You are on page 1of 50

Aerospace Engineering.

Aerospace Engineering Individual Investigative Project

Determination of the Structural Loads on Wind Turbine Rotor Blades


Matthew Harrison
May 2012

Supervisor: Dr R J Howell

Dissertation submitted to the University of Sheffield in partial fulfilment of the requirements for the degree of Master of Engineering

Abstract
Wind turbine technology is well established within the renewable energy market. Many customers seek to install new turbines with higher power output especially for offshore wind farm installations. However designers must ensure the structural integrity of these increasingly large and heavy wind turbine rotors without degrading their performance. Fluid-structure interaction analysis is a method currently being developed for use in solving aeroelastic design problems. It couples CFD and FEA computational tools to allow the simulation of dynamic engineering problems where the aerodynamic load influences the structural deformation and vice versa. This investigation seeks to develop and apply the method to the analysis of large wind turbine rotor blades. The results demonstrate the coupling between the aerodynamic load and structural deformation. They are also used to evaluate the performance of the turbine in terms of efficiency and power output for different operating conditions. The model used was greatly simplified for use in the first series of test resulting in some inaccuracies especially in the structural model.

Contents
Nomenclature Acknowledgements i ii

1. 2.

INTRODUCTION BACKGROUND AND LITERATURE REVIEW 2.1 2.2 Wind Energy Background Wind Energy Physical Concepts 2.2.1 Rankine-Froude: Actuator Disk Model 2.2.2 Blade Element Momentum Model Wind Turbine Aerodynamics 2.3.1 Blade Section Aerodynamics 2.3.2 Effect of Angle of Attack 2.3.3 Tip-Speed Ratio Wind Turbine Loads and Structures 2.4.1 Aerodynamic Loading 2.4.2 Inertial Loading 2.4.3 Gravitational Loading Literature Review

1 3 3 4 4 6 6 6 7 8 8 9 9 9 10 15 15 15 15 16 17 17 18 19 22 23 23 25 26 27 27 27 27 28 29 29 29 31

2.3

2.4

2.5 3.

METHODOLOGY 3.1 3.2 3.3 3.4 Introduction Selection of Blade Geometry 3.2.1 The Tvind Wind Turbine Selection of Computational Tools Geometry and Mesh Generation 3.4.1 Modelling the Blade Geometry 3.4.2 Modelling the Fluid Domain Geometry 3.4.3 Meshing the Fluid Domain 3.4.4 Meshing the Blade Structure CFD and FEA Solver Setup 3.5.1 CFD Solver Setup 3.5.2 FEA Solver Setup 3.5.3 FSI Interface Setup

3.5

4.

VERIFICATION AND VALIDATION 4.1 4.2 Introduction Simulation Verification 4.2.1 Residual Monitors 4.2.2 Point Monitors 4.2.3 Mesh Independence Results Validation 4.3.1 CFD Validation 4.3.2 FEA Validation

4.3

5.

RESULTS AND DISCUSSION 5.1 5.2 Introduction CFD Results 5.2.1 Physical Flow Features 5.2.2 Wind Turbine Efficiency 5.2.3 Wind Turbine Power Output FEA Results 5.3.1 Blade Deflection 5.3.2 Blade Stress

32 32 32 32 34 35 36 36 38 40 40 42 44

5.3

6.

CONCLUSIONS 6.1 Future Work

7.

REFERENCES CFD Mesh Statistics Histograms

Appendix 1:

Nomenclature
Rotor Area Power Coefficient Thrust Coefficient Drag Lift Mass Flow Rate Power Blade Radius Location Blade Radius Thrust Velocity Angle of Attack Pitch Angle Tip-Speed Ratio Density Rotor Angular Velocity

Acknowledgement
I would like to thank my project supervisor Dr. Rob Howell for his words of advice and guidance throughout the project. Thanks also to all friends and family who helped in editing and revising this report.

ii

1.

INTRODUCTION
At the turn of the millennium people around the world were becoming increasingly aware of the threat climate change posed to the planet. Today it is now widely accepted that human activities around the globe are affecting the average temperature of the Earths atmosphere and oceans. The main mechanism forcing climate change, known as the greenhouse effect, is driven by emissions of greenhouse gases, namely: carbon dioxide, nitrous oxide, halocarbons and methane. The primary human contribution to greenhouse gas emissions is from the creation of carbon dioxide through the burning of fossil fuels. Therefore, as countries seek to cut greenhouse gas emissions they look to increase the use of renewable energy and reduce their dependency on coal, oil and gas. Furthermore, the production of fossil-fuels is likely to enter decline in the near future as the worlds supply begins to run out. In the UK for example, North Sea oil production has halved over the last decade as described in the book by Strahan (2008). Amongst the ever increasing list of available renewable energy technologies, wind power has established itself at the forefront of the renewable energy sector. It has the advantage of being able to produce electricity with no carbon dioxide emissions at a production cost similar to those of fossil-fuel based generation. To meet the growing demand for wind power generation manufacturers are continually developing wind turbines of ever greater power output and size. For example, the Danish manufacturer Vestas offers turbines rated at 7MW with rotors 164m in diameter. The sheer scale of these turbines poses some tough design challenges for engineers in terms of aerodynamics and structural design. An aerodynamic load on a blade of such size will deform it placing a strain and stress on the structure. The aim of this final year project is to investigate the use of fluid-structure interaction (FSI) analysis as a tool for assessing the aerodynamic and structural performance of a wind turbine. In particular, it will seek to develop an understanding of the methodology required to successfully complete such an analysis, and how it could be used to aid the research and design of future wind 1

turbine systems. This will be achieved through applying the developed methodology in a simulation of a typical wind turbine blade. The report opens with a brief history and description of wind turbine technology before a more detailed discussion of the aerodynamic behaviour of wind turbines. A review of current research, methods and design tools used is covered in the literature review. Finally, the investigation and subsequent findings is given in the main body of the report.

2.
2.1

BACKGROUND AND LITERATURE REVIEW


Wind Energy Background
Wind power technology has been used since ancient times. However, it is only relatively recently that the power of the wind has been harnessed to generate electricity through the invention of wind turbines. An excellent account of the development of wind power as a technology from the earliest windmills through to modern day wind turbines is given in the book by Spera (1994). It also describes how wind turbine designers came to settle on the two most typical configurations seen today: the horizontal-axis wind turbine (HAWT), and the vertical-axis wind turbine (VAWT), examples of which can be seen in Figure.2.1. For large-scale wind turbine systems generating over 500kW the HAWT configuration is most common whereas the VAWT setup is usually applied to small-scale systems. This project is primarily concerned with HAWTs and will therefore not discuss the operation of VAWTs in the following sections; although many of the two configurations share the same principles of operation.

