See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/236678216

Aerodynamic Optimization and Open Field Testing of a 1 kW Vertical-Axis

Wind Turbine

Conference Paper · February 2013

CITATIONS READS

3 1,549

5 authors, including:

Gabriele Bedon Marco Raciti Castelli

Vattenfall AB University of Padova
28 PUBLICATIONS 449 CITATIONS 53 PUBLICATIONS 1,299 CITATIONS

SEE PROFILE SEE PROFILE

Uwe Schmidt Paulsen Luca Vita

Technical University of Denmark DNV GL
68 PUBLICATIONS 1,002 CITATIONS 12 PUBLICATIONS 264 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Gabriele Bedon on 31 March 2016.

The user has requested enhancement of the downloaded file.

 Aerodynamic Optimization and Open Field Testing
PO. ID
101
of a 1 kW Vertical Axis Wind Turbine
Gabriele Bedon1, Marco Raciti Castelli1, Uwe Schmidt Paulsen2, Luca Vita2, Ernesto Benini1
1University of Padua, Department of Industrial Engineering – 2DTU Wind Energy, Department of Wind Energy

Abstract Optimization Results

The design of a Darrieus wind turbine rotor is a complex process due to the non- The optimization Pareto front is reported in Figure 5. The individual represented by the
stationarity of the flow field inside the machine and the dynamic stall effects at rotor red cross is chosen for the following considerations. The chosen configuration presents
blades. The combination of a genetic algorithm with a Blade Element – Momentum (BE- the chord distribution shown in Figure 6.
M) code can be successfully adopted to improve the aerodynamic design of existing
wind turbines.
In this work, starting from the Venco Twister 1000-T vertical-axis wind turbine (VAWT),
considered as a baseline configuration, the WOMBAT (Weatherly Optimization Method
for Blades of Air Turbines) algorithm is adopted to provide the optimal chord distribution
for a SCSc rotor characterized by the same swept area.
The optimization is performed considering two objectives: the enhancement of the
power coefficient for a wind speed of 7 m/s - the wind speed that maximizes the power
production probability at the test site - and the maximization of the power coefficient for Fig. 5: Pareto front from WOMBAT. Fig. 6: Chord distribution for the chosen individual.
a wind speed of 12 m/s - the nominal wind speed of the baseline turbine.
The optimal chord distribution is not uniform:
An optimized three-bladed rotor is designed according to the geometrical constraints
imposed by the original turbine configuration and its aerodynamic performance is • The minimum value of the chord at the rotor top and bottom leads to a lower
investigated at the DTU Wind Energy test ground. aerodynamic resistance of the blade sections where the power production is very low.
• An increase of the chord length is registered towards the centre of the rotor, in order to
Case Study and Methodology maximize the aerodynamic load where the radius is increasing, thus enhancing the
power production.
The Venco Twister 1000-T, tested at DTU Wind Energy, is considered as a case study in
order to proceed subsequently with the experimental activity on the optimized prototype. •The successive chord reduction represents a compromise between the increase of the
Venco Twister 1000-T is a 2-meter high turbine characterized by three linear twisted aerodynamic load and the decrease of the interference factor, that would reduce the flow
blades placed at 1 meter from the rotational axis. A picture of the turbine is shown in velocity across the rotor section.
Figure 1. The test conducted at the DTU Wind Energy test site highlighted a maximum
The rotor power coefficient and the power production for different rotational speeds are
power coefficient for the total conversion of around 0.20, as shown in Figure 2.
shown in Figures 7 and 8 as a function of the free-stream wind speed.

Fig. 7: Rotor Power Coefficient. Fig. 8: Rotor Power Production.

Rotor Prototype
Fig. 1: Venco Twister 1000-T. Fig. 2: Twister 1000-T experimental CP curve. The rotor prototype is obtained by assembling several parts, designed in order to fit the
original generator geometry and preserve the blade optimized shape.
In the optimized configuration, a fixed thickness to chord ratio is maintained along the Rotor blades are manufactured adopting the rapid prototyping technique. They are
blade span, whereas the chord distribution is free to be optimized. The NACA 0015 made of ”Proto-plus”, a material with a density of 0.59 g/cm3 and a tensile strength of
profile is chosen, since it provides one of the highest performance in combination with a 40 MPa. Every blade is subdivided into three portions, in order to make their
variable chord distribution. The electrical generator needs to be maintained in the central manufacturing and final assembly easy. As can be seen from Figure 9, rotor blades are
position, due to the fact that the bearings are not designed to be stressed with a bending assembled with the metal reinforcement bars; epoxy resin is then sucked up from the
moment. The so-called ”cut SCS” (SCSc) variant is therefore adopted as optimization bottom to the top, in order to fill the holes between the steel and the blade parts,
geometry, shown in red in Figure 3. providing additional consistence.
The prototype, installed on the 11-meter high tower at Risø - DTU Wind Energy, test site
is shown in Figure 10.

