fluids

Article
Velocities in a Centrifugal PAT Operation:
Experiments and CFD Analyses
Mariana Simão 1, * ID , Modesto Pérez-Sánchez 2 ID , Armando Carravetta 3 ,
Petra López-Jiménez 2 ID and Helena M. Ramos 1 ID
1 Civil Engineering, Architecture and Georesources Department, CERIS, Instituto Superior Técnico,
Universidade de Lisboa, 1049-001 Lisboa, Portugal; hramos.ist@gmail.com
2 Hydraulic and Environmental Engineering Department, Universitat Politècnica de València,
46022 Valencia, Spain; mopesan1@upv.es (M.P.-S.); palopez@upv.es (P.L.-J.)
3 Department of Civil, Architectural and Environmental Engineering, Università di Napoli Federico II,
via Claudio 21, 80125 Napoli, Italy; arcarrav@unina.it
* Correspondence: m.c.madeira.simao@tecnico.ulisboa.pt; Tel.: +35-121-841-8151

Received: 27 October 2017; Accepted: 20 December 2017; Published: 27 December 2017

Abstract: Velocity profiles originated by a pump as turbine (PAT) were measured using an ultrasonic
doppler velocimetry (UDV). PAT behavior is influenced by the velocity data. The effect of the
rotational speed and the associated flow velocity variations were investigated. This research focuses,
particularly, on the velocity profiles achieved for different rotational speeds and discharge values
along the impeller since that is where the available hydraulic power is transformed into the mechanical
power. Comparisons were made between experimental test results and computational fluid dynamics
(CFD) simulations. The used CFD model was calibrated and validated using the same conditions
as the experimental facility. The numerical simulations showed good approximation with the
velocity measurements for different cross-sections along the PAT system. The application of this CFD
numerical model and experimental tests contributed to better understanding the system behavior
and to reach the best efficiency operating conditions. Improvements in the knowledge about the
hydrodynamic flow behavior associated with the velocity triangles contribute to improvements in
the PAT concept and operation.

Keywords: velocity profiles; PAT; CFD analyses; ultrasonic doppler velocimetry (UDV); experimental tests

1. Introduction
Small hydropower systems have become attractive for generating electricity, despite the cost
per kW of energy produced by these facilities that can be higher than the large hydroelectric power
plants [1]. In recent years, several researchers emphasized the importance of using a simple turbine
in order to reduce the cost of energy produced. The idea of using pumps as hydraulic turbines is an
attractive and important alternative, since pumps are relatively simple machines, their maintenance is
easy and readily available in most developing countries. From the economical point of view, it is often
stated that pumps working as turbines in the range of 1–500 kW allow capital payback periods of two
years or less which is considerably less than a conventional turbine [1].
Instead of dissipating excess pressure, energy recovery using micro hydro turbines or pumps as
turbines (PATs) have an increasing role as an effective way to control pressure level in water supply
systems [2]. According to [3,4], a PAT can reach a relatively high efficiency (up to 85%), compared to
traditional turbines [3], several case studies can be found regarding the application of PATs: In Murcia
(Spain), the maximum installation capacity in a main wall shear stress (WSS) with a flow rate of
300 m3 /h and a water head of 30 m, reach 100 kW using PATs [4,5]. In Germany, eight parallel PATs
with a total installation capacity of 300 kW were used in a WSS for pressure control and electricity

Fluids 2018, 3, 3; doi:10.3390/fluids3010003 www.mdpi.com/journal/fluids

Fluids 2018, 3, 3 2 of 21

generation in pipelines, where the generated electricity fed into local grid [6]. Others case studies
were enumerated by [7]; these cases showed the advantages of these machines in order to improve the
efficiency and sustainability of the water distribution systems.
Pumps operating as turbines, besides presenting higher head and discharge in the same rotational
speed, the hydraulic behavior rotates as a turbine changes. The flow field inside the volute is
Fluids 2017, 3, 3 2 of 21
three-dimensional with an important vortex pair created by the curvature of the flow passage and
by secondaryPATsflows,
with a which
total installation
are mainly capacity of 300 kW
generated bywere used in
the flow a WSS for
through pressure
guide control
vanes and This flow
[8–10].
electricity generation in pipelines, where the generated electricity fed into local grid [6]. Others case
structure creates a very difficult problem for the theoretical modeling and its corresponding numerical
studies were enumerated by [7]; these cases showed the advantages of these machines in order to
solution, which
improveneeds to be validated
the efficiency by detailed
and sustainability experimental
of the water data [8–10].
distribution systems.
Thus, since the PAT behavior is very complex and
Pumps operating as turbines, besides presenting higher head difficult to find a relationintothecover
and discharge same all pumps
rotational speed, the hydraulic behavior rotates as a turbine changes.
operating in reverse mode, the use of full computational fluid dynamics (CFD) to simulate theThe flow field inside the volute
is three-dimensional with an important vortex pair created by the curvature of the flow passage and
PAT performance can be a solution to overcome this problem. This research by using the setup of
by secondary flows, which are mainly generated by the flow through guide vanes [8–10]. This flow
the mathematical modela FloEFD
structure creates (Mentor
very difficult problem Graphics, Wilsonville,
for the theoretical OR, and
modeling USA) its performed
correspondingwith finite
volume method
numerical (FVM)
solution,with second-order
which spatial
needs to be validated by discretization scheme
detailed experimental of the convection terms in
data [8–10].
the governingThus, since thethe
equations PATCFDbehavior is very complex
workbench and difficult to findall
SolidWorks-FloEFD, a relation
geometryto cover
of all pumps including
a pump
operating in reverse mode, the use of full computational fluid dynamics (CFD) to simulate the PAT
inlet, volute, impeller, draft tube and outlet are simulated in reverse mode operation, as a turbine.
performance can be a solution to overcome this problem. This research by using the setup of the
The PAT facility system
mathematical wasFloEFD
model established
(MentoratGraphics,
the Instituto Superior
Wilsonville, OR, USA)Técnico, Universidade
performed with finite de Lisboa
and used for experimental verification of theoretical concepts and numerical
volume method (FVM) with second-order spatial discretization scheme of the convection terms results. Theinparameters
the governing equations the CFD workbench SolidWorks-FloEFD, all geometry
required to obtain the complete characteristic curves of the PAT were measured in the lab facility. of a pump including
Finally, allinlet, volute, impeller, draft tube and outlet are simulated in reverse mode operation, as a turbine.
theoretical, numerical and experimental results were compared and discussed. Moreover,
The PAT facility system was established at the Instituto Superior Técnico, Universidade de Lisboa
this research
and work
used for isexperimental
related to an experimental
verification study
of theoretical on the
concepts flow
and behavior
numerical and
results. Thevelocity
parameters profiles and
fields in a required
tested PAT.to obtain the complete characteristic curves of the PAT were measured in the lab facility.
Finally, all theoretical, numerical and experimental results were compared and discussed. Moreover,
2. Experimental Setup
this research work is related to an experimental study on the flow behavior and velocity profiles and
fields in a tested PAT.
2.1. Lab Facility Description
2. Experimental Setup
The hydraulic facility is composed of several components: (i) A loop high density polyethylene
2.1. Lab Facility Description
pipe (Figure 1); (ii) a pump that feeds the loop pipe system; (iii) a compressed air vessel to stabilize the
The hydraulic
pipe system pressure; (iv)facility is composed of several
an electromagnetic flow components:
meter to measure(i) A loopthe
highinstant
density flow;
polyethylene
(v) flow control
pipe (Figure 1); (ii) a pump that feeds the loop pipe system; (iii) a compressed air vessel to stabilize
valves along the pipe; (vi) pressure transducers, to register the pressure variations
the pipe system pressure; (iv) an electromagnetic flow meter to measure the instant flow; (v) flow
and a picoscope data
acquisition system
control connected
valves along the to a laptop;
pipe; (vii)transducers,
(vi) pressure a centrifugal microthe
to register PAT; (viii)variations
pressure a free surface
and a reservoir,
with a 90◦picoscope
triangulardata weir to calibrate
acquisition the discharge
system connected flow.
to a laptop; (vii) The nominal
a centrifugal external
micro diameter
PAT; (viii) a free (DN) is
surface reservoir, with a 90° triangular weir to calibrate the discharge flow.
50 mm with 100 m of length. The pressure inside the air vessel can oscillate between 20 and 35 mWc The nominal external
diameter (DN) is 50 mm with 100 m of length. The pressure inside the air vessel can oscillate
(meters ofbetween
water column).
20 and 35 mWc (meters of water column).

Figure 1. Experimental facility at Instituto Superior Técnico (IST) lab.

Figure 1. Experimental facility at Instituto Superior Técnico (IST) lab.
Fluids 2018, 3, 3 3 of 21

Two pressure transducers are located upstream and downstream of the PAT, respectively at points
A and B (Figure 1). The best efficiency point (BEP) of the PAT is characterized by a flow rate of 3.36 L/s,
with a head of 4 m, the efficiency of 60%, and the power of 80 W, for a rotational speed of 1020 rpm
(Table 1).

