Energy 93 (2015) 2116e2125

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

A novel twin-rotor radial-inow air turbine for oscillating-watercolumn wave energy converters

nio F.O. Falca
~o a, *, Lus M.C. Gato a, Joa
~o C.C. Henriques a, Joa
~o E. Borges a,
Anto
Bruno Pereiras b, Francisco Castro c
LAETA, IDMEC, Instituto Superior T

ecnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001 Lisboa, Portugal

n, Spain
Energy Department, University of Oviedo, Campus de Vieques, 33271 Gijo
c
Department of Energetic and Fluid Mechanics Engineering, University of Valladolid, Paseo del Cauce 59, E-47011 Valladolid, Spain
a

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 17 August 2015
Received in revised form
13 October 2015
Accepted 14 October 2015
Available online xxx

A novel air turbine for bidirectional ows in oscillating-water-column wave energy converters is presented and its performance is analyzed. The turbine is based on a pair of conventional radial-inow
rotors mounted on the same shaft, complemented by the corresponding guide vane rows, by a
curved-duct manifold arranged circumferentially in a period manner and by a two-position cylindrical
valve. Numerical values of the performance of the whole machine were obtained from published
experimental data of the ow through a conventional radial-inow gas turbine, together with CFD
(computational uid dynamics) results for aerodynamic losses in the curved duct manifold. Four
different geometries, combined with ve different sizes, of the curved-duct manifold were numerically
simulated. Windage losses, that occur at the inactive rotor and are inherent to the machine conception,
were found to be a major loss. A peak value of about 86% was obtained for the overall efciency of the
machine. Comparisons are presented between the new turbine and the biradial turbine (sliding guidevanes version), the latter being possibly the most efcient self-rectifying turbine model-tested so far.
The new turbine was found to be more efcient, both in peak instantaneous efciency and in maximum
average efciency in random waves, by a margin of about 8%.
2015 Elsevier Ltd. All rights reserved.

Keywords:
Air turbine
Wave energy
Oscillating water column
Numerical simulation
Aerodynamics

1. Introduction
The OWC (oscillating water column) device with air turbine is
arguably the simplest type of wave energy converter and, possibly
more than any other, has been object of extensive development
effort. It has a wide range of applications: isolated, integrated into
breakwaters, and in a variety of oating congurations. The success
of the OWC depends largely on the performance and reliability of
the air turbine. The use of self-rectifying turbines has the advantage
of not requiring a rectifying valve system. Several types of selfrectifying turbines have been proposed and developed, and in
some cases equipped prototypes tested under real sea conditions.
Most self-rectifying air turbines for wave energy conversion

* Corresponding author. Tel.: 351 218 417 927.

~o), luis.gato@
E-mail addresses: antonio.falcao@tecnico.ulisboa.pt (A.F.O. Falca
tecnico.ulisboa.pt
(L.M.C.
Gato),
joaochenriques@tecnico.ulisboa.pt
(J.C.C.
Henriques),
joao.teixeira.borges@tecnico.ulisboa.pt
(J.E.
Borges),
pereirasbruno@uniovi.es (B. Pereiras), castro@eis.uva.es (F. Castro).
http://dx.doi.org/10.1016/j.energy.2015.10.046
0360-5442/ 2015 Elsevier Ltd. All rights reserved.

proposed and tested so far are axial-ow machines of two basic

types: the Wells turbine and the impulse turbine. An extensive and
detailed review of Wells turbines was published in Ref. [1]. For the
impulse turbine see Ref. [2]. More recent reviews of self-rectifying
air turbines for OWCs can be found in Refs. [3e7].
Several variants of the Wells turbine have been developed:
without or with guide vanes, counter-rotating, bi-plane, multistage (see Ref. [5]). Peak efciencies up to about 75% were found
to be attainable in model testing of Wells turbines with sufciently
large models and Reynolds numbers. Regardless of the type of
Wells turbine, the curve of efciency versus pressure head is
characterized by a (more or less) sharp fall that occurs when the
angle of incidence at the rotor blades exceeds the stall-free limit.
The aerodynamic losses due to rotor blade stalling in practice
severely limit the range of operation of the Wells turbine and are its
main drawback. Because of its own conception, the rotational speed
(or more precisely the rotor blade tip speed) of the Wells turbine is
much higher as compared with self-rectifying impulse turbines in
identical application. This may lead to limitations, due to the need

