normalized with the tank averaged specific dissipation rate as
The experimental outline consisted of an acrylic SRT (T = 250 mm), ε* = ε/εT. Here, εT (W/kg) was calculated for each impeller from the

226
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 1. Geometrical description of the impellers: a) Regular pitched blade turbine (4PBT), b) U-grooved new turbine, c) V-grooved new turbine.

Fig. 2. Schematics of the ATR (a) and the laser array configuration (b).

power (P) obtained from the experimental torque measurements, per of the mass flux in a circular surface of radius R, projected 2 mm below
unit mass of water [3]. impeller blades. The pumping effectiveness relates the pumping of the
impeller per unit power consumption, is given by Eq. (7) [19].
2.4. Power and pumping number NQ
=
E
NP (7)
The power number and pumping number were calculated from Eqs.
(4) and (5).
P 2.5. Numerical modeling (Detached eddy simulation)
Np =
N 3D5 (4)
The philosophy of the DES, is to directly resolve the most energetic
Q eddies outside the boundary layer by a large eddy simulation (LES) that in
Nq =
ND3 (5) turn switches to a RANS based model within the boundary layer, allowing
In the DES model, P was the average of a torque monitor applied to relatively coarse grids in the near wall regions. The computational demands
the impeller walls. The experimental mass flow was obtained from the of the DES are larger than those of the RANS based models, but considerable
phase averaged axial velocity maps (average over all blade positions), less than the classical LES models. The viability of the DES model regarding
by integrating in the radial direction the velocity profile in a line lo- the predictions of vortex structures and other turbulent parameters in a ATR
cated 2 mm below the impeller blades, using Eq. (6) has been shown in the works of Gimbun et al. [20] and Chara et al. [21]. In
that sense, a numerical model of the ATR described previously was per-
R
Q= 2 Va rdr formed, and solutions were obtained by applying the finite volume method
(6)
solver of FLUENT® V.15.0. The adopted turbulence model was the DES with
0

Numerically, Q was the result of the average over the last 30 revolutions Spallart-Allmaras (S-A) to resolve the boundary layer. In the simulations,