Figure.2.1. Modern Day Wind Turbine Configurations: the Horizontal-Axis Wind Turbine (left) and the Vertical-Axis Wind Turbine (right)

2.2

Wind Energy Physical Concepts


The fundamental concept behind the operation of a wind turbine is to extract the kinetic energy of the wind to drive an electrical generator. Before considering the aerodynamics of a wind turbine in more depth it is a good point to provide an overview of some of the basic models used for wind turbine analysis.

2.2.1 Rankine Froude: Actuator Disk Model This model uses the principle for conservation of momentum to idealise rotational machinery such as propellers, helicopter rotors and wind turbines. Developed by W.J.M. Rankine and R.E. Froude in the 19th century the actuator disc model, or axial momentum theory, describes the main processes involved in extracting energy from the wind. It is described in numerous textbooks on rotational machinery including books specific to wind turbine applications such as the books by Spera (1994) and more recently Hansen (2000). In this idealisation of a wind turbine rotor a number of assumptions are made: The rotor is modelled as an infinitely thin disk. The loads and velocity distributions are uniform over the disk. The upstream and downstream pressures equal the freestream pressure. The model is purely a one-dimensional analysis. The fluid is incompressible.

If a control volume is defined by the outer streamlines of the streamtube that passes through the actuator disk as shown in Figure.2.2, then the wind speed reduces from the inlet to the outlet. From Newtons Second Law this change in momentum must be balanced by an external force which can only come from the actuator disk. Furthermore, by Newtons Third Law there must be an equal and opposite force on the actuator disk i.e. the force of the wind. Importantly the actuator disk theory shows that energy is extracted by a wind turbine through a step drop in pressure over the rotor itself with a net drop in wind speed.

Figure.2.2. Illustration of Control Volume for the Actuator Disk Model and the Corresponding Velocity and Pressure Profiles Along the Volume Wind speed decreases approaching the actuator disk and can be described by Bernoullis equation up until the disk where the step drop in pressure occurs. Pressure increases back to the freestream pressure downstream of the disk with a further loss in wind speed which may also be described by Bernoullis equation. Referring to Figure.2.2 expressions for the thrust and shaft power can be derived to be: 2.1 2.2 In the analysis of wind turbines there are two common non-dimensional coefficients used to indicate the efficiency: the power coefficient thrust coefficient , and the

. These are defined as the ratio of power generated to

power available, and thrust to kinetic energy of the wind respectively. 2.3

2.4

In 1919 Albert Betz showed using this idealised actuator disk model that there is a theoretical maximum amount of energy that may be extracted from the wind. A maximum power coefficient of is known as the Betz limit.

It means that for any conventional HAWT a maximum of 59.3% of the wind energy may be captured by a turbine. In reality a wind turbine will suffer from additional losses due to the effects of rotation in the wake. This is a key limitation of the actuator disk model which is purely a one-dimensional analysis. 2.2.2 Blade Element Momentum Model The blade element momentum (BEM) model seeks to build upon the actuator disk model by accounting for the geometry of the rotor. No longer is the rotor modelled as a solid disk but with its actual characteristics such as aerofoil section, chord length, aerodynamic twist and number of blades. Mathematically the blade element model is much more involved and comprehensive discussions can be found in the books by Spera (1994), Hansen (2000), and Walker (1997). It is a tool used extensively for the determination of the aerodynamic performance of wind turbine blades. As a computational tool it offers very good accuracy to cost making it widely used for research purposes and design codes.

2.3

Wind Turbine Aerodynamics

2.3.1 Blade Section Aerodynamics It has been shown that a wind turbine rotor extracts energy from the wind through a drop in pressure over the plane of the rotor. It will now be shown how this is achieved in reality with reference to Figure.2.3. During normal operation the wind turbine rotor will be rotating at a fixed velocity. This means a given section of a rotor blade will see a relative velocity made up of the incoming wind speed component and a component due to the rotational speed of that section of the blade. This relative velocity will approach the blade with an angle of attack . The angle between the chordline of the blade and the plane of of the blade section. A combination of the 6

rotation is known as the pitch

angle of attack and the curvature of the aerofoil itself results in an acceleration of the flow of the top (suction) surface of the blade. By Bernoullis principle an increased flow velocity results in a drop in pressure. A low pressure region develops over the top surface with a relative high pressure region on the lower (pressure) surface. This gives rise to aerodynamic forces which are commonly split into two components: lift (L) which acts perpendicular to the relative wind and drag (D) which acts parallel to the relative wind. Both of these forces act through the centre of pressure.

Figure.2.3. Velocity and Resulting Aerodynamic Force Vectors on a Typical Wind Turbine Aerofoil A major part of the lift force acts in the plane of rotation. When the complete blade is taken into account the result is a torque about the axis of rotation causing the blade to rotate. However, by the same mechanism, part of the drag force acts to counter the blades rotation. It is therefore important to maximise the lift to drag ratio to provide the best performance. This gives rise to a number of challenges in terms of aerodynamic design of rotor blade. Some of the key factors are discussed briefly here. 2.3.2 Effect of Angle of Attack For a given rotor speed the local blade section rotational velocity will

vary from the root to the tip of the blade. This will change the angle of attack of the relative velocity seen by the local blade section as shown in Figure.2.4.

Figure.2.4. Typical Velocity Vector Diagrams at 25%, 50% and 100% of Blade Radius Close to the root where the blade section is relatively slow moving the angle of attack increases significantly. Designers must account for this especially for large rotors where the variation can be significant. It may be accounted for by introducing additional pitch angle close to the root and using more suitable aerofoil section shapes. A modern turbine blade will therefore have some degree of blade twist and a range of aerofoils along its length to provide the optimal lift to drag ratio thus maximising the power coefficient at its specified operational speed. 2.3.3 Tip-Speed Ratio The tip-speed ratio wind speed i.e. 2.5 It is a value used widely in the assessment of wind turbine efficiency. Losses become a significant issue at low TSR values. Normally wind turbines are designed to operate at TSRs of 6-10. is the ratio of the tip speed of the blade to the incoming