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

to avoid shock waves or excessive centrifugal stresses, which is a

constraint in the more energetic wave climates, or in OWCs designs
characterized by large pressure heads like the spar-buoy OWC; in
such cases, mechanically more complex multi-stage Wells turbines
may be required as alternatives to impulse turbines [8].
The efciency of the self-rectifying axial-ow impulse turbine
with xed guide vanes is severely affected by the losses at the entry
to the downstream row of guide vanes. Peak efciencies measured
in model testing do not exceed about 50% (as compared with about
75% for the most efcient Wells turbines). On the other hand, the
efciency curves do not exhibit the sharp drop typical of most Wells
turbines. The performance of the axial-ow impulse turbine can
improve by about 10e15% if pivoting guide vanes are used instead
of xed ones. However, the mechanical complexity of this solution
has deterred its use. A way to reduce the aerodynamic losses at the
second row of guide vanes is to offset the guide vanes radially and
axially from the rotor in order to reduce the kinetic energy (and so
the losses due to stalling). This has been done in the HydroAir
turbine that equipped the Oceanlinx Mk3 multi-chamber oating
OWC prototype [7].
The so-called biradial turbine [9,10], Fig. 1, is an impulse turbine
that is symmetrical with respect to a plane perpendicular to its axis
of rotation. The ow into, and out of, the rotor is radial. The rotor is
surrounded by a pair of radial-ow guide-vane rows, each row
being connected to the corresponding rotor inlet/outlet by a duct
whose walls are at discs. In one of the versions of the biradial

Fig. 1. Biradial turbine: (a) cross sections; (b) perspective view. GV guide vanes;
RB rotor blades.

2117

turbine, the guide vane rows may be removed from, or inserted

into, the ow space by axially displacing the whole guide vane set,
so that the downstream guide vanes are prevented from
obstructing the ow coming out of the rotor. In this version, the
radial distance between the rotor and the guide vanes is small, see
Fig. 1. The measured peak efciency of a turbine model was about
79%, possibly the highest efciency of a self-rectifying air turbine
measured so far.
If, in a single-stage conventional turbine, with a row of guide
vanes followed by a bladed rotor, the sign of the pressure head is
changed (and the rotational speed is kept unaltered), the ow rate
(apart from changing sign) becomes substantially smaller (and the
turbine performance becomes very poor). This has led to the idea of
associating two identical conventional air turbines (turbines T1
and T2) in parallel to convert the pneumatic energy from an OWC,
such that, for a given pressure head situation, the ow sequence
guide-vanes-rotor-blades in turbine T1 is reversed with respect to
turbine T2 (see Refs. [11,12]). This topology is shown in Fig. 2, where
the twin turbines are of axial-ow type. With this arrangement, for
a given pressure head (independently of its sign), most of the ow
is admitted to one of the turbines (that is driven with good efciency) while a smaller fraction of the ow is admitted to the other
turbine (that is in reverse mode and operates at very low efciency). The two turbines can be coupled to a common electrical
generator (as in Fig. 2) or, alternately, each turbine is coupled to its
own generator. Since the turbines are not symmetrical, their rotor
blades need no longer to be symmetrical with respect to the midchord point, as appears to be the case in Fig. 2. Some positive degree of reaction may be convenient. Model testing of a
unidirectional-turbine pair in a rig capable of bi-directional oscillating air ow is reported in Ref. [12]. The turbine rotor diameter
was 165 mm and each turbine was coupled to its own generator. A
peak efciency of 0.6 was measured. The aerodynamic performance
of the twin unidirectional turbine conguration was numerically
simulated in detail in Ref. [13]. It was found that the ow rate
through the turbine in reverse mode is about one-third of the total
ow and produces a negative torque which reduces the system
efciency if the two turbines are directly connected to the same
electrical generator.
Descriptions of other types of air turbines for bidirectional ows
can be found in Refs. [5,7].
The paper proposes a new patented self-rectifying air turbine
[14]. The turbine is characterized by a twin rotor, i.e. two rows of
rotating blades axially offset from each other, mounted on the same
shaft, complemented by corresponding guide vanes, as in a conventional axial-ow or radial-ow turbine. The reciprocating air
ow between the OWC chamber and the atmosphere takes place as
unidirectional ow alternately through one or the other bladed set.
This is made possible by a double set of curved ducts arranged
circumferentially in a period manner, and by a two-position axially-

Fig. 2. Twin unidirectional impulse turbine topology [7].

