You are on page 1of 140

Probabilistic Dynamic Analysis of Offshore Wind Turbines under Random Wind and Wave Loading

Seyed Hesameddin Jamei


MSc Structural Engineering

University of Surrey Academic year 2010-11

Dissertation Supervisor: Professor Marios Chryssanthopoulos

Abstract
Offshore wind turbines are a great engineering solution to provide higher efficiency in producing electricity from wind. Due to the offshore conditions the loads induced to these structures are usually complex and require sophisticated engineering approaches to evaluate the structural responses. In this study an alternative method of incorporating probabilistic modelling of wind and wave loads into dynamic analysis of the support structure of an offshore wind turbine is presented. Accordingly a carefully selected modelling has been developed based on a moderate dynamic idealization which has been absent in other studies. The method is then accompanied by mathematical substrates which will be executed using suitable programmes developed in MATLAB. Consequently the results will be computed and the corresponding statistical properties such as the probability density function and probability distribution function of the maximum short term responses will be found. Finally the applicability of the results will be discussed and methodologies by which the secondary response can be achieved will be presented.

Contents 1 Introduction ............................................................................................................................. 3 1.1 History of Wind Turbines ................................................................................................ 3 1.2 Offshore Wind Turbines .................................................................................................. 6 1.2.1 Concepts and History ................................................................................................ 6 1.2.2 Main Components and General Typology ................................................................ 9 1.2.3 Construction of Monopile Offshore Wind Turbines ............................................... 12 1.2.4 Floating Offshore Wind Turbines ........................................................................... 18 1.3 Conclusions and Study Outlines .................................................................................... 21 2 Literature Review.................................................................................................................. 24 2.1 Various Mechanical Modelling by Different Researchers............................................. 24 2.2 Reliability Related Researches ...................................................................................... 29 2.2.1 Offshore Wind Farms and Turbulence Model for Fatigue Reliability (From the work of Srensen et al. 2008) .......................................................................................... 30 2.2.2 Long Term Fatigue Analysis of Multi-Planar Tubular Joints for Jacket-Type Offshore Wind Turbine (from the work of Moan et.al 2011) .......................................... 31 2.2.3 Long-term Fatigue Analysis of Offshore Fixed Wind Turbines based on Timedomain Simulations (from the work of Geo and Moan 2010) ......................................... 32 2.3 General Design Criterions for offshore Wind turbines (based on Det Norske Veritas 2004) .................................................................................................................................... 33 3 Probabilistic Dynamic Analysis of Offshore Wind Turbine Tower Structure ..................... 38 3.1 Analysis.......................................................................................................................... 38 3.1.1 General Analysis Strategy....................................................................................... 38 3.1.2 Probability Concepts ............................................................................................... 47 [1]

3.1.3 Structural Dynamic Concepts ................................................................................. 57 3.1.4 Detailed Analysis Approach ................................................................................... 65 3.1.5 MATLAB Code Description................................................................................... 74 3.2 Validation of Results and Methods ................................................................................ 87 3.3 Analysis Results ............................................................................................................. 95 3.3.1 Main results ............................................................................................................. 96 3.3.2 Initial results based on an inaccurate time stepping assumption .......................... 102 3.4 Discussion of Results ................................................................................................... 105 4 Conclusions ......................................................................................................................... 109 5 References ........................................................................................................................... 111 6 Appendices .......................................................................................................................... 118 6.1 MATLAB Code for the Main Analysis ....................................................................... 118 6.2 MATLAB Code to check the Implementation of the average acceleration method .... 123 6.3 MATLAB Code regarding the initial analysis based on inaccurate assumptions ....... 124 6.4 Simulation of Wind and Wave Forces Results ............................................................ 129 6.4.1 Wave Force vs. Time ............................................................................................ 129 6.4.2 Wind Force vs. Time............................................................................................. 130 6.4.3 Displacements vs. Time ........................................................................................ 131 6.4.4 Absolute Displacements vs. Time......................................................................... 133 6.4.5 PDF and CDF for 2 days hourly maximum displacements .................................. 135

[2]

1 INTRODUCTION 1.1 HISTORY OF WIND TURBINES

Wind can be described as air in motion. The motion of the air occurs due to uneven heating of the atmosphere from the sun. Such differences in temperature can produce pressure variations hence the motion of the air from one point to another or Wind (Simiu and Scanlan 1986). The presence of wind has imposed so many constraints to the world of structural engineering. It has a very important effect on the design of tall structures and can apply various actions locally and globally to any specific structure (Yang 2008). That is why national codes of structural design in different countries around the world usually have meticulous sections with regards to the consideration for the wind induced actions on the buildings and structures. Apart from these constraints, the energy from wind has been harvested by humans in order to assist him in different activities. Windmills can be considered as one the earliest attempts by man to use wind energy. According to Gipe (1995) the first windmill design goes back to 600 900 A.D. in ancient Persia. Such machines were based on a vertical axis system which worked by a dragging force produced by the wind movement:

Figure 1.1 Early Persian wind devices (Caroline Mawer 2011)

[3]

These sophisticated devices were used to grind wheat and other grains (Cheremisinoff 1979). The concept was brought to Europe and the earliest vertical axis windmills as described by White (1962) goes back to the early part of 12th century. Further improvements were made to the design of the windmills and horizontal axis windmills were introduced in the later part of 12th century in England. Windmills are referred to as the basis for wind turbines as wind turbines could be considered as windmills that convert wind energy into electricity (Cushman-Roisin 2011). According to Hardy (2010) one of the earliest attempts to use wind turbines to convert wind energy to electricity is known to be carried out by Professor James Blyth from Scotland in year 1887. His design was based on a vertical axis system and was developed to produce electricity in order to charge a storage device in his holiday cottage in Marykirk (Hardy, 2010). As it has been portrayed by Wyatt (1996) the development of the first horizontal axis wind turbine is undertaken in Yalta, USSR in the year 1931 and the design was based on a 100kw generator mounted on a 30 meter tall tower. Finally the first grid-connected wind turbine was constructed in 1951, Orkney Island, UK (Orkney Sustainable Energy 2010). Modern wind turbines are categorized based on their axis orientation. In general there are vertical axis wind turbines and horizontal axis wind turbines (Cushman-Roisin 2011). The following figures demonstrate some examples of them:

Figure 1.2 A vertical axis wind turbine (Energy Types 2011)

[4]

Figure 1.3 Horizontal Axis wind turbine (Generator blog 2011)

Furthermore wind turbines are categorized on the basis of their location of construction, onshore and offshore. The concept of producing energy is the same for both structures but they will be clearly different in design and construction. The following figure demonstrates examples of offshore Wind Turbines in an offshore wind farm:

Figure 1.4 Off shore Wind Turbines (My Eco Project 2010)

There also have been so many developments in introducing some very modern concepts in terms of the general design of wind turbines. The new generation of wind turbines can look different from what has been seen so far. For instance, the following is called Aerogenerator X which is planned to be built offshore and has been originally proposed by [5]

a British company called Wind Power Limited. This new design is anticipated to produce 1GW of power by the year 2020 (Firth 2010):

Figure 1.5 - Aerogenerator X conceptual Design (Wind Power 2010)

1.2 OFFSHORE WIND TURBINES

1.2.1 CONCEPTS AND HISTORY

Offshore construction has experienced a great deal of growth recently (Tan and Wichers 1990). This can be due to the major advances in engineering and construction technologies regarding such environments. Generally constructing an offshore structure can be costly and requires a great deal of sophistication. However building offshore can have some key benefits which can make up for the complexity of the design and construction. For instance, the structures that require a large amount of space and are difficult to be facilitated in land can be constructed offshore therefore certain advantages are achieved (Madsen 2010). Furthermore as it will be seen later, having some structures onshore can have loss of scenery and can damage the space arrangements of the built environment. Thus they tend to be built offshore to reduce this problem. As it was described in the above constructing certain structures offshore can have a positive impact on the built environment. This is also the case for wind turbines. Wind turbines are usually large structures and they need a considerable amount of space to function properly. It [6]

is common that wind turbines being built and placed in large numbers are known as wind farms (Nacfaire 1988). An example of a wind farm is shown below:

Figure 1.6 Example of an offshore Wind Farm (maritime journal 2011)

Wind farms require a large area of land and they habitually cannot be erected inside tight city spaces. Furthermore having a wind farm inside a city can have some noise pollution issues for the people living in the vicinity (Cushman-Roisin 2011).This can potentially make the planning more confined and thus can affect the energy production of a wind farm. Additionally, other problems can be present with regards to the loss of scenery in city environment similar to what was described in the above. These confinements in the onshore environment can make the use of offshore wind turbines more attractive in comparison to the onshore wind turbines. However as it was explained earlier, generally there is a much higher construction cost with regards to the offshore structures. This is indeed the case for offshore wind turbines too (Europes Energy Portal 2009). These structures are more difficult to build in comparison to their onshore counter parts. The higher cost can be due to the general idea of building in the sea which means the presence of waves and also in the case of shallow water offshore wind turbines, the higher cost of founding the structure on the sea bed (Robinson and Musial 2006). Furthermore the design of the structure will be more complex which will require a sophisticated technical approach (Camp et al. 2003). The general idea and pros and cons of building off shore wind turbines were described in the above. The main benefit of offshore wind turbines however can be considered to have better [7]

efficiency in electricity production. According to Ragheb (2010), there is higher wind velocity in the ocean giving the offshore wind turbine a higher advantage. This is the key concept which in fact makes them more preferred over their onshore counterparts. As it has been described by Henderson et al. (2003) the advantages and disadvantages of offshore wind turbines are summarized below: Advantages: i. ii. iii. iv. v. Availability of a larger area Higher wind velocity which increases the further they are from the coast Less turbulence in wind and thus higher efficiency in harvesting wind energy Less turbulence also means there will be a lower possibility of fatigue in the structure Wind shear reduces as moving further away from the shore and thus a shorter tower structure will be needed Disadvantages: i. ii. iii. More costly foundation More expensive to be connected to the national grid Higher maintenance costs

The Environmental and Energy study Institute (2010) suggests that the first offshore wind turbines were installed in Denmark in year 1991. This can be considered as the first practical step toward the commercial usage of these structures. With regards to the UK, the first offshore wind turbines were installed in the year 2003, 7 to 8 km off the off the north Wales coasts (Nixon 2008). According to Wind Tech International (2010) as of the year 2010, the UK has the highest offshore production of electricity from wind in the world with the total value of 1.3 GW. The same study shows that the rest of the worlds offshore electricity production is about 1.1GW. However according to Renewable Energy World (2011) the construction of offshore wind turbines in the European regions has increased by 51% in the year 2010. This demonstrates the importance of this method of energy production in the short future.

[8]

1.2.2 MAIN COMPONENTS AND GENERAL TYPOLOGY

As it was described in the above there are some major benefits of using offshore wind turbines in terms of energy production performance and impacts on the built environment. In this part the general typology of offshore wind turbines and their main mechanical components will be explored. Being an offshore structure means the structure will be built in the sea. But it is important to understand the actual position of the structure in terms of the water depth and the sea environment. It can be understood based on the work of Musial and Butterfield (2004) offshore wind turbines are categorized based on the depth of the water in which they are built: i. ii. iii. Shallow water offshore wind turbines Transitional Water Offshore wind turbines Deep Water offshore wind turbines

The following figure demonstrates the general structural arrangements for each of the above categories:

Figure 1.7- Different types of offshore wind turbines (Musial and Butterfield 2004)

The above figure demonstrates the general structural arrangements of an offshore wind turbine based on the depth of the water in which its built. It is observed from the figure that

[9]

as the depth of the water is increased some additional structural members are attached in order to improve the behaviour. According to the Det Norske Veritas (2004) there are some notable arrangements with regards to the supporting structures of offshore wind turbines some example of which have presented in the following figures. Alternative structural systems have been shown for the same water depth based on various engineering requirements:

Figure 1.8 a

Figure 1.8 b

Figure 1.8 c Figure 1.8-Some arrangements of shallow and transitional water wind turbines: a) Left: Jacket monopile hybrid, Right: Harvest jacket, b) Left: Monopile, Right: Supported Monopile, c) Left: Suction Bucket, Right: Lattice tower (Det Norske Veritas 2004)

[10]

And some main arrangements for deep water offshore wind turbines shown in the following figure:

Figure 1.9-Some arrangements of floating wind turbines: Left: Floater, Right: Tension leg platform

It is seen from the above that there can be so many structural systems for the supporting structure of an offshore wind turbine. Thus it is important to choose a suitable decision in opting for a structural system from which the highest engineering efficiency is achieved. On the other hand other factors also can be involved in the process of selecting a suitable structural system such as cost and construction simplicity based on the environment that the structure is planned to be erected. For instance the seabed profile can become a very influential parameter which can be suitable for a certain type of support structural system. In this paper mainly the fixed monpile offshore wind turbines are studied. The example of such a structure and its corresponding main components is illustrated in the following figure:

[11]

Figure 1.10-Different Components of a Monopile offshore wind turbine (Der Tempel 2006 p. 8)

The above figure will be used as a basic source of terminology of each of the structural parts. Especially when it comes to the analysis of a chosen structure where a wind turbine with given dimensions and a specific water depth is used.
1.2.3 CONSTRUCTION OF MONOPILE OFFSHORE WIND TURBINES

Design and analysis of any structure will require a great deal of understanding regarding different aspects of the life cycle of the structure. That means one needs to appreciate the different loading cases applied to the structure during the construction stages as well as the service stage of any particular structure. On that basis it is useful to understand the major construction stages for a typical offshore wind turbine. Furthermore to appreciate the critical financial aspects of the structure it can be beneficial to study the general construction procedure of it. This is because although learning the structural behaviour of an offshore wind turbine is very important in terms of a structural engineering point of view the whole structure can never be realized if it is not financially justified. This also provides a better [12]

consideration for financial characteristics of the structure at the design stages so that the structure will become much more financially optimised. In addition to those, exploring the different constructional stages can be extremely useful taking into account the potential hazards that can be present. This can also assist the designer to contemplate a suitable risk mitigating solution to avoid the possibility of such occurrences. Although there are many different types of offshore wind turbines, in this paper the general construction procedures of the monopile type offshore wind turbines are described. The monopile offshore wind turbines are constructed in shallow waters usually up to 30 meters deep. That means the structure can be fixed on the sea bed as there will be easier access available. Typically the construction of these types of structures starts with installation of the monopile foundation into the sea bed. According to Bilfinger Bergerm (2008) the foundation is divided into two main sections: Monopile Transitional Section

The project installation report published by Bilfinger Bergerm (2008) has explained the construction of Horns Rev 2 offshore wind farm monopile foundations. Considering this as a typical construction procedure for the foundation of a monopile offshore wind turbine, it has been used as a source of information in this paper. According to the above report the construction of the monopile foundation begins with the fabrication of steel parts. The steel tubes are fabricated uniquely for each of the wind turbines foundation. That means although the wind turbines of an offshore wind farm looks the same the pile foundation structures differ and this can have potential extra costs for the fabrications. Furthermore the transitional tower usually needs to have a platform attached to them and they can be constructed from steel or concrete. However according to Bilfinger Bergerm (2008) report, for the case of Horns Rev 2 offshore wind turbines concrete platforms are used. This is because the concrete is able to withstand healthier against the extreme corrosion and abrasion in an offshore environment. Furthermore it can be added that as the offshore wind turbines are built in less accessible environments the concrete can be beneficial as there will be less life time costs due to the maintenance of the structure. The platform then is grouted to the transitional section and is tested before being sent to the offshore site. The following photo shows the transitional parts and concrete platform intact:

[13]

Figure 1.11-Transiotional parts onshore (Bilfinger Bergerm 2008)

The above figure demonstrates the transitional piles connected to the concrete platforms. The importance of doing this operation onshore in an open space is because the large cranes can be utilized. It is also essential from a structural engineering point of view that the designer takes into account the wind stability of these structures onshore and to provide a safe solution to achieve the highest safety during the construction. It is also important to attain the highest level of safety for the way concrete platforms are connected into the transitional parts of the monopile foundations. Once the parts are built they are transported to the offshore site where the wind turbines are to be installed. The critical point is to appreciate the uniqueness of each monopile in terms of their general sizes and to allow suitable space for them in the transport vessels. It is essential to follow the planned pathway to the site based on the project schedule. The following figure illustrates the tubular members on an appropriate vessel for the offshore wind farm Horns Rev 2 (Bilfinger Bergerm, 2008):

[14]

Figure 1.12- Transitional Parts being loaded on the vessels (Bilfinger Bergerm 2008)

According to the fact sheet published by Scira Offshore Energy (2010) the general construction of the giant monopile structures and their transitional pieces is carried out by tubular structure fabrication specialists. It is important to arrange the construction and transportation of the steel members to the offshore site based on the suitable design and health and safety specifications. Once the steel parts are transported to the offshore site using a suitable form of transportation the giant monopiles have to be positioned into the seabed using applicable machinery. The report of the construction process published by Scira Offshore Energy (2010) has depicted that in order to execute the construction stages efficiently it was required to have a specific plant called the Svanen Crane. This crane is a self-powered heavy duty floating machine which was employed to insert the monopiles into the seabed as well as mounting each of the transitional pieces on the top of them. For that specific site the monopiles usually were inserted into the seabed for about 23 to 37 meters. The total height of each monopile was between 44 to 61 meters while the transitional pieces were around 22 meters (Scira Offshore Energy 2010). The sizes provided can give some general estimate to the dimensions of each component and proportion used for the foundation of an offshore wind turbine. Finally it is described by Bilfinger Bergerm (2008) that the foundations are completed by using the covering layer of scour fortification. The same work also explains that it was important to carry out the offshore works during the summer which was the original schedule of the construction for the case of Horns Rev 2. According to the video published by Seajacks UK (2010) once the construction of the monopiles and transitional pieces is completed the supporting tower will be installed on the [15]

top of each transitional piece on which they are connected using bolts. It is important to minimize the dynamic effect of waves during the construction process of the tower particularly at the stages where the precision lifting of the parts are required such as the final stage of assembling the blades. The effects of wind on the tall tower sections must be realized and the corresponding risk must be mitigated by considering a suitable engineering approach to stabilize these parts during the construction stages. The tower sections are transported and bolted at offshore locations where the existing transitional pieces imply the positioning of each of the wind turbines. It is essential to note that once the construction parts are transported to their corresponding planned installation positions the work must be completed for the sections that are present on the site which means a continuous working schedule for the construction team (Seajacks UK 2010). The following figures show the handling of the tower onshore before is taken to the site:

Figure 1.13-Handling the Tower onshore (Seajacks UK 2010)

[16]

An example of the installation process of an offshore wind turbine tower on a transitional part using a crane and guidance from the site crew is shown below:

Figure 1.14-Towers being positioned on the top of a transitional part (Seajacks UK 2010)

Finally once the supporting towers are installed the rotor and blades have to be transported to the site. Nacelle is mounted at the top and is fixed correctly into position using suitable bolts. The most precise assembling will be the lifting and instalment of the blades into the Nacelle. The procedure is executed slowly using cranes and the construction team guiding the bolts into the designed positions. It is very important that during the installation of the blades the effect of the wind and its swinging effects are taken into account and the construction works are executed based on suitable weather conditions (Seajacks UK 2010). The following demonstrates the above:

Figure 1.15 (a)

[17]

Figure 1.15 (b) Figure 1.15-Installing the blades: a) View from the bottom whilst the blades are connected. b) Close view of blade installation (Seajacks UK 2010)

1.2.4 FLOATING OFFSHORE WIND TURBINES

There has been extra emphasis on the monopile offshore wind turbines as these types have been used in practical modern applications. This is due to that fact that they are built in shallow water depth and their construction and design have been widely matured by the engineers and construction companies. However Lackner and Rotea (2010) suggest that there have been some recent motivation regarding the construction of offshore wind turbines in a deep sea environment. That means the general structural arrangements will be different to shallow water wind turbines and there will be major differences in the construction process. According to Heronemus (1972) the idea of a floating wind turbine was originally presented by professor William E. Heronemus at the University of Massachusetts in the year 1972. Furthermore according to Siegfriedsen et al. 2003 floating offshore wind turbines can be considered to provide the most efficient and most economical solutions. This however can be argued by undertaking an in depth gain-loss analysis in order to consider the influential factors.