227
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

the “Delayed” DES model was applied in order to preserve the S-A model 3. Results and discussion
through the boundary layers. Three similar grids were resolved in order to
account the three impeller models. For the 4PBT, the cell number count was 3.1. Dissipated power from torque
of 3.26 million cells, whereas for the U and V model, grids were composed
of 3.52 and 3.62 million cells respectively. To account blade thickness ef- The NP derived from torque measurements as function of Re is
fects, 5 uniformly distributed nodes were imposed along the blade thickness shown in Fig. 3. Both grooved impellers exhibit a decrease in the power
direction. Along the blade width, 40 equally spaced nodes were disposed, consumption in all the measured Re range compared with regular 4PBT.
and its length was discretized using approximately 80 nodal elements. The The mean NP value derived over the Re range was of 1.26 for the 4PBT,
maximum skewness deviations present in those grids were of 0.76 for the while for the U and V models values of 1.19 and 1.21 were obtained,
4PBT and U model, and 0.8 for the V model. As can be seen, all values were representing power consumption drops of 6% and 4% respectively,
below the 0.95 maximum value suggested in the grid generation guidelines compared with that of 4PBT. Taking into account that most of the re-
on the FLUENT manual [22]. For the pressure-velocity coupling, the actors work for long periods of time and are built to operate at in-
SIMPLE algorithm was applied. The spatial discretization of the momentum dustrial scale, savings of electrical energy could be obtained.
and turbulence equations was accomplished with the bounded central dif-
ferencing method being that the PRESTO scheme which accounts for 3.2. Experimental mean velocities, turbulent parameters and trailing
swirling and rotating flows was applied to the pressure [22]. The temporal vortices structures
term was discretized by the bounded second order implicit algorithm. The
non-slip boundary condition (BC) with standard log-law function was ap- The PIV non dimensional mean flow velocities, TKE and ε maps
plied to all walls while the symmetry BC was applied to the free surface resulting from the average over all angle positions (phase averaged) for
(zero stress wall). In all simulations, the constants of the DES model were set the three impellers is shown in Fig. 4. Here, the models are separated by
to the default values. A rotational region near the impeller (Zone 1) and a row, and columns refer to the flow variable. For all models, the axial
stationary in the rest of the domain (Zone 2), were created in order to velocity is slightly deflected outwards as a consequence of the blade
reproduce the impeller motion. Previous to the sliding mesh (SM) simula- angle. Also, a slight decrease in axial velocity (Va) inside the core of the
tion of the DES model, the multiple reference frame approach with the high gradient region located below the impeller was present for the
Realizable k-ε model was used as initial solution of the system. Once con- grooved impeller, being more evident for the V model. This decrease is
verged, pseudo-steady conditions were achieved by simulating 30 impeller related with redirections of the pumped flow induced by the wall folds,
revolutions, taking 3° of blade rotation (Δt = 1 × 10−3 s) with SM. which are especially visible in the radial velocity maps. Important in-
Finally, the results presented were obtained from statistics collected creases of radial velocity were obtained for the grooved models, in-
from 30 additional impeller revolutions, taking 1° of blade rotation dicating changes in z-vorticity (which is function of the axial and radial
(Δt = 3.333 × 10−4 s). The last temporal resolution is lower than the velocity derivatives). This particular issue will be discussed later.
Kolmogorov time scale K = v/ T calculated as ≈2.17 × 10−3 s for the The TKE maps showed an elongation in the zone of maximum
4PBT impeller, and 2.21 × 10−3 s for the U and V models. In τK, the mean magnitude (k / Vtip
2
=0.035) located near the blade tip for the U model.
specific dissipated power ( T ) was derived from torque measurements de- Although this high magnitude spot was decreased for the V model, the
pending of each impeller. overall high gradient region visible in the yellow color with
2
k / Vtip = 0.025 was extended. Further, for both grooved models, the
2.6. Trailing vortices identification formation of a distinctive second high magnitude TKE spot is present.
This was particularly evident for the V model. Thus, the U model were
The trailing vortices obtained by DES, were calculated using the capable to intensify the production of TKE near the blade tips, whereas
criterion based on an objective definition reported by Jeong and the V model extended its region of influence. The ε* maps showed that
Hussain [23] that states that any point belongs to a vortex core region if near the blade tips, most of the turbulent energy is dissipated and the
the second eigenvalue of the sum of the squared vorticity and strain grooves presence does not affect its location and mean magnitudes.
tensors Ω2 + S2 is negative (λ2 < 0). As undesired residual vorticity However, interesting small dissipation pockets are visible for the
appears when a threshold value of λ2 is close to zero, a value of grooved impellers, which are especially evident for the U model. In
λ2 = −12000 s−2 was selected in order to reconstruct the vortex re-
gions for DES. For the PIV measurements, the identification of the
trailing vortices was based on the simplified procedure reported by
Escudie and Liné [24]. In this procedure, four components of the ve-
locity gradient tensor are extracted directly from PIV measurements,
that is, two diagonal u , w and other two non-diagonal u , w , while the
x z z x
third diagonal is approximated from the continuity equation, that is
dv
dy
=
u
x (w
)
+ z . Thus the gradient velocity tensor is expressed as:
u u
0
x z
a 0 d
dv
0 0 = 0 (a + c ) 0
dy
w w e 0 c
0
x z (8)

Based on such tensor, the second eigenvalue of Ω2 + S2is given by:

a2 + 2de + c 2 + (a c )2 +(a + c + d + e ) 2
=
2
2 (9)

In Eq. (9), the vortex core contours are related to λ2 < 0. In the PIV
for the three impeller models depicted in Fig. 9, the contours were
shown at a λ2 = −400 s−2, in order to avoid the appearance of ex-
perimental noise reflected in false vortex core regions. Fig. 3. Power number vs Reynolds number curves from torque measurements.