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

2118

sliding cylindrical valve. Only the radial-inow rotor conguration

is analyzed here. Numerical values of the performance of the whole
machine are presented from published experimental data of the
ow through a conventional radial-inow gas turbine, combined
with CFD (computational uid dynamics) results for aerodynamic
losses in the curved-duct manifold. Four different geometries,
combined with ve different sizes, of the curved-duct manifold, are
analyzed.
The new turbine in presented in Section 2. This is followed, in
Section 3, by the description of the modelling of the ow through
the various parts of the machine. Numerical results are presented in
Section 4 for the overall performance of the new turbine, and
comparisons are made with the self-rectifying biradial air turbine.
Conclusion can be found in Section 5.
2. The new turbine
The efciency of the twin unidirectional turbine topology
shown in Fig. 2 is negatively affected by one of the twin turbines
absorbing a non-insignicant part of the total ow while
consuming some power from the generator (due to negative torque), as revealed by numerical modelling [13]. The new turbine
proposed here was conceived to avoid these shortcomings.
The turbine has two sets of rotor blades mounted on a common
shaft and axially offset from each other. Each set of rotor blades is
complemented by a set of guide vanes, as in a conventional unidirectional turbine. This whole set of rotor blades and guide vanes
may be regarded as forming two conventional single-stage turbines, T1 and T2, that may be of radial-ow type (Fig. 3) or axialow type (Fig. 4).
The turbine operates between spaces A and B at pressures pA
and pB respectively. These spaces are the OWC air chamber and the
atmosphere. When pA > pB, the air should ow only through bladeset T1. Conversely, when pA < pB, the air should ow only through
blade-set T2. This is made possible by a double set of curved ducts
arranged circumferentially and alternately open to space A and
space B, as represented in Figs. 3 and 4, and, in perspective, in Fig. 5.
Each rotor is connected to the corresponding set of curved ducts by
a bladeless space bounded by curved and plane walls of revolution,
as shown in Figs. 3 and 4. The radial extent of this bladeless space
allows it to act as a diffuser, recovering part of the kinetic energy at
rotor exit before ow entrance into the curved-duct manifold.
The conguration of the new turbine makes it possible to install
a two-position axially-sliding cylindrical valve that is operated to
prevent air from owing in the reverse direction (Fig. 6). This is
made easier by the relatively small size and stroke of the valve. The
valve actuator may be pneumatic, electrical or of other type.
If a three-position valve is used, as represented in Fig. 7, then air
can be prevented from owing between spaces A and B. If the valve
actuator is fast enough (opening and closing times not exceeding a
few tens of a second), then this provides a way of phase-controlling

Fig. 3. New turbine with radial-ow rotor conguration.

Fig. 4. New turbine with axial-ow rotor conguration.

the OWC plant by latching. Latching is known to be an effective way

of substantially increasing the amount of energy absorbed from the
waves by oscillating body devices. Only recently, with the emergence of new turbines like the biradial turbine, has phase control by
latching started to be seriously considered as applicable to OWC
converters [15,16]. This could be enhanced by the new turbine
proposed here.
3. Modelling
The efciency of the twin-rotor machine was investigated. Only
the radial-ow (not the axial-ow) conguration (represented in
Fig. 3) is considered here. A given geometry was adopted for the
turbine rotor. Four geometries of the curved-duct manifold were
analyzed. The radius ratio R3/R1 (Fig. 8) was allowed to vary in order
to investigate the effect of varying manifold-to-rotor size ratio upon
the performance of the machine. Naturally, the geometry of the
connecting axisymmetric duct has to be adequately adapted. It is
assumed that pA in space A is larger than the pressure pB in space B,
i.e. the pressure head available to the machine is pApB > 0.
Published experimental data for a conventional radial-inow
turbine were used. The tests had been carried out at low Mach
numbers, as under conditions representative of wave energy applications. The turbine geometry is typical of a highly efcient
radial-inow gas turbine.
CFD (computational uid dynamics) was employed to model the
ow and losses through the curved-duct manifold.
3.1. Radial-inow turbine
The turbine geometry, the test rig and the experimental techniques and results are described in detail in Refs. [17e19], where
the rotor is designated as the conventional impeller. The design
values of the specic speed Ns and specic diameter Ds (as usually
dened in turbomachinery literature, see e.g. Ref. [20]) were
Ns 0.6 and Ds 3.3. This gives a point in the well-known Balje
diagram for 90 radial-inow turbines [20,21] close to the highest
efciency zone. The rig was of blow-down type. The air was sucked
through a bellmouth into the turbine directly from the atmosphere.
The ow rate was obtained from the static pressure relative to atmosphere measured at the bellmouth throat. The work done by the
turbine was absorbed by an electrical motor-generator whose stator frame was cradle-mounted, allowing the measurement of the
torque with the help of a torque arm. The turbine rotor was
mounted directly onto the generator shaft.
As in most radial-inow gas turbines, the rotor was of open type
(no shroud), Fig. 9. Its main dimensions are shown in Fig. 10. The
radius at inlet was R1 155 mm and the inner and outer radii at
rotor outlet were R2i 33 mm and R2o 110 mm. The rotor had 17
blades with 90 angle at inlet. Close to the rotor entrance there was
an annular row of 31 streamlined guide vanes of 64 mm chord. The

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

2119

Fig. 5. Perspective representation of the curved-duct manifold, with a) trapezoidal, b) circular and c) elliptical exit sections.

Fig. 6. Two-position axially-sliding valve.

design ow angle at rotor entrance was 16 with respect to the

circumferential direction.
Fig. 11 gives a plot of the dimensionless power output P versus
the dimensionless ow rate F, which are dened as is usual (see
e.g. Ref. [20])

Pt
rU3 D5

Q
:
UD3

Fig. 8. Rotor, curved-duct manifold and axisymmetric connecting duct.