Table 1. Best efficiency point provided by the manufacturer.

Best Efficiency Point

Discharge 3.36 L/s
Head 4m
Efficiency 60%
Power 0.08 kW
Rotational speed 1020 rpm
Specific speed 51 rpm (m, kW)

Due to the laboratory’s operational restrictions, the characteristic values of the PAT are limited.
All tests were carried out for different steady state conditions depending on the opening degree of the
flow control valves.

2.2. Experimental Tests

Several tests were carried out for different flow values and consequently different rotational
speeds. Some tests are presented in Table 2 sorted by the rotational speed values. For each flow
rate the rotational speed of the PAT was measured by means of a digital tachometer. The pressure
values were recorded as previously described, and the velocity profiles were collected in different PAT
sections (S1–S4) using a 3000 series ultrasonic doppler velocimetry (UDV). This equipment allows the
registration of velocity profiles, in real-time, either in steady or unsteady state conditions. This device
is able to perform instantaneous measurements of fluid velocities at a number of points, along the
direction of emission of the ultrasonic signal, with remarkably high resolutions in space and time.
It is an interesting tool for turbulent flow analysis, where high spatial and temporal accuracies are
required for both instantaneous measurements and averaging operations. In each UDV measurement,
the velocity was captured in 100 different points in a total of 100 profiles. Details of the points (S1–S4)
where the measurements took place are identified in Figure 2.

Table 2. Experiments in the pump as turbine (PAT) for different rotational speeds with ultrasonic
doppler velocimetry (UDV) data acquisition from S1 to S4 points.

Section QA (L/s) N (rpm) HA (m) HB (m) ∆H (m) Ph (W) Torque (N·m) Pm (W) η (%)
S3 2.40 520 4.29 2.24 2.04 47.98 0.33 18.01 38
S4 2.35 600 4.19 2.17 2.02 46.52 0.32 20.11 43
S2 2.46 680 4.68 2.33 2.35 56.65 0.34 24.21 43
S1 2.40 780 4.71 1.74 2.98 70.09 0.35 28.18 40
S3 2.80 850 6.11 2.77 3.34 91.65 0.21 19.09 21
S2
2.80 880 6.15 2.7 3.44 94.39 0.37 34.10 36
S4
S1 2.70 1140 7.03 1.98 5.05 133.62 0.51 60.88 46

Figure 3a shows low velocity values in S3 at the draft tube induced by the existent vortex at the
exit of the impeller. The influence of the curves in the inlet and outlet pipe to/from the PAT is visible
in respective velocity profiles between S1 and S4 (Figure 3a,c,e,f). In Section S1 the velocity profile is
typical of a turbulent flow with a quasi-uniform velocity distribution, presenting only few fluctuations
(Figure 3d,h). The proximity of Section S1 to a 45◦ pipe branch also support this irregularity.
In order to collect the velocity profiles in Section S2, where the first velocity values were
recorded by the extruder of the curve (Figure 3c,f) with lower values, the influence of the curve
Fluids 2018, 3, 3 4 of 21

Fluids 2017, 3, 3 4 of 21
in the non-symmetric velocity distribution is visible. The velocity variation for this measure point S2
results
results from
from the
the effect
effect of
of the
the curvature,
curvature, and
and also
also from
from the
the effect
effect of
of the
the rotating
rotating flow
flow from
from S1
S1 to
to S2 due
S2 due
to
to the
the changes
changes of
of the
the flow
flow direction.
direction.

Fluids 2017, 3, 3 4 of 21

results from the effect of the curvature, and also from the effect of the rotating flow from S1 to S2 due
to the changes of the flow direction.

Figure
Figure 2.
2. Ultrasonic
Ultrasonic doppler
doppler velocimetry
velocimetry (UDV)
(UDV) measurement points.
measurement points.

As a result of the rotational speed of the PAT and the shape of the blades, the flow has a
As a result of the rotational speed of the PAT and the shape of the blades, the flow has a rotational
rotational behavior at the outlet of the runner and along the draft tube, i.e., where the measured
behavior at the outlet of the runner anddoppler
along the draft tube, i.e., where the measured point S3 is
point S3 is located (Figure Figure 2). Thus, the flow
2. Ultrasonic leaves the(UDV)
velocimetry impeller with points.
measurement a significant tangential velocity
located (Figure 2). Thus, the flow leaves the impeller with a significant tangential velocity in relation
in relation to the axial one. From the exit speedofofthethe runner and and along the draft the
tube,flowa has
strong turbulent
to the axial one.AsFrom a result of the
the exitrotational
of the runner andPAT along the shape
the draftof tube,the blades,
a strong a
turbulent vortex is
vortex is generated
rotational near
behaviortheatpipe’s
the outletaxis, where
of the runnertheandaxial
along velocity
the draft tube,of the
i.e., flow
where is thelower.
measured In Figure 3a,e,
generated near the
point valuespipe’s
S3 is located axis, where the axial velocity of the flow is lower. In Figure 3a,e, the low
the low velocity are(Figure
close2). toThus, the flow axis,
the pipe’s leaves where
the impellerthewith a significant
vortex core istangential
located. velocity
Additionally, the
velocity values are close
in relation to theto theone.
axial pipe’s
From axis,
the exitwhere the vortex
of the runner and along coretheisdraft
located. Additionally,
tube, a strong turbulent the velocity
velocity in the direction
vortex of the flow presents higher values fromthe the axis to the In periphery
Figure 3a,e, of the draft
in the direction ofis the
generated
flownear the pipe’s
presents axis, where
higher the axial
values from velocity
the of axis flow
to the is lower.
periphery of the draft tube,
tube, since near thevelocity
the low wallsvalues
the rotation
are close tovorticity is significantly
the pipe’s axis, where the vortexlower, enabling
core is located. higher the
Additionally, velocity values
since near the walls the rotation vorticity is significantly lower, enabling higher velocity values in the
in the mean velocity in the direction
flow direction. The of larger
the flow the
presents higher values of
cross-section from thethedraft
axis totube
the periphery
occupied of theby
draft
the vortex, the
mean flow direction. Thethelarger
tube, since near walls the therotation
cross-section
vorticity is of the draft
significantly tube
lower, occupied
enabling higherby the vortex,
velocity values the greater
greater the diametric
in the mean flowlength of the larger
direction. pipe (i.e., the axis-yof thein Figure 3a,e) where the the
lower velocity
the diametric length of the pipe The (i.e., thethe cross-section
axis-y in Figure draft
3a,e) tube occupied
where the by the vortex,
lower velocity values are
values are observed
greater thein the flow
diametric direction.
length of the pipe (i.e., the axis-y in Figure 3a,e) where the lower velocity
observed in values
the flow direction.
are observed in the flow direction.

Figure 3. Cont.
Fluids 2018, 3, 3 5 of 21
Fluids 2017, 3, 3 5 of 21

Figure 3. Velocity profiles along the pipe diameter measure with Ultrasonic doppler velocimetry
Velocity
Figure 3. (UDV), profiles
expressed as along the pipe
the distance diameter
between the UDVmeasure
probe andwith Ultrasonic
faraway pipe wall doppler velocimetry
in the radial
direction: as
(UDV), expressed (a) the
N = distance
520 rpm; (b) N = 600 rpm;
between (c) N = probe
the UDV 680 rpm; (d) faraway
and N = 780 rpm; (e) wall
pipe N = 850
inrpm;
the (f) N = direction:
radial
(a) N = 520880 rpm; (g) N = 880 rpm; (h) N = 1140 rpm.
rpm; (b) N = 600 rpm; (c) N = 680 rpm; (d) N = 780 rpm; (e) N = 850 rpm; (f) N = 880 rpm;
(g) N = 880The
rpm; N = 1140
(h)velocity
axial can rpm.
be virtually blocked, in cases where the cross-sectional area of the draft
tube, occupied by the vortex, is very significant. As a result of the rotational behavior of the flow and
the reduced pressure values that occur at the draft tube, the resulting velocities are also reduced,
The axial velocity can be virtually blocked, in cases where the cross-sectional area of the draft
which indicates an eventual reverse flow [11].
tube, occupiedInby the 3b,g,
Figure vortex, is very significant.
the measuring As a result
point S4 is already locatedofoutthe rotational
of the behavior
draft tube, of the flow and
so it is expected
the reduced pressure
that values
in this section the that occur
vorticity andat theturbulence
flow draft tube,
willthe resulting
decrease velocities
returning again to are also reduced, which
an approximate
indicates anuniform flow type.
eventual The flow
reverse differential
[11]. of velocities between the axis and the pipe wall in S4 is lower
than in S3. The non-uniformity in the velocity distribution showed in Figure 3b,g results from the
In Figure 3b,g, the measuring point S4 is already located out of the draft tube, so it is expected
vorticity that still attain S4, which also result from 45° junction of flow changed direction located
that in thisimmediately
section the vorticity
upstream and2).flow turbulence will decrease returning again to an approximate
(Figure
uniform flow type. The differential
Most of the velocity profilesof velocities between
recorded show the reduction
a strong axis andofthe thepipe
flowwall in S4
velocity in is
thelower than
adjacent area to the pipe walls, due to the tangential tension (drag tension) that
in S3. The non-uniformity in the velocity distribution showed in Figure 3b,g results from the vorticity occurs herein
increasing the resistance to the flow propagation.
that still attain S4, which also result from 45◦ junction of flow changed direction located immediately
upstream 3.(Figure 2). Simulations
Numerical
Most of the velocity profiles recorded show a strong reduction of the flow velocity in the adjacent
area to the3.1. Model Description
pipe walls, due to the tangential tension (drag tension) that occurs herein increasing the
resistance to the perform
To the computational simulation of the flow inside the selected PAT, it was necessary
flow propagation.
to build the 3D geometric model represented in Figure 4 based on the real PAT characteristics. For
that, the geometry of the PAT was created in Solidworks and then imported to the CFD model
3. Numerical Simulations
(FloEFD). The geometric model resulted from gathering several independent solid components,
with the most relevant parts (i.e., the impeller and the volute). The computational modeling was
3.1. Model Description
performed for the PAT system, between points A and B, located respectively upstream and
downstream
To perform of the PAT (Figuresimulation
the computational 4) being sections of connection
of the to any
flow inside thewater pipe system.
selected PAT, it was necessary
to
build the 3D geometric model represented in Figure 4 based on the real PAT characteristics. For that,
the geometry of the PAT was created in Solidworks and then imported to the CFD model (FloEFD).
The geometric model resulted from gathering several independent solid components, with the most
relevant parts (i.e., the impeller and the volute). The computational modeling was performed for
the PAT system, between points A and B, located respectively upstream and downstream of the PAT
(Figure 4) being sections of connection to any water pipe system.
Fluids 2018, 3, 3 6 of 21
Fluids 2017, 3, 3 6 of 21