Fig. 3: SCSc vs SCS geometry. Fig. 4: WOMBAT Algorithm.

Fig. 9: Connection of blade sections. Fig. 10: Installation at DTU Wind Energy test site.
The rotor is mounted on a permanent magnet generator equipped with eddy current
brakes. The optimization procedure will therefore consider the range between 0 and 270 Conclusions
rpm to find the best performance.
The aerodynamic design of a commercial vertical axis wind turbine is improved by
A total of 50 individuals are evaluated and evolved for 100 generations, using the adopting an optimization algorithm combined with a Blade Element - Momentum
WOMBAT algorithm, represented in Figure 4. The genes are considered as control point simulation code. Two target wind speeds are selected, in order to provide an optimized
coordinates for a spline curve that describe the trend of the chord. configuration suitable for the same working condition as the baseline configuration.
The fitness of the profile is calculated considering: The optimized rotor configuration results to be characterized by a variable chord
• the individual power coefficient at a wind speed equal to 7 m/s, considered as a design distribution, obtained in order to maximize the energy conversion at two design wind
wind speed. In fact, for this wind speed, the energy production probability for the Risø speeds. A complete simulation campaign is conducted, showing that rotor performances
test site is maximized. are highly increased with respect to the baseline configuration. However, an
experimental test is needed in order to confirm the predictions and to estimate the
• the power production at 12 m/s, considered as the maximum production. This value is aerodynamic and electrical losses, not included in the simulation model. For this reason,
the same declared by Venco for the original Twister 1000-T turbine configuration. a prototype obtained with the rapid prototyping technique is realized and installed on an
11-meter tower at the test site of DTU Wind Energy.

EWEA 2013, Vienna, Austria: Europe’s Premier Wind Energy Event

 AERODYNAMIC OPTIMIZATION AND OPEN FIELD
TESTING OF A 1 KW VERTICAL-AXIS WIND
TURBINE
Gabriele B edon, PhD Student at the Department of Industrial E ngineering, University of Padua,
Via Venezia 1, 35131 Padova, [email protected]
Marco Raciti Castelli, Research Associate at the Department of Industrial E ngineering, University
of Padua, Via Venezia 1, 35131 Padova, Italy, [email protected]
Uwe Schmidt Paulsen, Senior Scientist at DTU Wind E nergy, Department of Wind Energy,
Frederiksborgvej 399 Building 125, room S14 & 16, 4000 Roskilde, Denmark, [email protected]
Luca Vita, Post Doctoral Researcher at DTU Wind Energy, Department of Wind Energy,
Frederiksborgvej 399 Building 125, room S14 & 16, 4000 Roskilde, Denmark,
luca. [email protected]
Ernesto Benini, Associate Professor at the Department of Industrial Engineering, University of
Padua, Via Venezia 1, 35131 Padova, Italy, [email protected]

Abstract
The design of a Darrieus wind turbine rotor is a complex process due to the non-stationarity of
the flow field inside the machine and the dynamic stall effects at rotor blades. The combination of a
genetic algorithm with a Blade Element – Momentum (BE-M) code can be successfully adopted to
improve the aerodynamic design of existing wind turbines.
In this work, starting from the Venco Twister 1000-T vertical-axis wind turbine (VAW T),
considered as a baseline c onfiguration, the WOMBA T (W eatherly Optimization Method for Blades
of Air Turbines) algorithm is adopted to provide the optimal chord distribution for a SCSc rotor
characterized by the same swept area.
The optimization is performed considering two objectives: the enhancement of the power
coefficient for a wind speed of 7 m/s - the wind speed that maximizes the power production
probability at the test site - and the maximization of the power coefficient for a wind speed of 12 m/s
- the nominal wind speed of the baseline turbine.
An optimized three-bladed rotor is designed according to t he geometric al c onstraints imposed by
the original turbine configuration and its aerodynamic performanc e is investigated at the DTU Wind
Energy test ground.