(1)

Here Pt is the turbine power output, Q is the volume ow rate, U is

the rotational speed (in radians per unit time), D 2R1 is the rotor
outer diameter and r is the air density. Most of the measurements
were performed at rotational speeds about 3000 rpm. The experimental data are well approximated by the polynomial

PF 2:0356 F3 6:1343 F2  0:0606 F  0:0006;

(2)

whose curve is plotted in Fig. 11 together with the experimental

points.
So far, we have only considered the energy conversion in the air
ow process that takes place between the machine entrance and
the exit from the curved-duct manifold, with pressure head
pApB > 0. It should not be forgotten that, during that time,
windage losses take place at the other rotor mostly due to aerodynamic drag, even if the ow rate is zero as it should be if the

Fig. 9. The turbine rotor.

sliding valve is moved into the right position. Such losses take place
all the time, alternately at one and the other rotor, and should be
accounted for. Not much information on this effect is available,
since it is a situation of little practical interest for conventional

Fig. 7. Three-position axially-sliding valve.

2120

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

Fig. 12. Relative windage power loss versus dimensionless ow rate.


V2z
V2z U1 R1
1 ;

(3)


rV2q U1 R2
V2q
1 ;

(4)

p2 pA  p2 r1 U2 R2

1 :

(5)

The total-to-static efciency of the turbine is dened as (see e.g.

Ref. [20])
Fig. 10. Main dimensions of tested turbine rotor (in mm).

turbines. The experiments on the radial-inow turbine on whose

results the present analysis is based did not include torque measurements at zero ow rate. However, an estimate of such torque
can be made by extrapolation of the results plotted in Fig. 11. Such
results are well approximated by the polynomial relationship (2),
which gives P0 0.0006 for the dimensionless value of the turbine
power loss (at zero ow rate). The net power is Pnet(F) P(F)P0.
The relative windage power loss P0/P is plotted versus F in Fig. 12
for P0 0.0006. It is important at relatively small ow rates and
decreases rapidly with increasing F. It is about 5% under conditions
close to maximum efciency (see Section 4).
Detailed radial traverses were made with a three-hole probe at
the rotor exit, close to the rotor trailing edges, for the measurements of the static pressure, axial velocity and circumferential
(swirl) velocity. These measurements were performed for ve
values of the dimensionless ow rate: F1 0.03503, F2 0.04156,
F3 0.04795, F4 0.05353 and F5 0.06061. The results, in
dimensionless form, for the axial and circumferential components
 and V  , and for the pressure
of the absolute ow velocity, V2z
2q
difference p2 (pA is the pressure in the space surrounding the
turbine entrance) versus the radial coordinate r* r/R1, are shown
in Figs. 13e15. It is

Fig. 11. Dimensionless plot of turbine power output versus ow rate.

hts Pt Q 1 pA  p2 1
av ;

(6)

where

Z
pA  p2 av

R2o
R2i

pA  p2 rV2z rr dr
Z

R2o
R2i

(7)
V2z rr dr

is the radially averaged pressure difference weighted by the

elementary ow rate. The total-to-static efciency hts was
computed for the ve dimensionless values of the ow rate, F1 to
F5. The results are plotted in Fig. 16, together with a polynomial
interpolation curve. The maximum value, hts 0.906, occurs for
F 0.046.
3.2. Curved-duct manifold
What is essentially new in the turbine is the arrangement of
curved ducts shown in Figs. 3e5. It is important to investigate the

 V U1 R1 at rotor exit

Fig. 13. Dimensionless plot of axial velocity component V2z
2z
1
versus radial coordinate r* r/R1.

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

 rV U1 R2 at
Fig. 14. Dimensionless plot of circumferential velocity component V2q
2q
1
rotor exit versus radial coordinate r* r/R1.

Fig. 15. Dimensionless plot of pressure p2 pA  p2 r1 U2 R2

1 at rotor exit versus
radial coordinate r* r/R1.

losses in that part of the machine in comparison with the power

output of the turbine.
Since the curved-duct manifold is circumferentially periodic,
only one part of the total ow space needs to be analyzed. Fig. 17
represents the computation domain. There are Z ducts open to
space A and equal number open to space B. The entrance into the

Fig. 16. Total-to-static efciency hts of the radial-inow turbine versus ow rate coefcient F.