Figure 4. Geometric
Geometric model
model for
for computational
computational fluid
fluid dynamics
dynamics (CFD)
(CFD) analyses: (a)
(a) pump
pump as turbine
(PAT) impeller; (b)
(PAT) impeller; (b) PAT
PATvolute.
volute.

The elementsused
The elements usedtotosupport
support
thethe construction
construction of impeller
of the the impeller geometry
geometry of theof
PAT thewere
PATaswere as
follows:
follows:
 Values of
Values of the
the characteristic
characteristicparameters
parametersofofthe
theBEP
BEPofofthethe
PAT, supplied
PAT, by the
supplied by manufacturer and
the manufacturer
presented in Table 1.
and presented in Table 1.
 Longitudinal and
Longitudinal and transversal
transversal sections
sections of
of the
the PAT
PAT device,
device, supplied
supplied by
by the
the manufacturer.
manufacturer.
The
The simulations
simulations were
were performed
performed considering
considering an
an incompressible
incompressible flow,
flow, with constant properties
and using the k-ε model as the turbulent model. The flow inside the suction pipe is described by the
Navier–Stokes–Reynolds Average for thethe k-ε
k-ε model
model [12]:
[12]:

∂ρ + ∇ ∙
+ ∇·(ρU ) =
= 00
(1)
(1)
∂t
∂ρU + ∇ ∙ − ∇ ∙ ∇U = ∇ + ∇ ∙  ∇U + T (2)
+ ∇·(ρU ·U ) − ∇· µe f f ∇U = ∇ p0 + ∇· µe f f ∇U + B (2)
∂t
where is the effective viscosity for turbulence, is the sum of forces, and is the modified
where µ f f is the effective viscosity for turbulence, B is the sum of forces, and p0 is the modified
pressureegiven by [13]:
pressure given by [13]:
22
p0 =
= p+
+ ρk (3)
(3)
33
= µ + µt
µe f f = (4)
(4)
k-ε model
modelassumes
assumes that thethe
that turbulent viscosity
turbulent is related
viscosity to the turbulent
is related kinetic energy
to the turbulent andenergy
kinetic dissipation
and
through thethrough
dissipation relation:the relation:
k2
µt = Cµ ρ (5)
= ε (5)
The values of k and ε come directly from the transport equations for turbulent kinetic energy and
The values of k and ε come directly from the transport equations for turbulent kinetic energy
dissipation range of turbulence [14]:
and dissipation range of turbulence [14]:
  
∂ρk µt
+ ∇·(ρUk) = ∇· µ + ∇k + P − ρε (6)
∂t + ∇ ∙ =∇∙ + σk ∇ + k− (6)
  
∂ρε µt ε
+ ∇·(ρUε) = ∇· µ + ∇ε + (Cε1 Pk − Cε2 ρε) (7)
∂t + ∇ ∙ =∇∙ + σε ∇ + k − (7)

where , and , and are constants and is the production of turbulence due to the
forces of viscosity which is modeled as follows [13]:
Fluids 2018, 3, 3 7 of 21

where Cε1 , and Cε2 , σk and σε are constants and Pk is the production of turbulence due to the forces of
viscosity which is modeled as follows [13]:
  2
Pk = µt ∇U · ∇U + ∇U T − ∇·U (3µt ∇·U + ρk ) + Pkb (8)
3
For an incompressible flow, ∇U is small, and the second term on the right does not contribute
significantly [14].

3.2. Mesh
After defining and bringing in the geometric model, an automatic generation of the initial
calculation mesh was conducted, by specifying values for the parameters used to control the mesh
resolution. These parameters need to be adequate to the characteristics of the geometric models to
obtain results with a satisfactory level of accuracy, without using significant computational resources.
Considering the complexity of the rotor geometry, it is advantageous to refine the cells in the local
region of the computational domain relative to the rotor, using the definition of an initial local mesh [15].
The mesh generation process begins with the definition of the rectangular computational
domain. Three sets of orthogonal planes to the Cartesian coordinate system were defined inside
of it. The intersection of these planes defines the set of rectangular cells that form the base mesh.
In the next stage of the process, the geometry inside the computational domain, i.e., the PAT system,
is captured [16]. The program crosses the mesh cells sequentially and for each cell it parses the
geometrical configuration inside the cell. Followed by a refinement criterion:

 If Csplit > ε split , the cell has to be subdivided. The value of indicator Csplit (i.e., refinement
indicator value) is calculated individually for each cell according to an algorithm that controls
the type of refinement. The value of ε split is a constant, defined automatically and represents the
refinement threshold value. Setting refinement levels individually for each criterion makes it
possible to influence the total number of cells that will be generated [15].

Additionally, the refinement level of the cell is the number of times it has been refined relative to
the initial cell in the base mesh. In SolidWorks Flow Simulation the maximum number of refinement
levels is limited to seven, so that the smallest cell size is 1/128 of the original cell in the base mesh
(level 0) [16].
An initial mesh that best fits the geometric model is obtained, leading to results that more
accurately translate the flow dynamics. The values for the parameters that govern the procedure of the
CFD model were defined for the automatic generation of the initial local calculation mesh, according
to [15]. The CFD model automatically specifies the remaining parameters that govern the initial, local
and global mesh generation. Since the flow field inside the PAT presents significant complexity induced
by the rotational movement of the rotor and the associated geometry, the initial calculation mesh was
also refined. Hence, different values were allocated to the parameters that govern the procedure of the
CFD model, for the initial calculation mesh, to the solution during the calculation process.
To ensure a proper and acceptable accuracy of the results, the discretization mesh undergo a Mesh
Sensitivity Analysis (MSA), in which the density of cells is gradually increased and the solution must be
re-assessed until results demonstrate sufficient independence on the mesh size [16]. The discretization
of the mesh was performed by assuming an initial coarse mesh with the total number of cells equal to
25,668, followed by a mesh refinement process. This process stopped when the difference in the head
drop (i.e., in point B) between consecutive meshes was less than 1%, which occurred for mesh with
102,274 cells (Table 3).
Fluids 2018, 3, 3 8 of 21

Table 3. Summary of Mesh Sensitivity Analysis (MSA).

Mesh Number of Fluid Mesh Cells Number of Solid Mesh Cells HB (m) Error (%) Duration
Mesh 1 25,668 15,400 5.10 0.15 h
Mesh 2 50,381 30,229 2.30 55% 0.29 h
Mesh 3 100,691 60,088 1.42 38% 1.23 h
Mesh 4 101,936 61,162 1.73 22% 1.52 h
Mesh 5 102,152 61,291 1.68 3% 2.28 h
Mesh 6 102,274 61,364 1.67 1% 2.33 h

3.3. Computational Fluid Dynamics (CFD) Results

In the CFD model, the boundary conditions were consigned to sections A and B (Table 2),
as the inlet and outlet measured points of the PAT system, respectively. The volume of the flow
rate, corresponding to each test, was allocated to the inlet condition. As for the outlet condition,
the pressure recorded in the transducer, obtained at point B, was set. For the rotation of the impeller,
the corresponding rotational speed obtained experimentally was also considered. Table 4 shows the
operating and boundary conditions given to each of the tests and the respective numerical results.