1. Introduction and Background

The purpos e of this work is to increase the rot or performances in terms of power production and
power coefficient of a real vertical axis wind turbine (the V enco Twister 1000-T) by changing the
blade aerody namic design.
The aerodynamic design optimization of a vertical axis wind turbine is a complex process due to
the difficulties in obt aining fast and reliable performance estimation. Apart from expensive
experimental tests, which can be c onducted on the physical models, the performances of a wind
turbine can be preliminary estimated by means of numeric al methods. The most important
numerical approaches available in the literature are:
Computational Fluid Dy namics (CFD) codes, which are based on a numerical approximation of
the Navier-Stokes equations and can provide a reliable description of the turbine behavior.
However the required computational time and effort are high [1].
Codes based on the Vortex Model: this approach also allows obtaining a reliable performance
 estimation [2]. Anyway it still requires a considerable computational time to c onverge,
especially at low tip speed ratios, when complex vortexes are generated [3].
Codes based on the BE-M Theory: equating the blade element force on the rotor to the change
in fluid momentum, a very fast prediction of the rotor performance is obtained [3–5]. On the
other hand, the results provided by the algorithm do not include much additional data, including
detailed air flow information.
These methods can be coupled with an optimization algorithm, capable of evaluating the
aerodynamic performance of the rotor geometry. Genetic algorithms have been developed to find
the optimal solution in complex spaces efficiently [6,7] and are widely used for solving engineering
problems: this leads to an iterative loop t hat should provide an improved design configuration. This
approach was followed by several authors. Bourguet et al. [8] adopted a CFD code coupled wit h a
genetic algorithm to find the optimal airfoil shape to increase both the nominal power production
and the aerodynamic efficiency while reducing the weight of the blade. The optimal blade profile
resulted to be very close to the NACA 0025 geometry. Carrigan et al. [9] matched an evolutionary
algorithm and a bi-dimensional CFD simulation for the simulation of a H-Darrieus rot or equipped
with NA CA 0015 airfoils. The optimal configuration resulted to have an increase in efficiency of
around 6%, achieved with an increase of 58% in rot or blade thickness and a reduction of 40% in
rotor solidity.
Being the aim of this work the maximization of the rotor performanc e in terms of power production
and power coefficient, a simulation tool based on the BE-M Theory is selected. This allows the
evaluation of a large number of blade geometries in a limited computational time, widening the
investigation space considered by the algorithm. The adopted BE-M code is developed by Raciti
Castelli et al. [10] and is based on the double multiple streamt ube model developed by Strickland
[3] and Paraschivoiu [4,5]. Moreover, dynamic stall models bas ed on the Gormont-Strickland and
Gormont-B erg studies [11] are included, as well as a blade finite aspect ratio correction from Viterna
and Corrigan [12]. This BE-M code is coupled with the genetic algorithm bas ed on the studies of
Deb [13], provided as gamultiobj function in the Matlab suite, obtaining the WOMBA T algorithm, first
presented by Bedon et al. [14].

2. Case Study and Methodology

The V enco Twister 1000-T, tested at DTU Wind Energy [15], is considered as a case study in
order to proceed subsequently with t he ex perimental activity on the optimized prototype. Venco
Twister 1000-T is a 2-meter high turbine characterized by three linear twisted blades placed at 1
meter from the rotational axis. The three blades are connected by spokes wit h the generator,
placed in corres pondence to the rot ational axis. A picture of the turbine is shown in Figure 1.
The test conducted at the DTU Wind Energy test site highlighted a maximum power coefficient for
the total conversion of around 0.20, as shown in Figure 2 [17].
The adopted blade geomet ric paramet ers are quite difficult to measure. However, it can be
observed that the blade is characteriz ed by a symmet ric profile, with uniform chord and thickness
along the whole length. Moreover, it is difficult to establish the influence of the spoke design (quite
wide and thick) on t he power production. The numerical simulation of the Venco rotor would
therefore require a consistent amount of work on several geometric parameters that need to be
optimized. A more simple baseline geometry is thus adopted, nevertheless maint aining the same
swept area of Twister 1000-T, equal to 3.61 m².
 Fig. 1: Venco Twister 1000-T VAW T [16].