[18]

Butterfield et al. (2005) describe that floating offshore wind turbines have some benefits and some draw backs in comparison with the current popular monopile type structures: Advantages:

No need to construct an underwater foundation Higher exposure to offshore wind Better public acceptance in terms of the loss of scenery as the machines are further away from the shore Lower environmental impacts

Disadvantages:

Greater distance from the land and thus additional cost for the power distribution system Additional cost of the floating platform More complexity in design

Furthermore according to Henderson et al. (2002) the major technological challenges for such structures will be as below: Large Dimensions: as the structure will be quite large that means large amounts of materials will be required which can potentially leads to an uneconomical structure. Stability: it is essential to appreciate the combination of turbine thrust and overturning moment to the structure and provide an optimised solution to mitigate it. Wave Loads: these can be high and requires large structural members to be able to resist them. Resonance: it is important to analyse and design the structure for the effect of resonance which can cause fatigue failure to the structure. Anchorage: these elements are very expensive and to optimise their applicability is a great challenge. Some of the parameters in the above can become magnified in accordance to a specific projects financial and environmental constraints. According to Henderson et al. (2002) another major benefit of using floating offshore wind turbines is its independence from the seabed conditions as some countries such as Japan has a limited area of shallow water which can be suitable for monopile offshore turbines. [19]

The analysis of floating wind turbines can become complex without having knowledge of the possible structural arrangements. Reviewing the works done by different researchers and engineers, one can observe a large range of conceptual propositions with regards to the floating wind turbines. Clearly the most optimised concept is the one that provides the best solution in terms of performance efficiency, structural behaviour, economical factors as well as the construction procedure and technologies. Again these factors can become more critical on the basis of a particular project. As it is presented by Butterfield et al. (2005) there have been three major conceptual designs with regards to floating offshore wind turbines. These are mostly categorized on basis of the structural solution by which they achieve their stability in a remote deep sea environment: Type 1 Ballast Stabilized: in this type the platforms gain their stability using a ballast weight hung underneath a central buoyancy tank. This will provide correcting moment and good inertia resistance against the overturning and destabilizing actions. Type 2 Mooring Line Stabilized: the stability is achieved using tension legs which are used to counter the destabilizing actions. Type 3 Buoyancy Stabilized: it is based on taking advantage of distributed buoyancy provided from the weight of the removed water by the platform. The following figure demonstrates the three types described (Butterfield et al. 2005):

Figure 1.16- Three design concepts of Floating wind turbines (Butterfield et al. 2005)

[20]

The table below has rationally listed the pros and cons of each concept based on different technological and design factors (Butterfield et al. 2005):

Table 1.1-Comparing the three concepts in brief (Butterfield et al. 2005)

1.3 CONCLUSIONS AND STUDY OUTLINES

The application of renewable energy resources has experienced a reasonable growth within recent times. The ever increasing cost of energy and the rise of world population have accelerated the use of such energy resources even further (Ghouri 2007). Wind energy hasnt been an exception. As the science and technology advance there will be more sophistication in engineering devices. Thus today engineers can come up with more ambitious solutions so that more energy can be generated as a result of more optimized application of materials and methodologies. Offshore wind turbines are a fine example of such advances in the engineering fields.

[21]

Analysis and design of offshore wind turbines is a challenging task. There are so many different parts covering a wide range of engineering disciplines from offshore engineering, mechanical engineering to structural engineering and more. As for structural engineering, there is a variety of components to be dealt with. For example; foundation, tower and blades which each have their own corresponding fields of study. However the parameters involved in studying the behaviour of such structural components are quite different in comparison to a conventional structure (for instance a typical tower structure in a city environment). The actions on the structure are a mixture of different environmental loadings and mainly include wave, wind, Ice and seismic (Det Norske Veritas, 2004) and their effectiveness generally depends on the component of interest. Clearly the choice of the structural analysis can be governed by the required level of study and the type of results needed as well as the study constraints existing at the time of analysis. Such constraints can be due to the lack of experimental information or computational limitation as well as variety of other factors (Saouma 2004). The point is to appreciate such constraints and allow for those when it comes to the choice of analysis. With regards to this research, the structure of interests will be the supporting tower of a fixed monopile offshore wind turbine. Towers are subjected to wind and wave loading simultaneously. Many researches have dealt with this structure using different methods based on the objective of the study. In this study the probabilistic concepts involved in the nature of wind and waves will be incorporated into the structural dynamic procedure and the response of the structure will be evaluated accordingly. This was quite important as clearly there is a large amount of uncertainty in the loads induced by wave and wind to the offshore structure due to the nature of these two actions. The probabilistic modelling has been based on the joint distribution of wind pressure and wave properties given by Ditlevsen (2002) from which the results will be simulated using Monte Carlo simulation. The dynamic modelling of the structure has been based on a moderately selected two degrees of freedom system as most of the researches have either considered a complex finite element model or a simple single degree of freedom model and thus the absence of a moderate modelling was apparent. The structure is modelled and analysed using the well-known mathematical software MATLAB. The results have been validated using SAP2000 or have been compared with text books to observe the effectiveness of methodologies. The final results are then used to determine the displacements and some of their statistical properties. Finally the applicability of the

[22]

responses has been discussed and their significance concerning the design of the structure is observed.

[23]

2 LITERATURE REVIEW 2.1 VARIOUS MECHANICAL MODELLING BY DIFFERENT RESEARCHERS

In order to undertake any analysis with regards to the tower structure of offshore wind turbines it is compulsory to determine a suitable mechanical model. The suitable modelling type is determined based on its applicability to the purpose of this dissertation. However it is worth mentioning some examples of a mechanical modelling carried out for onshore wind turbines. The following study is undertaken to optimize the tower of an onshore wind turbine for cost and thus there is much more concentration on the tower structure. The model has been used by Uys et al (2007) and is clearly for onshore wind turbines which there will be subject to no loading induced from ocean waves:

Figure 2.1-Wind Loads and bending moments on each shell segments (Uys et al 2007)

The above figure demonstrates the tower structure of an onshore wind turbine divided into the number of circular shell elements with chosen dimensions and physical properties. The elements are installed on top of each other and thus form the tower which itself will be in a vertical cantilever arrangement. The dynamic loadings are considered to act on the structure in three levels and the nacelle and blades are replaced with corresponding loads such as vertical load and others which are applied at the top of the tower. Figure 2.1(c) [24]

demonstrates bending moment in the tower aligned with the tower made from shell elements on the left. In addition to this work, another research done by Lavassas et al (2003) has used the 3 dimensional finite elements modelling for analysis and design of the tower structure of an onshore wind turbine. Second research is carried out by AlHamaydeh and Hussain in 2010. Similarly, the study is based on optimisation of the tower structure of an onshore wind turbine. The research has considered the soil structure interaction and dynamicity of the tower in order to optimise the tower. The following figure demonstrates the mechanical model of the structure based on the above:

Figure 2.2-single degree of freedom system (Al Hamaydeh and Hussain 2010)

Similar to the previous study, the flexibility of the foundation is considered by introducing rotational stiffness which is represented by the spring . The lumped mass on the top has

been placed so that the inertia loads are taken into account. As it has been described in the figure, the system is simplified as a single degree of freedom system. The rotational and translational displacements also have been realized. The rest of this chapter will deal with the representation of different mechanical modelling for offshore wind turbines. The mechanical modelling for offshore wind turbines is versatile. The differences in modelling can be traced within the objective of each research for instance some research has been based on investigating the foundation and others which have dealt with the tower and so on. Majority of the researches have considered a dynamic model for the tower structure. The reason for that could be based on the fact that the offshore wind towers are subject to simultaneously induced wind and wave loadings that vary with the time. Furthermore the [25]

differences exist in the method by which the dynamical model is represented. Some have used simple single degrees of freedom and others have applied multi degrees of freedom idealization. Many of them also have dealt with the soil structure interaction based on such models. One of the most recent published papers with regards to offshore wind turbines is a research done by Bhattacharya and Adhikari in 2011. In this paper the soil structure interaction of the offshore wind turbines have been investigated based on the dynamic behaviour of the structure. The paper has proposed and applied the following idealization for the dynamic analysis. The effect of a flexible foundation and the inertia have been taken into account by allowing the springs at the bottom for rotational and translational stiffness and a lumped mass at the top of the structure. The equation of motion is then constructed on the basis of EulerBernoulli beam equation. The arrangement has been depicted in the following figure:

Figure 2.3- Process of idealization from the structure to a mass and spring system (Bhattacharya and Adhikari, 2011)

The structural modelling in the above is taking into account the vertical action to the structure denoted P. Furthermore the structure will be subject to dynamic wave from the sea or as well as dynamic wind induced on the top of it. Again the flexibility of the foundation has been considered by introducing springs rotationally and translational. The dynamic behaviour of the foundation of these structures are of interest as they tend to be much more complex in comparison to both onshore wind turbines and offshore oil platforms (Zaaijer, 2006). The following figure demonstrates the modelling of the foundation inside the water based on the study of Zaaijer in the year 2006:

[26]

Figure 2.4 Foundation models for pile structures (Zaaijer 2006)

The above figure shows a proposed modelling for a pile foundation of an offshore wind turbine. The figure illustrates 4 different models in order to achieve various components of the analysis. The first model from the left is an idealized formation of soil-structure interaction taking into account the stiffness of the soil. The second model is a vertical cantilever in order to determine the effective length of the pile and the third is configured in order to evaluate the stiffness matrix. The last model similar to the ones investigated by previous researchers considerers the translational and rotational stiffness by allowing springs in appropriate positions of the structure. The approach is more focused on the soil-structure interaction and the dynamic behaviour of the foundation however the tower is not modelled in detail. Neither of the above modelling so far has considered the interaction with the water below the sea level. A report carried out by Petersen et al (2010) has provided a useful modelling based on the effect of water around the monopile structure shown below:

[27]

Figure 2.5- Modelling with lumped parameters approach to the foundation and fluid structure interaction (Petersen et al 2010)

Clearly the stiffness and damping effect of the soil on the sea bed is allowed. Furthermore there have been some considerations with regards to the damping effect of the sea water around the pile related to the horizontal load acting on the structure from the left. The structure is modelled as a multi degrees of freedom system and lumped masses are considered. In the figure as the wave load is acting opposite to the wind load which is induced at the top, the effect of wind is taken into account by providing damping at the position of catching wind. However the model hasnt considered the detailed dynamic load from the wind when it is applied at the same direction to the dynamic wave loading. According to Davison and Owens (2003) any structure that is subject to excessive cycle loading can suffer from fatigue failure. This is visibly the case for offshore wind turbines when they are under the load fluctuations from wind and wave loading. All of the modelling presented so far has considered the dynamic behaviour of the structure and this can be considered as a higher evidential possibility of cycle loading to the structure. The report of Petersen et al (2010) has investigated the fatigue behaviour of offshore wind turbines. The analysis is undertaken to assess the fatigue life of the structure using corresponding fatigue stress and number of cycles curve:

[28]

Figure 2.6 Strategy of Fatigue life assessment (Petersen et al. 2010)

The above shows a general model that suggests the procedure of evaluating the fatigue life estimation curve in principal. The model evidently has not considered the dynamic load induced by sea waves and has only taken into account the dynamic wind loading acting at the top of the structure. The analogy has then considered the dynamic stress produced at the base of the tower where it has reached the sea bed and the fatigue life estimation has been done accordingly. Additionally, the general modelling seems to be simply ignoring the soilstructure interaction and fluid-structure interaction below the sea level.
2.2 RELIABILITY RELATED RESEARCHES

Structural reliability deals with almost every structure that is subject to uncertainties. Such uncertainties can be present in both loading and material properties (Ditlevsen and Madsen 1996). Wind turbines in general and offshore versions of them in particular can be considered as structures that contain such uncertainties in great depth. It is evident that wind and wave actions on the structure can include some major variability in terms of their magnitude and direction and an example of this can be found in the work of Masuda et al. (1998) in which the statistics of wave and wind related to Eastern Tsushima Strait is investigated. It is explained in this work that for instance the significant wave height and period showed a sinusoidal variability. Furthermore both actions are not fully independent and a larger wind force in a coastal environment can translate into larger wave induced loads on an offshore structure. Therefore the usability of probabilistic engineering in offshore wind turbines becomes very important. That means all the components of the structure which are exposed to wind and wave loads may be suitable for reliability/probabilistic analysis and design. For instance, one of the key components that are subjected to both wave load and wind load is the

[29]

supporting tower of an offshore wind turbine. There have been a number of reliability related studies carried out by various researchers with regards to different parameters involved in the probabilistic analysis and design of the offshore wind turbines structural components. Many of the studies carried out are related to the fatigue life of blades as their role is very significant in the power production of the wind turbine. Blades are exposed to the large and complex wind load effects (Ronold et al. 1999) especially for the offshore units as there will be increases in the wind speed in the sea environment (Simiu and Scanlan 1986). There have been also other studies regarding the fatigue effect on the other parts of the structure for example welded connections of the supporting structure.
2.2.1 OFFSHORE WIND FARMS AND TURBULENCE MODEL FOR FATIGUE RELIABILITY (FROM THE WORK OF SRENSEN ET AL. 2008)

The work has focused on the modelling of the turbulence in offshore wind farms with 50 to a 100 wind turbines. These are more critical for extreme loading and fatigue limit states. In these wind farms due to the significant increase in turbulence intensity the fatigue life is reduced considerably. The increase in the turbulence intensity is known to be based on a slight decrease in the mean wind velocity behind the wind turbine upon formation of wake. This has been the main motive behind the study. The study begins with an evaluation on the change in turbulence intensity using the data collected from a Danish site and then the equivalent fatigue load is found for both stochastic and deterministic part. The stochastic part refers to ambient turbulence and the Eigenfrequencies of the structure and deterministic part concerns the sinusoidal motion of the blades with frequency equal to revolution of the rotor of an offshore wind turbine. Furthermore appropriate probabilistic modelling of the fatigue failure is determined. The study proposes a slight un- conservative effective turbulence model which can be used in design and analysis. Examples also have been provided in order to demonstrate the applicability of such a model. However the model is observed to be slightly conservative where both parts of the fatigue loading are stochastic.

[30]

2.2.2 LONG TERM FATIGUE ANALYSIS OF MULTI-PLANAR TUBULAR JOINTS FOR JACKET-TYPE OFFSHORE WIND TURBINE (FROM THE WORK OF MOAN ET.AL 2011)

The study has considered long term fatigue analysis of welded multi-planar tubular joints for fixed jacket type offshore wind turbines. It is particularly about the design of the offshore wind turbines for the water depth of 70m in North Sea. Joint probabilistic model of wind velocity, critical wave height and spectral peak period has been evaluated comprehensively. Furthermore three principal load cases have been dealt with. Firstly, wind loads only, secondly, wave load only and thirdly, load case based on combination of wave and wind load. The work is continued by obtaining dynamic response of a jacket type offshore wind turbine based on the given probabilistic modelling of the load and corresponding load cases as it was described in the above. However the dynamic analysis is divided into three sections which are dynamic response analyses of the environmental loads, aerodynamic and hydrodynamic and finally structural response of the structure. At the next stage long term analysis of hot spot stress ranges have been carried out. The reason for this is described as its inclusiveness of other stress raising effects which is done by utilization of suitable stress magnification factors to nominal brace stresses. Using this approach the long term fatigue analysis is undertaken and various contour plots of fatigue damage for wave load only has been obtained. Additionally, 3d contour plots of fatigue damage at specific joints have been evaluated. However there have been some major assumptions on which the general procedure of the paper has been established:

1- Wind and waves always are induced in the same direction and their phasing is probabilistic and is not correlated with each other. 2- Based on the tubular joints in a North Sea Jacket support structure 3- Hot-spot computation is founded on the linear superposition method

[31]

2.2.3 LONG-TERM FATIGUE ANALYSIS OF OFFSHORE FIXED WIND TURBINES BASED ON TIME-DOMAIN SIMULATIONS (FROM THE WORK OF GEO AND MOAN 2010)

This study is more or less similar to the previous with the difference of having an equivalent fixed structure for the analysis of the previously explored jacket-type. This is done in order to reduce the computation complexity of the jacket type structure which due to having so many slender elements can become very time consuming. Furthermore it is described that the time consuming stages can be due to the requirement of the fatigue analysis for numerous wind and wave conditions which must be considered. The joint distribution of one hour mean wind velocity for 10m reference elevation and significant wave height and spectral peak period is used analytically. Moreover a structural analysis is carried out for 20 simulations per 10 min intervals. The results for the short term analysis then is gathered and rain-flow cycle counting is undertaken and average short-term distribution of the counted response cycles are estimated. The long term distribution of the response cycles is determined by summing the contribution of each short term condition based on the probability at which it occurs. Once the long term distribution of cycles is found the long term fatigue damage for time T has been formulated as below: Where K and m are the material parameters of the SN-curve where given. N is the number of cycles to failure under constant stress range S and has been is the long

term stress range distribution. In conclusion the paper has found that the fatigue damage regarding the overturning moment is governed by the wind load and furthermore the wave load also contribute to base shear related effects. The study explains that in order to check the influence of difference between the angle direction of wind and wave the long term distribution of wind and wave directions is needed. However it has been concluded that such parameters effect can be insignificant.