2.4

Wind Turbine Loads and Structures


A wind turbine is subjected to wide range of loads and must be built to withstand them for a lifetime of operation which may last decades. This text focuses on the rotor and its blades but does not cover additional components that go into a complete wind turbine system such as the tower, drive shafts and generator housing. More detailed descriptions of all parts can be found in the books mentioned previously or through many of the papers referenced in the 8

literature review. For a wind turbine rotor there are three key sources of loading: aerodynamic loading, inertial loading, and gravitational loading. 2.4.1 Aerodynamic Loading The wind creates a pressure loading on the blades of a wind turbine rotor as discussed in the previous sections. In addition to the aerodynamic forces delivering toque to the rotor, there is a significant component of these forces acting to bend the blade in the direction of the flow. It is therefore common to attribute these loads to different bending moments around the root of the blade. The edgewise bending moment describes the loading that results from the forces delivering torque to the blade, whereas the flapwise bending moment describes the loading that results from the forces acting normal to the flow. Although a typical analysis of wind turbine performance will only consider a uniform wind flow hitting the rotor, in reality the wind is turbulent and for large turbines will contain some degree of wind shear. The local terrain, trees and vegetation all have an impact on the incoming wind profile. These factors increase the complexity of any analysis as they would require a time-dependent variation of aerodynamic loading. 2.4.2 Inertial Loading There are two main sources of inertial loads; the first being loads created during the acceleration and deceleration of the rotor; the second being from the centrifugal force on the blades. Centrifugal forces can be used to the advantage of the wind turbine designer by mounting the blades at an angle (coning angle) which allows the centrifugal force to counteract the flapwise bending moment. 2.4.3 Gravitational Loading On a HAWT the blades are also subjected to a sinusoidal loading as each blade passes from the top of its rotation to the bottom and back again. This creates alternating periods of tensile and compressive stresses in the blade structure which provides designers with an undesirable fatigue problem. Given that a typical wind turbine will endure many millions of cycles during its lifetime it is important to estimate the number of cycles to failure in the design phase. 9

2.5

Literature Review
There is a vast range of studies that have been conducted on wind turbines, many of which focus on individual aspects covered in the overview of wind turbines given above. This literature review concentrates on the previous work concerning the use of computational modelling of wind turbines and the application of CFD and FEA tools in their analyses. An overview of techniques used for the aerodynamic and aeroelastic analysis of wind turbines is given in the paper by Hansen et al (2006). It highlights the importance of the coupling between the aerodynamic loads and the structural behaviour of the blades. The blade element model is still a key tool in the analysis of wind turbines. It is concluded that the development of new aeroelastic design codes is driven by the design trends of new wind turbines which demand new capability of the codes. This is demonstrated in the aims of this project where the increasing size of todays wind turbines requires the development of methods for coupling their aerodynamic and structural design. Up-scaling of wind turbines is likely to be predominating as the off-shore sector becomes more established. In terms of aerodynamic analysis of wind turbines there have been several papers that seek to validate the use of CFD through comparison with the theoretical models. The study by Hartwanger and Horvat (2008) found that the use of a high resolution mesh with a suitable turbulence model provided an excellent match between CFD and 2D blade section experimental data. An interesting comparison was made with the basic actuator disk model which showed good agreement for low wind speed cases. This can be expected as greater losses occur at higher wind speeds through wake vorticity and tip losses. Importantly the study employed the commercial ANSYS CFX solver and the XFOIL panel code, both of which are available at The University of Sheffield. The methodology used for the CFD analysis in this study was a good source of information for the CFD part of this final year project, in particular the use of the CFX Frozen-Rotor method. There are a range of design codes and software available from both commercial and open sources for use in wind turbine analysis. Throughout the initial stages 10

of the project research was undertaken into which tools could potentially be used for the FSI study as either the primary analysis tool or as an additional verification tool. A compiled list of commonly used design codes is given in the thesis by Ahlstrom (2005). Some of the specialist codes and software considered for the project included: the commercial BLADED software package from GL Garrad Hassan wind energy consultancy; the FAST code developed by the National Renewable Energy Laboratory (NREL) and Oregon State University; the PHASTAS programs from the ECN Wind Energy of the Netherlands Energy Research Foundation; and HAWC developed at Ris DTU Laboratory for Sustainable Energy in Denmark. A recent investigation into the use of FSI analysis is detailed in the two-part paper by Bazilevs et al (2010). As one of the first fully-coupled FSI analysis of a complete wind turbine rotor the paper is an important reference for future developments in this field. It was conducted on the well-researched NREW 5MW offshore baseline turbine and made use of the FAST design code for the aerodynamic modelling of the turbine. In the first part which covers the geometry and mesh development alongside the fluid flow solution, the importance of 3D modelling and simulation were highlighted. For example, it can be seen in Figure.2.5 how losses at the tip can affect the torque distribution along the blade span.

11

Figure.2.5. Patches Along the Blade (top) and the Aerodynamic Torque Contribution from Each Patch (bottom), Reproduced from Bazilevs et al (2010) The study also took on the challenge of modelling the composite materials used in the blade structure although their final model did not include more complicated features such as shear webs and spar caps. The set of images in Figure.2.6 show an excellent illustration of a wind turbine blade deflecting under the aerodynamic load and its subsequent affect on its aerodynamic performance. Although the flow remains fully attached on the pressure surface, there is clear variation in the flow separation point on the suction surface at each of the different time intervals.

12

Figure.2.6. Isocontours of Relative Wind Speeds for Different Time Instances Superimposed onto a Moving Blade. (a) t=0.7s; (b) t=1.2s; (c) t=2.0s; (d) t=4.5s Reproduced from Bazilevs et al (2010) In 2011 another study took place which applied FSI analysis which compared the performance of a rigid and flexible blade structure also of a 5MW turbine. The report by D.Kim and Y.Kim (2011) details an approach using their own FSI program called FSIPRO3D developed previously by the authors. The results show clear differences between the rigid and flexible blade model under an aeroelastic analysis. The power output of the flexible blade was 0.5MW lower than the rigid model under the same flow conditions. A study by Maheri et al (2006) developed a method for the analysis of bendtwist adaptive wind turbine rotor blades. The blades are considered to be subject to an induced twist which varies with the rotor speed and wind conditions. It is stated that the use of commercial FEA and CFD software for the analysis of such blades is both cumbersome and inefficient. The problems associated with the FSI methodology are also highlighted such as the requirement of an interface between the two solvers and the challenges of accurate mesh

13

generation. In a similar manner to the investigation by D.Kim and Y.Kim, a unique design code called WTAB was developed in attempt to overcome these problems. It is stated that the code which combines a single FEA solution with analytical approach for determining the blade twist and a method for adapting the blade topology, is much faster than a typical FSI solution. However, the aerodynamic load calculator is only based on the BEM model, which does not offer the accuracy required to resolve the complicated flow characteristics associated with larger wind turbine blades, as discussed in the later study by Bazilevs et al (2010). Several previous studies conducted at The University of Sheffield were of particular relevance for this project. The thesis by Vignaux (2011) looked into the application of FSI to an inflatable wing for micro air vehicles. It contains a discussion of the different options available in terms of the choice of the computational model. A fully coupled model is capable of solving the fluid and structural equations simultaneously but cannot be used easily on large meshes due to the high computational demands. In comparison a loosely coupled model solves the two sets of equations separately mapping the results through an interface of the fluid and structural models. The exchange of data between the two models usually takes place after each solution run converges. In a closely coupled solution the exchange takes place after every time step but requires remeshing after each step, which for large models can be a time-consuming process. Another recent mini-report by Chisman (2011) contains a valuable description of the mesh development process for a wind turbine CFD analysis using the ANSYS Workbench software package available at the university.