Table 4. Results obtained in computational fluid dynamics (CFD) simulations.

Q (L/s) N (rpm) HA (m) HB (m) ∆H (m) Torque (N·m) w (rad/s) Ph (W) Pm (W) η (%)
2.40 520 4.80 2.33 2.47 0.39 54.45 58.13 21.24 37
2.35 600 4.41 2.57 1.83 0.28 62.83 42.24 17.59 42
2.46 680 4.76 2.37 2.39 0.35 71.21 57.55 24.92 43
2.40 780 4.92 1.85 3.07 0.36 81.68 72.10 29.41 41
2.80 850 6.71 3.18 3.53 0.35 89.01 96.83 31.15 32
2.80 880 7.85 2.88 4.97 0.62 92.15 136.46 57.14 42
3.36 1020 6.83 1.67 5.06 1.13 97.39 185.56 110.05 60
2.70 1140 7.34 2.12 5.22 0.55 119.38 138.05 65.66 48
4.17 1275 9.61 1.69 7.76 0.86 133.52 317.21 114.83 36
4.91 1500 15.29 3.82 11.24 1.30 157.08 540.79 204.20 38

Figure 5 shows the velocity streamlines and the velocities field in planes that intersect the
geometric model. The hydrodynamic of the flow that occurs near point B is characterized by the
presence of curves and elbows at upstream. This figure shows a reduction in the velocity magnitude,
upstream and along the impeller, caused by the energy promoted by the flow to the impeller. The vortex
that forms at the exit of the rotor and that extends downstream thereof is associated with turbulence,
and hydrodynamic instability, which effects result in pressure fluctuations and efficiency losses.
In Figure 5, the velocity streamlines distribution confirms the rotation of the flow along the draft
tube, and the distribution of the tangential velocity shows increasing values from the axis to the wall
surface of the pipe. The reduced flow velocity values that occurred in the vortex core (located near the
axis of the pipe) that is formed downstream of the rotor, and the marked velocity that reaches the wall
of the draft tube due to the rotationally of the flow, are more visible in Figure 5d,f,g. The rotational
behavior, affected by the rotational speed of the impeller and the shape of the blades is also verified.
A decrease of the tangential velocity values from the periphery to the shaft of the runner confirming
the centrifugal force effect is shown which arises with the rotor rotation increase. The flow enters
the volute and exerts a radial force on the rotor, thereby imparting a certain rotational speed to the
impeller. Consequently, the runner rotation continuously varying the direction of flow along the
impeller, becomes to axial direction at the exit.
The results show that for the same number of velocity vectors, and as the speed ratio increases,
high concentration of velocity vectors migrates from the outside of the blade to the core side.
The complex flow near the impeller inlet varies with the rotation, inducing different performances.
Hence, the complex vortices cause additional vibrations, turbulences and consequently additional
dissipative effects [17,18].
Fluids 2018, 3, 3 9 of 21
Fluids 2017, 3, 3 9 of 21

(a)

(b)

(c)

(d)

(e)

Figure 5. Cont.
Fluids 2018, 3, 3 10 of 21
Fluids 2017, 3, 3 10 of 21

(f)

(g)
Figure 5. Velocity values for different rotational speeds (m/s): Streamlines (on the left); velocities
Figure Velocity
field in5.the PAT (onvalues for different
the right). rotational
(a) N = 600 rpm; (b)speeds (m/s):
N = 780 Streamlines
rpm; (c) (on (d)
N = 880 rpm; theNleft); velocities
= 1020 rpm; (e)
field in the PAT (on the right). (a) N = 600 rpm;
N = 1140 rpm; (f) N = 1250 rpm; (g) N = 1500 rpm. (b) N = 780 rpm; (c) N = 880 rpm; (d) N = 1020 rpm;
(e) N = 1140 rpm; (f) N = 1250 rpm; (g) N = 1500 rpm.
4. Experimental and Computational Fluid Dynamics (CFD) Comparisons
4. Experimental and Computational Fluid Dynamics (CFD) Comparisons
4.1. Velocity Profiles
4.1. Velocity Profiles
Experimental tests were used to calibrate and validate the CFD model. Therefore, the
hydrodynamic oftests
Experimental wereinside
the flow used tothecalibrate
PAT was andsimulated
validate the forCFD model.operating
the same Therefore,conditions
the hydrodynamic
as the lab
ofconditions,
the flow inside
according to the following methodology: (1) Obtaining the distribution ofaccording
the PAT was simulated for the same operating conditions as the lab conditions, physical
toparameters
the following methodology: (1) Obtaining the distribution of physical parameters
describing the hydrodynamic of the flow in planes that intersect the representative describing the
hydrodynamic
geometric model of the flowPAT;
of the in planes that intersect
(2) recording the representative
the velocity variation ingeometric model
the measured of thewhere
points PAT; (2)
the
recording the velocity variation in the measured points where the velocity
velocity profiles were recorded in the direction of the UDV velocity values; (3) comparing the profiles were recorded in
the direction
velocity of the
profiles UDV velocity
between values;tests
experimental (3) comparing the velocity
and the numerical profiles between experimental
simulations.
tests and the numerical simulations.
From the computational simulation for different rotation speeds, the velocity profiles were
From in
obtained thethecomputational
four measure points simulation
(Figure for6).different
Compared rotation
with the speeds, the velocity
experimental profiles
velocity were
profile, the
obtained in the four measure points (Figure 6). Compared with the
CFD model shows a trend of velocity variation similar to the one observed with UDV. Since the experimental velocity profile,
the CFD model
variation of theshows
physicala trend of velocity
parameters thatvariation similarthe
characterizes to hydrodynamics
the one observedofwith the UDV. Since the
flow inside any
variation of the physical parameters that characterizes the hydrodynamics
hydraulic device depends on the geometry, it is necessary to adequately simulate the shape of the of the flow inside any
hydraulic device
device. Thus, thedepends on the geometry,
major differences it is necessary
can be related to adequately
to the limited simulate
data to perform thethe shape ofofthe
geometry the
device. Thus,
PAT installation.the major differences can be related to the limited data to perform the geometry of the
PAT installation.
Figure 6 shows some irregularities in the distribution of velocity values. Both velocity profiles
Figure 6ofshows
are typical someflows.
turbulent irregularities in the distribution of velocity values. Both velocity profiles are
typicalDueof turbulent flows.
to the proximity of S3 to the turbulent flow in the PAT rotor (see Figure 2), an irregular
Due to the proximity
velocity distribution of S3 to
is expected in the
this turbulent flow in
section (Figure 6a).the
The PAT rotor (see Figure
experimental velocities2), profile
an irregular
shows
velocity
this expected irregularity noticeably, and the CFD results also allow the identification of the shows
distribution is expected in this section (Figure 6a). The experimental velocities profile typical
this
flowexpected
patternirregularity
of the draftnoticeably,
tube flow.and Thethemean
CFD velocity
results also allow
values the identification
relative to the velocityof thediagrams
typical
flow pattern
obtained of the draft tube
experimentally andflow. The mean velocity
computationally values
are 1.43 relative
m/s and 1.49tom/s
therespectively
velocity diagrams obtained
for values of the
experimentally and computationally are 1.43 m/s and 1.49 m/s respectively
flow around 2.4 L/s. Thus, the difference in the mean velocity between experiments and simulations for values of the flow
are quite similar.
Fluids 2018, 3, 3 11 of 21

around 2.4 L/s. Thus, the difference in the mean velocity between experiments and simulations are
quite similar.
Fluids 2017, 3, 3 11 of 21

(a) S3 (b) S4

(c) S2 (d) S1
Figure 6. Velocity profiles (m/s) obtained in experimental tests and CFD simulations: (a) N = 520 rpm;
Figure 6. Velocity profiles (m/s) obtained in experimental tests and CFD simulations: (a) N = 520 rpm;
(b) N = 600 rpm; (c) N = 680 rpm; (d) N = 780 rpm.
(b) N = 600 rpm; (c) N = 680 rpm; (d) N = 780 rpm.
In Figure 7, for higher rotational speed values the behavior is similar. Additionally, the
non-uniform velocity distribution is visible due to change of direction of the pipe flow. The velocity
In Figure 7, for higher
is affected rotational
by the viscous speed stresses
tangential values occurred
the behavior is similar.
near the Additionally,
pipe wall. the non-uniform
Thus, the velocity
velocity distribution is visible due to change of direction of the pipe flow. The velocity
distribution that was obtained through the CFD model shows the same tendency of the measured is affected by
flow behavior verified in the experimental setup. The mean velocity values relative
the viscous tangential stresses occurred near the pipe wall. Thus, the velocity distribution that wasto the velocity
distribution obtained experimentally and by numerical simulation were 1.87 m/s and 1.83 m/s
obtained through the CFD model shows the same tendency of the measured flow behavior verified
respectively for a mean flow of 3 L/s.
in the experimental setup.
The results The
for the mean
target velocity
operating values
conditions arerelative to the
summarized velocity
in Table 5. Thedistribution
results show obtained
experimentally and by numerical
good correlation between the simulation were
experimental data1.87
and m/s andmodel.
the CFD 1.83 m/s respectively
The main discrepancyforwas
a mean flow
of 3 L/s. encountered for the torque measurement predictions, since the CFD overestimate the torque
extracted by the turbine, and consequently the turbine efficiency, since the scale effects are not
The results for the target operating conditions are summarized in Table 5. The results show good
possible to reproduce with the CFD model. This overestimation is caused by the bearing losses,
correlationwhich
between themodeled
are not experimental
in the CFD data and the
analyses. CFD
The totalmodel. The
estimated mainlosses
bearing discrepancy
representedwas encountered
about
for the torque measurement predictions, since the CFD overestimate the torque extracted by the turbine,
and consequently the turbine efficiency, since the scale effects are not possible to reproduce with the
CFD model. This overestimation is caused by the bearing losses, which are not modeled in the CFD
analyses. The total estimated bearing losses represented about 11.36% of the discrepancy (in average) in
the shaft power between the experimental data and the numerical predictions, as in previous research
suggested by [17].
 Fluids 2017, 3, 3 12 of 21