[32]

2.3 GENERAL DESIGN CRITERIONS FOR OFFSHORE WIND TURBINES (BASED ON DET NORSKE VERITAS 2004)

In this section the general design considerations for offshore wind turbines will be discussed. Here the study temporarily moves from engineering analysis to the field of structural design. This can provide a further understanding on some key consideration with regards to the design of these structures which can justify the areas at which the analyses can focus. The information in this section will be in accordance to Det Norske Veritas (2004) and is Danish standard for designing offshore wind turbine tower structures. It has been mentioned in the above document that the design life of offshore wind turbines is about 20 years and the design data must be based on a good understanding of the site conditions. Different Limit States are considered for the design of offshore wind turbines which have been described below: Ultimate Limit States (ULS): considers the maximum load carrying capacity of the structure (eg. Buckling failure) Fatigue Limit States(FLS): considers the structural failure as a result of the influence of cyclic loading (eg. failure due to repeated loads) Accidental Limit States (ALS): it is regarding the structural impair or failure as a result of accidental loads (eg. Impact load from ships) Serviceability Limit States (SLS): considers the certain conditions regarding the normal functioning of the structure or durability (eg. Excessive deflection ) The design procedures have been categorized based on their corresponding methodologies and approaches as following: Design based on LRFD method: in this method the target safety is achieved by using characteristic values for load and resistance Design based on direct simulations of combined load effect of concurrent load processes: this is an attractive option when the wind turbine structure is subject to two or more concurrent load processes for instance wind and wave. So the characteristic load values are simulated instead of being linearly combined same as the LRFD method. [33]

Design based on testing: this method considers the mechanical properties such as load effect and material resistance by means of observation from a real structure. These results then will be used in analytical methods in order to design the structure.

Probability based design: using this approach, the structural components are designed so that their probability of failure remains below a given probability.

This document allows for the wind load modelling for every 10 minutes sea state. That means the mean value for wind speed is assumed to be constant during this period. This allows evaluating the probability distribution function of the short term wind parameters by multiplying the corresponding value for every 10 minute stationary periods. A similar approach have been presented for the wave loading from which the long term distribution of wave forces can be evaluated with respect to wave height, wave periods and other wave parameters. The other loading considered in addition to the above is Ice loading. There has been given information about the water level which can be used to determine the sizes of rotor and tower for extreme conditions. Offshore wind turbines are built on sea bed. That means it is important to study the site condition in that respect. The document has given the list of soil investigation required for an offshore wind turbine as below:
Geology survey for the site Topography survey of the seabed Investigation of relation between soil borings and in-situ testing Laboratory testing to provide soil sampling In-situ test such as cone penetration

Furthermore it is important to appreciate other site conditions such as seismicity, salinity (the water relation to corrosion), temperature, marine growth, sea activities traffics for example fishery transportation etc., disposed matters, existing pipelines and cables. In order to

establish any design approach one needs to appreciate the correct load combinations. The following has been given by Det Norske Veritas (2004):

[34]

Table 2.1-Load Combinations (Det Norske Veritas 2004 p.67)

The above table suggests the load combination for ultimate limit states and accidental limit states for design of an offshore wind turbine. There have been five load combinations considered in which the wave, wind, current and ice loading as well as water level is present. The period given for each parameters in years represent the 100 years reoccurring value. As it is observed the number of years is different for each load combination. It is also noted that Ice loading is only considered in the last two load combinations where at one of them the wave load is absent and another the 10 years reoccurring value is taken. Furthermore for the first two combinations the reoccurring values for wind and wave load have been taken differently in which one is less than the other at each combination. For fatigue limit state the load value is mentioned to be used must be based on the design life fluctuations which must be comprised of the inputs from wind, wave, current, ice and water level. In order to appreciate the reoccurrence period the table below is given to present some guidance in terms of a statistical representation of them:
Table 2.2-Statistical description of reoccurrence period (Det Norske Veritas 2004 p. 47)

[35]

The application of different characteristic values for each limit states has been indicated in the table below:
Table 2.3-Different Limit States (Det Norske Veritas 2004 p.46)

The definition of permanent and variable loads are given as: Permanent Loads: self-weight of structural components, internal and external pressure which has a permanent nature and etc. Variable loads: loads from personnel, crane operational loads, loads from installations; stored materials, life boats etc. Environmental loads: wind, wave, current and ice and snow induced loads, tidal effects, earthquake loads etc. With regards to wind loads there are some indirect loads that must be considered which are induced from the wind turbine operation. Some examples of these loads are gravitational loads applied to the rotor, centrifugal forces due to rotation of the blades and many more. It is seen from the above table that different descriptions have been provided for various limit states with regards to the types of applied loading. In addition to the above the following table also has been provided in order to determine the correct partial safety factors to be applied in design of the structure:

[36]

Table 2.4-Determining the correct partial safety factors (Det Norske Veritas 2004 p.73)

The factor

must be determined from the Clauses 103 and 104 provided in the document

itself. Having the above framework one should be able to determine correct load combinations together with right factors to design the structure based on an up-to-date load modelling.

[37]

3 PROBABILISTIC DYNAMIC ANALYSIS OF OFFSHORE WIND TURBINE TOWER STRUCTURE 3.1 ANALYSIS

3.1.1 GENERAL ANALYSIS STRATEGY

The uncertain natures of wind and waves together with the dynamicity of such actions has led researchers to use different approaches with regards to analysis of the various parts related to offshore wind turbines. However most of the works tackle the problem either by using the simplest mechanical modelling or a very complex finite element approach as described in chapter 2. Accordingly there wasnt a study to deal with the problem moderately by considering a system more than a single degree of freedom but yet not taking a heavy duty finite element method at the same time. This was the motive of the analysis approach that is followed in this study. Such framework is followed and the analysis approach deals with various important aspects so that their corresponding influence in structural response is considered. The objective of the analysis is to determine the structural displacements of the supporting tower of a fixed Monopile Structure in a shallow water zone under simultaneous probabilistic wind and wave loads. These loads are assumed to act in the same direction and must be evaluated with respect to the time indicating the dynamicity of the system. The approach is not based on a complicated finite element analysis but the effort has been made in order to capture important parameters involved in the behaviour of such structures sufficiently by simulating the values of actions. Consider the following as the structure of interest:

Figure 3.1-A typical offshore wind turbine of interest

[38]

The above shows a typical offshore wind turbine based on a monopile structure. Now based on the description presented in the above, the main factors that have been dealt with regarding the study of offshore wind turbines are randomness in actions as well as the dynamic nature of the system. In an offshore environment wind and wave loads can be considered as processes that contain uncertainty in their nature and these actions are related so the higher the wind the stronger the waves (Novak et al. 2001). Considering wave load as a random value which is statistically dependent to wind, one should be able to simulate the value for wind and wave given the wind and wave probability distribution functions and corresponding parameters. These values can then be utilized in order to evaluate the response of the structure by application of a suitable dynamic analysis. Consider the following two degrees of freedom system as an idealization of a support structure of an offshore wind turbine shown in the previous figure:

Figure 3.2-Idealization into 2 degrees of freedom system

The first degree of freedom corresponds to the random wave loading and the second degree of freedom to the random wind loading. The above actions are induced simultaneously and their values are random based on their corresponding distribution functions. The equation of the system can be written as following:

[39]

Where and and and


st

(3.1)

are displacement responses for the 1st and 2nd degrees of freedom and are time varying wave and wind loads respectively. The parameters

correspond to tower mass and the hub mass of the offshore wind turbine of and are present to take into account the damping properties
nd

interest. Furthermore regarding 1 and 2

degree of freedom. The nature of the dynamic loading for which the

variation of loads in time is established will be in accordance to a concept called sea state. A similar idea had been used in the previous section where the design of tower structure was presented in which the wind and wave modelling was based on 10 minute stationary conditions where the mean value for wind and wave loads for this period was considered (Det Norske Veritas 2004). In this section a similar approach is applied. Thus the sea state can be defined as a period of time in which the mean value of wind and wave load can be considered to be a constant value. In this study a 10 minutes stationary period has been allowed. This means that in any sea state with a period of 10 minutes the statistical values which are the mean and standard deviation of the wind and waves are assumed to remain constant. So in every 10 minutes there will be a new mean value for wind and wave loading with a corresponding assumed standard deviation. This is required to simulate forces for intervals within any sea states called time steps which also are required for dynamic integration. This is an effort to appreciate the variability of these loads within any given sea state. Thus they can show some variation though small during each sea states This is very important as it will also assist the analysis to further study the dynamicity of the system due to the fact that a constant load that is applied for a period of 10 minutes could overshadow the requirement for dynamic analysis (though a correct judgment can be made after knowing the actual natural frequencies of the system). It is seen from this method that both probabilistic and dynamic aspects of the behaviour of the structure can be taken into account and the time varying response can be evaluated for any given time. It is important to note that the structure can be analysed for any time span based on the simulation of wind and wave actions at any time steps. However due to the computational constraints which mean that analysis of the structure for a very large time span can be computationally intensive, thus a reasonably short time span is considered for example 2 days. This short time analysis can be used to evaluate the distribution for long time responses by using a similar method used by Moan (2011) which was briefly presented in previous sections. The distribution of short term of responses will be [40]

obtained by establishing the probability distribution function of maximum displacement per chosen interval which has been taken as a 1 hour interval for this study. The displacements that are considered are the maximum values of differential displacements between first and second degrees of freedom and also the displacements of the first degree of freedom which all will be varying with time. In order to elaborate on the approach further the following diagrams demonstrate the procedure of simulation which will be used in this study:

Figure 3.3-General analysis strategy

The above figure represents the strategy used for simulating different random values and contribution of each to the various aspects of the problem. As it is shown in the above, the [41]

first graph corresponding to first stage represents the simulation of mean value of an action which could be wind or wave within any sea state period of 10 minutes. This graph also shows the probabilistic modelling of the actions which in this study will be based on the work of Ditlevsen (2002) from which the mean value of wind and wave properties for every 10 minutes sea state period is simulated. Second graph shows the second stage of simulation of the wind and wave loading based on the already found mean from the first stage and a given standard deviation at each time step. Time steps are used for the evaluation of responses from the equations of motion by means of numerical integration which will be described in the later sections. The time step for this study is assumed to be 0.5 seconds so that at each time step a new wind and wave value will be simulated based on the mean that is found in the initial simulation from sea states and a given standard deviation. This stage in addition to being relative to the probabilistic nature of the actions but also it corresponds to the dynamicity of them. The final graph is the resultant of all the previous simulations and contains the contribution of both dynamic and the probabilistic nature of the system. As it is shown in the final graph, the forces theoretically must show some micro random behaviour for each sea state of 10 minutes and some major change once each sea state is completed as the new mean value is generated at that stage. Once wind and wave loads are simulated based on the process explained in the above, the procedure will be continued by solving the structural dynamic problem based on a two degrees of freedom system where the numerical values of actions applied to each degree of freedom have been given previously. The problem can be approached by the modal analysis of the structure and solving the eigenvalue of the system to evaluate the natural frequencies as well as the modal shape coefficients. The results will be used and the simultaneous ordinary differential equations are uncoupled by transformation of the variables after which the total matrix of displacement with respect to time will be obtained by the superposition of the results from each mode. The numerical integration method used in this study will be the average acceleration method which is a standard method that is found in most structural dynamic text books. The application of this method has been validated in this study by solving a problem from Chopra (2007) and the results are compared in order to check whether the numerical integration code is valid. Finally the probability distribution functions of maximum displacements for short time analysis have been plotted and the methodology to extend the results to long term distribution of displacement is discussed. The following diagram demonstrates the different stages of the main analysis algorithmically: [42]

Assign Geometrical and Mechanical Properties

Deteremine Mass and Stiffness Matrices

Find Natural frequencies and Modal matrix

Assume Dampling

Decide for the Analysis period

Simulate wind pressure, wavelength and height and period for each seastate

Find mean values of wave and wind forces for each sea states

Decide for the time step size and obtain its number inside each seastate

simulate wave and wind forces based on the theirs mean and a std for each time step

evaluate the displacements by numerical integration

Figure 3.4-The analysis procedure

It is also worth mentioning that the above process was established based on experimenting with alternative methods beforehand. Initially the structure had been analysed differently based on simulation of wind and wave loadings from the modelling proposed by Ditlevsen (2002) for each of the time steps which were equal to 0.01 seconds. Although this was based on incorrect assumptions it provided some results and has been shown in a latter section. The assumptions were not true as the probability density functions for simulating mean value for [43]

wind and wave properties which had been used from the work of Ditlevsen (2002) were based on the initial assumption that the sea state must be equal to 10 minutes which wasnt consistent with the analysis where the sea state had been taken to be equal to the integration time step equal to 0.01 seconds. Such inconsistency resulted in redoing the analysis based on the completely new methodology and procedure where the correct assumption of the 10 minutes sea state was applied. Both analyses have been provided in the later sections in order to demonstrate the total progress in which the final results were achieved. The main final analysis results have been provided in the beginning of the section 3.1.7 which is based on correct assumptions and has been justified. The other analysis results are also provided at the end part of the same section. In order to apply the final analysis method and measure its effectiveness, it is necessary to observe its application to a realistic structure with specified mechanical and geometrical values. This is done to evaluate the numerical analysis so that further observation can be made with regards to the analytical tools and methods presented in this report. Thus the following structural arrangement has been used for carrying out the analysis numerically and the results were recoded for further study:

[44]

35m

75m

25m

Figure 3.5-Selected structure for numerical analysis (Not to scale)

[45]

Table 3.1-Mechanical and environmental properties regarding the structure used for the numerical analysis

Property Mass of the Tower Mass of the Hub Thickness of the Tube Material Elastic Modulus Average Diameter of the Tube Height of the Hub from Sea Level Height Inside the Water Total Tower Height Length of each Blade Maximum Width of each Blade

Value 100000 kg 80000 kg 0.01 m 200 GPa (Steel) 3.25m 75m 25m 100m 35m 5m(used to find the tributary area and calculating the wind drag force from wind pressure)

Air Density Sea Water Density Blade Coefficient of Drag Tower Coefficient of Drag Tower Coefficient of Inertia Damping Ratio from Water Damping Ratio from Wind Fixity to the Seabed Direction of the Actions

1.2 kg/m3 1020 kg/m3 0.80 (conservative) 1.20 1.50 10% (assumed) 5% (assumed) Considered fully fixed at the bottom Acting in the same direction as each other

The results from the main analysis have been validated using various means. As it was mentioned previously the numerical integration method is validated by solving an example from Chopra (2007) and the results are compared and checked for consistency. The displacements are checked with comparison to the equal static load on the structure and they are compared and corresponding points are discussed.

[46]

3.1.2 PROBABILITY CONCEPTS 3.1.2.1 DETERMINISTIC ANALYSIS VS. PROBABILISTIC ANALYSIS

The engineering world has experienced a great deal of development during the last century. Here structural engineering hasnt been excluded. Modern societies demand more optimised, intelligent and sophisticated engineering accomplishments. Today structural engineering fields have to carefully provide solutions with regards to the degree of safety and reliability at which the structure must function. Additionally it is important to care for the materials and cost and not to generously over design the structural components. Therefore modern engineering is about establishment of a very much quantitative approach in which most of the possible parameters are contemplated. Some of the bases of this approach could be found in the mathematical science where the probabilistic concepts reside. These concepts provide means to deal with residual uncertainties within the world where the structure must perform (Hart 1982). Such uncertainties involved in every aspect of engineering parameters, for instance the uncertainty in the magnitude of the actions on the structure or uncertainty within the actual material strength (Ditlevsen and Madsen 1996). The point is to deal with these realities quantitatively and consider them during the process of engineering decision making. Such an opportunity would have been overlooked by the traditional deterministic approaches where the concentration would be simplifying assumption so that the engineer could provide a solution. However today the development of modern engineering tools such as advanced computers has eased the introduction of the new generation of engineering analysis and design where the so many important factors including the uncertainties and probabilistic nature of different engineering parameters are incorporated within the process. Traditionally deterministic analyses would treat the different engineering parameters as a single value. For example during the design of a structural component the value of live load would have been taken as a single number which would be the average value based on codes of design or engineering experience at the time. Furthermore the material behaviours are over simplified, for instance with regards to concrete the compressive strength is taken is a single value representing its average in real life. Such an approach has an advantage of simplicity and at the same time for some not very important structures, a more elaborate method would be unnecessary. However deterministic methods disregard the variability of parameters and dont deal with them in a solid quantitative way where the effect of such variability will remain within a controllable engineering range. Furthermore there can be a significant level [47]

of uncertainty with the real time values of such parameters, for example according to Rao (1992) a range of maximum wind velocity values which have been collected for a number of years is presented in the table below:
Table 3.2-Variation of annual maximum wind velocity (Rao 1992)
Year 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 Max Wind Velocity(Mile/h) 82.4 76.9 70.2 91.8 76.1 72.7 58.6 85.3 92.5 68.5

It is clearly seen that there is a noticeable distribution of values within the data. This has resulted in the existence of uncertainty inside the wind speed data and as a consequence in the magnitude of wind induced actions for which a structure must be analysed and designed. Furthermore there could be similar uncertainties regarding the direction of actions induced by winds which can be overlooked in traditional deterministic approaches. In addition to those, the material properties can demonstrate the similar variable nature concerning engineering specifications. The following figure demonstrates the variability of concrete elastic modulus and its compressive strength (Tomosawa and Noguchi 1993):

[48]

Figure 3.6 (a)

Figure 3.6 (b)

The randomness within the above which can be understood by the variability of parameters is another reason to appreciate the lack of accuracy regarding the deterministic engineering methods. These shortages added to the misunderstanding of the variable nature of loads can result in overly simplified methods where some key realties are over looked. The problem is that using deterministic methods will deprive the engineering from opportunity to quantitatively measure the uncertainties and incorporate such knowledge to the engineering [49]

decision making. For instance, not knowing the level of uncertainty it will not be possible to evaluate the level of risk thus it cant be known if a component is over designed or under designed. However this problem has dealt with the probabilistic approaches where such variability is appreciated and considered within the process. This can demonstrate the significance of these approaches. Generally, Probabilistic methods are based on considering not just the simple average value of parameters but also taking their level distribution into account (Rao 1992). That means for instance the mechanical analysis in designing a steel member will have to receive the average value as well as the measure of spread of data from this value. This can increase the accuracy of the process and provide further confidence in the engineering solution.
3.1.2.2 RANDOM VARIABLES AND PROBABILITY DENSITY FUNCTIONS

The importance of probabilistic approaches was explained in the previous section. In order to present an in depth method of probabilistic analysis applied in this study it is necessary to provide some mathematical substrates regarding the main concepts of probability and its corresponding properties. This will be done in this section where various basics of the probability theory will be presented. Random Variables According to (Soong 1981) probability mathematics is utilized to appreciate the various mathematical arrangements concerning randomness such as uncertain nature of a mechanical system. This can be considered more or less as tools that can quantify the randomness of a parameter. Based on the explanation provided by Rao (1992), an event can be defined as a possible outcome of an experimentation which is done to produce some results and on that basis a random variable is a variable that can accept inputs of random events. So if Random variable then it can take inputs of random event is a

. In this book random variables are

denoted by capital letters and random events with small letters. The reason this variable is called random is because it is impossible to predict its exact value at any time before the outcome of the event is revealed Smith (1986). Furthermore random variables also called Stochastic Variables which implies to a variable that takes value stochastically (Spiegel and Murray 2000). Random variables are categorized in two (Hart 1982):

[50]

i-

Discrete Random Variables: where it can only have isolated values

ii-

Continuous Random Variables: here the random variable can take a real

value within a particular range.