2121

domain is a segment of circular cylindrical surface of radius R3 and

width b3, with area A3 2pR3b3/Z. The following conguration was
adopted: Z 9, b3 R3/3, area A3 0:2327 R23 . Four geometries
were numerically modelled for the curved ducts, having exit sections of trapezoidal, rectangular (with rounded corners), elliptic
and circular shapes. They are represented, together with their area
A4, in Fig. 18.
CFD (computational uid dynamics) based on the numerical
integration of the RANS (Reynolds-averaged NaviereStokes)
equations was employed to simulate the ow through the curvedduct manifold. In the computations, it was assumed that the
entrance ow at radial coordinate r R3 is axisymmetric with
uniform radial velocity V3r, uniform circumferential velocity V3q and
zero axial velocity V3z 0. The angle of the entrance velocity vector
with the radial direction is g arctan(V3q/V3r). PQRS and the homologous surface (Fig. 19) are open surfaces introduced to bound
the circumferentially periodic ow space. The computations were
performed for R3 0.450 m, V3r 20 m/s and g 0 , 15 , 30 , 45 .
The ow simulation was done with FLUENT v12, which employs
the nite-volume numerical method for solving the RANS (Reynolds-averaged NaviereStokes) equations by using a segregated
solver. The mesh was structured, totally composed of 3  105
hexahedrical cells and was built in GAMBIT 2.4. Some details can be
seen in Fig. 19, that also shows the imposed boundary conditions:
velocity at inlet and pressure at outlet. The realizable k turbulence model was used with the standard wall function. The y
values were in the correct range, over 95% of the cells were in the
interval 30  y  150. The time-dependent term was approximated by a second-order implicit scheme. The pressureevelocity
coupling was performed with the aid of the SIMPLE algorithm. The
higher order MUSCL (Monotone Upwind Scheme for Conservation
Laws) was used for convective term discretization and the classical
central difference approximation for diffusion terms. The residuals
were set to 105, reached after approximately 2000 iterations.
The loss (including exit kinetic energy loss) per unit mass of air
at the curved-duct manifold is L r1(p03pB), where
p03 p3 rV32 =2 is the total pressure at section 3 and pB is the exit
pressure. We dene a loss coefcient at the curved-duct manifold
as

2L
:
2
V3r

(8)

Fig. 17. Element of curved-duct manifold geometry adopted in the computations

(trapezoidal exit section). Dimensions in mm.

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

2122

Fig. 18. The four shapes and areas of the curved-duct-manifold exit sections, at the
same scale.

Fig. 20. Loss coefcient at the curved-duct manifold versus angle of incidence g for
the four exit section geometries (g 0 means purely radial inlet ow).

duct manifold inlet is approximately invariant with axial coordinate y. We may write in dimensionless form

V3q
Y

R1 
V X;
R3 2q

(9)

1 2
1 2
1 2
1 2
p3 V3r
Y V3q
Y p2 X V2z
X
V X:
2
2
2
2X 2 2q

(10)

 V U1 R1 , V  V U1 R1 and

Here, it is X x/R1, Y y/R1, V3r
3r
3q
1
3q
1
p3 pA  p3 r1 U2 R2
.
For
given geometry and given condi1
 Y as a function of X
tions at rotor exit, Eqs. (9) and (10) yield V3r
and p3. The continuity of ow rate between stream surfaces allows
us to write

dY

 X
R1 V2z
X dX:

R3 V3r YX

(11)

A relationship between variables X and Y may be obtained by

integration

R1
R3

Fig. 19. Computation domain, mesh details and boundary conditions.

ZX
X2i

 X
V2z
X

V3r YX

dX;

(12)

where X2i R2i/R1. Setting the upper limit of integration

X X2o R2o/R1, we nd
The computed loss coefcient is plotted versus the angle of
incidence g in Fig. 20.

b3

R21
R3

3.3. Flow in the connecting duct

We analyze the ow in the duct, formed by walls of revolution,
that connects the rotor and the curved-duct manifold (Fig. 8). The
ow is assumed axisymmetric. We introduce radial coordinate r
and axial coordinate z, and consider two points on a generic stream
surface of revolution: point P2 at the exit from the rotor (r x,
z 0) and point P3 at the entrance to the curved-duct manifold
(r R3, z ay). The velocity components are V2r 0, V2z(x), V2q(x)
at point P2, and V3r(y), V3z 0, V3q(y) at point P3.
We further assume ideal uid ow (viscous losses are neglected), so that total pressure, as well as angular momentum per unit
mass, are constant over a stream surface. Fig. 15 shows that the
pressure at rotor exit is almost invariant with the radial coordinate.
It is not unreasonable to assume that the pressure p3 at the curved-

ZX2o
X2i

 X
V2z
X

V3r YX

dX:

(13)

The correct value of pressure p3 satises Eq. (13) for the adopted
width b3.

3.4. Overall performance

The overall efciency of the machine is

hAB

Pnet
;
FJAB

(14)

where P and F are dened as in Eq. (1), the subscript net means
that windage power losses were subtracted (see Section 3.1), and

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

Z
pB

R2o
R2i

2123

b
p B xV2z xx dx
R2o

R2i

(16)

V2z xx dx

In dimensionless form, we may write

2 X
2 X
V 2 X V2q
p X XV3r
c

 2z
;

J AB X  2
4
8
8
8X 2

(17)

and, for the average value,

Z
JAB

R2o =R1
R2i =R1


c
XX dX
J AB XV2z

R2o =R1

R2i =R1

Fig. 21. Curves of overall efciency hAB versus ow rate coefcient F for R3/
R1 0.8,1.0,1.2,1.4,1.6. The exit section of the curved-duct manifold is circular. The
curve of total-to-static efciency hts of the radial-inow turbine alone is also shown.
All curves are corrected for windage losses.