228
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 4. Phase averaged non dimensional velocities, turbulent kinetic energy and dissipation rate derived for the three impellers. The x-y axis are x/T and y/T.

order to gain knowledge about the angular evolution of the turbulent effect was moderated for the V model, although the attenuation remained.
parameters, Figs. 5 and 6 show the non-dimensional TKE and ε obtained As the blade is rotated to 30°, the near wall high dissipation region grew for
for several angles. In such figures, columns refer to the resolved angle all impellers models and the attenuation of the down impeller semicircular
whereas rows to the different impellers. For the 4PBT impeller, the high regions continued.
gradient region of k / Vtip
2
= 0.035 located below the blade captured at 0° This growth continued at 45° where now, a single high dissipation re-
remained for all positions, although for the 75° plane, it was elongated gion is present, as consequence of the merging of the semicircular spots to
along the radial direction. In addition, its magnitude was a decreased those detached from the blades. For the V model, the lower spot almost
for the 60° and 75° maps, indicating some sort of intermittency. At 30°, disappeared. From this point, the shape of this peculiar single and larger
a second spot is detached from the blade tip, and remained for the structure remained, and it was attenuated and separated again at 75°. As
subsequent angles. Thus, two unique regions of high TKE production resume, the evolution of the dissipation starts with semicircular shapes of
are present for the 4PBT in the measured planes. The 0° plane resolved high magnitude that merge up as the blades rotate with those detached
for the U model shows two spots of high TKE which in subsequent from the walls, forming then one solely structure that is subsequently at-
positions join the other detached from the tip, forming a single zone tenuated and separated in a cyclic manner. Another interesting feature is
which reach its maximum at 45° and it is preserved in the other posi- that between the 30° and 45° planes, the maximum dissipation rate regions
tions. For the V model, the two spots at 0° have less magnitude in of ε* = 20 covered the largest extension for all angular planes. Fig. 7 shows
comparison to the U model and are attenuated for the other angular the mean dissipation rate as function of the angular plane for the three
positions. Also, the tip spot which appears from 30° is smaller compared models. Comparing the 4PBT and U model curves, there are no important
with the other cases. differences in all angular positions with respect to the U model. However,
The dissipation rate is a significant parameter in mixing, heat transfer for the V model, an increase of the mean dissipation rate is present from 20°
and chemical reaction. Further, its evolution as function of the angle, could and a maximum value of 6.5 was found at 30°. This represent an increase of
be helpful to define points of injection for multiphase mixing. For this 12% of the mean dissipation in comparison of the maximum value of 5.8
parameter (Fig. 6), the 4PBT model shows a single high magnitude region of obtained for the 4PBT and U model at the same angular plane. Thus, the V
ε* = 20 at 0°, which resembled the semicircular high TKE spot obtained at grooves which have sharp peaks, promote higher levels of dissipation in the
the same angle. In this plane, the U model does not present important discharge plane entailing less power to operate.
differences, contrasting with the V model that showed a second high dis- To gain insight into the shape and in-plane distribution of the highly
sipation region which is even larger than that located below the blade tip. rotational regions, Fig. 8 shows the dimensionless z-vorticity contours of the
At 15°, below the U grooves are formed new high dissipation regions, which three turbines at different angles. This dimensionless parameter was ob-
interacted with the single one present at 0°, which is now attenuated. This tained by dividing the z vorticity between the blade passage frequency

229
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 5. Non dimensional turbulent kinetic energy maps for different angles. The x-y axis are x/T and y/T.

Fig. 6. Non dimensional dissipation rate maps for different angles. The x-y axis are x/T and y/T.