14

3.
3.1

METHODOLOGY
Introduction
The main aim of this investigation was to develop a methodology for the application of fluid-structure interaction analysis on wind turbine rotor blades. A large proportion of the project time was allocated to allow a full understanding of the processes and software which may be used for such an investigation. A number of preliminary studies were conducted and are detailed in the interim report for the project by Harrison (2011). The following chapter details the subsequent work in terms of research and development of an initial blade model and its simulation in the CFD and FEA solvers.

3.2

Selection of Blade Geometry


It was decided that a rotor blade design of radius greater than 20m should be used in the investigation. This would allow easier demonstration of the aerostructural coupling and its importance in wind turbine aeroelastic design. Initially there were some difficulties in sourcing suitable existing geometry or even raw data for wind turbine planforms and aerofoils. Although some planforms and their aerofoils were found, sourcing the actual aerofoil profile data to use in the model development was particularly difficult. This can be expected given the commercial sensitivity of most large modern wind turbines of the scale required in this project. A solution was found in the form of an old but rather better documented wind turbine design.

3.2.1 The Tvind Wind Turbine In 1975 a group of students and teachers from the town of Tvind in Denmark began constructing a large wind turbine in response to the escalating oil and gas prices following the 1973 oil crisis. As detailed in the paper by Carlin et al (2001) the final configuration was a three-bladed, downwind rotor, 54m in diameter mounted on a 53m high concrete tower. When construction had finished in 1978 it was the largest wind turbine in the world and was rated at 900kW (although it was capable of generating significantly more). It continues to operate today and its rotor has completed over 100 million revolutions.

15

Data for the planform of the blades of the Tvind turbine including aerofoil sections, twist and taper was taken from page.213 of the book by Le Gourirs (1982). Its blades used three different aerofoils form the National Advisory Committee for Aeronautics (NACA) 5-digit series of aerofoils. Since the time of the Tvind turbine construction many aerofoils have been developed especially designed for use on wind turbine blades. However, as discussed previously, profile data for such aerofoils is harder to come by in comparison to the 5-digit series which are freely available from numerous sources. Table.3.1. list the data used in the construction of the blade profile. Aerofoil Circular NACA 23035 NACA 23024 NACA 23012 Pitch Angle () N/A 36 10 -3 Chord (m) 1.40 2.13 2.13 0.70 Section Location r/R 0.000 0.113 0.347 1.000

Table.3.1. Tvind Wind Turbine Blade Profile Data

3.3

Selection of Computational Tools


Some of the third-party design codes mentioned in the literature review were considered for use in this project. The availability of some of the codes was limited and the commercial design packages were ruled out due to their cost. It was also desirable that a well established CFD solver was selected to determine the fluid flow field. Therefore the computational tools were selected from those available on The University of Sheffield research computing cluster ICEBERG. As of the 2011/12 academic year, the university made available ANSYS Workbench v.13, a platform which brings together well established CFD and FEA solvers in a relatively user-friendly project schematic user interface. The advantages of this platform include its support for multi-physics problems; its bi-directional support for all major CAD tools; and its integration of several mesh generation tools. Given these advantages it was decided that the project should be developed in this environment. The two different CFD solvers available for use in ANSYS Workbench are Fluent and CFX. Both are well known and well developed CFD solvers so the selection of the solver was based upon its support for FSI analysis. It was discovered that although both solvers could be used for one-way FSI analysis it was only the CFX solver which had built-in support for two-way FSI analysis at 16

the current time. Fluent would require the coding of a user-defined function to perform a two-way analysis which would require a lot of additional time to develop. Considering that CFX solver offered the option for future development of a two-way FSI analysis it was selected for use in this project. Out of the meshing options available in ANSYS Workbench the built-in ANSYS Meshing tool was selected. It is a generic meshing tool which may be adapted for generating either fluid flow simulation meshes or structural meshes using its physic preference settings. The use of the ICEM meshing tool was considered for the fluid domain mesh. It is more advanced then the ANSYS meshing tool and offers a number of features for creating highly structured meshes for CFD simulations. However, its use would have required significant time commitment in order to use it effectively and it was found that the ANSYS Meshing tool offered acceptable control over the accuracy of its meshes. The single option for use in the structural simulation was the ANSYS Mechanical FEA solver. Finally, SolidWorks was selected to create the CAD model. It is widely used piece of CAD software and is well supported by the ANSYS Workbench environment. A summary of the selected software and their uses within the project are given in Table.3.2. Software SolidWorks 2008 ANSYS Design Modeller ANSYS Meshing ANSYS CFX-Pre/CFX-Post ANSYS Mechanical Application in Project CAD modelling CAD integration and fluid domain modelling. Mesh generation for both fluid domain and structural model CFD simulation/post-processing FEA simulation

Table.3.2. Computational Tools Used in Project

3.4

Geometry and Mesh Generation


For reasons explained in a later section only one third of the rotor would make up the simulation models. Therefore, the model consisted of a single rotor blade and a 120 section of the hub.

3.4.1 Modelling the Blade Geometry A single blade of the Tvind wind turbine rotor was modelled in the SolidWorks CAD program. Sketch planes were created at each of the four section locations 17

along the blade given in Table.3.1. The corresponding blade cross-sections were then sketched on these planes with the associated twist and taper. No geometric data was found on the hub of the Tvind rotor so a fillet radius was applied to both the upwind and downwind section to give a typical rotor hub surface. This inaccuracy in the model should have little impact on the analysis given the minimal torque generated in this region of the rotor. The blade surface was created using a surface loft which would allow a thickness to be applied at the meshing stage. The hub was modelled as a solid 120 section around the axis of rotation before the application of the edge fillets. Although the hub was modelled as a solid body in SolidWorks only its surfaces were passed to the meshing software again to allow thickness specification at the meshing stage.

Figure.3.1. CAD Drawing of Tvind Wind Turbine Rotor Blade 3.4.2 Modelling the Fluid Domain Geometry The completed blade geometry was imported into ANSYS Design Modeller for the modelling of the fluid domain geometry. The fluid domain was created 18

using several blocks to allow control over the mesh refinement at the meshing stage. As this part of the process was based upon trial and error a number of different arrangements of these blocks were considered before the arrangement shown in Figure.3.2 was settled upon. It consists of the blade region from which the blade volume was cut; the transitional region made from block surrounding the blade region; and both an upstream and downstream block.