Probability Density Functions and Cumulative Distribution Functions According to Rao (1992) consider as a continuous random variable and function P to

represent the probability of an event then the probability density function (PDF) of it is given as: (3.2) Where the probability that the random variable will be obtained as below: In the above, (3.3) will be less than or equal to a value say

is termed as the probability distribution function or cumulative . According to (Spiegel and Murray 2000) the

distribution function of random variable

following properties are given for probability density function: -

Furthermore it can be evaluated that from Smith (1986) the following correspondence exists between probability density function and probability distribution function: The following figure demonstrates this relationship in a graphical manner: (3.4)

[51]

Figure 3.7 a) Probability density function b) Probability distribution function

In the above graph one can see the relation between the areas underneath the curve for the values less that is correlated with the value in the vertical axis of the graph of the

probability distribution function. The vertical axis here shows the probability of a given event in the corresponding location along the horizontal axis. Note that the curve reaches the value of 1 in infinity. The concept of probability distribution function and its relative properties will be utilized in this study at the stage where simulation of a random variable based on a given density function is required. Normal and Lognormal random Variables Once the definition of probability distribution and probability density functions and their corresponding mathematical properties is known, one can explore some very famous specific types of these functions which can have a wide range of application in the analysis procedure which will be descried later. In any engineering application mainly the nature of distribution that is chosen is governed by the following observations (Rao 1992): the type of problem the particular hypothesis with regards to a specific distribution geometrical nature and characteristics of a plotted set of data which can resemble a particular distribution computation constraints for instance a required level of simplicity for the calculation

[52]

Two main distributions utilized in this study are Normal distribution and Lognormal distribution. According to Soong (1981) these two are very close in terms of their mathematical properties. Normal distribution which is also known as Gaussian Distribution which shows a famous symmetrical bell shaped curve. The normal distribution can be specified by two very important parameters only which are and

which correspond to the statistical properties of a set of data. According to Spiegel and Murray (2000) the statistical mean can be defined as an average value of all the values inside a given data set. Furthermore it describes the mean using a function called which is defined below: Where is a probability density function and (3.5) is a function of random variable .

Therefore the statistical mean is defined as following expression based on the expectation function (Hart 1982): Where (3.6) . Throughout this study the mean

denotes the mean value for the random variable

value will be represented by a Normal distribution is

. Another important concept which also will be used to define . Spiegel and Murray 2000 portray that this

concept will represent the level of variability of the data. In other words, it is used to quantify the level of spread of given data from its central tendency which is its mean value. The standard deviation is also defined as the square root of . Based on the work of Hart

(1982) the Variance is expressed in terms of expectation function as below: Where (3.7)

is the standard deviation of a given set of data. Having studied the above concepts

one can express the probability density function of a normal random variable as below (Hart 1982): (3.8)

[53]

Where

and

are mean and standard deviation in accordance to the descriptions in the

above. Furthermore the probability density function of a Lognormal Distribution is given as (Soong 1981): (3.9) are different from standard deviation and mean value explained in the above. and respectively and are defined

Here and

These are called

to express the probability density function of a lognormal distribution in the formation of the above formula. These parameters must not be confused with similarly denoted parameters previously (Walpole and Myers 1989).
3.1.2.3 MONTE CARLO SIMULATION

The Monte Carlo method is founded on the idea of simulating the actual result based generating random numbers. This method has a very wide range of applications in science, econometrics and financial forecasting (Siddal 1983). As far as engineering applications are concerned Monte Carlo methods can be considered to be the best and the most accurate as other probabilistic methods can become extremely complex for the problem with high level on non-linearity and complicated mechanical concepts (Hart 1982). However there is price to pay and that is because the Monte Carlo simulation can be computationally intensive. Metropolis (1987) suggests that the Monte Carlo method was first introduced by Stanislaw Ulam, a Polish-American mathematician in 1947, based on his concept of Lucky Numbers. This idea later contributed to solving the problem of neutron multiplication in fission devices during the course of the Second World War. According to Ditlevsen and Madsen (1996), historically, the process of generating random numbers used to be undertaken using mechanical means such as the roulette machine. However by introduction of the modern age computers this procedure has been updated by application of computer programmes that utilizes certain algorithms that produce the so called pseudo random numbers which is within the interval . According to the same work, the process is

carried out by using complex mathematical formulas where the numbers are generated in a deterministic manner but they show chaotic nature. This process is carried out using MATLAB in this study. According to (Hart 1982), to simulate random numbers based on

Monte Carlo methods the following stages are followed:

[54]

Definition of the input domains Generating random numbers in accordance to any distribution of interest Use of the generated numbers to carry out the deterministic operations Continue the above stages for the number of loops required Store the results for all of the loops Compute the statistical properties from the gathered data The most relevant stage in the above is where the random numbers based on their corresponding distributions are generated. This has some mathematical significance in order to carry out such an operation correctly. The stage by which a random number is generated in accordance to its distribution function is presented below (Ditlevsen and Madsen 1996): Utilize an algorithm that generates pseudo random number Consider these numbers to imply to the probability of an event as they are between 0 to 1 Use the inverse of probability distribution function to calculate the corresponding probability value:

The value

is the required generated number

The above steps are very general however they will establish a framework for the details of analysis used in this study regarding the simulation of the random loads for wind and wave. Thus its applicability will be explored in greater depth in the next sections. The following diagram also graphically demonstrates the general algorithm by which Monte Carlo simulations are carried out in this report:

[55]

Figure 3.8- General illustration of the Monte Carlo simulation approach

The above diagram demonstrates the process of the Monte Carlo simulation in a more computer friendly approach. In this report the same approach will be deployed in order to carry out the Monte Carlo method for the evaluation of the statistical properties of the displacement responses for a short term analysis. There will be applications of computer coding in MATLAB where such tasks are undertaken. However as it was explained earlier, the major limitation of Monte Carlo methods is their computational intensiveness. Therefore in carrying out this analysis care must be taken so that the highest computation efficiency is assured. This can be possible by avoiding the repetition of unnecessary computations and to provide the robust memory use so that the analysis is done in the highest level of computational effectiveness.

[56]

3.1.3 STRUCTURAL DYNAMIC CONCEPTS 3.1.3.1 APPLICATION OF STRUCTURAL DYNAMICS

Different concepts of Structural mechanics will be needed for different problems and this makes the approaches for analysis of structures subjective to the problems of interest. The choice of approach can also be affected by other factors such as complexity, time and the tools that are available at the time of analysis. Sometimes a basic linear analysis can be suitable and satisfy the requirements of the problem whereas for other problems this approach might be too simplified and might not represent the actual physical arrangements of the structure. So there are various parameters that can potentially influence the process of selecting a suitable structural analysis approach. This choice can also be to select from a Static approach or Dynamic approach which again can be determined based on the problem and its corresponding objectives and constraints. In addition to the above, one can achieve a better view by surveying the works of other researchers and make an analytical judgement. The review of the structural mechanic modelling for offshore wind turbines carried out by different researches was presented previously in this study which can provide a better overview in order to opt for a suitable mechanical modelling approach. According to Paz and Leigh (2004) the natural phenomenon and human activities impose actions that have time-dependant variability. Walking by a human on a bridge or a moving vehicle all has time-dependant changeability. Furthermore, an earthquake as a natural phenomenon will apply the ground acceleration that varies with time and thus it induces time varying load to a structure. The time varying nature mainly contributes to the system of dynamicity which suggests the need for dynamic analysis of the structure. According to Humar (2002) the nature of structural responses subject to a dynamic load can be treated as a rigid-body dynamic where it undergoes motion without deformation and non-rigid body i.e. includes deformation, that undergoes motion and deformation. The structural engineering mostly deals with the latter which means the structure will have motion and deformation. The same work also suggests that the non-rigid body motion is usually of an oscillatory nature in which the structure vibrates about an arrangement of stable equilibrium. The following figure demonstrates the oscillations of a frame structure due to a dynamic wind:

[57]

Figure 3.9-Vibration of a framed structure about its equilibrium position

It is seen from the above that the structural members oscillate around the static stability configuration. It is observed that the beams and columns have deformed in different orientations i.e. laterally and rotationally. Genta (1999) suggests that the dynamic behaviour of structure can be of significant importance as the dynamic responses of the structure can cause excessive stress in the structural components, structural fatigue failure and unpleasant effect for the users of the structure. With regards to offshore wind turbines, both wind and wave loadings are extremely time dependant. This can be the main reason as most researches have also used structural dynamic modelling for their works. Especially for the works done regarding fatigue that the dynamic responses will be used to evaluate fatigue curves (Davison and Owens 2003). Thus in this report the structure has been analysed dynamically and time varying responses for different assigned points are evaluated.

[58]

3.1.3.2 ASSEMBLAGE OF THE EQUATION OF MOTION

Equation of Motion for a Single Degree of Freedom System According to Chopra (2007) the simplest case regarding the equation of motion can be considered as a single degree of freedom system. This system is usually idealized as the motion followed by a pulley car as below:

Figure 3.10-A single degree of freedom system

The above demonstrates the forced damped vibration for a single degree of freedom system. In the above figure following parameters are used and defined (Paz and Leigh 2004): : is external time varying force lumped mass of the single degree of freedom stiffness of the system which equals to the force required to cause a unit displacement damping which represents the energy dissipation of the system represents the velocity which is equal to

According to Humar (2002) the equation of motion can be defined using D'Alembert's principle providing the condition of dynamic equilibrium as a conceptual substrate as following: (3.10)

Where

is function of displacement with respect to time. The above equation is also

represented in the following style conveniently: (3.11)

[59]

The equation in the above is a 2nd order ordinary differential equation and can be solved using any analytical or numerical methods to evaluate the function displacements and its variation with respect to time. In this study as the idea is to compute the displacement response with respect to time for an offshore wind turbine subject to simultaneous random actions from wind and waves, the solution will be carried out by numerical integration methods the description of which will be followed in later sections. Some further key concepts which will be used in this study is and

of vibration and their relationships. This is also in contrast to and from the following (Chopra 2007): . The natural frequency can be found

(3.12)

Where

is the natural frequency and

denotes the natural period of vibration.

Furthermore the period of excitation and excitation frequency relates to the external excitations i.e. the external dynamic forces. The above is used for determining the corresponding values for a single degree of freedom system where the system is idealized by a lump mass and stiffness related to the direction of the degree of freedom. Although this will not be the case for the problem that deals with the analysis approach followed by this study but it will show its significance as the more complex problem can be simplified in the number of single degree of freedom problems. The results then will be superimposed so the total response of the system is evaluated. Equation of Motion for Multiple Degrees Of Freedom System An offshore wind turbine of interest will be subject to random wind and wave loading. That means that the point of actions for these two actions can be different and thus the motion will be defined by more than one degrees of freedom system. Thus in this section the basic principle of multiple degrees of freedom system will be explored. Consider the following arrangement:

[60]

mN

kN,cN

m3

k3, c3
m2

k2,c2
m1

k1, c1

Figure 3.11-A Multiple degrees of freedom system

According to Paz and Leigh (2004) to define the motion of the system it is required to use simultaneous equation corresponding to each degree of freedom and thus can be represented as a matrix form: [ Where: [ : mass matrix containing the mass related to each degree of freedom [ : matrix of damping [ : stiffness matrix : vector of external forces : vector of displacement for each degree of freedom [ [ (3.13)

The above Equation can be solved simultaneously to evaluate the total response of the structure for each degree of freedom. The process of solving this will be presented in the following section.
3.1.3.3 SOLVING THE EQUATION OF MOTION

The process of solving the equation of motion can be done analytically or numerically as it has been suggested by Humar (2002). The Analytical technique will not be used in this study as it doesnt seem to provide the required application with the simulated random loadings. Thus the main procedure will be based on numerical integration methods termed as Average [61]

Acceleration Method provided in Chopra (2007).

There are indeed other techniques

available to solve these equations numerically. It is worth mentioning that in order to be able to apply the average acceleration method to multiple degrees of systems; one has to uncouple the equations so that they can be converted to a number of single degree of freedom systems. The coupled system of equations indicates the square matrix of damping or stiffness could be non-diagonal meaning they have non-zero elements among their non-diagonal elements and certain methods are applied to eliminate this complication. The results can then be obtained for each degree of freedom separately and the final results will be evaluated by modal superposition of the displacements from all the modes (i.e. degree of freedom). The process of uncoupling has also been described in this section. Finally, the average acceleration method in this study will be carried out in MATLAB using suitable codes. As it was mentioned earlier from equation 3.13 the formulation of equation of motion for a multiple degrees of freedom system is expressed in matrix form as below: [ [ [

According to Chopra (2007) to uncouple, consider the following matrix called that contains the modal coefficients to define the mode at which the system vibrates. Any vibration for the column in this matrix corresponds to the coefficients related to a Mode of the structure which concerns the degree of freedom as

well. This matrix is evaluated once the natural frequencies of the vibration of the system are determined. According to Humar (2002) these values can be evaluated using the following expression: [ Where the and (3.14) correspond to the natural frequency and vector of modal coefficients for between normalized modes

mode, then the so called

imply that: [ [ The value for will be obtained by solving the equation below: (3.15) (3.16)

[62]

(3.17) will be computed and the results together with the equation

For each mode a separate

3.15 can provide the solution for the elements of the modal matrix. Upon computation of the natural frequencies and modal matrix, consider substituting displacement so that [ [ [ the equation of motion will be rewritten (Paz and Leigh 2002): [ [ : [ [ [ [ [ [ (3.19) [ [ (3.18)

Now pre multiply by [ [ [ [ [

Now consider the expression below: (3.20) Where and are damping ratio and mass of the degree of freedom. Consider for that mode

dividing both sides of the equation 3.19 for each mode by transformed for each mode n: Let then the above can be rewritten as:

which is equal to unity according to equation 3.15, the following expression will be obtained

(3.21)

(3.22)

The above is the single degree of freedom equation and can be solved separately and the results has to be transformed back using equation [ and then the final results will

be obtained by superimposing the results for each mode. For the solution of a single degree of freedom the Average Acceleration Method will be used. introduction of a suitable time step ( The Method includes

) which will be used to evaluate the results by

numerical integration. Firstly the initial values are computed which will be based on the problem itself. For instance if the structure is considered to be at rest before carrying out the dynamic analysis then the initial displacements and velocities must be equated to zero. The following formulas are given for the initial value calculation (Chopra 2007):

[63]

(3.23) (3.24) (3.25) (3.26)

Where in the case of equation 3.22: (3.27) (3.28) (3.29) And is the assumed time step for numerical integration described in the above. Once the

above initial values are computed the following formulas are applied in order to compute the values of displacements at any given time step : (3.30) (3.31) (3.32) (3.33) (3.34) times.

The formulations of the above must be carried out repetitively for

That means for a time step equal to 0.5(s) and period of analysis equal to 10 seconds, one has to carry out the above from . This process is carried out using suitable

codes in MATLAB in this study. The code will be described and validated in latter sections.