(ωrot = 33.3 Hz). The nature of the rotational direction, shape and number the vortex located near tip. Here, the third vortex located below the high
of vortices identified is quite different between models. For example, at 0°, positive rotational region is vanished. Five vortices were found for the V
the 4PBT impeller exhibits three main rotational regions. Near the hub, an model impeller, and the positive vortices presented more strength. The near
elongated vortex is present, whereas near the blade tip, there are one vortex hub vortices conformed three well defined regions, whereas those vortices
with high vorticity and another with decreased magnitude is located below. located near tip were preserved. At 15°, new vortices derived from the
The rotational direction of the two last vortices is opposite to that of the grooved surfaces are visible, which then in next angles, evolves and interact
near hub. For the U model, the vortices detached from the hub form and with the near tip and hub vortices. This situation is notorious at 30° where
form a peculiar semicircular region, which rotates in opposite direction to the near wall vorticity region has detached for all impellers and now

230
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

3.3. PIV profiles of TKE and ɛ in the impeller discharge

Figs. 10 and 11 show the plots of non-dimensional TKE and ɛ* extracted

from two radial lines located in the impeller discharge at the PIV angles of
0°, 30° and 60°. The first line “L1” and the second one “L2”, were plotted
5 mm and 28 mm below the impeller hub, respectively. The locations were
chosen in order to analyze the high production and dissipation spots in the
immediate vicinity of the impellers, and those formed downwards (see 1
and 2 in Fig. 5). In L1 and 0°, negligible differences were found for both,
TKE and ɛ* profiles between the three models. At 30°, the PBT and V model
TKE profiles were slightly attenuated while the U model profile presented
two distinctive peaks, the main located at r/T ≈ 0.17 of k / Vtip
2
= 0.045 and
a second one located at r/T ≈ 0.9 of k / Vtip
2
= 0.028. The existence of such
secondary high TKE peak is connected with the new vortices created in the
immediacy of the impeller induced by grooves. The ɛ* was intensified im-
portantly in this angle, as levels reached between 50 and 70 units. The main
peaks of the grooved models were located almost at the same r/T in com-
parison to the TKE ones, and their maximum magnitudes surpassed that of
Fig. 7. Mean angle resolved values of dissipation rate for the impellers. the PBT45 for about 40% for the U model and 30% for the V one. At 60°
both TKE and ɛ* presented a similar profile of that of 0°. However, an in-
crease of maximum TKE is present for the U model in comparison with the
interacts with the previous vortices located near blade hub and tips. Four other impellers. At the same angle, the ɛ* magnitudes of the three models
regions were found for the 4PBT and at least six for the U and V models. In were larger than those of 0° (ɛ* ≈ 48 at 60° vs ɛ* ≈ 30 at 0°). The latter
the next planes, vortices start to vanish and maps return to the distribution suggests that starting at 0°, in the subsequent angles there are not only
obtained at 0°. Again, in the plane located at 30°, most of the turbulent changes in the vortices shapes as discussed later, but the dissipation is in-
parameters are maxima, and vorticity reaches higher levels. This is not tensified.
surprising, as the high production and dissipation rates have been found to At the 0° plane, the TKE L2 profiles of the U and V models of Fig. 11
be immersed in high vorticity regions [25]. The presence and shape of new presented increased values in comparison of those of the 4PBT for most of
high rotational regions which define the trailing vortices were confirmed by the positions, specially from r/T = 0.08 for the U model, and r/T = 0.05 for
the 2 criterion of Fig. 9. Again, at the full blade position, for the 4PBT the V. The latter is connected with the increase of the extent of the sec-
model, the dominant vortex is located near the blade tip. ondary high TKE region induced by grooves, discussed in Section 3.2. The
The grooves induced more vortices in this position. It is noticeable that maximum peaks of grooved models presented an increase of about 20% in
at the following position, the born of new vortex zones takes place near magnitude in comparison with the regular 4PBT. In the 30° and 60° planes,
walls. At 30° such new regions are mature and start to spread out in the the increases were sustained, although maximum peaks locations were
radial direction. Finally, at 75° starts the attenuation process. Thus, there is slightly shifted as consequence of the vortex movements. For the dissipation
an important relation between highly rotational zones and grooves pre- rate, the increases were visible larger for the V model, especially at 0° in
sence, which could be used in future tailored impellers design.