Figure.3.2. Fluid Domain Geometry Dimensions (side view) Careful attention was paid to the dimensions of the fluid domain geometry to ensure that the boundaries of the domain were an appropriate distance away from the blade region. The boundaries should be far enough away that the boundary conditions may be justified but not so far that the mesh is inefficient. The upwind region extends a distance equal to one blade radius (1R) ahead of the blade whereas the downwind region extends a distance equal to three times the blade radius (3R) behind the blade. 3.4.3 Meshing the Fluid Domain Once the fluid domain dimensions had been locked in the meshing process could begin. This part of the project was particularly time consuming as the correct balance of size controls and mesh methods needed to be ascertained in order to provide a mesh of suitable quality. Several trial meshes were run in the CFX solver in order to check the performance of the mesh in terms of mesh independence and computational efficiency. 19

The first stage was to set the global sizing controls. In ANSYS Meshing the default settings were set to CFD with CFX as the chosen solve. This automatically sets a number of preset global sizing controls including: Transition, Smoothing, and Span Angle Centre. Transition governs the rate of growth between adjacent elements which should ideally be as low as possible for a CFD mesh. It was increased slightly from its default to reduce the element count in the blade region which under default settings was in excess of 1million. The smoothing setting governs the number of smoothing iterations used in the meshing algorithm whereas the span angle centre governs the maximum angle of curvature which may be covered by a single element side. Both of these were set to their highest default settings (high and fine respectively) to provide the best accuracy. Finally, within the global controls the Advance Sizing Function was turned on. This feature seeks to refine the mesh based upon the degree of curvature of the surfaces, in this case around the blade surfaces placing the highest concentration of elements on the leading and trailing edges of the blade. However, even under the current settings the resulting mesh was considered too coarse. In order to refine the mesh, local mesh controls were applied to the different regions. A patch conforming tetrahedron method was used on the blade region and transitional region. Tetrahedron elements provide good matching to the relatively complex blade and hub geometry at the expense of an increased element count. A sweep method using hexahedral elements was used for the upwind and downwind regions with a bias placed towards the central regions. A summary of the methods and size controls applied is given in Table.3.3. Region Blade Region Transitional Region Upwind Region Downwind Region Method Patch Conforming (Tetrahedron) Patch Conforming (Tetrahedron) Sweep (Hexahedron) Sweep (Hexahedron) Max Size (m) 1.5 1.5 6.0 6.0 Bias 2.0 2.0

Table.3.3. Meshing Methods and Size Controls by Regions

20

Figure.3.3. Fluid Domain Mesh: Side View (top), Isometric View (bottom left), and Enlarged Blade Region (bottom right) Match controls were applied to all the faces with rotational symmetry as can be seen in the bottom right image of Figure.3.3. This would allow boundary conditions of rotational symmetry to be applied in the setup of the CFD solver. ANSYS Meshing also contains the Inflation feature which permits elements to grow in a specified number of layers from a surface. This was applied to the blade and hub surface where it would provide a better resolution of the boundary layer in the CFD simulation. Statistics for each of the mesh refinements were monitored and are listed for the final mesh in Table.3.4. The entire fluid domain contains 803629 elements with an average element quality of 0.94 (on a scale of 0 to 1), an average aspect ratio of 1.39, and importantly an average skewness of 0.29. As described in the ANSYS Meshing User Guide, the element skewness is a key measure of the quality of the mesh since it indicates how close an element is to its ideal shape. The guide states that average skewness values between 0-0.25 indicate an 21

excellent mesh and values between 0.25-0.5 indicate a good mesh. The average value of 0.29 was taken as an acceptable value for the final mesh. Histograms of the mesh statistics can be found in Appendix 1. Region Elements Nodes Element Quality (Average) 0.73 0.83 0.94 0.94 0.76 Aspect Ratio (Average) 2.28 1.89 1.40 1.39 2.17 Skewness (Average) 0.32 0.24 0.07 0.07 0.29

Blade Region Transitional Region Upwind Region Downwind Region Fluid Domain

662702 74058 16201 50668 803629

183170 15372 18198 54756 268032

Table.3.4. CFD Mesh Statistics by Region In summary the fluid domain mesh generation was a lengthy process using trial and error to find the best set of controls that give the mesh required. The final mesh had acceptable statistics despite being unstructured. 3.4.4 Meshing the Blade Structure In comparison the mesh generation for the blade structure was a much simpler process. The fluid domain was suppressed leaving just the imported blade surface geometry. All of the surfaces were grouped into a single part to ensure conformal meshing of the blade and hub. A uniform quadrilateral/triangle mesh method control was applied with a maximum element size of 0.075m. This produced a hollow blade structure of quadrilateral SHELL 181 elements. Figure.3.4 shows the final structural mesh with the associated mesh statistics given in Table.3.5.

22

Figure.3.4. Structural Mesh: Blade Surface (left); Hub (top right) and Blade Cross-Section (bottom right) Region Elements Nodes Element Quality (Average) 0.96 Aspect Ratio (Average) 1.05 Skewness (Average) 0.04

Blade Structure

20396

20219

Table.3.5. Structural Mesh Statistics

3.5

CFD and FEA Solver Setup


With the CFD and Structural meshes now finalized the next stage of the project focused on the modelling of the physics of the problem including the determination of appropriate boundary conditions. This stage of the project also involved creating an interface for the FSI analysis.

3.5.1 CFD Solver Setup The fluid domain mesh was imported into the CFX-Pre CFD pre-processor. As seen in the previous sections only one third of the rotor was modelled since it has a rotational symmetry with each blade having the same flow field if tower interference is neglected. Not only does this cut the simulation run time 23

significantly it also lends itself to the use of the Frozen Rotor Model (FRM). The FRM simulates the rotational motion of the blade by placing it in a secondary domain with a rotating frame of reference. For this model the Blade Region (see Figure.3.2/Figure.3.5) was set as the rotational domain with a Fluid-Fluid interface between the Blade Region and the Transitional Region. FRM analysis is one of several multi-stage models that may be used for modelling turbomachinery. It offers the advantage of being relatively simple to set up and being computationally efficient for steady-state problems. It is unable to resolve transient problems but does provide a good initial solution from which to run transient simulations if required. The boundary conditions applied to the fluid domain are illustrated in Figure.3.5. The leading upstream face was set to an inlet boundary with a uniform velocity profile. The far end downstream face was set to an outlet boundary with a uniform pressure profile with the ambient reference pressure of 1.0atm. The entire curved outer surface of the fluid domain was set as an outflow again with a uniform pressure profile set at 1.0atm. Unlike the inlet and outlet boundary conditions which only allow fluid to pass in and out of the domain respectively, the outflow condition allows flow out of and into the domain. To ensure the rotational symmetry of the model a periodic boundary condition was applied to the matched faces of the domain. This would mean any flow leaving one face would re-enter the domain at the other. Finally, a no-slip wall boundary condition was applied to the blade and hub surface. Additionally theses surface were given a rotational motion equal to that of the surrounding rotational domain (i.e. a relative velocity of zero).