[64]

3.1.4 DETAILED ANALYSIS APPROACH

The outlines of the analysis were introduced in the section regarding the analysis strategy in order to provide a brief overview of the approach. But in this section the strategy will be detailed based on the corresponding mathematical and mechanical concepts. The relative mathematical formula will be provided and specific structural arrangement and its consequent mechanical modelling and calculations will be presented parametrically. Consider the following wind turbine with given parameters such as the one presented in Table 3.1 subject to simultaneous dynamic wind and wave loadings:

Figure 3.12 a Typical Wind turbine

[65]

The structure is idealized as a two degree of freedom system as following:

Figure 3.13 Idealization into a 2 degrees of freedom system

Where

correspond to the mass of the tower and hub, and

and

are

stiffness of the bottom and top part of the tower and

are damping assigned to the and are time varying

1st and 2nd degrees of freedom respectively. Furthermore

wave and wind loads. This is a two degrees of freedom system based on equation 3.13 and thus the system of simultaneous ordinary differential equations is written in matrix form as following: [ [ [

According to the equation 3.21 transforming the above using modal matrix the following will be obtained:

[66]

(3.35)

The above equation is in the form of: (3.36)

The system of equation for the two degrees of freedom is written as:

(3.37)

And thus it can be solved using the numerical integration method provided in the previous section. The key issue here is the external forces as they dont have any specified function. Thus the problem turns into a simulation problem as the value of is required at any

given time step so that the numerical integration will be feasible. To achieve this it is required to apply a suitable probability distribution for wind and wave load. This is undertaken by using the stochastic modelling of joint wind and wave loads from the work of Ditlevse (2002) on the basis of which all of the probabilistic and simulation methods are established. In this study the idea of 10 minute sea states are considered where the mean values of wind and wave loads are considered to be constant. The distribution of corresponding wind and wave data is then introduced based on each 10 minute sea states. Furthermore as waves has been considered to be statistically and mechanically dependant on wind, in order to evaluate the wave height one needs to have a value for wind and applying it to the relative distribution. According to Ditlevsen (2002) the following expression for probability distribution function of wind pressure has been provided: | (3.38)

Where and

is random mean wind pressure, is given based on the following expression: [

(specified by Ditlevsen (2002))

(3.39)

Where

is the height above the sea at the point of interest from the sea surface. For example

for an offshore wind turbine which its hub is a 100 meters above the sea level the z value [67]

must be substituted by a 100 inside the above so that the corresponding

is obtained. The

equation 3.38 is a probability distribution function and as it was described in section 3.1.2 by inversing that and having the corresponding probability of an event which is between 0 and 1, one is able to find the corresponding value for wind pressure by solving for as following: ( | | ) . The inversion is carried out

(3.40)

In the simulation methods carried out in MATLAB this process is undertaken and wind pressure is evaluated for each simulating loop. A pseudo random number is generated and has been substituted for the parameter | in the above equation and having the is simulated which also will be used to

other parameters the corresponding value for

evaluate the wind force induced at top of the tower at each sea state i based on the blade area and relative drag coefficient. The mean value for the wind force for each sea state denoted by 1990): (3.41) Where represents the simulated mean wind pressure for each sea state, is the total is the is

and is calculated from the following familiar fluid dynamic expression (Mott

area of blades and is taken as three times of the area of a single blade, and

coefficient of drag which in this study conservatively has been taken to be 0.8. According to Novak et al. (2001) in order to evaluate the load induced by waves one needs to have relative wave period and height. The maximum load induced from wave that is comprised of inertia and drag components normal to a cylinder will be then obtained from the following expression based on their geometrical mean (Dean and Dalrymple 1984): ( Where and ) (3.42)

correspond to maximum drag and inertia forces respectively and . The drag and inertia

represents the load induced by waves for a given sea state

components are evaluated based on the amplitude of their relative periodic time varying wave expressions and can be obtained from the formulations provided below (Novak et al. 2001):

[68]

(3.43) (3.44) The above are the maximum drag and inertia forces Where coefficient respectively, and denote drag and inertia is wave height, and

represents the density of the sea water,

is the diameter of the cylinder of interest that in the case of this study the average diameter of the tower tube structure. The values of 1.2 and 1.5 respectively. The expressions for 2001): (3.45) And (3.46) In the above equations, and is known as the wave circular frequency, is wave number and represents the depth of and and in this study have taken to be equal to are given as following (Novak et al.

are wavelength and wave period respectively and finally

the water. This value in the case of offshore wind turbines is the height of the tower that is inside the water. In the above equation the value of wave length can be evaluated from the formula based on the sea conditions at the sea state of interest and can be obtained from the formulation below where the value for is computed as where g and T are

acceleration of gravity and wave period respectively and denotes the water depth (She 2010) : And (3.48) From the above formulas and having the wave period and wave height it will be feasible to determine the total force induced by wave to the offshore wind turbine tower structure. The values of wave period and height are taken to be completely probabilistic for which the suitable distribution will be utilized. These values are required to be simulated for any given [69]
3

(3.47)

sea state in order to contribute to the probabilistic aspects of the analysis and allow for the presence of uncertainty in the state of wave and wind. To simulate the mean value of the total force applied to the structure due to waves (which is statistically dependant to the wind pressure as well) the corresponding parameters for the equations 3.43 and 3.44 must be evaluated for each sea state. Firstly it is possible to carry out simulation for the wave period based on the following distribution provided by Ditlevsen (2002): ( ) (3.49) denotes

The above demonstrates the cumulative distribution of wave period. The function

the standard normal probability which includes mean and standard deviation equal to 0 and 1 respectively. It is possible to inversely calculate the wave period related to each simulated probability based on the generation of pseudo random numbers. To achieve this one has to calculate the inverse of the above distribution i.e. solve for normal probability function to both sides it will be seen: ( ) (3.50) . Apply the inverse of standard

Thus it can be obtained: ( ) can be computed for each sea state i as: (3.52) (3.51)

Hence the wave period


( )

In the above equation the

can be computed using the invers standard probability function from the above it can be possible to

that has been provided in MATLAB. Finding the

compute the wave height for the sea state using the joint probability distribution of wave period, wave height and wind pressure. To undertake this, the joint distribution function of wave height conditioned to a wind pressure and wave period is given by Ditlevsen (2002) as bellow:

[70]

| ( ( [

| | | ) (3.53) )

Where: [ ( ) ] (3.54) (3.55) ( In which ) and and (3.56) were

given previously. From the above equation in order to simulate the wave height , it is necessary to carry out some algebraic operations. Consider the equation 3.53, it is seen that by having the values inside the brackets which can be simulated as well as the specified constants, the only remaining unknown will be which has been expressed in terms of in

the equation 3.54. Apply the inverse of the standard normal probability function to both sides it is clearly seen: ( Hence ( And ( ) (3.59) ) (3.58) [ ( ) ]) (3.57)

Now re apply the inverse of the standard normal probabilities function to both sides: ( ) (3.60)

[71]

And thus the value for ( )

can be expressed as: (3.61)

From the above equation it will be feasible to simulate the wave height for each sea state. Firstly, the value for 3.53: | (3.62) correspond to the sea state i is simulated based on the equation

Rearranging the above: Replacing simulated | and ( | ) (3.63)

term in the above by a generated pseudo random number, and from the equations 3.55 and 3.56 for each sea state the corresponding , it is then possible to compute value from the

value for that state is obtained. Having

equation 3.61 with regards to each sea state. This value together with wave period and length are applied in equation 3.42 to produce the mean value of maximum wave load within any specific sea state. It is seen the dependency of wave properties to wind pressure has been taken into account by using the equation 3.53. It is worth noting that the current process will enable the simulation of the Mean value for this force. Whereas in order to use this value for each sea state in the dynamic analysis, there will be additional simulation correspond to each time step regarding numerical integration of the equations of motion termed as secondary simulation. This secondary simulation will be based on the primarily simulated mean values of the sea state and specified standard deviations. So the total of forces simulated will be equal to the number of sea states multiplied by the number of time steps for integration. Having the vector of forces generated probabilistically from the above based on presented methodologies; the equation of motions for two degrees of freedom system can be solved numerically using the values of forces for each time step. For instance consider and correspond to the final vectors of simulated wave and wind forces based on the

above then in order to apply them in the equation of motion for the two degrees of freedom system one needs to evaluate the forces based on the contribution of both modes of vibrations as below (Paz and Leigh 2004): [72]

{ { Where vibration and

} } and

(3.64) (3.65) are the modal vectors for the 1st and 2nd mode of represent the element in the vectors and . Thus the

and

compact version of the equations will be then obtained as below:

(3.66)

The number of time steps which is also related to the chosen number of sea states for the analysis will be determined based on the required accuracy of the computation as well as the computational constraints. Computational constraints also indicate that using inefficient analytical assumptions can cause the computations to become extremely time consuming and thus ineffective. In this study the maximum analysis length has been taken to 2 days. That equates to minutes which include number of sea states.

That indicates that there will be a necessity to primarily simulate the mean value for wave and wind forces at least 288 times. Furthermore, considering the time step of 0.5s for the dynamic analysis the total number of 600/0.5=1200 secondary simulations will be carried out based on the previously obtained mean values and a specified standard deviation with regards to each sea state. Thus the total number of integration steps required for the short term dynamic analysis will be equal to 288 x 1200 = 345600. The total displacement responses for the short term dynamic analysis will be evaluated from the above procedure. In order to enhance that to the long term analysis one needs to assemble suitable probabilistic properties of such information. To do that first the maximum displacement values for every 1 hour of analysis is obtained for each degree of freedom and is recorded accordingly. The total number of 48 maximum values will be found for the period of analysis explained in the above for each degree of freedom. Having this data, the probability distribution function and probability density function of them is determined using curve fitting methods. Thus probability of the displacement value being less or greater than any value can be found by having those probabilistic properties explained. This information can be used to calculate failure probability for any specified value of displacement threshold.

[73]

3.1.5 MATLAB CODE DESCRIPTION

As it has been mentioned before, in this study the mathematical programme MATLAB has been used to carry out the main calculations. In this section the developed MATLAB code that is used in this study will be explained. As the computation could be very intensive, care has been taken regarding the programming efficiency and to minimize the unwanted calculation wherever it was necessary. The various part of the code will be explored and the full code has also been put in the appendix section of this report. The first section of the code involves the initialization of the different parameters required for the analysis:
Clc %Clear the screen

clear all; %Clear the memory

%2DOF system stochastic dynamic analysis of an offshore wind turbine %subject to correlated wind and wave dynamic loading

%Aerodynamic and Hydrodynamic Coeffs CdragWind=0.8; CdragWave=1.2; %Conservative For Cylinder CinertiaWave=1.5;

%Enter the air and water density (kg/m3) AirDens=1.2; SeaWaterDens=1020;

%Enter a Blade Sizes (m) bladeLength=35; bladeWidth=5; bladeArea=bladeLength*bladeWidth; %to be used for Wind Force Calculation

%Enter the tower height htower=100;

%Enter the bottom width and top width wb=3.5; wt=3;

[74]

%Enter the thickness of the tower tube tt=.01;

%Enter the water depth dw=25;

%Enter the tower mass mt=100000;%kg

%Enter the hub mass mh=80000;%kg

As it has been also described in the code description itself, the computation constants are specified which also represent the problem. Items such as hydrodynamic and aerodynamic coefficients regarding the wave and wind load as well as the structural properties such as the size of the tower tube structure at the top and the bottom and the masses correspond to the tower and the hub. Another important factor is the depth of the water which will play a very important role in other stages of the programme. The following piece of code will compute the average second moment of area for the section of the tower structure as it will not be constant from bottom to the top:

%second momoent of areas for bottom and top part(m4) wm =wb-2*dw/htower*(wb-wt)/2; dob=wb; dib=wb-2*tt; dom=wm; dim=wm-2*tt; dot=wt; dit=wt-2*tt; im =3.14/64*(dom^4-dim^4); %2nd moment of area at the point of water level ibw=3.14/64*(dob^4-dib^4); %2nd moment of area at the base itw=3.14/64*(dot^4-dit^4); %2nd moment of area at the top ib =(im+ibw)/2; it =(im+itw)/2; %average 2nd moment of area for bottom member %average 2nd moment of area for top member

[75]

The second moment of area formulation for the tubes is calculated as: (3.67) Because the applied modelling is based on a two degrees of freedom system the structure has been modelled as:

Figure 3.14-Structural modelling

And the corresponding stiffness properties must be evaluated for the bottom member and top member separately. In the above code an average value of cross section has been used for each bottom and top members by finding for the top and the bottom and then to divide their

summation by 2. This calculation must be done only once to enable the evaluation of the stiffness matrix. The following section finds the stiffness matrix as below:
%stiffness matrix assemblage E=200*10^9; %Elastic modulus in N/m2 hb=dw; ht=htower-hb; kb=12*E*ib/hb^3; kt=12*E*it/ht^3; k=[kb+kt,-kt;-kt,kt];

[76]

Using the values for the second moment areas calculated previously and considering the elastic modulus of steel equal to 200 Giga Pascal, the stiffness coefficients are calculated as below Chopra (2007): (3.68) and are elastic modulus and second moment of area of the related member and will be equal to the water depth specified

Where

denotes its height. For the bottom member

previously and for the top member will be equal to the total height of the tower minus the height of the bottom member. The total stiffness matrix is then evaluated from the following arrangements: [ ] (3.69)

The following code will establish the matrix of masses as below: [ ] (3.70)

And then using the previously computed stiffness matrix it solves the eigenvalue problem to determine the natural frequencies of the structure:
%mass matrix assemblage (kg) m=[mt,0;0,mh];

%Evaluating the Spectral matrix syms xx; fn=solve((k(1,1)-m(1,1)*xx)*(k(2,2)-m(2,2)*xx)-k(1,2)*k(2,1));%natural Frequency fn=double(fn); Sp=[fn(2),0;0,fn(1)];%Assemblage of the Spectral Matrix(for w^2)

The following parabolic equation is solved: (( ) ) ( ) ( ) (3.71)

[77]

The above equation represents the following determinant which has been equated to zero: |[ [ | (3.72)

The spectral matrix is then constructed based on the following: [ Where and ] (3.73) denote the natural frequency for each mode. The next section is very

important where the natural modes of vibration based on evaluation of the corresponding matrix elements are obtained:
%Evaluating the Modal matrix w1=(fn(2))^.5; w2=(fn(1))^.5; ModeCoef1=(m(1,1)*w1^2-k(1,1))/k(1,2); TempModeAnswer=double(solve(m(2,2)*(ModeCoef1*xx)^2+m(1,1)*xx^2-1)); Modal(1,1)=TempModeAnswer(1); Modal(2,1)=Modal(1,1)*ModeCoef1;

ModeCoef2=(m(1,1)*w2^2-k(1,1))/k(1,2); TempModeAnswer=double(solve(m(1,1)*xx^2+(ModeCoef2*xx)^2*m(2,2)-1)); Modal(1,2)=TempModeAnswer(1); Modal(2,2)=Modal(1,2)*ModeCoef2;

Here the values so called ModeCoef1 and ModeCoef2 related to each modal vector are defined. These are used to generalize the formulation on finding the modal coefficients based on the described concepts in the previous section regarding mechanical concepts. On that basis the element of modal matrix will be obtained one by one by using the above procedure. Thus the ModeCoef1 and ModeCoef2parameteres can be defined similarly from the expression below:
( )

(3.74)

[78]

Having

the

above

parameters

the

following

equations

which

represent

the

between normalized modes will be solved using the relative values from the previous sections: (3.75) (3.76) Where { } and { } . The above equations are solved separately and vectors

of modal coefficients are obtained. To elaborate further on the above equations, consider the 1st mode, the equation above will be based on solving the equation below simultaneously with an equation from the eigenvalue problem used to find the natural frequency of the problem similar to the procedure described in equation 3.15: { } [ { } (3.77) assembled, in the next section the damping matrix will be

Having the modal matrix [

found and the diagonal mass, stiffness and damping matrix are evaluated:
%Assembling the matrix of damping coefficients zeta1=.10; zeta2=.05; w1=(fn(2))^.5; w2=(fn(1))^.5; aCoef = 2*w1*w2/(w2^2-w1^2)*[w2,-w1;-1/w2,1/w1]*[zeta1;zeta2]; c=aCoef(1)*m+aCoef(2)*k; %Matrix of Damping

%Matrix Diagonalization mm=Modal'*m*Modal; cc=Modal'*c*Modal; kk=Modal'*k*Modal;

For the structure used in this analysis the damping ratio corresponding 1st and second degrees of freedom are considered to be and . These values are assumed just to

demonstrate the analysis approach however their exact values must be found from [79]

experimentation or a dedicated research. According to Chopra (2007) the matrix of damping coefficients can be defined based on the mass and stiffness matrix as below: (3.78) Where and can be obtained from the following matrix expression:

]{ }

(3.79)

This procedure has been applied in the code. Finally the diagonal matrices of mass, stiffness and damping are obtained by using the modal matrix: [ [ [ Where [ [ [ [ [ is transposed of the matrix [ [ [ (3.80) (3.81) (3.82) . At this stage the simulation of the parameters

corresponding wind and wave loads are begun. The procedure has been described in the previous section in details. Here the procedure has been codified:

q0=(5*log(100*75))^2; al=.7; be=2.5; for i=1:1000 t(i)=exp(norminv(rand)*.17+log(6.8)); Wavedelta(i)=dw/(9.81*(t(i))^2); if Wavedelta(i)<.08 WaveLenth(i)=9.81*(t(i))^2*(Wavedelta(i)13.5*(Wavedelta(i))^2+60*(Wavedelta(i))^3)^.5; else WaveLenth(i)=9.81*(t(i))^2/(2*3.14); end q(i)=133.33333*log(1-rand)+q0; y(i)=(log(t(i))-log(6.8))/.17; z(i)=norminv(1-exp(-.0075*q(i))); x(i)=0.14142*norminv(rand)+0.7*y(i)+0.7*z(i); h(i)=norminv((normcdf(x(i))+1)/2)*be+al;

[80]

end

%Converting Wind Pressure and Wave Loads to Newtonian Force for i=1:1000

F2i(i)=bladeArea*q(i)*CdragWind*3; induced by Wind kWave(i)=2*3.14/WaveLenth(i); %Computing Wave Number sigma(i)=2*3.14/t(i); A1Wave(i)=(sigma(i))^2/kWave(i);

%Force

ADwave(i)=A1Wave(i)*(sinh(2*kWave(i)*dw)+2*kWave(i)*dw)/(sinh(kWave(i)*dw)) ^2; F1drag(i)=CdragWave*SeaWaterDens/32*(h(i))^2*wb*ADwave(i); Drag Component F1iner(i)=CinertiaWave*SeaWaterDens*3.14*wb^2/8*A1Wave(i)*h(i); %Wave inertia Component F1i(i)=((F1drag(i))^2+(F1iner(i))^2)^.5; Induced Force %Wave %Wave

end

As it is seen in the above code, firstly the initial values are assigned to the relative variables. Then the programme will run 1000 loops and in each single loop values of wind pressure, wave height and period based on the procedure explained in the previous section using equations 3.40, 3.52, 3.54, 3.55, 3.56, 3.61 and 3.63. The reason the total number of loops are equal to 1000 is to provide flexibility in the number of sea states chosen for the analysis time span and one can change the number of loops without constantly changing the number of loops for the wave height and other parameters. Additionally, the extra number of loops will have no effect on the results of the analysis. The different built-in MATLAB functions are also used here. For instance, the function norminv finds the inverse of standard normal probability which had been shown as in the previous section regarding the analyses

framework. Once the wave height and wave period as well as wind pressure are simulated for each sea state, the second loops apply the data to simulate the forces and store them inside [81]

vectors to be used for the next stage of the analysis. The next stage is quite important as it uses to the mean value generated previously and simulates forces for each time step of integration using those values and specified standard deviation of 0.1 x Mean (i.e. coefficient of variation equal to 0.1):
for j=1:144*2; % no. of 10 mins required

for i=1200*j-1199:j*1201 F1(i)=abs(normrnd(F1i(j),.1*F1i(j))); F2(i)=abs(normrnd(F2i(j),.1*F2i(j))); F(1,i)=Modal(1,1)*F1(i)+Modal(2,1)*F2(i); %Modal Transformation 1st DOF F(2,i)=Modal(1,2)*F1(i)+Modal(2,2)*F2(i); %Modal Transformation 2nd DOF end end %%Force Simulation completed

In the above piece of code, so called nested loops are used. That indicates the application of one set of loops inside another one. Here the first loop which can be termed as the parent loop will run for 288 times which is equal to the total number of sea states i.e. total number of 10 minutes required for the analysis. At each parent loop which also correspond to each sea state, the second loop or termed as sub loops will be run for 1201 times where the forces are generated at each sub loop based on the mean values and standard deviation as it was described earlier. The loops are connected with each other as 1 increase in the parent loop will be translated using a mathematical mechanism by which each set of 1201 values are assigned inside the main force vectors in the correct position. For instance, for j=1 the sub loops will run for 1201 times and the location of the simulated values inside the main vector will be defined by 1200 X 1 1900 to 1201 X 1 that means from 1 to 1201 and so on. The extra 1 for simulation for each of the time steps are due to the vector sizes compatibility with the numerical integration procedure used in the next stages. At each loop also the transformation is in accordance to the equations 3.64 and 3.65 are carried out for each of the simulated forces and the results are stored in the vectors which will be used for numerical integration using the average acceleration method. [82]