Fig. 8. Representative angle resolved z-vorticity maps for the three models.

231
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 9. Comparison of the contour lines of the trailing vortices obtained at different angles for the three cases (λ2 = −400 s−2).

comparison to the 4PBT classical model and the U model. At 30°, although 3.4. Tridimensional vortical structures from DES results and validation
the increases remained, the overall magnitudes of the three profiles were (DES vs PIV profiles)
attenuated in comparison of those of L1, as L2 is not so close of the impeller.
In the 60° plane, the U model values were as high as those of the V model. The power, pumping and pumping effectiveness numbers derived ex-
Thus, the grooves presence intensified the dissipation rate in regions away perimentally and numerically are shown in Table 1. Numerical NP was
from the impeller. obtained from a torque monitor applied to impeller walls. Good agreement

Fig. 10. PIV turbulent profiles along L1. The first row of figures refers to non-dimensional TKE, the second row, refers to the non-dimensional ɛ.

232
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 11. PIV turbulent profiles along L2. First row of figures refers to non-dimensional TKE, the second row, refers to the non-dimensional ɛ.

was present for all parameters as differences do not surpass 5% in com- elongated region was decomposed in several regions as consequence of the
parison to measurements. Also, reduction in power consumption was also grooved walls for the U and V models, and the DES simulations adequately
present in the numerical method. Although the pumping number decreased captured such feature. At the 60° plane, the vortex merging and formation
for the grooved models, the overall pumping effectiveness was slightly in- of two counter-rotating vortical spots are visible for the three impeller
creased. Thus, the DES model was capable to fairly reproduce the tenden- models. Thus, the presence and evolution of the near impeller vortices as
cies presented experimentally. This is a very noticeable feature considering function of the angle was captured by the DES model. It is interesting to
that this model had not been tested in blade walls with small perturbations. note that between the grooved blades, vorticity is intensified by the U and V
Fig. 12 shows the comparison of numerical and experimental z-vorticity models. A three dimensional depict of the trailing vortices was obtained by
profiles for the new design impellers, at the planes that represent the full DES through the λ2 < 0 criterion, as shown in Fig. 14. Comparing the
blade coincidence, at 0°, and at 45°, where high vorticity gradients are grooved models with the 4PBT, there is an intensification of the vortices
present. The line was located 10 mm below the hub. This flow parameter presence on the blade walls. Peculiar vortical regions were detected in the
was selected as the numerical visualization of vortices presented in Fig. 14, grooves valleys, and vortices that separated from the upper and lower blade
relies on spatial derivatives of velocity components. Again, the measur- faces were larger than those of the 4PBT. The latter suggests that grooves
ements are in good agreement with numerical results for both grooved intensifies turbulence production and dissipation in the whole blades.
models. Thus, the DES model is not only useful to reproduce trailing vor- The turbulence intensity field (T.I.) is another measure of the fluctua-
tices of radial flow turbines as pointed out by Gimbun et al. [20] and Chara tions present in a flow field, related with mixing capabilities of impellers. As
et al. [21], but also is useful to reproduce those present in axial pumping shown in Fig. 15, there are high gradients of T.I. in the upper blade regions
devices, even where grooved walls are used. (last third of the blade) and near their tips for the three studied models.
The numerical predictions of z-vorticity at 0°, 30° and 60° planes for the However, the grooved models were capable to elongate radially the region
three impeller models are shown in Fig. 13. The regions enclosed by the located in the upper zone of the blades, where T.I. = 0.4. Another inter-
dashed rectangles were not measured by PIV as the impeller blade inter- esting feature is that the high gradient region of the near blade tip discharge
fered with the laser sheet during the measurements. By comparing Fig. 13 increased their magnitudes. This, is related with the new vortex regions
with Fig. 8 at the 0° plane, the two main vortical spots, one positive near formed in the near blade positions discussed early.
blade tip and the another negative near the hub were adequately re-
produced by the numerical simulation of the three models. 4. Conclusions
In the 30° plane, the elongation of the vortex region located near hub
where ωz/ωrot = −3 was present in the 4PBT numerical model. This The assessment of the mean turbulent flow parameters of turbines
with the new design grooved blades was assessed by experimental
Table 1 measurements of torque, angle resolved PIV and DES. Compared with
Power, pumping and circulation number for the three impellers at 500 rpm. the regular 4PBT, reductions in the power consumption of 6% with the
NP Nq
U grooved and 4% with the V grooved new models were reached in a
E
wide Re range. The phase averaged maps showed important increases
Exp CFD Exp CFD Exp CFD of radial velocities as consequence of the grooved walls. U model was
capable to intensify the production of TKE in the blade tips, and the V
4PBT 1.26 1.28 0.79 0.75 0.62 0.59
model increased its region of influence. From several angle resolved
U 1.20 1.22 0.72 0.73 0.60 0.61
V 1.20 1.22 0.77 0.73 0.64 0.61 maps, the mean value of obtained as function of the angular position
was larger for the V model than those of the U model and the regular