Fig. 2: Venco Twister 1000-T experimental power coefficient curve [17].

A fixed thickness to chord ratio is maintained along the blade span, whereas the chord
distribution is free t o be optimized. Considering the results obtained by Bedon [18], the NA CA 0015
profile is chosen, sinc e it provides one of the highest performance in combination with a variable
chord distribution.
The simulation algorithm was validat ed with turbines adopting blades of Troposkien shape [19]:
this design allows limiting blade inertial loads [20]. Finally, a variable chord distribution is adopted,
because of the difference in blade section radius.
The electrical generat or needed to be maintained in the central position, due to the fact that the
bearings are not designed to be stressed with a bending moment. The so-called ”cut S CS” (SCSc)
variant [21] is therefore adopted as optimization geometry. This solution is realized by cutting a 20%
in SCS height, 10% at the top and 10% at the bottom extremities. In order to have a fair
comparis on, the same swept area is maintained, by increasing the maximum radius.
A comparison between the original SCS and the SCSc midline geometries is shown in Figure 3.
 Fig. 3: Comparison bet ween the classical SCS geometry (blue line) and the proposed S CSc (red
line) one.

The S CSc shape requires connecting arms at bot h top and lower sections. Connecting arms are
anyway provided in the original Venco configuration. The same s wept area of Twister 1000-T, 3.61
2
m , is maintained. The main geometrical parameters of the optimized rotor configuration are
reported in Table 1.

H [m] 1.96
R [m] 1.18
N 3
Blade Shape SCSc
Table 1: Main geometric al features of the optimized rotor configuration

The rotor is mount ed on a permanent magnet generator equipped wit h eddy current brakes. The
maximum rot ational speed without their activation is 270 rpm [22]. The optimization procedure will
therefore consider the range between 0 and 270 rpm to find the best performance.
A total of 50 individuals are evaluated and evolved for 100 generations. The number of genes ng
provided by the genetic algorithm is equal to half of the number of vertical divisions. The genes are
considered as control point coordinates for a spline curve (ng/2 abscissas and ng/2 ordinates) that
describe the trend of the chord.
The gamultiobj algorit hm tends to a solution that minimizes the fit ness value. The fitness of t he
profile is thus calculated considering the following expressions:

where:
is the individual power coefficient at a wind speed equal to 7 m/s, considered as a
design wind speed. In fact, for this wind speed, the energy production probability for the Ris ø
test site is maximized [23,24].
 is the power production at 12 m/s, considered as the maximum production. This value
is the same declared by Venco for the original Twister 1000-T turbine configuration [15].

3. Optimization Results
The WOMBA T algorithm provides a Pareto front containing several optimum individuals. The
Pareto front is reported in Figure 4.

Fig. 4: Pareto front from the WOMBA T optimization algorithm. The red cross is the individual
chosen for the analysis.

The individual represented by the red cross is chosen for the following considerations. In fact,
individuals characterized by higher power c oefficients present a sensibly lower power production,
due to the curve dec reasing slope. On the other side, individuals with a slightly higher power
production present a sensibly decrease in power coefficients.
The chosen configuration presents the chord distribution shown in Figure 5.

Fig. 5: Chord distribution for the chosen individual.

The optimal chord distribution, considering a fixed airfoil thickness/chord ratio, is therefore not
uniform. The minimum value of t he chord at the rotor top and bottom leads to a lower aerodynamic
resistance of the blade sections where the power production is very low. An increase of the chord
length is registered towards the centre of the rotor, in order t o maximize the aerodynamic load
where the radius is increasing, thus enhancing the power production. The successive chord
reduction represents a compromise between the increase of the aerodynamic load and the
decrease of the interference factor, that would reduce the flow velocity across the rotor section.
A complete simulation of the optimal individual is conducted, in order to have a complete
overview of the rotor perform ance. The rotor power coefficient and the power production for
different rotational speeds are shown in Figures 6 and 7 as a function of the free-stream wind
speed.

Fig. 6: Rotor power coefficient for several angular velocities, as a function of the unperturbed
wind speed.

Fig. 7: Rotor power production for several angular velocities, as a function of the free-stream wind
speed.