In this method one needs to assign the values for initial displacements, accelerations and velocities for both degrees of freedom. The code undertaking this is given below:
%Numerical Integration Using Average Acceleration Method %Computing the corresponding values

dt=.5;

k_hat1=kk(1,1)+2/dt*cc(1,1)+4/dt^2*mm(1,1); a1=4/dt*mm(1,1)+2*cc(1,1); b1=2*mm(1,1);

k_hat2=kk(2,2)+2/dt*cc(2,2)+4/dt^2*mm(2,2); a2=4/dt*mm(2,2)+2*cc(2,2); b2=2*mm(2,2); %Initializing Dynamic Properties u1(1)=0; uD1(1)=0; uDD1(1)=0;

u2(1)=0; uD2(1)=0; uDD2(1)=0;

In the above code firstly the size of the time step used for integration has been specified to 0.5 seconds. This also corresponds to the number of loops used in the previous part where the secondary simulations were carried out. The initial values have been allocated to the relative variables. The mathematical formulations of those were provided in the previous section regarding mechanical concepts where the process of integration was presented:

[83]

%First degree of freedom p1=F(1,:); delta_p1(1)=p1(2)-p1(1); delta_p_hat1(1)=delta_p1(1)+a1*uD1(1)+b1*uDD1(1); delta_u1(1)=delta_p_hat1(1)/k_hat1; delta_uD1(1)=2/dt*delta_u1(1)-2*uD1(1); delta_uDD1(1)=4/dt^2*(delta_u1(1)-dt*uD1(1))-2*uDD1(1);

%Second degree of freedom p2=F(2,:); delta_p2(1)=p2(2)-p2(1); delta_p_hat2(1)=delta_p2(1)+a2*uD2(1)+b2*uDD2(1); delta_u2(1)=delta_p_hat2(1)/k_hat2; delta_uD2(1)=2/dt*delta_u2(1)-2*uD2(1); delta_uDD2(1)=4/dt^2*(delta_u2(1)-dt*uD2(1))-2*uDD2(1);

The above code carries out the analysis for the first value of displacements, velocities and accelerations. This is prerequisite before entering the main loop of integration. The

corresponding formulation for different parameters was presented in section 3.1.3. Finally the main loops start based on the following codes:

for i=2:2*144*1200

uD1(i)=uD1(i-1)+delta_uD1(i-1); uDD1(i)=uDD1(i-1)+delta_uDD1(i-1); delta_p1(i)=p1(i+1)-p1(i); delta_p_hat1(i)=delta_p1(i)+a1*uD1(i)+b1*uDD1(i); delta_u1(i)=delta_p_hat1(i)/k_hat1; delta_uD1(i)=2/dt*delta_u1(i)-2*uD1(i); delta_uDD1(i)=4/dt^2*(delta_u1(i)-dt*uD1(i))-2*uDD1(i);

u1(i)=u1(i-1)+delta_u1(i-1);

end

[84]

for i=2:2*144*1200

uD2(i)=uD2(i-1)+delta_uD2(i-1); uDD2(i)=uDD2(i-1)+delta_uDD2(i-1); delta_p2(i)=p2(i+1)-p2(i); delta_p_hat2(i)=delta_p2(i)+a2*uD2(i)+b2*uDD2(i); delta_u2(i)=delta_p_hat2(i)/k_hat2; delta_uD2(i)=2/dt*delta_u2(i)-2*uD2(i); delta_uDD2(i)=4/dt^2*(delta_u2(i)-dt*uD2(i))-2*uDD2(i);

u2(i)=u2(i-1)+delta_u2(i-1);

end

Both loops run for 345600 times which translates to the period of analysis of two days. The results are found for each mode separately and in sequence and stored in the relative vectors. Then the results ate transformed back using modal vectors and then are superimposed to evaluate the final matrix that contains the displacement responses of the structure for two days with respect to time. This is achieved using the following code:
U1st=Modal(:,1)*u1; U2nd=Modal(:,2)*u2;

Utot=U1st+U2nd;

Disp1=Utot(1,:); Disp12=abs(Utot(2,:)-Utot(1,:));

The matrix Utot in the above contains all the displacement information of both degrees of freedom with respect to time. Additionally the vectors Disp1 and Disp12 include the displacement for the first degree of freedom and absolute value of the differential displacement for the second and first degrees of freedom with respect to time. In order to be able to extract the probabilistic and statistic weight of the above information and then enable their application for future analysis one can assemble the distribution of hourly maximum displacements for a period of two days. The following code was used to determine the

[85]

matrices that contain the maximum values of the absolute displacement and differential displacements per hour for the period of two days for each degree of freedom:
%Finding the max displaecments in 2 days and producing the distribution of %maximum value per hour for the total of two days clc DispAbs1=abs(Disp1); kk=1;

for i=1:7200:345888-(7200-1)

jj=1; for j=i:i+7200-1

Disp1Interval(jj)=DispAbs1(j); %putting the displacements in a given interval Disp21Interval(jj)=Disp12(j); jj=jj+1;

end MaxAbsDisp1(kk)=max(Disp1Interval); MaxAbsDisp21(kk)=max(Disp21Interval); kk=kk+1; end

The code uses the nested loop where the parent loop will run from 1 with the increment of 7200 to 338689 to cover all the hours existing in the period of analysis. The sub loop will be run 7200 times for every loop done by the parent for command to check the data and place correctly in order to place inside the vectors Disp1Interval and Disp21Interval. Once the sub loops for every parent loop is completed the maximum displacement for every hour will be obtained and placed inside another vectors namely MaxAbsDisp1 and MaxAbsDisp21 for each of the degrees of freedom. Here there have been wide usages of increment-able variables which are kk and jj which manage the process of placing variables inside the vectors. The analysis is completed here for which the results will be shown in section 3.3.

[86]

3.2 VALIDATION OF RESULTS AND METHODS

The full analysis approach and the computer code that was used to apply the approach to a given structure based on the specified conditions were presented previously. It was also observed that the analysis was based on probabilistic modelling provided by Ditlevsen (2002) from which the relevant parameters that contribute to the wind and wave corresponding to each sea states and the time steps constituents were simulated. Subsequently these values were stored inside the vectors of forces and the dynamic analysis based on the numerical integration using the average acceleration method were utilized to evaluate the vectors of displacement responses with respect to time. However in order to use the results confidently one needs to validate the approach based on selected results in comparison to some other methods. This will provide further assurance that the approach as well as the implantation of the approach using computer coding is justified. In this section the objective is to use the other possible analysis and compare the results from the existing analysis and critically observe the presence of differences. The difference then should be investigated to check the causes and to draw a conclusion regarding effectiveness of the approach presented in this study. The comparisons have been done based two aspects of the analysis, firstly the implantation of the numerical integration regarding the dynamic analysis is validated and then the displacement values are compared with the case where the same loads applied to the structure statically in order to check the difference between the maximum displacement from the static analysis and the mean displacement from the dynamic analysis. In order to validate the computer implantation of the numerical integration for the average acceleration method, the code is used to carry out the analysis for another equation for which the results have already been computed and presented in the book of Chopra (2007). Then the computed results are compared with ones that have been given. Consider the example 5.3 in the page 178 of Chopra (2007) where the initial values for mass and stiffness are provided. The results from the computation using the average acceleration method have been presented in a tabulated form in page 179. The results are shown below:

[87]

Table 3.3-Results from Chopra (2007) Page 178-179

It is seen that the analysis is done for the period of 1 second using time steps equal to 0.1 seconds. The forces corresponding to each time step are given in the next column. The full results are shown accordingly. The same vectors of forces based on the same value of time steps are used in the following MATLAB code in which the integration part is identical to the one that was used for the analysis approach:
clear all clc

dt=.1; k=10; m=.2533; c=.1592; k_hat2=k+2/dt*c+4/dt^2*m; a2=4/dt*m+2*c; b2=2*m;

u2(1)=0; uD2(1)=0; uDD2(1)=0;

p2=[0,5,8.6602,10,8.6603,5,0,0,0,0,0,0,0,0,0]; delta_p2(1)=p2(2)-p2(1); delta_p_hat2(1)=delta_p2(1)+a2*uD2(1)+b2*uDD2(1); delta_u2(1)=delta_p_hat2(1)/k_hat2; delta_uD2(1)=2/dt*delta_u2(1)-2*uD2(1);

[88]

delta_uDD2(1)=4/dt^2*(delta_u2(1)-dt*uD2(1))-2*uDD2(1);

for i=2:11

uD2(i)=uD2(i-1)+delta_uD2(i-1); uDD2(i)=uDD2(i-1)+delta_uDD2(i-1); delta_p2(i)=p2(i+1)-p2(i); delta_p_hat2(i)=delta_p2(i)+a2*uD2(i)+b2*uDD2(i); delta_u2(i)=delta_p_hat2(i)/k_hat2; delta_uD2(i)=2/dt*delta_u2(i)-2*uD2(i); delta_uDD2(i)=4/dt^2*(delta_u2(i)-dt*uD2(i))-2*uDD2(i);

u2(i)=u2(i-1)+delta_u2(i-1);

end

It is seen that the vector of force values and the time step duration is similar to the example in the text book. Now the results are computed and presented below:
Table 3.4-Results from the MATLAB Code

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

0 0.0436665968001118 0.232616041706599 0.612061262014571 1.08252288660969 1.43092576138115 1.42304972919922 0.962160399708597 0.190788705076606 -0.604332843286676 -1.14412140982012

[89]

The results are almost identical to each other. To quantify the difference, the results have been placed inside an excel spread sheet were the differences in terms of percentage are calculated based on the formula below: || | | | | || (3.83)

The spread sheet can be found below:


Table 3.5-Comparing the both results

Time 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

u from Chopra 0.0437 0.2326 0.6121 1.0825 1.4309 1.4231 0.9622 0.1908 -0.6044 -1.1442

u from MATLAB 0.043660 0.23262 0.61206 1.08252 1.43093 1.42305 0.96216 0.19079 -0.60433 -1.14412

Error (%) 0.091533 0.006897 0.006329 0.002114 0.0018 0.003532 0.004116 0.00592 0.011111 0.006869

As its clearly observed above the error is very low as the maximum error is found to be around 0.09% which indicates the accuracy of the code and the implementation of the method in MATLAB. Subsequently the structure is modelled in SAP2000 as a vertical cantilever with a tubular cross section and a fixed bottom. The loadings are randomly chosen mean values for wave and wind force and the similar geometrical and mechanical properties as below where the loads are applied statically and the maximum displacement of the top of the structure were recorded:

[90]

Figure 3.15-Modelling in SAP2000 under static loads for a step of integration

[91]

Figure 3.16-Modelling 3d View

The unit of the forces shown in the above is Newton. The analysis is completed and the maximum displacement recorded at the top is shown in the following figure:

[92]

Figure 3.17-Analysed shape and deformation

The maximum displacement at the top is equal to 8 meters based on the given mechanical properties and loadings. The mean value of displacement recorded from dynamic analysis is found to be:

There is a significant difference between the two results. However this can be considered to be justified as regarding the static case where there will be enough time for the structure to deform. This is not the case for dynamic loads as they are applied for a very short period of time of 0.5 seconds and indicating a very low excitation period with respect to the natural period of the structure and thus the structure will not have enough time to deform fully indicating that the system is inertia driven (Molenaar and Dijkstra 1999). The natural frequency of the first mode of vibration is found from the code and formulation as equal to and is well below the excitation frequency which is equal:

[93]

Based on the study presented by Molenaar and Dijkstra (1999) the following figure demonstrates the nature of the structural response based on the relation between excitation frequency and the natural frequency of the structure:

Figure 3.18(a)-Excitation frequency much smaller than natural frequency

Figure 3.18(b)-Excitation frequency close to natural frequency (resonance)

Figure 3.18(c)-Excitation frequency well above the natural frequency

Figure 3.18-Response behaviour based on the relation of excitation frequency to natural frequency (Molenaar and Dijkstra 1999) (Red Curves denote response, Blue Curves denote Excitation force)

[94]

The above graphs are based on a harmonic excitation of a single degree of freedom system and are used to demonstrate the relation between the response behaviour and the natural and excitation frequency. The first graph indicates the situation that the excitation frequency is well below that of natural frequency of the structure which will result in a quasi-static response. Thus the behaviour of the structure will be as if the load is applied statically. Figure b represents a situation in which the natural frequency and the frequency of excitation are close so there will be a much larger response suggesting the state of resonance. Figure c represents a similar case to the structure analysed in this study where the natural frequency is much less than the excitation frequency. It is seen that the response will be much smaller than the static response hence the situation is considered to be dominated by inertia. Thus it is observed that the big difference in the static and dynamic displacements have mechanical explanation which also implies that the study evidently has achieved the results that represents the physical reality with regards to the dynamic behaviour of the structure and therefore based on the above concepts, the results seem to be valid.
3.3 ANALYSIS RESULTS

The detailed description of the analysis procedure was presented in section 3.1.4. Furthermore the methodologies used for analysis were validated and the effectiveness of them was observed. In this section the results of the analysis will be presented based on the selected structural arrangement described in section 3.1.1. The results will include the time history of forces and responses of the structure for each degrees of freedom separately, differential displacements and finally the statistical frequency histograms. Accordingly, the probability density functions of the maximum hourly values of displacements for the period of 2 days will be provided. There will be two sets of results, the first set which also is the main result is based on the correct assumption of load simulation time steps and the second is based on inaccurate assumptions regarding the time step of the integration. The latter is presented in order to solely demonstrate the progress by which the final approach was developed and the effective methods were achieved. Clearly the first set of results is regarding the main approach which has been developed based on the correct assumption of time steps of integration. Additionally the units are in Newton and seconds, and the largest version of the diagrams has also been placed in the appendices.

[95]

3.3.1 MAIN RESULTS

In this section the main analysis results which has been on the basis of correct assumptions regarding time steps are demonstrated. These results firstly include the time history of forces regarding the 1st and 2nd degrees of freedom:

Figure 3.19- The variations of Wave force with respect to time

Figure 3.20-The variations of Wind Force with respect to time

[96]

The figures 3.19 and 3.20 present the simulated values of wave and wind forces respectively corresponding to each time step. The following are the time history displacement responses based on the above forces:

Figure 3.21-Displacement response of the first degree of freedom with respect to time

Figure 3.22- Displacement response of the 2nd degree of freedom with respect to time

[97]

The variation of displacement for each degree of freedom with respect to time has been demonstrated. Furthermore the variation of differential displacement with respect to time is given in the following graph:

Figure 3.23- Differential Displacement response (

It is noted that there is not much difference between the displacement response of the 2nd degree of freedom and the differential displacement of the two degrees of freedom; this is because the magnitude of the displacement for the first degree of freedom is much smaller than the second one and thus the difference will become visibly minor. The following diagrams demonstrate the absolute values for the displacement with respect to time. These will be used to evaluate the distribution of the maximum hourly displacements:

[98]

Figure 3.24-Absolute Displacement response values of the first degree of freedom with respect to time

Figure 3.25- Absolute differential Displacement response values of with respect to time

The distribution of maximum hourly values will be evaluated from the histogram representing the frequencies of the data. The following demonstrate the histograms of the maximum hourly displacement:

[99]

PDF

U1 Figure 3.26-Histogram and corresponding curve of the maximum hourly values of 1st degree of freedom displacements

PDF

U12 Figure 3.27-Histogram and corresponding curve of the maximum hourly values of differential displacements

Furthermore the cumulative distribution functions have also been evaluated in accordance to the results:

[100]

CDF

Figure 3.28- Cumulative distribution curve of the maximum hourly values of 1 st degree of freedom displacements

U1

CDF

U12 Figure 3.29- Cumulative distribution curve of the maximum hourly values of differential displacements

It is also observed that the suitable distribution curves have been fitted and plotted on the frequency histograms and thus the probability density functions of those can be evaluated using the MATLAB curve fitting module as below. As the results are positive the lognormal [101]

density functions have been established based on the relative statistical properties. These values have been determined by MATLAB automatically once the curve fitting has been completed: Probability density function of maximum hourly displacements values for the analysis of a period of 2 days regarding the first degree of freedom:
3

Probability density function of maximum hourly values for the analysis of a period of 2 days regarding the differential displacement values:
3

This will conclude the results regarding the main analysis for this section. The next part will present some force and displacement results which had been achieved previously based on an inaccurate assumption. These have been placed here to demonstrate the general progress in which the above results have been evaluated.
3.3.2 INITIAL RESULTS BASED ON AN INACCURATE TIME STEPPING ASSUMPTION

Originally the integration had been carried out based on the assumption of producing different mean values of wind and wave forces for each time step. That means the analysis had been undertaken based on presuming that each sea state corresponded to a time step. This however was inaccurate as the probabilistic modelling for loads had been achieved based on the work of Ditlevsen (2002) in which the sea state had been considered to be equal to 10 minutes. This was a contradiction as the dynamic integration was based on the time stepping period of less than a second which also had a relation with the natural frequency of the structure. Once this contradiction was discovered the new method of analysis was introduced where the sea state was assumed to be 10 minutes and the load fluctuations within these sea state was considered probabilistically which produced the results that was presented previously. In this section however the results from the initial approach is provided so that the progress of the development of the suitable approach is observed.