233
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 12. Comparison between DES and PIV profiles of non-dimensional z-vorticity for the 2 grooved models. a) At 0° plane, b) at 45° plane.

Fig. 13. DES results of non-dimensional z vorticity at the 0°, 30° and 60° planes. The x-y axis are x/T and y/T.

4PBT. Thus, the sharp peaks of the V model induced an intensification vortices were present in the upper and lower grooved walls particularly
of the dissipated power in the discharge region. Two particular inter- at intermediate angular positions, which are related with the in-
mittent higher production and dissipation regions were present for the tensification of the production and dissipation rate. The DES simula-
new design models as function of the angle. Also, new well defined tions also predicted that the power consumption decreases, and flow

234
S.A. Martínez-Delgadillo et al. Chemical Engineering Journal 358 (2019) 225–235

Fig. 14. Trailing vortices structures obtained by DES simulation (λ2 = −12000 s−2).

Fig. 15. DES Turbulence intensity maps obtained for the three impeller models at the 0° plane.

visualization under the λ2 criterion showed the location of such new on the performance of six-bladed disk turbine impellers, Ind. Eng. Chem. Res. 39 (1999)
flow structures in the grooves valleys and peaks. Considering the easi- 203–213.
[8] D. Ankamma Rao, P. Sivashanmugam, Experimental and CFD simulation studies on
ness of manufacture, these kind of grooved blades could be re- power consumption in mixing using energy saving turbine agitator, J. Ind. Eng. Chem. 16
commended for long operating conditions mixing. (2010) 157–161.
[9] K. Steiros, P.J.K. Bruce, O.R.H. Buxton, J.C. Vassilicos, Power consumption and form drag
of regular and fractal-shaped turbines in a stirred tank, AIChE J. 61 (2014) 1–15.
Acknowledgements [10] S. Başbuğ, G. Papadakis, J.C. Vassilicos, Reduced power consumption in stirred vessels by
means of fractal impellers, AIChE J. 64 (2018) 1485–1499.
[11] H.C. Lim, S.J. Lee, Flow control of circular cylinders with longitudinal grooved surfaces,
Alejandro Alonzo-Garcia is grateful with the Centro de Ingeniería y AIAA J. 40 (2002) 2027–2036.
Desarrollo Industrial for the facilities provided during the accomplish- [12] M.J. Walsh, Turbulent boundary layer drag reduction using riblets, in: AlAA 20th Aerosp.
Sci. Meet., January 11–14, Orlando, FL, USA, 1982, pp. 1–8.
ment of this work. Financial supports of this work by the Consejo
[13] D.W. Bechert, M. Bruse, W. Hage, R. Meyer, Fluid mechanics of biological surfaces and
Nacional de Ciencia y Tecnología (Proy. No. CB-2011/169786) are their technological application, Naturwissenschaften 87 (2000) 157–171.
gratefully acknowledged. Thanks are also expressed to the unknown [14] K.V. Sharp, R.J. Adrian, PIV Study of small-scale flow structure around a Rushton turbine,
AIChE J. 47 (2001) 766–778.