JAB

pA  pB
:
rU2 D2

(15)

Here, as above, pA is the pressure in space A surrounding the turbine inlet and pB is the pressure in space B at the machine exit.
In the calculations, we assume that the inlet pressure pA, the
ow rate Q and the rotational speed U are xed, and compute the
value of the exit pressure pB. Note that the dimensionless coefcient of losses in the curved-duct manifold, plotted in Fig. 20, was
computed assuming uniform values of the radial and circumferential components of the velocity at duct inlet. This is not the case:
the values of V3r and V3q are functions of the axial coordinate y, and
so is the angle g arctan(V3q/V3r). This difculty is circumvented by
taking the loss coefcient also as a function of y (and consequently
of the radial coordinate x), the same happening with the computed
value of the exit pressure pB, which we denote as b
p B x. Finally, we
compute an average value for pB by integration, taking the
elementary ow rate as a weight function

:

V2z
XX

(18)

dX

4. Numerical results
4.1. Performance curves
Numerical results were obtained for the performance of the
whole machine operating between inlet pressure pA and outlet
pressure pB. Results for the overall efciency hAB versus ow rate
coefcient F are shown in Fig. 21 for the circular exit section and in
Fig. 22 for the rectangular exit section. In both gures, curves are
shown for ve values of the ratio R3/R1 0.8,1.0,1.2,1.4, 1.6. In each
gure, the curve of the total-to-static efciency hts of the radialinow turbine alone (corrected for windage power losses) is also
shown for comparison. We recall that, when radius R3 is changed,
all dimensions of the curved-duct manifold vary proportionally, i.e.
the geometry (apart from size) is conserved.
A curve of the overall pressure head coefcient JAB versus the
ow rate coefcient F is shown in Fig. 23 for the case of the curvedduct manifold with circular exit section and radius ratio R3/R1 1.2.
Unlike self-rectifying turbines whose rotor is symmetrical with
respect to a plane perpendicular to the axis of rotation (as is the case of
the Wells turbine, axial-ow impulse turbines and the biradial turbine), in the present case the overall pressure difference is not expected to vanish at zero ow rate. This is because of the centrifugal
force effect on the uid in the rotor that persists even at zero ow rate.
4.2. Comparison with biradial turbine
The new turbine and the biradial turbine are both radial-ow
turbines for bidirectional ows that have a two-position element

Fig. 22. As in Fig. 21, for rectangular exit section of the curved-duct manifold.

Fig. 23. Dimensionless plot of overall pressure head JAB versus ow rate F for the
curved-duct manifold with circular exit section and radius ratio R3/R1 1.2.

2124

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

that is moved in translation whenever the ow changes its direction. The element is a cylindrical valve in the new turbine and the
twin set of guide vanes in the biradial turbine. It is interesting to
compare the performance of both turbines, one of the reasons
being that the peak efciency, measured in model testing, of the
biradial turbine seems to exceed that of any other self-rectifying air
turbine. Information from model testing of the biradial turbine is
available in Refs. [7,10]. In the comparison, the curved-duct manifold of the new turbine has a circular exit section and R3/R1 1.2.
Fig. 24 shows that the peak efciency of the new turbine exceeds
that of the biradial turbine by about 8%; this occurs at a lower value
of the ow rate coefcient Fhmax. In the same gure, the efciencies
are also plotted versus the ratio F/Fhmax, which provides an easier
comparison of the two turbines close to their best efciency
conditions.
It is also of interest to know the ratios of rotor diameter, Dnew/
Dbirad, and rotational speed, Unew/Ubirad, of the new turbine and the
biradial turbine, if the two turbines are to operate at peak efciency
under identical conditions of ow rate Q and pressure difference
pApB. The values are Fnew;hmax 0:0517, Jnew;hmax 0:264 for the
new turbine with circular exit section and R3/R1 1.2, and
Fbirad;hmax 0:0963, Jbirad;hmax 0:338 for the biradial turbine.
From the denitions of F and J, we easily nd Dnew/Dbirad 1.28,
Unew/Ubirad 0.88. For an identical application, the new turbine has
a larger rotor and rotates at a lower rotational speed, as compared
with the biradial turbine.
4.3. Performance in irregular waves
The turbine is to be mounted on an OWC wave energy converter
and operate under real random wave conditions. A reasonable
assumption is to assume that the oscillation in air pressure head
Dp pApB is a Gaussian process with variance s2p , whose probability density function is (see Ref. [22])

1
fp Dp p exp
2psp

!
Dp2
 2 :
2sp

(19)

The instantaneous power output of the turbine versus the

pressure head is assumed known in dimensionless form as
Pnet fP(JAB). From the denitions of JAB and P, we nd the
following relationship between the turbine net power output Pt,net
and the pressure head Dp


Pt;net rU3 D5 fP

Dp

rU2 D2

(20)

The averaged value of the turbine power output is (see Ref. [22])

Z
P t;net

fp DpPt;net DpdDp

(21)

or

rU3 D5
P t;net p
2psp

Z
exp

! 