These curves represent the estimation of t he rotor performance, obt ained from the BE-M
algorithm. The real performance is however reduced by the presence of the spokes and by the
efficiency of the energy conversion performed by the electrical generator, the rectifier and the
inverter.
4. Rotor Prototype
The rotor prototype is obtained by assembling several parts , designed in order to fit the original
generator geometry and preserve the blade optimized shape.
In t he original configuration, the spoke connections are shifted azimut hally, in order to connect
the twisted blades, as can be seen in Figure 8. On the other side, the prototype configuration
provides a blade whose medium line is entirely contained in a plane. A disk, presenting the spoke
links in the correct position, is therefore designed and installed on top of the electrical generator,
using the existing bolts. With reference to Figure 8, the disk is represented in red, whereas the
generator is reproduced in whit e color.

Fig. 8: Connecting disk (red) installed on the original Venco generat or (white).

Six spokes, obtained from a 10 mm steel plate and manufactured by laser-cut in order to obtain
the desired geometry, are connected to the correct links. Their width is varying between 105 mm at
the generator side and 80 mm at the blade side, for a total length of 900 mm.
Spokes are intended not to be in contact with the blades: six metal components, the so-called ”in-
blades”, are designed in order to be fixed on t he terminal portion of the spoke, as well as to be part
of the blade. These components, represented in Figure 9, are obtained from a 8 mm steel plate and
manufactured by laser-c ut.

Fig. 9: Connection component between the blade and the spoke (”In-blade”).
 The in-blade components present one rectangular and two circular holes. In the rectangular
opening, t wo 1-mm thick steel plates are inserted. These plates permit to lock the orientation of the
blade with respect to the spoke. In eac h circular hole, four 1-mm diameter steel wires, providing
strength to the whole assembly, are inserted.
Rotor blades are manufactured adopting the rapid prototyping technique. They are made of
3
”Proto-plus” [25], a material with a density of 0. 59 g/cm and a tensile strength of 40 MP a. E very
blade is subdivided into three portions, in order to make their manufacturing and final assembly
easy. As can be seen from Figure 10, rotor blades are assembled with the metal reinforcement
bars; epoxy resin is then sucked up from the bottom to the top, in order to fill the holes bet ween the
steel and the blade parts, providing additional consistence.

Fig. 10: Connection assembly of the in-blade components and blade sections, inserting the steel
reinforcements.

Steel strings are added in order to reduce the blade deformation and ensure an operational
stability. Such strings have a diameter of 2 mm and do not compromise t he aerodynamic efficiency
excessively.
The prototype, installed on t he 11-meter high tower at Risø - DTU Wind Energy, test site is
shown in Figure 11.
 Fig. 11: Installation at Ris ø - DTU Wind Energy test site.

5. Conclusions
The aerodynamic design of a commercial vertical axis wind turbine is improved by adopting an
optimization algorithm combined with a Blade Element - Momentum simulation code. Two target
wind speeds are selected, in order to provide an optimized configuration suitable for the same
working condition as the baseline configuration.
The optimized rotor configuration res ults to be characterized by a variable chord distribution,
obtained in order to maximize the energy conversion at two design wind speeds. A complete
simulation campaign is conducted, showing t hat rotor performances are highly increased with
respect to the baseline configuration. However, an experimental test is needed in order to confirm
the predictions and to estimate the aerodynamic and electrical losses, not included in the simulation
model. For this reason, a prototype obtained with t he rapid prototyping technique is realized and
installed on an 11-meter tower at the test site of DTU Wind Energy. Further work is required in order
to complete the experimental campaign of measurements on the prot otype.