[102]

The following shows the force simulations for the analysis:

Figure 3.30- The Variations of Wave force with respect to time (based on the initial incorrect assumption)

Figure 3.31- The Variations of Wind force with respect to time (based on the intial incorrect assumption)

[103]

The chosen displacement responses for the 1st and 2nd degree of freedom are shown here:

Figure 3.32- Displacement response of the first degree of freedom with respect to time (based on the intial incorrect

assumptions)

Figure 3.33- Displacement response of the 2nd degree of freedom with respect to time (based on the intial incorrect

assumptions)

[104]

There is noticeably less fluctuations within the responses based on the initial incorrect assumptions in comparison to the results from the main analysis which is based on a 10 minute sea state consideration. This shows that the main analysis has taken into account the correct sea state assignment and thus the higher level of fluctuation is observed accordingly.
3.4 DISCUSSION OF RESULTS

The results for the main analysis were produced in the previous section. The results included time history Reponses for all the degrees of freedom individually as well as their differential time history of lateral responses and their corresponding histograms and as a result the probability density functions of the maximum hourly displacement values for the period of two days which were presented in mathematical form. Accordingly, the results contain the contribution from the probabilistic wind and wave load simulated primarily on the basis of which the load fluctuations during each sea state were produced. This can be seen from figures 3.19 and 3.20 that there are evidently a higher number of small fluctuations than larger fluctuations. This suggests that the forces have been simulated in accordance to the analysis strategy in which each sea state corresponds to a mean value for wind and wave but every time step contains a simulated value of force based on the sea state mean and a given standard deviation. Therefore the micro fluctuations are apparent. Furthermore there are implications regarding the dynamic behaviour which suggested that the two degrees of freedom system is dominated by inertia due to the much larger excitation frequency in comparison to the natural frequency of the structure. This can demonstrate the importance of the natural frequency of the structure and its relation to the excitation frequency. On that basis, one can use the analysis to determine the suitable mass and stiffness so that amplification of the responses due to the resonance is avoided. Accordingly this can be an important factor for reducing the deformation due to the structural vibration by determining the suitable natural frequency of the structure. This is an important achievement when it comes to the design of the tower structure. it can also enable the design process to incorporate a realistic stochastic wave and wind modelling into the dynamic analysis so that the analysis can further converge to the representation of the realistic offshore conditions. Additional significant parameters that were achieved from the analysis were probabilistic distributions of hourly maximum displacements for the period of two days. This data can be used based on a similar approach used by Moan et al. (2011) to be extended to the long term by considering a displacement threshold as follows. Considering the structural design [105]

requires the structure to experience maximum deformation less that the value

. The

probability density functions of the maximum displacement have been evaluated from the analysis. Thus consider the probability of failure for the 2 day period as: (3.84) Where is the probability of failure and is based on the failure criteria of that a denotes the probability density function of

displacement exceeds a certain value and

hourly maximum displacements that can be for the 1st or 2nd degree of freedom dependant on the objective of the computation. Now say it is desired to compute the probability of failure for a period of number of 2 days. can be defined by the required period for instance a and is equal to 15. Therefore it can

required period of 30 days this value will be be seen:


N Times

(3.85)

Thus the

statement can be replaced by union operator


N Times

which will result in:

(3.86)

And from De Morgan's laws it is seen (Spiegel and Murray 2000):


N Times

(3.87)

[106]

As the probability of exceedance for each 2 days are independent from each other then it is shown:
N Times

(3.88)

Then from equation 3.84:

(3.89)

The above equation can be used to determine the probability of exceedance in any required long term period. This can be used in determining the maximum stress induced due to displacement or to design the structure by determining suitable stiffness and mass so any displacement in a given period of time is avoided. The value will contain contribution from the presence of uncertainty in an offshore environment and dynamic behaviour of the structure. The analysis result can also be used in an estimation of the fatigue life of the tower structure similar to the approach used by Moan et al. (2011) by counting the amplitudes and gather the suitable data in order to generate the corresponding fatigue curves (Davison and Owens 2003). The results can be used to determine the short term and long term fatigue life. Accordingly the Same method as the above also can be utilized to determine the response of the structure for different periods. Another key point to note in here is the time period in which the analysis is carried out. This can define the number of sea states required by finding the number of 10 minutes within the required period and hence effects the number of Monte Carlo simulation loops for the mean value of forces. Additionally this can increase or decrease the length integration. For instance for the analysis of a period of 2 days the integration will be carried out for 10 X 6 X 24 X 2 X 1200. The number 1200 corresponds to the number of time steps which exist in any sea state and it can also be changed which again affect the length of integration but doesnt have any influence on the number of Monte Carlo loops that generate the mean value of forces. Both values, i.e. analysis period and size of the time step can influence concurrently the total length of the computation and can be used as a determining factor regarding the analysis. A lengthy computation can make the analysis less effective as it could be the case that further [107]

simulations may not provide additional noticeable accuracy in the results. Thus at the stage of determining these factors one has to incorporate the computation time and the level of accuracy within an optimized volume of computation.

[108]

4 CONCLUSIONS

The use of Wind energy as a renewable source of energy has attracted attention from scientists and engineers for decades. New methods have been introduced to efficiently harvest Wind energy such as onshore and offshore wind turbines. Offshore wind turbines are a smart solution as they can be exposed to more constant wind velocity in an open sea environment and can minimize the use of land in urban regions. However they usually require a much more elaborate engineering approach to understand their structural behaviour and responses. Being subject to offshore conditions these structures experience a combination of various environmental actions. Wind and wave loads in particular are amongst these actions which also contain a considerable amount of uncertainty regarding their magnitudes and directions. Furthermore these actions are not independent from each other for example a higher wind velocity can result in a higher load induced by waves to the structure. In addition to those, wind and wave actions are very much time dependant and their variations with time must be considered for studying the structural behaviour of offshore wind turbines. In this report the behaviour of the supporting tower structure of offshore wind turbines has been studied. The structure was considered to be subject to simultaneous wind and wave dynamic actions based on a suitable probabilistic modelling of various contributing parameters corresponding to each load. The loads were simulated using the Monte Carlo method and the short time analysis was carried out accordingly thus the displacement responses with respect to time were evaluated. Finally suitable probabilistic distributions of maximum values were established and the methodologies by which the short term results can be extended to longer term results were explored. The structural modelling applied in this study is carefully selected so that it is not based on a very complex finite element approach and at the same time is not as simple as a single degree of freedom system. This was done because most of the works carried out in previous studies didnt include that modelling approach. Throughout the study there was an extensive application of computer coding in the mathematical software MATLAB where suitable codes were developed. The codes can be used to model wind and wave loading for a desired period of time which can be usable for other studies. Furthermore the code can also be applied for the analysis of any structure to undertake a sensitivity study corresponding to each factor. Moreover, the results from the analysis can be used for the design of the tower structure as well as estimating the short term and long term structural responses. Additionally by establishing any failure criteria it will be possible to obtain probability of failure for a given period of time and reliability of the [109]

structure accordingly based on the probabilistic distributions computed in this study. Upon accomplishment of this study the importance of the presence of uncertainty and its influence on the behaviour of the offshore wind turbine tower structure was observed. It also became evident that offshore structures are subject to a very probabilistic environment which must not be over looked in any engineering analysis. Furthermore it was understood that the dynamic response can differ to quasi-static response based on the natural frequency of the structure and its relation to the frequency of excitation. For future works, it is recommended to carry out the design of the tower structure based on the presented analysis results and make a comparison to an alternative approach and observe the results. Furthermore the future work can include optimization of the structure based on the dynamic response and its relation to frequency of excitation and natural frequency of the structure to achieve the highest structural efficiency.

[110]

5 REFERENCES

Al Hamaydeh, M., Hussain, S., (2010). Optimized frequency-based foundation design for wind turbine towers utilizing soilstructure interaction. Journal of the Franklin Institute. Volume 348, Issue 7, September 2011, Pages 14701487 Bhattacharya, S., Adhikari, S., (2011). Experimental validation of soilstructure interaction of offshore wind turbines. Soil Dynamics and Earthquake Engineering Volume 31, Issues 5-6. Bilfinger Bergerm, (2008). Horns Rev 2 Offshore Wind farm Monopile Foundations. [Online]. Available at: http://www.fzbau.de/C1257130005050D5/vwContentByKey/W27KFKHE534VENSDE/$FI LE/Horns%20Rev%202_BB.pdf [Accessed: 24 June 2011] Butterfield, S., Musial, W., Jonkman, J., Sclavounos, P., Wayman, E. (2005). Engineering Challenges for Floating Offshore Wind Turbines. Copenhagen Offshore Wind 2005 Conference and Expedition Proceedings, 25-28 October 2005, Copenhagen, Denmark. Camp, T. R. et al., (2003). Design methods for offshore wind turbines at exposed sites. [Online]. Available at: http://www.offshorewindenergy.org/reports/report_020.pdf [Accessed: 22 June 2011] Caroline Mawer (2011). Mills driven by the wind. [Online]. Available from: http://www.carolinemawer.co.uk/blog/?p=1469 [Accessed: 22 June 2011] Cheremisinoff, N.P. (1979). Fundamentals of Wind Energy 2nd Ed. Ann Arbor, Science Publisher Inc. Chopra A. K., (2007). Dynamics of Structures: Theory and Applications to Earthquake Engineering, 3rd Edition. Prentice-Hall: Upper Saddle River. Cushman-Roisin, B. (2011). Wind Energy Systems. [Online]. 2011. Available from http://engineering.dartmouth.edu/~cushman/courses/engs44/wind.pdf. [Accessed: 22 June 2011] Davison, B., Owens, G.V., (2003). Steel Designers Manual 6th Ed., Blackwell Publishing.

[111]

Dean R.G. and Dalrymple R.A., (1984). Water Wave Mechanics for Engineers and Scientists. Prentice-Hall. Der Tempel J. V., (2006). Design of support structures for offshore wind turbines. A Thesis Submitted in fulfilment of the Requirements of Delft University for the Degree of Doctor of Philosophy. Published and distributed by author in cooperation with Offshore engineering: Delft and DUWIND: Delft. Det Norske Veritas, (2004). Offshore Standard DNV-OS-J101 Design of Offshore Wind Turbine Structures (Det Norske Veritas, Hvik, Norway,) Ditlevsen, O. (2002). Stochastic model for joint wave and wind loads on offshore structures. Structural Safety, 24:139-163. Ditlevsen, O.,Madsen, H. O., (1996). Structural Reliability Methods. Chichester: John Wiley & Sons. Energy Types (2011). Wind Mill Types. [Online]. Available from: http://www.energy-types.info/about/renewable-sources/wind/windmilltypes [Accessed: 22 June 2011] Environmental and Energy study Institute, (2010). Offshore wind energy. [Online]. Available at: http://www.eesi.org/files/offshore_wind_101310.pdf [Accessed: 24June 2011] Europes Energy Portal. (2009). Europe's onshore and offshore wind energy potential. [Online]. 2009 http://www.energy.eu/publications/a07.pdf [Accessed: 22 June 2011] Firth, N. (2010). The giant sycamore of the seas: British engineers design 900ft offshore wind turbine which mimics tree's seeds. dailymail. [Online]. 30th of July. Available at http://www.dailymail.co.uk/sciencetech/article1297713/Giant-offshore-wind-farm-mimics-sycamore-seed-joins-racedevelop-generation-turbine.html [Accessed: 23 June 2011] Gao Z, Moan T. (2010). Long-term fatigue analysis of offshore fixed wind turbines based on time-domain simulations. In: Proceedings of PRADS. 2010. Generator blog (2010). What affect the efficiency of wind generators. [Online]. Available from: http://www.aboutgenerator.com/2010/07/what-affects-theefficiency-of-wind-generator/ [Accessed: 20 Aug 2011] Genta, G.,(1999). Vibration of Structures and Machines, Springer-Verlag: New York (third Ed). Ghouri, S. S., (2007). Global energy outlook. [Online]. Available at: http://www.worldenergy.org/documents/p000616.pdf [Accessed: 23 June 2011] [112]

Gipe, P., 1995. Wind Energy Comes of Age. Wiley, New York. Hardy, C (2010). Renewable energy and role of Marykirk's James Blyth. The Courier. D. C. Thomson & Co. Ltd. Available from http://www.thecourier.co.uk/Community/Heritage-andHistory/article/2332/renewable-energy-and-role-of-marykirk-s-jamesblyth.html. [Accessed: 22 June 2011] Hart, Gary C., 1982. Uncertainty analysis, loads, and safety in structural engineering. Englewood Cliffs; London: Prentice-Hall.

Henderson, A., Morgan, C., Smith, B., Srensen, H. C., Barthelmie, R.,and Boesmans, B. (2003). Offshore wind power: A major new source of energy for Europe. Intl. J. Environment and Sustainable Development, 1(4), 356370. Henderson, A.R., Leutz, R., and Fujii, T., (2002). Potential for Floating Offshore Wind Energy in Japanese Waters. Proceedings of The Twelfth (2002) International Offshore and Polar Engineering Conference, Kitakyushu, Japan, May 2631, 2002, ISBN: 1-880653-58-3; ISSN: 1098-6189. Heronemus, W. E., (1972). Pollution-Free Energy from Offshore Winds, 8th Annual Conference and Exposition Marine Technology Society, Washington D.C., September 11-13. Humar, J.L., (2002). Dynamics of Structures, 2nd Edition. A.A. Balkema Publishers: Lisse. Lackner MA, Rotea MA., (2010). Structural control of floating wind turbines. Mechatronics, doi:10.1016/j.mechatronics.2010.11.007 Lavassas, I., Nikolaidis, G., Zervas, P., Efthimiou, E., Doudoumis, I., and Baniotopoulos, C. (2003). Analysis and design of the prototype of a steel 1-MW wind turbine tower. Engineering Structures, 25(8), pp. 10971106. Madsen, P. H. (2010), State of art design practices for offshore wind farms. [Online]. 2010. Available from http://www.sintef.no/project/Nowitech/Wind_presentations/Madsen,%20P .H.,%20Ris%C3%B8.pdf. [Accessed: 23 June 2011] Maritime Journal (2011). Photos. [Online]. Available from: http://www.maritimejournal.com/__data/assets/image/0008/174671/New s_Wind.jpg [Accessed: 20 June 2011]

[113]

Masuda, A., Kusaba, T., Marubayashi, K. and Ishibashi, M., (1998). Statistics of Wind and Waves off Tsuyazaki, Fukuoka, in the Eastern Tsushima Strait, Annual Report of Oceanographic Data at the Tsushima Station, Inst. Res. Applied Mech., Kyushu Univ.,pp207-236. Metropolis, N. (1987). The Beginning of the Monte Carlo Method. Los Alamos Science[online],125-130,Available at: http://jackman.stanford.edu/mcmc/metropolis1.pdf [Accessed 20 April 2010] Moan T., Dong W. and Gao Z., (2011). Long-term fatigue analysis of multiplanar tubular joints for jacket-type offshore wind turbine in time domain, Engineering Structures Volume 33, Issue 6, June 2011, Pages 2002-2014. Molenaar, D.P., Dijkstra S.j., (1999). Modelling the structural dynamics of flexible wind turbines. In Proceedings of European Wind Energy Conference and Exhibition, Acropolis Convention Centre, Nice, France Mott R.L. (1990). Applied Fluid Mechanics. Macmillian Inc. Musial, W., Butterfield, S., (2004). Future for offshore wind energy in the United States, conference paper preprint. Report Number NREL/CP-50036313, National Renewable Energy Laboratory, Golden, Colorado. Available at: http://www.nrel.gov/docs/fy04osti/36313.pdf. [Accessed: 26 June 2011] My Eco Project (2010). Offshore Wind Farm photo. [Online]. Available from: http://myecoproject.org/wp-content/uploads/2010/10/offshore-windturbines.jpg [Accessed: 20 Aug 2011] Nacfaire, H. (1988). Grid Connected Wind Turbines. London and New York: Elsevier Applied Science. Nixon, N. (2008). Timeline: The history of wind power. Guardian. [Online]. 17th of October 2008. Available athttp://www.guardian.co.uk/environment/2008/oct/17/wind-powerrenewable-energy [Accessed: 23 June 2011] Novak, P., Moffat, A.I.B., Nalluri, C., and Narayanan, R. (2001). Hydraulic Structures. 3rd edition. Spon Press, London, UK. Orkney Sustainable Energy (2010). Costa Head Experimental Wind Turbine. Orkney Sustainable Energy Website. Orkney Sustainable Energy Ltd. Available from http://www.orkneywind.co.uk/costa.html. [Accessed: 22 June 2011]

[114]

Paz M., Leigh W., (2004). Structural Dynamics Theory and Computation, Kluwer Academic Publishing Groups. Petersen et al. (2010). Evaluate The Effect of Turbine Period of Vibration Requirements on Structural Design Parameters: Technical Report of Findings. [Online]. Available at: http://www.boemre.gov/tarprojects/651/M10PC00066-8VibrationFinal.pdf [Accessed: 27 June 2011]
Ragheb, M., (2010). Offshore wind farm sitting. [Online]. Available at: https://netfiles.uiuc.edu/mragheb/www/NPRE%20475%20Wind%20Power%20S ystems/Offshore%20Wind%20Farms%20Siting.pdf [Accessed: 23June 2011]

Rao, S., 1992. Reliability-based design. London: McGraw-Hill. Renewable Energy World (2011). Offshore wind booming in Europe.[Online]. Available at: http://www.renewableenergyworld.com/rea/news/article/2011/01/offsho re-wind-booming-in-europe??cmpid=WNL-Friday-January21-2011 [Accessed: 23 June 2011] Robinson, M. and Musial W., (2006). Offshore wind technology overview. [Online]. 2006 Available at: http://www.nrel.gov/docs/gen/fy07/40462.pdf [Accessed: 22 June 2011] Ronold, K.O., J. Wedel-Heinen, and C.J. Christensen (1999). Reliability-based fatigue design of wind turbine rotor blades. Engineering Structures, 21, 1101 1114. Saouma. V.,(2004). Lecture notes in Structural Engineering. University of Colorado. Available at: http://ceae.colorado.edu/~saouma/LectureNotes/se.pdf. [Accessed: 26 June 2011] Scira Offshore Energy, (2010). Sheringham Shoal Offshore Wind Farm Fact Sheet. [Online]. Available at: http://www.scira.co.uk/NewsDownloads/Downloads/Sheringham%20Shoal %20Factsheet%20Sept%202010.pdf[Accessed: 24 June 2011] Seajacks UK, (2010). Offshore Wind Turbine Installation (parts 1 and 2). [Online video]. 29 September 2010. Available from: http://www.youtube.com/user/SeajacksUK. [Accessed: 25 June 2011] She, K. (2010). Wave and Coastal Engineering. [Lecture notes]. Brighton University, 2010. Siddall, J. 1983. Probabilistic engineering design: principles and applications. New York : M. Dekker

[115]

Siegfriedsen, Lehnhoff, & Prehn, (2003). Aerodynamic engineering, GmbH Presented at Offshore Wind Energy in the Mediterranean and other European Seas Conference, April 10-12, Naples, Italy. Simiu E. and Scanlan H., (1986). Wind effects on structures: an introduction to wind engineering. 2nd Ed. New York: Wiley. Smith, G. N., 1986. Probability and statistics in civil engineering: an introduction. London: Collins Professional and Technical. Soong, T. T., 1981.Probabilistic modelling and analysis in science and engineering. New York ; Chichester : Wiley Srensen, J. D., Frandsenb S., and Tarp-Johansen N., (2008). Effective turbulence models and fatigue reliability in wind farms. Probabilistic Engineering Mechanics, vol. 23, pp. 531538, 2008. Srensen, J. D., Frandsenb S., and Tarp-Johansen N., (2008). Effective turbulence models and fatigue reliability in wind farms. Probabilistic Engineering Mechanics, vol. 23, pp. 531538, 2008. Spiegel J., Murray R., 2000. Schaum's outline of theory and problems of probability and statistics. 2nd ed. New York; London: McGraw-Hill. Tan, S.G. and Wichers J.E.W. (1990), Hydrodynamic developments in offshore engineering, Marine & Offshore Technology Schip & Werf, 57ste Jaargang, nr. 4, April 1990 Tomosawa, F., and Noguchi, T., 1993. Relationship between Compressive Strength and Modulus of Elasticity of High-Strength Concrete, Proceedings of the Third International Concrete, V. 2, Lillehammer, Norway, pp. 1247-1254. Symposium on Utilization of High-Strength Uys, P.E., Farkasb, J., Jarmaib, K. & van Tondera, F. (2006). Optimisation of a steel tower for a wind turbine structure, Journal of Engineering Structures, doi:10.1016/j.engstruct.08.011. Walpole, R., Myers, R., (1989). Probability and statistics for engineers and scientists.4th ed. London: Macmillan. White, L. (1962), Medieval Technology and Social Change, Oxford University Press, Oxford. Wind Power (2010). Aerogenerator X Concept. [Online]. Available from: http://www.windpower.ltd.uk/ [Accessed: 20 June 2011]

[116]

Wind Tech International (2010). UK reaches 5GW milestone. [Online]. Available at: http://www.windtech-international.com/industry-news/ukreached-5gw-milestone. [Accessed: 23 June 2011] Wyatt, A.(1996) Electric Power: Challenges and Choices, Book Press Ltd., Toronto. Yang, T. (2008). Behaviour of plastic design of steel structures [Online]. Available from http://peer.berkeley.edu/~yang/courses/ce248/CE248_CN_Wind_loads.pdf [Accessed: 20 June 2011] Zaaijer, M. B., (2006). Foundation modeling to assess dynamic behaviour of offshore wind turbines. Applied Ocean Research. Vol. 28, pp. 45-57.