Figure.3.5. CFX-Pre Setup for CFD Simulation with Boundary Conditions

24

CFX-Pre offers a number of predefined fluids for use in simulations and air at 25C was selected as the default fluid used throughout the model. The reference pressure was set to 1.0atm. The k- turbulence model was used to approximate the turbulence in the simulation. It is a well established turbulence model based on the Reynolds Averaged Navier-Stokes (RANS) equations. The Large Eddy Simulation (LES) and Direct Numerical Simulation (DES) turbulence models were considered to be too time-consuming to be implemented in this study. The k- turbulence model required the specification of a boundary condition in terms of turbulence intensity at each of the external boundaries. The values of k and are scaled by the turbulence intensity and velocity at the respective boundary. Low turbulence intensity (1%) value was used on all boundaries as this is most representative of external flows. Medium and high intensity values tend to be reserved for internal flows. CFX solves the Navier-Stokes equation in their conservative form using a coupled solver. In the solver control high resolution was selected for the advection scheme and the turbulence numerics to ensure the best accuracy in the solver. The convergence criterion was set to an RMS value of and

the maximum number of iterations set to 1000 which allowed all simulations to reach the convergence criterion. 3.5.2 FEA Solver Setup The structural model was setup using ANSYS Mechanical. Before any loads or boundary conditions were applied the material model had to be specified. In the paper by Carlin (2001) it states that the blades of the Tvind wind turbine were made from glass-fibre reinforced epoxy (GFRP). Properties of E-Glass GFRP were taken from the materials textbook reference by Callister (2003). A single constraint was placed upon the model using a fixed support at the base of the hub. This would allow the blade to freely deform under the imported aerodynamic loading which is discussed in the next section. The FEA simulation neglected inertial and gravitational loads because of the adoption of a steady state CFD analysis. This is a large simplification of the problem but 25

allows the independent analysis of the aerodynamic loading on the structural model providing the groundwork for future transient studies. 3.5.3 FSI Interface Setup As discussed previously ANSYS Workbench offers a user interface with a project overview which allows the setup of a fluid-structure interaction analysis. Both the CFD and FEA analysis are contained within their own system blocks. An FSI interface between the two blocks is created by linking the CFD solution cell with the FEA setup cell as shown in Figure.3.6. This allows the CFX Solver to output the results of the simulation to ANSYS Mechanical where the import load feature is made available. As the wall boundary used in the CFX setup conforms to the blade surface of the structural model ANSYS Mechanical is able to interpolate the results onto the structural mesh. This places a pressure loading on each element of the structural mesh thus representing the aerodynamics loading on the blade.

Figure.3.6. ANSYS Workbench Project Schematic for FSI Analysis

26

4.
4.1

VERIFICATION AND VALIDATION


Introduction
An important step in any computer simulation used in engineering analysis is to verify and validate the simulation and its results. This is due to the large number of sources for errors in both CFD and FEA simulations. Verification techniques check the implementation of a simulation, whereas validation techniques seek to ensure the results resemble the actual physics of the problem.

4.2

Simulation Verification

4.2.1 Residual Monitors The first stage of CFD simulation verification is to ascertain whether or not the governing equations have been solved i.e. if the solution has converged. CFD solvers usually contain a feature which allows the user to monitor the normalised residuals of the equations, an example of which is shown in Figure.4.1. This provides the best indication of solution convergence and may be specified in the solver control as described in the CFD Solver Setup section. It means that the user can effectively control the accuracy to which the solver will complete its solution run. Although specifying a low convergence criterion is likely to ensure a converged solution, it is important to verify this by running a simulation to an even lower criterion. To check the solutions of the simulations were independent of the convergence criterion a simulation was run at three different convergence criteria. It was found that the blade torque and velocity values did not change as convergence criterion was lowered indicating solution convergence.

27

1.00E+00

1.00E-01
Variable Value 1.00E-02 1.00E-03 1.00E-04 1.00E-05 1.00E-06 0 100 200 300 400 Iterations RMS P-Mass RMS U-Mom RMS V-Mom RMS W-Mom 500 600 700

Figure.4.1. Residuals Monitor for a CFD Solution 4.2.2 Point Monitors In addition to monitoring the residuals during each solver run, several point monitors were placed in the fluid domain. These were located at a 20m radius to the axis of rotation at distances 1, 5, 10 and 20m downwind of the rotor plane. The wake region within which the monitors are located will see the greatest change in the flow field with each solver iteration. By monitoring the absolute pressure and velocity at these points there will be another method by which to verify the solution has converged. An example is shown in Figure.4.2.
1.014E+05 Absolute Pressure (Pa) 1.014E+05 1.013E+05 1.013E+05 1.012E+05 1.012E+05 0 100 200 300 400 500 600 700 Iterations Monitor (1m Downwind) Monitor (10m Downwind) Monitor (5m Downwind) Monitor (20m Downwind)

Figure.4.2. Point Monitors (Absolute Pressure) for a CFD Solution 28

4.2.3 Mesh Independence A final important verification of the simulations concerns the mesh fidelity and its effect on the solution. Ideally with unlimited computing power a dense mesh with no limitation on the number of elements could be used to ensure the solution is mesh independent. However, in reality computing power is limited and the mesh quality was limited due to use of an unstructured mesh. For these reasons the solution was not expected to be independent of the mesh quality and it was accepted the solution accuracy could be improved. Despite this the mesh was of sufficient quality to provide qualitative results as demonstrated in the following discussion.

4.3

Results Validation

4.3.1 CFD Validation The CFD solution was validated by visualising the flow field of the first simulation using contour and vector plots in CFX-Post. In the example shown in Figure.4.3 the stagnation point on the lower leading edge and the resolved boundary layer can be seen both coloured in blue indicating regions of low velocity. Typical aerofoil section characteristics are observed as the flow is accelerated to a higher velocity over the suction surface and is coloured in red. Although the CFD solution could be validated qualitatively, the quantitative validation of the results was much more challenging. Fortunately data for the Tvind wind turbine was found in an old conference paper published in the International Symposium on Wind Energy Systems (1980). The power coefficient could be calculated from the data obtained from the CFD simulation and compared to the data shown in Figure.4.4. Details of how the power coefficient is covered in the next chapter. The results show the power coefficient has some degree of accuracy but is not truly representative of the real-world results. This is likely to be due to the simplifications and assumptions made in the simulation.