Nomenclature
c [m] airfoil chord
CP [-] power coefficient
h [m] blade element height
H [m] rotor height
N [-] blade number
ng [-] number of genes
P [W] power production
R [m] rotor maximum radius
U∞ [m/s] free-stream mean air velocity
 References
[1] Racit i Castelli M, Englaro A, Benini E. The Darrieus Wind Turbine: Proposal for a New Performance
Prediction Model Based on CFD. Energy 2011; 36(8): 4919–4934.
[2] Strickland J, Webster B, Nguyen T. A Vortex Model of the Darrieus Turbine: An Analytical and
Experimental Study. Journal of Flu ids Engineering 1979; SA ND79-705.
[3] Strickland J. The Darrieus Turbine: A Performance Predict ion Model Using Multip le Streamtub es.
Sandia Nat ional Laboratories Report 1975; SA ND75-043.
[4] Paraschivoiu I. Double-mu ltip le Streamtube Model for Darrieus in Turb ines. Wind Turbine Dynamics
1981.
[5] Paraschivoiu I. Double Mult iple Streamtube Model with Recent Imp rovements. Journal of Energy 1983.
[6] Holland JH. Adaptation in Natural and Art ificial Systems. Ann Arbour: The University of M ichigan
Press 1975.
[7] Holland JH. Adaptation in Natural and Art ificial Systems: An Introductory Analysis with Applications to
Biology, Control, and Artificial Intelligence. Camb ridge, MA, USA: M IT Press 1992.
[8] Bourguet R, Martinat G, Harran G, Braza M. Aerodynamic Multi-criteria Shape Opt imization of VAWT
Blade Profile by Viscous Approach. Wind Energy 2007: 215–219.
[9] Carrigan T, Dennis B, Han Z. Aerodynamic Shape Optimizat ion of a Vert ical-A xis Wind Turbine Using
Differential Evolution. ISRN Renewable Energy 2012: 1– 40.
[10] Racit i Castelli M, Fedrigo A, Benin i E. Effect of Dynamic Stall , Finite Aspect Ratio and Streamtube
Expansion on VAWT Performance Predict ion Using the BE-M Model. International Journal o f Engineering
and Physical Sciences 2012; 6: 237–249.
[11] Massé B. Description De Deu x Programmes D’ordinateur Pour Le Calcul Des Performances Et Des
Charges Aérodynamiques Pour Les Éoliennes à Axe Vert ical. IREQ-2379 1981.
[12] Viterna LA, Corrigan RD. Fixed Pitch Rotor Performance of Large Horizontal A xis Wind Turbines.
Annual Review of Flu id Mechanics 1982; 15(1): 69–85.
[13] Deb K. Mult i-objective Optimization Using Evolutionary Algorith ms. John Wiley & Sons 2001.
[14] Bedon G, Raciti Castelli M, Benin i E. Optimization of a Darrieus Vertical-A xis Wind Turbine Using
Blade Element Momentum Theory and Evolut ionary Algorith m. Accepted by Renewable Energy 2012.
[15] Diznabi B. Test Plan of Duration Test IEC-61400-2, Danish Standard & Performance Test IEC 61400-
12-1 for VENCO-Twister-1000-T. Risø Report, Draft 2012.
[16] VencoPower. Http://www.vencopower.com. Website Accessed 06/2012 n.d.
[17] Friis P, Sch midt Paulsen U. Venco Project Status Report 06-Dec-2011. Risø Report, Draft 2011.
[18] Bedon G, Raciti Castelli M, Benin i E. Chord Optimizat ion for a Darrieus Vertical A xis Wind Turbine
Using WOMBAT A lgorith m. Sub mitted to Renewab le Energy 2012.
[19] Bedon G, Raciti Castelli M, Benin i E. Nu merical Validation of a Blade Element-Momentu m Algorithm
Based on Hybrid Airfoil Polars for a 2-m Darrieus Wind Turbine. International Journal of Pure and Applied
Sciences and Technology 2012; 12(1): 1–6.
[20] Nord mark A, Essén H. The Skipping Rope Curve. European Journal of Physics 2007; 28(2): 241–247.
[21] Bedon G, Raciti Castelli M, Benin i E. Proposal for an Opt imal Chord Distribution of a Small Size
Darrieus VAWT Blade. Sub mitted to Engineering with Co mputers 2012.
[22] Bedon G, M ijatovic N, Træholt C, Sch midt Pau lsen U, Vita L, Test I, Co mponents S, Unit V, Controller
S, Boy W. Mechanical to Electrical Power Conversion in Venco Twister 1000-T Turbine. DTU Wind Energy
Report 2012; (I-0010 (EN)).
[23] Diznabi B. Duration Test. Risø Report, Draft 2012.
[24] Diznabi B. Risø Site Layout and Class. Risø Report, Draft 2012.
[25] Protoservice. http://www.p rotoservice.it/Scheda_tecnica_ProtoService.pdf. Website Accessed 06/2012
n.d.

View publication stats