Dp2
Dp
dDp:
f
P
2s2p
rU2 D2

(22)

This can be written in dimensionless form as

1
Pnet p
2psJ

Z Z
exp
 

!
J2
fP JdJ;
2s2J

(23)

where P is the averaged value of P (or, equivalently, the dimensionless value of P t ) and

sJ

sp
rU2 D2

(24)

is the dimensionless value of sp.

The average efciency of the turbine is dened as
hAB P t;net =P avai Pnet =Pavai , where P avai is the time-average of
the power Pavai QDp available to the turbine, and Pavai is its
dimensionless form. The values of P avai and Pavai can be obtained
by integration with respect to Dp and J, respectively, as explained
above for P t;net and Pnet . In the integrations, it is assumed that fP is
an even function, and the innite limits of integration are replaced
by nite ones, account being taken of the fast decay of the exponential function. It should be noted that extrapolated values of JAB
versus F were used in the integrations.
Results for the average efciency of the new turbine in random
waves, together with values for the biradial turbine, are shown in
Fig. 25. As in the comparison above for instantaneous values, the
curved-duct manifold of the new turbine has a circular exit section
and R3/R1 1.2. The maximum value of the average efciency of the
new turbine is 79.6% and occurs for sJ z 0.22. It exceeds the
corresponding value of the biradial turbine by about 8%.

5. Conclusions

Fig. 24. Efciency versus ow rate for the new turbine and for the biradial turbine
with sliding guide vanes. The dotted lines are results from extrapolation. Above: efciency versus ow rate coefcient F. Below: efciency versus ow rate ratio F=Fhmax ,
where subscript hmax means maximum efciency conditions.

A new air turbine for bidirectional ow applications in wave

energy conversion was presented. The turbine is based on a pair of
conventional radial-inow gas turbines known to be highly efcient turbomachines. This pair is complemented by a curved-duct
manifold and a two-position valve. The numerical values of the
performance of the whole machine are based on available experimental results for the gas turbine performance, together with CFD
results for aerodynamic losses in the curved duct manifold. Four

~o et al. / Energy 93 (2015) 2116e2125

A.F.O. Falca

2125

rather than the much heavier and mechanically more complex twin
set of guide vanes.
Acknowledgements
This work was funded by the Portuguese Foundation for the
Science and Technology (FCT) through IDMEC, under LAETA PestOE/EME/LA0022 and contracts PTDC/EME-MFE/103524/2008 and
PTDC/EME-MFE/111763/2009, by Project Offshore Test Station, KIC
InnoEnergy, European Institute of Technology, and by the
European Union's Horizon 2020 program, project WETFEET,
under grant No. 646436. Author JCCH was supported by FCT
researcher grant no. IF/01457/2014.
References

Fig. 25. Average efciency hAB in irregular versus rms of pressure head for the new
turbine and for the biradial turbine with sliding guide vanes. Above: average efciency
versus rms sJ of coefcient J. Below: average efciency versus ratio sJ =sJhmax , where
subscript hmax means conditions at maximum value of hAB .

different geometries for the curved-duct manifold were numerically simulated. This was combined with ve different sizes. As a
consequence of detailed measured ow velocity and pressure data
at gas turbine rotor exit being available only over a limited range of
ow rates, results for the overall efciency of the whole turbine
could be computed only over the same range of ow rates.
Extrapolation was used to compute average performance of the
turbine in irregular waves.
Windage losses, that occur at the inactive rotor and are inherent
to the machine conception, were found to be a major loss, reducing
the overall efciency close to best efciency point by about 5%. The
curved-duct manifold with circular cross section is the most efcient of the tested geometries, because of its largest area, and
presumably also because the rounded shape avoids secondary ow
losses at corners. As should be expected, the overall performance
depends on the size of the manifold, represented by the radius ratio
R3/R1. It is only marginally improved for R3/R1 > 1.4, the range R3/
R1 1.0 to 1.2 possibly being a good compromise between aerodynamic performance, overall machine size and cost.
Comparisons were presented between the new turbine (with
R3/R1 1.2 and circular exit section) and the biradial turbine
(sliding guide-vanes version), the latter being possibly the most
efcient self-rectifying turbine model-tested so far. The new turbine was found to be more efcient, both in peak instantaneous
efciency and in maximum average efciency in irregular waves, by
a margin of about 8%.
For identical applications (same pressure head and ow rate),
the new machine is larger than the biradial turbine, in terms of
rotor diameter (Dnew/Dbirad 1.28) and of overall diameter, and its
rotational speed is lower (Unew/Ubirad 0.88). Apart from the
higher efciency, the new machine has the advantage of the
moving element being a simple axially-sliding cylindrical valve