29

Figure.4.3. Velocity Contour Plot of Blade Section at 20m Radius

0.45 0.40 0.35

0.30
Cp 0.25 0.20 0.15 0.10 0.05 0.00 0.00 2.00 4.00 Real-World Data CFD Result 6.00 8.00 10.00

Figure.4.4. Power Coefficient vs. Tip Speed Ratio from CFD Simulation and Real World Study. Real-World Data taken from Paper Published in International Symposium on Wind Energy Systems (1980)

30

4.3.2 FEA Validation Validation of the structural model was achieved by comparing the maximum deflection value with an estimate calculated using classical beam theory. The second moment of area for an elliptical section was used to produce an estimated deflection for a 27m long blade under the flapwise bending moment result from the simulation.
Second Moment of Area (m4) 0.0143 Torque (Nm) 391392.00 Estimated Deflection (m) 0.0995 FEA Deflection (m) 0.0284

Table.4.1. Flapwise Deflection Estimated from Beam Theory and Calculated from FEA Simulation In a similar manner to the CFD validation the FEA simulation has produced a result of reasonable accuracy and at least of the correct order of magnitude. In both cases the discrepancies can be attributed to the simplifications used in the models and possibly at some degree to the simulations themselves. This was expected as the main focus of this investigation was to develop an understanding of the FSI process rather than quantitative accuracy.

31

5.
5.1

RESULTS AND DISCUSSION


Introduction
The results in the FSI analysis included data from both the CFD and FEA solutions. A discussion will be made on both set of results and the relationship between them. The main focus of the discussions will centre on the physical accuracy of the analysis rather than the quantitative accuracy for the reasons described in the previous section.

5.2

CFD Results
This section will focus on the physical flow features observed in the fluid domain solution and their impact on the aerodynamic loading of the blade. It will also be shown how post-processing can be used to evaluate the efficiency of the turbine rotor. Although several simulations were run the description of the physical flow features present results from one simulation since the basic features were common to all results.

5.2.1 Physical Flow Features Pressure contour plots were used to visualise how the pressure distribution varied along the span of the blade. As shown in Figure.5.1 the sections closest to the tip of the blade give the greatest drop in pressure over the suction surface. Close to the root of the blade there is little variation in pressure suggesting that there is little power produced from this region of the blade. This distribution is as expected with the aerodynamic load increasing from a minimal value at the root to a maximum close to the blade tip. Another feature of particular interest is what happens to the flow in the wake of the rotor. The use of periodic boundary condition in the CFD Simulation allowed the full rotor to be visualised using a periodic transform of the solution in the CFX post-processor. The wake was visualised using a velocity contour plot on a plane located 5m downwind of the rotor plane as shown in Figure.5.2. It shows regions of relatively slow moving air swept by the blade. Averaged over the entire plane indicates an overall decrease in wind speed demonstrating the extraction of kinetic energy by the rotor.

32

Figure.5.1. Pressure Contour Plots on Sections Located 5, 10, 15, 20 and 25m from the Axis of Rotation

Figure.5.2. Velocity Contour Plot of the Wake on a Plane Located 5m Downwind of the Rotor Plane

33

Figure.5.3. Pressure Variation along the Centreline of the Fluid Domain Furthermore, the basic principle of energy extraction for a wind turbine can be shown to be true using a plot of the pressure along the centreline of the fluid domain as shown in Figure.5.3. It shows a step drop in pressure over the plane of the rotor (located at Z=0) matching the description given by the RankineFroude Actuator Disk Model shown in Figure.2.2. 5.2.2 Wind Turbine Efficiency Five simulations were run at different TSRs in order to gauge its influence on the efficiency of the rotor. This was achieved by holding the rotational speed of the rotor constant and varying the wind speed. Torque around the axis of rotation generated by the aerodynamic load on the blade was extracted using the

34

function calculator in CFX-Post. Power output of a single blade could therefore be calculated using Equation 5.1, and for the entire rotor using Equation 5.2. 5.1 5.2 The expression for the coefficient of pressure given in Equation 2.3 could then be used to give an indication of the rotor efficiency. The results for the different TSRs are shown in Figure.5.4.
0.35 0.30 0.25 0.20 Cp

0.15
0.10 0.05 0.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00

Figure.5.4. Power Coefficient of the Rotor Operating at Different TSRs The graph suggests that the rotor can extract the kinetic energy from the wind most efficiently whilst operating at a TSR of approximately 5.7. It also shows how efficiency is lost at lower TSR due to the blades not travelling fast enough to capture as much energy as possible from the wind. Equally at TSRs greater than the optimum each blade is travelling too fast and suffers losses as it begins to pass through the turbulent wake from the previous blade. Figure.5.4 highlights the importance of selecting the appropriate operating speed and gearbox design for the wind turbine system. 5.2.3 Wind Turbine Power Output A second sequence of three simulations was also conducted to investigate the effect of wind speed on power output. A constant TSR of 6 was maintained by increasing both the wind speed and rotational speed by appropriate amounts. 35

500 450 400 350 300 Protor 250 (kW) 200 150 100 50 0 0.00 2.00 4.00 6.00 Wind Speed (m/s) 8.00 10.00 12.00

Figure.5.5. Power Output for Different Wind Speeds at a Constant TSR Power output was calculated for the rotor as described above and the results were as expected. Higher wind speeds produced the most torque on the rotor blade thus providing greater power output.

5.3

FEA Results
The results of the FEA analysis are presented in the form of deflection and stress plots of the whole blade. It should once again be noted that the model only considers the aerodynamic loading provided by the CFD solution and does not include centrifugal or inertial loads.

5.3.1 Blade Deflection The Ansys Mechanical post-processor allows the deflected model to be displayed alongside the undeformed (unloaded) model of the blade. Figure.5.6 illustrated the blades deflection scaled up by a factor of 50.

36

Figure.5.6. Deformed and Undeformed Blade Model: Side View (top left), Top View (top right), and Leading Edge View (bottom) The blade acts in a conventional fashion with the greatest deflection occurring at the tip. There is also some degree of twisting of the blade structure which is also typical of a large wind turbine blade. Unfortunately due to the large simplifications made in the structural model the deflection magnitude is not as large as would normally be expected. For all simulations the maximum deflection was in the range of 2 6cm. A clear correlation was found between the aerodynamic load and the deflection produced in the blade as shown in Figure.5.7. Higher wind speeds not only produce greater torque and power output but also work to bend the blades in the direction of the wind. This is also demonstrated in Figure.5.8 which shows more energy goes into bending the blades at lower TSRs. 37

30.0000 Maximum von Mises Stress (MPa) 25.0000 20.0000 15.0000 10.0000 5.0000 0.0000 0.00 5.00 10.00 Wind Speed (m/s) Max Stress (MPa) Max Deflection (m) 15.00