reviewers of this work for their valuable comments. [15] F.R. Khan, C.D. Rielly, D.A.R. Brown, Angle-resolved stereo-PIV measurements close to a
down-pumping pitched-blade turbine, Chem. Eng. Sci. 61 (2006) 2799–2806.
References [16] J. Sheng, H. Meng, R.O. Fox, A large PIV method for turbulence dissipation, Chem. Eng.
Sci. 55 (2000) 4423–4434.
[17] F.R. Khan, Investigation of turbulent flows and instabilities in a stirred vessel using
[1] D. Mao, L. Feng, K.A.I. Wang, Y. Li, The mean flow field generated by a pitched blade particle image velocimetry, Thesis, Loughborough University, 2005.
turbine: changes in the circulation pattern due to impeller geometry, Can. J. Chem. Eng. [18] J. Meyers, P. Sagaut, On the model coefficients for the standard and the variational multi-
75 (1997) 307–316. scale Smagorinsky model, J. Fluid Mech. 569 (2006) 287–319.
[2] Z. Jaworski, K.N. Dyster, A.W. Nienow, The effect of size, location and pumping direction [19] J. Aubin, P. Mavros, D.F. Fletcher, J. Bertrand, C. Xuereb, Effect of axial agitator con-
of pitched blade turbine impellers onflow patterns: LDA measurements and CFD predic- figuration (up-pumping, down-pumping, reverse rotation) on flow patterns generated in
tions, Chem. Eng. Res. Des. 79 (2001) 887–894. stirred vessels, Trans IChemE 79 (2001) 845–856.
[3] D. Chapple, S.M. Kresta, A. Wall, A. Afacan, The effect of impeller and tank geometry on [20] J. Gimbun, C.D. Rielly, Z.K. Nagy, J.J. Derksen, Detached eddy simulation of the liquid
power number for a pitched blade turbine, Trans IChemE 80 (2002) 364–372. mixing in stirred tanks, AIChE J. (2011) 1–18.
[4] G. Montante, A. Bakker, A. Paglianti, F. Magelli, Effect of the shaft eccentricity on the [21] Z. Chara, B. Kysela, J. Konfrst, I. Fort, Study of fluid flow in baffled vessels stirred by a
hydrodynamics of unbaffled stirred tanks, Chem. Eng. Sci. 61 (2006) 2807–2814. Rushton standard impeller, Appl. Math. Comput. 272 (2016) 614–628.
[5] C.Y. Ge, J.J. Wang, X.P. Gu, L.F. Feng, CFD simulation and PIV measurement of the flow [22] Ansys, User’s Guide, Central Resource Park, 10 Cavendish Court, Lebanon, USA, 2009.
field generated by modified pitched blade turbine impellers, Chem. Eng. Res. Des. 92 [23] J. Jeong, F. Hussain, On the identification of a vortex, J. Fluid Mech. 285 (1995) 69–94.
(2014) 1027–1036. [24] R. Escudie, A. Liné, A simplified procedure to identify trailing vortices generated by a
[6] T. Kumaresan, J.B. Joshi, Effect of impeller design on the flow pattern and mixing in Rushton turbine, AIChE J. 53 (2007) 523–526.
stirred tanks, Chem. Eng. J. 115 (2006) 173–193, https://doi.org/10.1016/j.cej.2005.10. [25] A. Delafosse, J. Morchain, P. Guiraud, A. Liné, Trailing vortices generated by a Rushton
002. turbine: assessment of URANS and large Eddy simulations, Chem. Eng. Res. Des. 87
[7] J.M.T. Vasconcelos, S.C.P. Orvalho, A.M.A.F. Rodrigues, S.S. Alves, Effect of blade shape (2009) 401–411.

235