[1] Raghunathan S. The Wells air turbine for wave energy conversion. Prog Aerosp
Sci 1995;31(4):335e86. http://dx.doi.org/10.1016/0376-0421(95)00001-F.
[2] Setoguchi T, Santhakumar S, Maeda H, Takao M, Kaneko K. A review of impulse turbines for wave energy conversion. Renew Energy 2001;23(2):
261e92. http://dx.doi.org/10.1016/S0960-1481(00)00175-0.
[3] Setoguchi T, Takao M. Current status of self rectifying air turbines for wave
energy conversion. Energy Convers Manage 2006;47(15e16):2382e96.
http://dx.doi.org/10.1016/j.enconman.2005.11.013.
[4] Curran R, Folley M. Air turbine design for OWCs. In: Cruz J, editor. Ocean wave
energy e current status and future perspectives. Springer-Verlag; 2008.
p. 189e219.
[5] Falc~
ao AFO, Gato LMC. Air turbines. In: Sayigh A, editor. Comprehensive
renewable energy, vol. 8, Ocean energy. Elsevier; 2012. p. 111e49. http://
dx.doi.org/10.1016/B978-0-08-087872-0.00805-2.
[6] Starzmann R. Aero-acoustic analysis of Wells turbines for ocean wave energy
t Siegen; 2012. Fortschrittconversion. PhD thesis. Germany: Universita
Berichte VDI Reihe 7, Nr 500, Dsseldorf, VDI Verlag.
[7] Falc~
ao AF de O, Henriques JCC. Oscillating-water-column wave energy converters and air turbines: a review. Renew Energy 2015 (in press), http://dx.
doi.org/10.1016/j.renene.2015.07.086.
[8] Falc~
ao AFO, Henriques JCC, Gato LMC, Gomes RPF. Air turbine choice and optimization for oating oscillating-water-column wave energy converter. Ocean
Eng 2014;75:148e56. http://dx.doi.org/10.1016/j.oceaneng.2013.10.019.
[9] Falc~
ao AFO, Gato LMC, Nunes EPAS. A novel radial self-rectifying air turbine
for use in wave energy converters. Renew Energy 2013;50:289e98. http://
dx.doi.org/10.1016/j.renene.2012.06.050.
[10] Falc~
ao AFO, Gato LMC, Nunes EPAS. A novel radial self-rectifying air turbine
for use in wave energy converters. Part 2. Results from model testing. Renew
Energy 2013;53:159e64. http://dx.doi.org/10.1016/j.renene.2012.11.018.
[11] Jayashankar V, Anand S, Geetha T, Santhakumar S, Kumar VJ, Ravindran M,
et al. A twin unidirectional impulse turbine topology for OWC based wave
energy plants. Renew Energy 2009;34(3):692e8. http://dx.doi.org/10.1016/
j.renene.2008.05.028.
[12] Mala K, Jayaraj J, Jayashankar V, Muruganandam T, Santhakumar S,
Ravindran M, et al. A twin unidirectional impulse turbine topology for OWC
based wave energy plants e experimental validation and scaling. Renew
Energy 2011;36(1):307e14. http://dx.doi.org/10.1016/j.renene.2010.06.043.
[13] Pereiras B, Valdez P, Castro F. Numerical analysis of a unidirectional axial
turbine for twin turbine conguration. Appl Ocean Res 2014;47:1e8. http://
dx.doi.org/10.1016/j.apor.2014.03.003.
[14] Falc~
ao AFO, Gato LMC, Henriques JCC. Air turbine for applications in wave
energy conversion. World Patent No. WO/2014/185806. 2014.
[15] Henriques JCC, Falc~
ao AFO, Gomes RPF, Gato LMC. Latching control of an OWC
spar-buoy wave energy converter in regular waves. J Offshore Mech Arct Eng
2013;135:56e64. http://dx.doi.org/10.1115/1.4007595.
~o AFO, Gomes RPF. Latching control of a oating
[16] Henriques JCC, Chong JC, Falca
oscillating water column wave energy converter in irregular waves. In: Proc
33rd Int. Conf. Ocean Offshore Arctic Eng., San Francisco, USA; 2014.
[17] Borges JE. Three-dimensional design of turbomachinery. PhD thesis. UK:
University of Cambridge; 1986.
[18] Borges JE. A three-dimensional inverse method for turbomachinery: part I e
theory. J Turbomach Trans ASME 1990;112(3):346e54. http://dx.doi.org/
10.1115/1.2927666.
[19] Borges JE. A three-dimensional inverse method for turbomachinery: part II e
experimental verication. J Turbomach Trans ASME 1990;112(3):355e61.
http://dx.doi.org/10.1115/1.2927667.
[20] Korpela SA. Principles of turbomachinery. Hoboken, NJ: Wiley; 2011.
[21] Balje OE. Turbomachines. New York: Wiley; 1981.
[22] Falc~
ao AF de O, Rodrigues RJA. Stochastic modelling of OWC wave power plant
performance. Appl Ocean Res 2002;24(2):59e71. http://dx.doi.org/10.1016/
S0141-1187(02)00022-6.