0.060 Maximum Deflection (m) 0.050 0.040 0.030 0.020 0.010 0.000 20.00

Figure.5.7. Deflection of Blade Structure and Associated Stress due to Increasing Aerodynamic Load

0.060
0.050 0.040 0.030 0.020 0.010 0.000 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00

Maximum Deflection (m)

Figure.5.8. Deflection of Blade Structure at Different TSRs 5.3.2 Blade Stress Deflection of the blade structure places a strain on its skin leading to stress concentrations in certain areas. ANSYS Mechanical is able to evaluate the stress in a number of ways, although the most common method used in engineering analysis is to plot the von Mises stress. This plot would allow design engineers to determine if any stress concentrations would result in the material yielding 38

and potentially leading to failure of the structure. Figure.5.9 shows the von Mises stress plot for the blade under the same loading and deformation shown above. High stresses occur at the root of the blade particularly at the joint with the hub. Typically these regions are strengthened significantly in wind turbine blades to transmit the torque effectively. Again given the major simplifications made little interest was taken in the stress values; the main focus in demonstrating a working FSI solution. Figure.5.7 also shows the stress increasing with the applied aerodynamic load and importantly in proportion to the deflection in the blade.

Figure.5.9. von Mises Stress Plots: Side View (top left), Front View (top right) and Leading Edge View (bottom) 39

6.

CONCLUSION
The project succeeded in developing an FSI methodology for use in wind turbine analysis. A one-way FSI study was conducted on the Danish Tvind wind turbine and produced results which were used to assess its aerodynamic and structural performance. It was shown how the physics of wind turbine operation naturally lead to a coupling between the aerodynamic and structural behaviour. As larger wind turbines are developed there will be a requirement for more detailed and stringent analysis of this coupling. The advantages of using FSI analysis include: a more accurate solution of the aerodynamics compared with traditional methods; integration of both aerodynamic and structural engineering disciplines; and streamlining of the final design and analysis process. Its disadvantages are due to the complications involved in implementing the analysis. Particularly the need for an accurate mesh for the CFD simulation and the demands it creates in terms of computing power. It is perhaps a method most suited to the analysis of the final design of a wind turbine rather than the design development phase due to the time it takes to accurately set up and run an FSI simulation. The results demonstrated the coupling of the fluid flow and the reaction of the structure. They highlighted the requirement to integrate the analysis of both factors and not treat them as separate individual problems. The simulation results gave reasonable accuracy for the efficiency and performance of the turbine, matching real-world data. However, due to lack of available data and time the large simplifications made in the structural model had a significant impact on the accuracy of the structural characteristics. Deflections and stresses were judged to be too low for the applied aerodynamic loading.

6.1

Future Work
Due to the time constraints of the project an unstructured CFD mesh was adopted. With additional time a more accurate structured mesh would be made for the fluid domain using specialist meshing software such as ICEM CFD or Pointwise. 40

The results demonstrated the inaccuracies in the structural model. Future studies would seek to acquire more detailed information regarding blade structure with the aim to develop an accurate blade model. Perhaps a collaborative project with industry would be the ultimate goal in securing reliable sources of information in the development of the blade model. The next stages for the development of the FSI method would seek to include gravitational loads and inertial loads to fully demonstrate the abilities of FSI analysis. This would require the use of a transient two-way FSI approach where the deformation of the blade is mapped back onto the CFD mesh. After each time-step both the CFD and FEA simulations would pass the relevant data to each solver in an iterative process until both solvers are satisfied. Clearly this would be highly intensive computationally and extra attention should be paid to the accuracy of the simulation setup.

41

7.

REFERENCES
Ahlstrom, A., 2005. Aeroelastic Simulation of Wind Turbine Dynamics. PhD Thesis, Royal Institute of Technology, Stockholm, Sweden. ANSYS Inc, 2010. ANSYS Meshing User Guide. [online] Available at: http://www1.ansys.com/customer/content/documentation/130/wb_msh.pdf [12/04/2011]. Bazilevs, Y., et al, 2010. 3D Simulation of Wind Turbine Rotors at Full Scale. Part I: Geometry Modelling and Aerodynamics. International Journal for Numerical Methods in Fluids, vol. 65, pp. 236-253. Bazilevs, Y., et al, 2010. 3D Simulation of Wind Turbine Rotors at Full Scale. Part II: Fluid-Structure Interaction Modelling with Composite Blades. International Journal for Numerical Methods in Fluids, vol. 65, pp. 236-253. Callister, W.D., 2003. Materials Science and Engineering: An Introduction. 6th ed. Chichester: Wiley. Carlin, P.W., et al, 2001. The History and State of the Art of Variable-Speed Wind Turbine Technology. NREL/TP-500-28607, National Renewable Energy Laboratory. Chisman, L., 2012. Using ANSYS Engineering Simulation Software to Validate Design Assumptions in Sustainable Energy. E-Futures Mini Project Report, The University of Sheffield. Hansen, M.O.L., 2000. Aerodynamics of Wind Turbines. London: James & James. Hansen, M.O.L., et al, 2006. State of the Art in Wind Turbine Aerodynamics and Aeroelastics. Progress in Aerospace Sciences, vol. 42, pp. 285-330. Harrison, M.L., 2011. Interim Report: The Determination of the Structural Loads on Wind Turbine Rotor Blades, MEng Aerospace Engineering Individual Investigative Project, The University of Sheffield. Hartwanger, D., and Horvat, A., 2008. 3D Modelling of a Wind Turbine Using CFD. United Kingdom: NAFEMS Conference. 1980. International Symposium on Wind Energy Systems. 3rd ed. Cranfield: BHRA Fluid Engineering. Kim, D-H., and Kim, Y-H., 2011. Performance Prediction of a 5MW Wind Turbine Blade Considering Aeroelastic Effect. World Academy of Science, Engineering and Technology, vol. 81, pp. 771-775. Le Gourieres, D., 1982. Wind Power Plants, Oxford: Pergamon.

42

Maheri, A., et al, 2007. Combined Analytical/FEA-based Coupled Aero Structure Simulation of a Wind Turbine with Bend-Twist Adaptive Blades. Renewable Energy, vol. 32, pp. 916-930. Spera, D.A., 1994. Wind Turbine Technology: Fundamental Concepts of Wind Turbine Engineering. New York: ASME Press. Strahan, D., 2008. The Last Oil Shock: A Survival Guide to the Imminent Extinction of Petroleum Man. London: John Murray Publishers. Vignaux, C., 2011. Developing a Methodology for Modelling an Inflatable Wing. PhD Thesis, The University of Sheffield. Walker, J.F., 1997. Wind Energy Technology. Chichester: John Wiley & Sons.

43

Appendix 1: CFD Mesh Statistics Histograms


Element Skewness

Element Quality

44

You might also like