Fluids 2018, 11.36%

3, 3 of the discrepancy (in average) in the shaft power between the experimental data and the 12 of 21
numerical predictions, as in previous research suggested by [17].

(a) S3 (b) S4

(c) S2 (d) S1
Figure 7. Velocity profiles (m/s) obtained through experimental tests and CFD simulations: (a) N =
Figure 7. Velocity profiles (m/s) obtained through experimental tests and CFD simulations: (a) N = 850 rpm;
850 rpm; (b) N = 880 rpm; (c) N = 880 rpm; (d) N = 1140 rpm.
(b) N = 880 rpm; (c) N = 880 rpm; (d) N = 1140 rpm.
Table 5. Results obtained in experimental tests and CFD model for different rotational speeds.

EXP CFD
N (rpm) Q (L/s) Ph Pm Torque η Ph Pm Torque η
Table 5. Results obtained
∆h (m)in experimental tests and CFD model
∆h (m) for different rotational speeds.
(W) (W) (N·m) (%) (W) (W) (N·m) (%)
520 2.40 2.04 47.98 18.01 0.33 38 2.47 58.13 21.24 0.39 37
600 2.35 2.02 EXP
46.52 20.11 0.32 43 1.83 42.24 17.59 CFD0.28 42
N (rpm) Q (L/s)
680 ∆h2.46
(m) Ph2.35
(W) 56.65
Pm (W) 24.21
Torque (N0.34
·m) η (%)43 2.39
∆h (m) 57.55
Ph (W) 24.92
Pm (W)0.35 Torque
43 (N·m) η (%)
780 2.40 2.98 70.09 28.18 0.35 40 3.07 72.10 29.41 0.36 41
520 2.40
850 2.04
2.80 47.98
3.34 18.01
91.65 19.090.33 0.21 38 21 2.47
3.53 58.13
96.83 31.15 21.24 0.35 0.39
32 37
600 2.35
880 2.02
2.80 46.52
3.44 20.11
94.39 34.100.32 0.37 43 36 1.83
4.97 42.24 57.1417.59 0.62
136.46 0.28
42 42
680 2.46
1020 2.35
3.36 56.65
6.30 24.21
218.00 104.000.34 1.21 43 60 2.39
5.06 57.55 110.0524.92 1.13
185.56 0.35
60 43
780 2.40
1140 2.98
2.70 70.09
5.05 28.18
133.62 60.880.35 0.51 40 46 3.07
5.22 72.10 65.6629.41 0.55
138.05 0.36
48 41
850 2.80
1275 3.34
4.17 91.65
7.06 19.09
292.42 112.000.21 0.84 21 38 3.53
7.76 96.83 114.8331.15 0.86
317.21 0.35
36 32
880 2.80
1500 3.44
4.91 94.39
10.23 34.10
498.84 181.000.37 1.15 36 36 4.97
11.24 136.46 204.2057.14 1.30
540.79 0.62
38 42
1020 3.36 6.30 218.00 104.00 1.21 60 5.06 185.56 110.05 1.13 60
1140 2.70 5.05 133.62 60.88 0.51 46 5.22 138.05 65.66 0.55 48
1275 4.17 7.06 292.42 112.00 0.84 38 7.76 317.21 114.83 0.86 36
1500 4.91 10.23 498.84 181.00 1.15 36 11.24 540.79 204.20 1.30 38

The efficiency values present the same range between 30% and 60% both for low (<850) or high
(>850) rotational speed values.

4.2. Triangle of Velocities

The distribution of the velocity vectors as the flow enters and leaves the impeller in a PAT
(Figure 8) is studied as a fluid structure interaction problem between static and movable parts [19–21].
The velocity vectors of a PAT can be decomposed into: (a)The absolute velocity (V ) of the flow at a
4.2. Triangle of Velocities
The distribution of the velocity vectors as the flow enters and leaves the impeller in a PAT
(Figure 8) is studied as a fluid structure interaction problem between static and movable parts [19–
21]. The
Fluids velocity
2018, 3, 3 vectors of a PAT can be decomposed into: (a)The absolute velocity of the 13
flowof 21
at a point, i.e., the velocity relative to a stationary point at radius r; (b) and the relative velocity of the
flow , i.e., the velocity of the fluid relative to the impeller. and are the tangential and the
point,components
radial i.e., the velocity
of relative to a stationary
, respectively. = 2 point is at
theradius r; (b) and
peripheral the relative
velocity of the velocity
impellerofitself,
the flow
at
( W ) , i.e., the velocity of the fluid relative to the impeller.
radius r with rotational speed n. The subscripts 1 and 2 represent V t and V are the tangential and the radial
ther inlet and outlet of the impeller,
components of V, respectively. C = 2πrn is the peripheral velocity of the impeller itself, at radius r with
respectively.
rotational speed n. The subscripts 1 and 2 represent the inlet and outlet of the impeller, respectively.

Figure 8. Inlet and outlet velocity triangles of the PAT impeller.

Figure 8. Inlet and outlet velocity triangles of the PAT impeller.

Todesign
To designthethevelocity
velocityvectors,
vectors,ititisisassumed
assumedthatthatthe
theflow
flowpasses
passesthrough
throughthetheimpeller,
impeller,and
andthe
the
relative velocity traces a path which is everywhere congruent to the blade, according
relative velocity traces a path which is everywhere congruent to the blade, according to [22–25]. At to [22–25]. At any
radius
any r inr the
radius impeller,
in the thethe
impeller, flow
flowdescribes
describesa tangential
a tangentialcomponent velocityVt , corresponding
componentofofvelocity , correspondingto
a tangential momentum
to a tangential momentum ρQV . The moment of this flux to the impeller
t . The moment of this flux to the impeller axis is axis is ρQV t r. Thetorque
. The torque
changeofofmoment
change momentfluxfluxbetween
betweenradius
radius(entrance
(entranceimpeller radius)r1r1and
impellerradius) and(exit
(exitimpeller radius)r2r,2 ,
impellerradius)
respectively the [26]:
respectively the [26]:
Torque = ρQVt1 r1 − ρQVt2 r2 (9)
= − (9)
The flow across a runner is characterized by the three types of velocities (i.e., the absolute velocity
The flow across a runner is characterized by the three types of velocities (i.e., the absolute
of the water (V), the relative velocity (W) guided by the blades orientation and the tangential velocity
velocity of the water (V), the relative velocity (W) guided by the blades orientation and the tangential
(C) imposed by the rotation of the impeller, a comparison with classical turbines velocity triangles
velocity (C) imposed by the rotation of the impeller, a comparison with classical turbines velocity
at inlet of the impeller (Figure 9a), with the studied PAT, is visible the differences in the peripheral
triangles at inlet of the impeller (Figure 9a), with the studied PAT, is visible the differences in the
velocity due to the smaller inertia of the PAT and the PAT blades configuration. This difference is
peripheral velocity due to the smaller inertia of the PAT and the PAT blades configuration. This
responsible for a lower efficiency value of PATs against the lower specific speeds Francis turbines.
difference is responsible for a lower efficiency value of PATs against the lower specific speeds
This difference will affect the dynamic behavior of both machines.
Francis turbines. This difference will affect the dynamic behavior of both machines.
Analyzing the velocities field of the flow in the radial PAT impeller the V and C components
Analyzing the velocities field of the flow in the radial PAT impeller the V and C components at
at each particle position are visible (Figure 9b). The inlet and outlet velocity vectors of V and W
each particle position are visible (Figure 9b). The inlet and outlet velocity vectors of V and W
components can be observed in Figure 10a,b.
components can be observed in Figure 10a,b.
According to the CFD results, the inlet and outlet triangle of velocities in the PAT is also obtained
giving information about the momentum and the rotation effect of the fluid at the impeller entrance
and exit.
 Fluids 2017, 3, 3 14 of 21

According to the CFD results, the inlet and outlet triangle of velocities in the PAT is also
Fluids obtained
2018, 3, 3 giving information about the momentum and the rotation effect of the fluid at the impeller
14 of 21
entrance and exit.

(a)

(b)
Figure 9. Velocity components: (a) Classical turbines inlet triangle of velocities for different Ns (m,
Figure 9. Velocity components: (a) Classical turbines inlet triangle of velocities for different Ns (m, kW)
kW) adapted from [26]; (b) radial PAT flow impeller for Ns = 51 rpm (m, kW), in front and cross
adapted from [26]; (b) radial PAT flow impeller for Ns = 51 rpm (m, kW), in front and cross section
section view. Reproduced with permission from [26].
view. Reproduced with permission from [26].
4.3. Dissipative Effects
4.3. Dissipative Effects
Patterns of wall shear stress (WSS) influence the progression of the flow resistance. WSS,
defined as
Patterns ofthe
wall tangential force(WSS)
shear stress imposed on the pipe
influence or device wallofby
the progression the
the flow,
flow plays an important
resistance. WSS, defined
as therole in the dissipation
tangential effects.on
force imposed Furthermore,
the pipe orextremely
device wall highbyvalues of WSS
the flow, are associated
plays an important withrole
fluidin the
structure interface interaction [21] responsible for the resistance to the flow progression and for wear
dissipation effects. Furthermore, extremely high values of WSS are associated with fluid structure
of material inducing friction, separation zones, pressure variation associated to the velocity
interface interaction [21] responsible for the resistance to the flow progression and for wear of material
exchange, visible in the path lines, towards to initial erosion process. In a fully developed steady
inducing
uniformfriction,
regime, separation zones,ofpressure
the equilibrium forces showsvariation
that theassociated
average wallto shear
the velocity
stress is exchange, visible
related to the
in thefriction
path lines, towards to initial erosion process. In a fully developed steady
slope. When the accelerations are too great for the quasi-steady hypothesis to be used, the uniform regime,
the equilibrium
instantaneousofvalue forces shows
of the wall that
shearthe average
stress, wall shear
τw, requires a morestress
preciseisevaluation,
related toconsidering
the friction theslope.
Wheneffects
the accelerations
of the variation areintoo
timegreat
of thefor the quasi-steady
velocity fields. Hencehypothesis
the analysesto be used,
used the 3Dthe instantaneous
velocities and
valuethe
of Reynolds
the wall stress
shearvalues
stress,byτw, using the CFD
requires modelprecise
a more coupledevaluation,
k-ε turbulence model.
considering the effects of the
variation The turbulence
in time of the model
velocity used in thisHence
fields. study isthetheanalyses
robust and wellthe
used known realizable k-e
3D velocities andmodel. The
the Reynolds
term realizable means that the model satisfies
stress values by using the CFD model coupled k-ε turbulence model. certain mathematical constraints of the Reynolds
stresses, consistent with the physics of turbulent flows. In comparison to the standard k-e model, the
The turbulence model used in this study is the robust and well known realizable k-e model.
realizable k-e contains an alternative formulation for the turbulent kinetic energy, and a modified
The term realizable means that the model satisfies certain mathematical constraints of the Reynolds
transport equation for the dissipation rate, that has been derived from an exact equation for the
stresses, consistent
transport with the physics
of the mean-square of turbulent
vorticity fluctuation.flows. In comparison
In the current to the standard
study, a rotor–stator k-e model,
simulation is
the realizable k-e contains an alternative formulation for the turbulent kinetic
performed with the entire turbine assembly. The flow prediction in the impeller and draft tube energy, and a modified
is
transport equation for the dissipation rate, that has been derived from an exact equation for the
transport of the mean-square vorticity fluctuation. In the current study, a rotor–stator simulation
is performed with the entire turbine assembly. The flow prediction in the impeller and draft tube
is sensitive to the boundary conditions applied and the turbulence model selected. Swirl plays a
vital role in the transformation of kinetic energy into pressure energy essentially in the draft tube,
preventing the flow separation at the cone wall. The CFD results fit well with the UDV measurements
 Fluids 2017, 3, 3 15 of 21
Fluids 2018, 3, 3 15 of 21
sensitive to the boundary conditions applied and the turbulence model selected. Swirl plays a vital
role in the transformation of kinetic energy into pressure energy essentially in the draft tube,
(see Figures 6 and
preventing the7)flow
alongseparation
the PAT system
at thefor the all
cone turbulence
wall. The CFDused model.
results fit In thewith
well present
therotor–stator
UDV
simulation, the entire turbine flow domain was used.
measurements (see Figures 6 and 7) along the PAT system for the all turbulence used model. In the
present rotor–stator simulation, the entire turbine flow domain was used.

(a) (b)

(c)
Figure 10. Velocity triangles obtained by CFD simulation for the best efficiency point (BEP): (a)
Figure 10. Velocity triangles obtained by CFD simulation for the best efficiency point (BEP): (a) Absolute
Absolute fluid path; (b) relative fluid path; (c) triangle of velocities at inlet and outlet of the impeller.
fluid path; (b) relative fluid path; (c) triangle of velocities at inlet and outlet of the impeller.
The small wakes due to the counter rotating below the runner due to the blade-hub clearances,
or the
The smallflow jet coming
wakes dueouttoofthe
thecounter
blade-hub with high
rotating belowvelocity, are captured
the runner due tointhethe blade-hub
CFD simulations.
clearances,
However, flows with a strong streamline curvature that happens in curves,
or the flow jet coming out of the blade-hub with high velocity, are captured in the CFD simulations. in the interaction rotor–
stator, and
However, flows in with
the draft tube need
a strong special concerns
streamline curvature since the happens
that model caninpresent
curves, somein inability to
the interaction
characterize all the real effects. Figure 11 presents the swirling strength of the vortex core region at
rotor–stator, and in the draft tube need special concerns since the model can present some inability to
the impeller exit in the draft tube and in all flow direction changes.
characterize all the real effects. Figure 11 presents the swirling strength of the vortex core region at the
In order to analyze the effect of the jets, the velocity and turbulent quantities were observed
impeller
fromexitthe in the draft tube
simulations at theand in all flow tube
impeller-draft direction changes.
interface. A separation tendency appears on the
In order to analyze
rotor–stator interface theandeffect of the jets, the
on impeller-draft velocity
tube andThe
transition. turbulent quantities
circumferential were observed
distribution of the from
the simulations
energy affects at the
theflow
impeller-draft
near the runnertube increasing
interface. the
A separation
turbulence tendency appears
and associated on the effects.
dissipative rotor–stator
interface and on
The radial impeller-draft
wall-shear stress istube transition.
presented. Thewall-shear
Higher circumferential distribution
stress is of theinenergy
obtained locally the outaffects
surfaces of the impeller. The flow is completely attached until
the flow near the runner increasing the turbulence and associated dissipative effects. The the end of the runner. Theradial
high-velocity
wall-shear stress is jetspresented.
in pipe curvesHigher andwall-shear
the draft tube
stresscreate a lowerlocally
is obtained pressure region.
in the out The lowerof the
surfaces
impeller. The flow is completely attached until the end of the runner. The high-velocity jets to
pressure allows bending the streamlines toward the swirl, despite the centrifugal force imposed in pipe
the fluid in the impeller and at the transition to the draft tube. The CFD model accounts for the effect
curves and the draft tube create a lower pressure region. The lower pressure allows bending the
streamlines toward the swirl, despite the centrifugal force imposed to the fluid in the impeller and at
the transition to the draft tube. The CFD model accounts for the effect of the rapid strain, the strong
streamline curvature and the swirling motion (vorticity), which are visible in the simulation results.
 Fluids 2017, 3, 3 16 of 21

Fluids of the3,rapid
2018, 3 strain, the strong streamline curvature and the swirling motion (vorticity), which are16 of 21
visible in the simulation results.

(a)

(b)
Figure 11. Values of velocity, shear stress and turbulence in several parts of the PAT: (a) N = 810 rpm
Figure 11. Values of velocity, shear stress and turbulence in several parts of the PAT: (a) N = 810 rpm
and Q = 2.9 L/s; (b) N = 1500 rpm and Q = 4.6 L/s.
and Q = 2.9 L/s; (b) N = 1500 rpm and Q = 4.6 L/s.
5. Characteristic Curves
5. Characteristic Curves
The specific speed plays an important role for selecting the type of turbine. It is a quantity
derived fromspeed
The specific dimensional analysis,
plays an defined
important roleasforthe speed ofthe
selecting a turbine
type of which is identical
turbine. in shape,
It is a quantity derived
from dimensional analysis, defined as the speed of a turbine which is identical in shape, geometrical
size, blade angles, with the actual turbine, but of such a size that it will develop unit horse power when
working under a unit head [11–14,22,27,28]. The main characteristic parameters of a turbo-machine
can be converted into non-dimensional parameters, such as:
Fluids Fluids 3, 3 3, 3
2018,2017, 17 of 2117 of 21

geometrical size, blade angles, with the actual turbine, but of such a size that it will develop unit
horse power when working under a unit head [11–14,22,27,28]. The main characteristic parameters
√
of a turbo-machine can be converted into non-dimensional parameters, such as:
N P
Specific speed—Ns Ns,P =
H 5/4
(m, kW)
√
Specific speed—Ns , = ⁄ (m, kW)√ (10)
Ns,Q = NH 3/4Q (m, m3 /s) (10)
, = Q⁄ (m, m3/s)
Discharge number ϕ = (11)
Discharge numberND3 = (11)
gH
Head number Ψ = 2 2 (12)
Head number = N D (12)

where where
H (m),H (m), N (rpm),
N (rpm), Q (m
Q (m 3/s), P (kW) and D (m) are the head, rotational speed, flow rate, power,
3 /s), P (kW) and D (m) are the head, rotational speed, flow rate, power,
and impeller diameter respectively.
and impeller diameter respectively.
Therefore, is a characteristic parameter [19] concerning the turbine type, the runner shape
Therefore, Ns is a characteristic parameter [19] concerning the turbine type, the runner shape and
and its associated dynamic behavior. According to [17,18] and [23,29] the performance
its associated dynamic behavior. According to [17,18] and [23,29] the performance characteristics of
characteristics of 9 PATs with different specific speeds (Ns) were considered (Figure 12a), namely for
9 PATs with different
different specific
specific speeds speeds
(i.e., 21 rpm,(Ns)
24.5were
rpm,considered
35 rpm, 36.4(Figure 12a),
rpm, 39.4 rpm,namely for46.4
45.2 rpm, different specific
rpm, 61.3
speeds (i.e., 21 rpm,
rpm and 79.1 rpm). 24.5 rpm, 35 rpm, 36.4 rpm, 39.4 rpm, 45.2 rpm, 46.4 rpm, 61.3 rpm and 79.1 rpm).

(a)

(b)

Figure 12. Cont.

Fluids 2018, 3, 3 18 of 21
Fluids 2017, 3, 3 18 of 21

3.50

70
3.00

2.50 60

2.00
50

η (%)
H/Ho
1.50
40

1.00

30
0.50

0.00 20
0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50

Q/Qo
H/Ho (CFD) H/Ho (Ns cruve) H/ho (EXP)
Eff (CFD) Eff (Ns curve) Eff (EXP)

(c)
Figure 12. Performance curves and selection of a PAT. (a) Characteristic curves for different Ns [19];
Figure(b)12. Performance
comparison curves
between andexperimental
CFD, selection offora PAT. (a) Characteristic
the tested PAT and Nscurves
curve: for different
Variation Ns [19];
of the
efficiency and
(b) comparison headCFD,
between number with the for
experimental discharge
the testednumber;
PAT and(c) Ns
dimensionless headofand
curve: Variation flow
the efficiency
characteristic
and head numbercurves. Reproduced
with the dischargewith permission
number; from [19].
(c) dimensionless head and flow characteristic curves.
Reproduced with permission from [19].
Figure 12a is representative of different PATs tested in laboratory [26,30–32]. Knowing the
specific speed of the analyzed PAT at IST, which is 21 rpm (n, m3/s) or 51 rpm (m, kW), with a
Figure
nominal12a is representative
rotation speed of 1020ofrpm,
different PATs tested
a comparison in laboratory
is made [26,30–32].
with the available Knowing
Ns curves the specific
of available
speedPATs.
of the analyzed PAT at IST, which is 21 rpm (n, m 3 /s) or 51 rpm (m, kW), with a nominal
The maximum efficiency obtained for the PAT at IST and by the CFD model was below than
rotation
70%,speed
due toof the1020 rpm, aturbulence
additional comparison is made
between rotorwith the available
and stator Ns curves
edges inducing of available
small leaks, and duePATs.
to the microefficiency
The maximum scale of the small PAT
obtained turnPAT
for the this at
effects
IST andpredominant,
by the CFD reducing
model thewas
efficiency
below of the 70%,
than
machine. Figure 12b,c present comparisons between CFD, experimental for the
due to the additional turbulence between rotor and stator edges inducing small leaks, and due to thetested PAT and with
microthe Ns available
scale of the smallcurvePAT
in (a).
turn this effects predominant, reducing the efficiency of the machine.
Figure 12b,c present comparisons between CFD, experimental for the tested PAT and with the Ns
6. Conclusions
available curve in (a).
Experimental and CFD simulations were performed for detailed flow analysis to study the
velocity
6. Conclusions variation in a PAT with the specific speed of 21 rpm (m, m3/s). For the velocity profiles
analysis, a UDV 3000 series technique was performed. Despite the difficulty of the probe access and
Experimental
the complexityand CFDinstrumentation
of the simulations were performed
adjustment, thefor detailed
flow field inflowtheanalysis
region of to study
the PAT thewas
velocity
variation in a PAT with 3
characterized. The the specific
obtained speed of 21data
experimental rpmand(m, CFD
m /s). For the
results velocity
allowed profiles analysis,
to understand the flowa UDV
3000 series
behaviortechnique
along the was performed.
PAT, inside theDespite
volute,the difficulty
crossing the of the probe
impeller, access
in the draftand theand,
tube, complexity
most of
interesting, wasadjustment,
the instrumentation the analysisthe offlow
the field
velocity vectors
in the regionalongof thethe
PAT blades for different operating
was characterized. The obtained
experimental data and CFD results allowed to understand the flow behaviorobtained
conditions associated to each flow discharge, head and rotational speed. The along the data provides
PAT, inside the
volute,a good base the
crossing for better characterizations
impeller, the best
in the draft tube, flow
and, conditions,
most the main
interesting, wasloss
themechanisms
analysis ofthrough
the velocity
the velocity fields between the rotating and static parts of a runner. The analysis to the velocity fields
vectors along the blades for different operating conditions associated to each flow discharge, head
of the flow in the radial PAT impeller allowed the characterization of the triangle of velocity
and rotational speed. The obtained data provides a good base for better characterizations the best
components (i.e., V, W, C) at each particle position at inlet and outlet of the impeller. From the CFD
flow conditions, the main loss mechanisms through the velocity fields between the rotating and static
simulation and for different rotational speeds, the velocity profiles were obtained in different
parts measured
of a runner. Theofanalysis
points the PAT.to Thethevelocity
velocity fieldsfrom
profiles of theCFD flow in the radial
simulations showPAT impeller
a trend allowed the
of the velocity
characterization of the triangle of velocity components (i.e., V, W, C) at each particle
variation similar to the one observed with the UDV in the experimental facility. Since the variation position at inlet
of and
outlettheof the impeller.
physical From the
parameters CFD
that simulationthe
characterizes andhydrodynamic
for different rotational
of the flow speeds,
insidetheanyvelocity
hydraulicprofiles
were turbo-machine, adequatemeasured
obtained in different shape of the impeller
points wasPAT.
of the then The
imposed.
velocity profiles from CFD simulations
show a trend CFDof analyses allowed
the velocity the estimation
variation nottoonly
similar the the
onetriangle
observed of velocities
with theatUDV inlet and
in theoutlet of the
experimental
facility. Since the variation of the physical parameters that characterizes the hydrodynamic oftube,
PAT impeller, but also the effect of the blades’ orientation, the rotational flow at the draft the flow
inside any hydraulic turbo-machine, adequate shape of the impeller was then imposed.
CFD analyses allowed the estimation not only the triangle of velocities at inlet and outlet of
the PAT impeller, but also the effect of the blades’ orientation, the rotational flow at the draft tube,
Fluids 2018, 3, 3 19 of 21

which induces additional efficiency losses in the PAT system, when compared with classical turbines
which are geometrically more adequate to this flow direction.
Analyzing the differences in individual quantities impacting efficiency leads to the conclusion
that the largest disparity is due to differences in computed and measured torque (or shaft power) on
the PAT shaft. All the computed variants yielded values of torque significantly higher than the value
actually measured. The difference ranges from as little as 3% to a whopping 40%. As a result of the
torque overestimation, the numerical simulation also foresaw the turbine efficiency as a function of the
configuration of the velocity triangles. Aside from mechanical losses, other factors that can cause the
torque discrepancy are small geometry differences between CAD and the real PAT, most significantly
through the scale effect of frictions and bearings, the surface roughness of the test model (since CFD
analyses were performed with hydraulically smooth surfaces), eventually entrapped air micro bubbles
in the pipe loop and inside the PAT, carrying additional turbulence and swirl effects. The scale
effects namely friction, bearings and flow impact, flow separation and vortex may cause the torque
calculation inaccuracy in the numerical simulation, which lead to the numerical efficiency greater than
the experimental data. The majority of the momentum losses (~12%) is induced by bends, the inlet
and outlet of the impeller, along the blades, and the change of flow direction due to the viscous drag
on these surfaces, inducing rotational flow characteristics with additional instability. A rotor–stator
simulation was performed showing swirls due to the transformation of the kinetic energy into pressure.
The swirling strength of the vortex core region at the transition from the impeller to the draft tube and
in all flow direction changes induce dissipative effects occurred in this PAT system. The circumferential
distribution of the energy affects the flow near the impeller, increasing the turbulence and consequently
the associate head loss effects. An effort should be done to minimize the losses due to the shear stress,
turbulence, and leaks with specific jets and rotating flow and vortex formation in PATs. Nevertheless,
the overall predicted torque is relatively accurate for the turbulent model tested with an increase in the
accuracy upon successive grid refinements.
Overall, the selection of a suitable turbulence model is of paramount importance regarding the
accuracy of expected results. Under such circumstances, the choice must be guided by computer
experience, parameters of the generated mesh, and the boundary (initial) conditions being set.

Acknowledgments: The authors wish to thank to the project Reducing Energy Dependency in Atlantic Area
Water Networks (REDAWN) EAPA_198/2016 from INTERREG Atlantic Area Programme 2014–2020 and CERIS
(CEHIDRO-IST) and the Hydraulic Laboratory, for the support in the conceptual developments and the
experiments on PATs. The authors also thank the program to support the academic career of the faculty of
the Universitat Politècnica de València 2016/2017 in the project “Maximization of the global efficiency in PATs in
laboratory facility” of the second author.
Author Contributions: The author Helena M. Ramos contributed to the idea, the revision of the document,
supervising the whole research, and in particular in analyses of the head drop in each system component.
Mariana Simão conceived and developed the CFD modeling and the analysis of the results between CFD and
experiments. Modesto Pérez-Sánchez, Petra López-Jiménez and Armando Carravetta supervised the whole
research and revised the final document.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Yang, S.; Derakhshan, S.; Kong, F.-Y. Theoretical, numerical and experimental prediction of pump as turbine
performance. Renew. Energy 2012, 48, 507–513. [CrossRef]
2. Coelho, B.; Andrade-Campos, A. Efficiency achievement in water supply systems—A review. Renew. Sustain.
Energy Rev. 2014, 30, 59–84. [CrossRef]
3. Ramos, H.M.; Mello, M.; De, P.K. Clean power in water supply systems as a sustainable solution:
From planning to practical implementation. Water Sci. Technol. Water Supply 2010, 10, 39–49. [CrossRef]
4. Kaunda, C.S.; Kimambo, C.Z.; Nielsen, T.K. A technical discussion on microhydropower technology and its
turbines. Renew. Sustain. Energy Rev. 2014, 35, 445–459. [CrossRef]
Fluids 2018, 3, 3 20 of 21

5. Pérez García, J.; Cortés Marco, A.; Nevado Santos, S. Use of centrifugal pumps operating as turbines for
energy recovery in water distribution networks. Two case study. Adv. Mater. Res. Trans. Tech. Publ. 2010,
107, 87–92. [CrossRef]
6. Jain, S.V.; Patel, R.N. Investigations on pump running in turbine mode: A review of the state-of-the-art.
Renew. Sustain. Energy Rev. 2014, 30, 841–868. [CrossRef]
7. Pérez-Sánchez, M.; Sánchez-Romero, F.J.; Ramos, H.M.; López-Jiménez, P.A. Energy recovery in existing
water networks: Towards greater sustainability. Water 2017, 9, 97. [CrossRef]
8. Mesquita, A.L.A.; Ciocan, G.D. Experimental analysis of the flow between stay and guide vanes of a
pump-turbine in pumping mode. J. Braz. Soc. Mech. Sci. 1999, 21, 580–588. [CrossRef]
9. Simão, M.; Ramos, H.M. Hydrodynamic and performance of low power turbines: Conception, modelling
and experimental tests. Int. J. Energy Environ. 2010, 1, 431–444.
10. Ramos, H.M.; Simão, M.; Borga, A. Experiments and CFD analyses for a new reaction microhydro propeller
with five blades. J. Energy Eng. 2013, 139, 109–117. [CrossRef]
11. Wang, B.; Hellman, D.-H. Turbulent 3D flows near the impeller of a mixed-flow pump. Hydraul. Mach. Cavitation
1996, 176, 1044–1052.
12. Ramos, H.; Borga, A. Pumps as turbines: An unconventional solution to energy production. Urban Water
1999, 1, 261–263. [CrossRef]
13. Nataraj, M.; Singh, R.R. Analyzing pump impeller for performance evaluation using RSM and CFD.
Desalin. Water Treat. 2013. [CrossRef]
14. Castro, L.; Urquiza, G.; Adamkowski, A.; Reggio, M. Experimental and numerical simulations predictions
comparison of power and efficiency in hydraulic turbine. Model. Simul. Eng. 2011, 2011, 146054. [CrossRef]
15. MG Corporation. Flow Simulation—Technical Reference, USA: Solid Works 2011; EUA, Ed.; Mentor Graphics
Corporation: Wilsonville, OR, USA, 2011.
16. Abilgaziyev, A.; Nogerbek, N.; Rojas-Solórzano, L. Design optimization of an oil-air catch can separation
system. J. Transp. Technol. 2015, 5, 247–262. [CrossRef]
17. Derakhshan, S.; Nourbakhsh, A. Experimental study of characteristic curves of centrifugal pumps working
as turbines in different specific speeds. Exp. Therm. Fluid Sci. 2008, 32, 800–807. [CrossRef]
18. Pérez-Sánchez, M.; Simão, M.; López-Jiménez, P.A.; Ramos, H.M. CFD analyses and experiments in a PAT
modeling: Pressure variation and system efficiency. Fluids 2017, 2, 51. [CrossRef]
19. Carravetta, A.; Derakhshan Houreh, S.; Ramos, H.M. Pumps as Turbines, Fundamentals and Applications;
under edition; Springer International Publishing: Basel, Switzerland, 2017.
20. Ramos, H.M.; Simão, M.; Borga, A. CFD and experimental study in the optimization of an energy converter
for low heads. Energy Sci. Technol. 2012, 4, 69–84.
21. Simao, M.; Mora-Rodriguez, J.; Ramos, H.M. Computational dynamic models and experiments in the
fluid-structure interaction of pipe systems. Can. J. Civ. Eng. 2016, 43, 60–72. [CrossRef]
22. Ramos, H.; Almeida, A.B. Parametric analysis of water-hammer effects in small hydro schemes.
J. Hydraul. Eng. 2002, 128, 689–696. [CrossRef]
23. Singh, P. Optimization of Internal Hydraulics and of System Design for Pumps as Turbines with Field
Implementation and Evaluation. Ph.D. Thesis, University of Karlsruhe, Karlsruhe, Germany, 2005.
24. Rawal, S.; Kshirsagar, J.T. Numerical simulation on a pump operating in a turbine mode. In Proceedings of
the 23rd International Pump Users Symposium, Houston, TX, USA, 5–8 March 2007.
25. Nautiyal, H.; Kumar, V. Reverse running pumps analytical, experimental and computational study: A review.
Renew. Sustain. Energy Rev. 2010, 14, 2059–2067. [CrossRef]
26. Ramos, H. (Ed.) Guidelines for Design of Small Hydropower Plants; WREAN (Western Regional Energy Agency
and Network) and DED (Department of Economic Development—Energy Division): Belfast, North Ireland,
2000; ISBN 972-96346-4-5.
27. Eisele, K.; Zhang, Z.; Casey, M.V.; Gülich, J.; Schachenmann, A. Flow analysis in a pump diffuser-part1:
LDA and PTV measurements of the unsteady flow. ASME J. Fluids Eng. 1997, 119, 968–977. [CrossRef]
28. Ramos, H.; Almeida, A.B. Dynamic orifice model on waterhammer analysis of high or medium heads of
small hydropower schemes. J. Hydraul. Res. 2001, 39, 429–436. [CrossRef]
29. Singh, P.; Kumar, H.H.; Praveen, R.; Shankarappa, C. Laboratory Testing of Centrifugal Pumps as Turbines;
BEng Final Year Project Report; Department of Mechanical Engineering, The National Institute of
Engineering: Mysore, India, 1998.
Fluids 2018, 3, 3 21 of 21

30. Su, X.; Huang, S.; Zhang, X.; Yang, S. Numerical research on unsteady flow rate characteristics of pump as
turbine. Renew. Energy 2016, 94, 488–495. [CrossRef]
31. Cao, S.; Goulas, A.; Wu, Y.; Tsukamoto, H.; Peng, G. Three-Dimensional Turbulent Flow in a Centrifugal
Pump Impeller Under Design and Off-Design Operating Conditions. In Proceedings of the ASME Fluids
Engineering Division, FEDSM-6872, San Francisco, CA, USA, 18–23 July 1999.
32. Yang, S.; Kong, F.; Fu, J. Numerical research on effects of splitter blades to the influence of pump as turbine.
Int. J. Rotat. Mach. 2012, 2012, 123093.

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).