[117]

6 APPENDICES 6.1 MATLAB CODE FOR THE MAIN ANALYSIS


clc clear all; %2DOF system stochastic dynamic analysis of an offshore wind turbine %subject to corrolated wind and wave dynamic loading %Aerodynamic and Hydrodynamic Coeffs CdragWind=0.8; CdragWave=1.2; %Conservative For Cylinder CinertiaWave=1.5; %Enter the air and water density (kg/m3) AirDens=1.2; SeaWaterDens=1020; %Enter a Blade Sizes (m) bladeLength=35; bladeWidth=5; bladeArea=bladeLength*bladeWidth; %to be used for Wind Force Calculation %Enter the tower height htower=100; %Enter the bottom width and top width wb=3.5; wt=3; %Enter the thikness of the tower tube tt=.01; %Enter the water depth dw=25; %Enter the tower mass mt=100000;%kg %Enter the hub mass mh=80000;%kg %Computation of the engineering properties %second momoent of areas for bottom and top part(m4) wm =wb-2*dw/htower*(wb-wt)/2; dob=wb; dib=wb-2*tt; dom=wm; dim=wm-2*tt; dot=wt; dit=wt-2*tt; im =3.14/64*(dom^4-dim^4); %2nd moment of area at the point of water level ibw=3.14/64*(dob^4-dib^4); %2nd moment of area at the base itw=3.14/64*(dot^4-dit^4); %2nd moment of area at the top ib =(im+ibw)/2; %average 2nd moment of area for bottom member

[118]

it =(im+itw)/2;

%average 2nd moment of area for top member

%stiffness matrix assemblage E=200*10^9; %Elastic modulus in N/m2 hb=dw; ht=htower-hb; kb=12*E*ib/hb^3; kt=12*E*it/ht^3; k=[kb+kt,-kt;-kt,kt]; %mass matrix assemblage (kg) m=[mt,0;0,mh]; %Evaluating the Spectral matrix syms xx; fn=solve((k(1,1)-m(1,1)*xx)*(k(2,2)-m(2,2)*xx)-k(1,2)*k(2,1));%natural Frequency fn=double(fn); Sp=[fn(2),0;0,fn(1)];%Assemblage of the Spectral Matrix(for w^2) %Evaluating the Modal matrix w1=(fn(2))^.5; w2=(fn(1))^.5; ModeCoef1=(m(1,1)*w1^2-k(1,1))/k(1,2); TempModeAnswer=double(solve(m(2,2)*(ModeCoef1*xx)^2+m(1,1)*xx^2-1)); Modal(1,1)=TempModeAnswer(1); Modal(2,1)=Modal(1,1)*ModeCoef1; ModeCoef2=(m(1,1)*w2^2-k(1,1))/k(1,2); TempModeAnswer=double(solve(m(1,1)*xx^2+(ModeCoef2*xx)^2*m(2,2)-1)); Modal(1,2)=TempModeAnswer(1); Modal(2,2)=Modal(1,2)*ModeCoef2; %Assembling the matrix of damping coefficients zeta1=.10; zeta2=.05; w1=(fn(2))^.5; w2=(fn(1))^.5; aCoef = 2*w1*w2/(w2^2-w1^2)*[w2,-w1;-1/w2,1/w1]*[zeta1;zeta2]; c=aCoef(1)*m+aCoef(2)*k; %Matrix of Damping %Matrix Diagonalization mm=Modal'*m*Modal; cc=Modal'*c*Modal; kk=Modal'*k*Modal; %%Monte Carlo Loops Begins Here

%%Force Loop Starts %Simulating the Wind and then Wave which is dependant to the Wind q0=(5*log(100*75))^2; al=.7; be=2.5; for i=1:1000 t(i)=exp(norminv(rand)*.17+log(6.8)); Wavedelta(i)=dw/(9.81*(t(i))^2); if Wavedelta(i)<.08

[119]

WaveLenth(i)=9.81*(t(i))^2*(Wavedelta(i)13.5*(Wavedelta(i))^2+60*(Wavedelta(i))^3)^.5; else WaveLenth(i)=9.81*(t(i))^2/(2*3.14); end q(i)=133.33333*log(1-rand)+q0; y(i)=(log(t(i))-log(6.8))/.17; z(i)=norminv(1-exp(-.0075*q(i))); x(i)=0.14142*norminv(rand)+0.7*y(i)+0.7*z(i); h(i)=norminv((normcdf(x(i))+1)/2)*be+al; end %Converting Wind Pressure and Wave Loads to Newtonian Force for i=1:1000

F2i(i)=bladeArea*q(i)*CdragWind*3; induced by Wind kWave(i)=2*3.14/WaveLenth(i); %Computing Wave Number sigma(i)=2*3.14/t(i); A1Wave(i)=(sigma(i))^2/kWave(i);

%Force

ADwave(i)=A1Wave(i)*(sinh(2*kWave(i)*dw)+2*kWave(i)*dw)/(sinh(kWave(i)*dw)) ^2; F1drag(i)=CdragWave*SeaWaterDens/32*(h(i))^2*wb*ADwave(i); %Wave Drag Component F1iner(i)=CinertiaWave*SeaWaterDens*3.14*wb^2/8*A1Wave(i)*h(i); %Wave inertia Component F1i(i)=((F1drag(i))^2+(F1iner(i))^2)^.5; %Wave Induced Force end for j=1:144*2; % no. of 10 mins required for i=1200*j-1199:j*1201 F1(i)=abs(normrnd(F1i(j),.1*F1i(j))); F2(i)=abs(normrnd(F2i(j),.1*F2i(j))); F(1,i)=Modal(1,1)*F1(i)+Modal(2,1)*F2(i); %Modal Transformation 1st DOF F(2,i)=Modal(1,2)*F1(i)+Modal(2,2)*F2(i); %Modal Transformation 2nd DOF end end %%Force Simulation completed

%Numerical Integration Using Average Acceleration Method %Computing the corresponding values dt=.5; k_hat1=kk(1,1)+2/dt*cc(1,1)+4/dt^2*mm(1,1); a1=4/dt*mm(1,1)+2*cc(1,1); b1=2*mm(1,1); k_hat2=kk(2,2)+2/dt*cc(2,2)+4/dt^2*mm(2,2); a2=4/dt*mm(2,2)+2*cc(2,2);

[120]

b2=2*mm(2,2); %k_hat1=w1^2+2/dt*2*zeta1*w1+4/dt^2*1; %k_hat2=w2^2+2/dt*2*zeta2*w2+4/dt^2*1; %a1=4/dt*1+2*2*zeta1*w1; %a2=4/dt*1+2*2*zeta2*w2; %b1=2; %b2=2; %1st DOF Analysis %Initializing Dynamic Properties u1(1)=0; uD1(1)=0; uDD1(1)=0;

p1=F(1,:); delta_p1(1)=p1(2)-p1(1); delta_p_hat1(1)=delta_p1(1)+a1*uD1(1)+b1*uDD1(1); delta_u1(1)=delta_p_hat1(1)/k_hat1; delta_uD1(1)=2/dt*delta_u1(1)-2*uD1(1); delta_uDD1(1)=4/dt^2*(delta_u1(1)-dt*uD1(1))-2*uDD1(1);

for i=2:2*144*1200 %No. of .5 seconds exist inside the whole analysis span = 1 day at the moment = j * 1200 uD1(i)=uD1(i-1)+delta_uD1(i-1); uDD1(i)=uDD1(i-1)+delta_uDD1(i-1); delta_p1(i)=p1(i+1)-p1(i); delta_p_hat1(i)=delta_p1(i)+a1*uD1(i)+b1*uDD1(i); delta_u1(i)=delta_p_hat1(i)/k_hat1; delta_uD1(i)=2/dt*delta_u1(i)-2*uD1(i); delta_uDD1(i)=4/dt^2*(delta_u1(i)-dt*uD1(i))-2*uDD1(i); u1(i)=u1(i-1)+delta_u1(i-1);

end %2nd DOF Analysis %Initializing the Dynamic Properties u2(1)=0; uD2(1)=0; uDD2(1)=0;

p2=F(2,:); delta_p2(1)=p2(2)-p2(1); delta_p_hat2(1)=delta_p2(1)+a2*uD2(1)+b2*uDD2(1); delta_u2(1)=delta_p_hat2(1)/k_hat2; delta_uD2(1)=2/dt*delta_u2(1)-2*uD2(1); delta_uDD2(1)=4/dt^2*(delta_u2(1)-dt*uD2(1))-2*uDD2(1);

for i=2:2*144*1200

[121]

uD2(i)=uD2(i-1)+delta_uD2(i-1); uDD2(i)=uDD2(i-1)+delta_uDD2(i-1); delta_p2(i)=p2(i+1)-p2(i); delta_p_hat2(i)=delta_p2(i)+a2*uD2(i)+b2*uDD2(i); delta_u2(i)=delta_p_hat2(i)/k_hat2; delta_uD2(i)=2/dt*delta_u2(i)-2*uD2(i); delta_uDD2(i)=4/dt^2*(delta_u2(i)-dt*uD2(i))-2*uDD2(i); u2(i)=u2(i-1)+delta_u2(i-1); end U1st=Modal(:,1)*u1; U2nd=Modal(:,2)*u2; Utot=U1st+U2nd; Disp1=Utot(1,:); Disp12=abs(Utot(2,:)-Utot(1,:)); %Finding the max displacements in 2 days and producing the distribution of %maximum value per hour for the total of two days clc DispAbs1=abs(Disp1); kk=1; for i=1:7200:345888-(7200-1) jj=1; for j=i:i+7200-1 Disp1Interval(jj)=DispAbs1(j); %putting the displacements in a given interval Disp21Interval(jj)=Disp12(j); jj=jj+1;

end MaxAbsDisp1(kk)=max(Disp1Interval); MaxAbsDisp21(kk)=max(Disp21Interval); kk=kk+1; end

[122]

6.2 MATLAB CODE TO CHECK THE IMPLEMENTATION OF THE AVERAGE ACCELERATION METHOD
clear all clc dt=.1; k=10; m=.2533; c=.1592; k_hat2=k+2/dt*c+4/dt^2*m; a2=4/dt*m+2*c; b2=2*m; u2(1)=0; uD2(1)=0; uDD2(1)=0;

p2=[0,5,8.6602,10,8.6603,5,0,0,0,0,0,0,0,0,0]; delta_p2(1)=p2(2)-p2(1); delta_p_hat2(1)=delta_p2(1)+a2*uD2(1)+b2*uDD2(1); delta_u2(1)=delta_p_hat2(1)/k_hat2; delta_uD2(1)=2/dt*delta_u2(1)-2*uD2(1); delta_uDD2(1)=4/dt^2*(delta_u2(1)-dt*uD2(1))-2*uDD2(1);

for i=2:11 uD2(i)=uD2(i-1)+delta_uD2(i-1); uDD2(i)=uDD2(i-1)+delta_uDD2(i-1); delta_p2(i)=p2(i+1)-p2(i); delta_p_hat2(i)=delta_p2(i)+a2*uD2(i)+b2*uDD2(i); delta_u2(i)=delta_p_hat2(i)/k_hat2; delta_uD2(i)=2/dt*delta_u2(i)-2*uD2(i); delta_uDD2(i)=4/dt^2*(delta_u2(i)-dt*uD2(i))-2*uDD2(i); u2(i)=u2(i-1)+delta_u2(i-1); end

[123]

6.3 MATLAB CODE REGARDING THE INITIAL ANALYSIS BASED ON INACCURATE ASSUMPTIONS
clc clear all; %2DOF system stochastic dynamic analysis of an offshore wind turbine %subject to corrolated wind and wave dynamic loading %Number of Monte Carlo Loops (must be an Even number) NoLoops=1000; %Aerodynamic and Hydrodynamic Coeffs CdragWind=0.8; CdragWave=1.2; %Conservative For Cylinder CinertiaWave=1.5; %Enter the air and water density (kg/m3) AirDens=1.2; SeaWaterDens=1020; %Enter a Blade Sizes (m) bladeLength=35; bladeWidth=5; bladeArea=bladeLength*bladeWidth; %to be used for Wind Force Calculation %Enter the tower height htower=100; %Enter the bottom width and top width wb=3.5; wt=3; %Enter the thikness of the tower tube tt=.01; %Enter the water depth dw=25; %Enter the tower mass mt=100000;%kg %Enter the hub mass mh=80000;%kg %Computation of the engineering properties %second momoent of areas for bottom and top part(m4) wm =wb-2*dw/htower*(wb-wt)/2; dob=wb; dib=wb-2*tt; dom=wm; dim=wm-2*tt; dot=wt; dit=wt-2*tt; im =3.14/64*(dom^4-dim^4); %2nd moment of area at the point of water level ibw=3.14/64*(dob^4-dib^4); %2nd moment of area at the base itw=3.14/64*(dot^4-dit^4); %2nd moment of area at the top

[124]

ib =(im+ibw)/2; it =(im+itw)/2;

%average 2nd moment of area for bottom member %average 2nd moment of area for top member

%stiffness matrix assemblage E=200*10^9; %Elastic modulus in N/m2 hb=dw; ht=htower-hb; kb=12*E*ib/hb^3; kt=12*E*it/ht^3; k=[kb+kt,-kt;-kt,kt]; %mass matrix assemblage (kg) m=[mt,0;0,mh]; %Evaluating the Spectral matrix syms xx; fn=solve((k(1,1)-m(1,1)*xx)*(k(2,2)-m(2,2)*xx)-k(1,2)*k(2,1));%natural Frequency fn=double(fn); Sp=[fn(2),0;0,fn(1)];%Assemblage of the Spectral Matrix(for w^2) %Evaluating the Modal matrix w1=(fn(2))^.5; w2=(fn(1))^.5; ModeCoef1=(m(1,1)*w1^2-k(1,1))/k(1,2); TempModeAnswer=double(solve(m(2,2)*(ModeCoef1*xx)^2+m(1,1)*xx^2-1)); Modal(1,1)=TempModeAnswer(1); Modal(2,1)=Modal(1,1)*ModeCoef1; ModeCoef2=(m(1,1)*w2^2-k(1,1))/k(1,2); TempModeAnswer=double(solve(m(1,1)*xx^2+(ModeCoef2*xx)^2*m(2,2)-1)); Modal(1,2)=TempModeAnswer(1); Modal(2,2)=Modal(1,2)*ModeCoef2; %Assembling the matrix of damping coefficients zeta1=.10; zeta2=.05; w1=(fn(2))^.5; w2=(fn(1))^.5; aCoef = 2*w1*w2/(w2^2-w1^2)*[w2,-w1;-1/w2,1/w1]*[zeta1;zeta2]; c=aCoef(1)*m+aCoef(2)*k; %Matrix of Damping %Matrix Diagonalization mm=Modal'*m*Modal; cc=Modal'*c*Modal; kk=Modal'*k*Modal; %%Monte Carlo Loops Begins Here for j=0:2:NoLoops

%%Force Loop Starts %Simulating the Wind and then Wave which is dependant to the Wind q0=(5*log(100*75))^2; al=.7; be=2.5; for i=1:1002 t(i)=exp(norminv(rand)*.17+log(6.8)); Wavedelta(i)=dw/(9.81*(t(i))^2);

[125]

if Wavedelta(i)<.08 WaveLenth(i)=9.81*(t(i))^2*(Wavedelta(i)13.5*(Wavedelta(i))^2+60*(Wavedelta(i))^3)^.5; else WaveLenth(i)=9.81*(t(i))^2/(2*3.14); end q(i)=133.33333*log(1-rand)+q0; y(i)=(log(t(i))-log(6.8))/.17; z(i)=norminv(1-exp(-.0075*q(i))); x(i)=0.14142*norminv(rand)+0.7*y(i)+0.7*z(i); h(i)=norminv((normcdf(x(i))+1)/2)*be+al; end %Converting Wind Pressure and Wave Loads to Newtonian Force for i=1:1002

F2(i)=bladeArea*q(i)*CdragWind*3; induced by Wind kWave(i)=2*3.14/WaveLenth(i); %Computing Wave Number sigma(i)=2*3.14/t(i); A1Wave(i)=(sigma(i))^2/kWave(i);

%Force

ADwave(i)=A1Wave(i)*(sinh(2*kWave(i)*dw)+2*kWave(i)*dw)/(sinh(kWave(i)*dw)) ^2; F1drag(i)=CdragWave*SeaWaterDens/32*(h(i))^2*wb*ADwave(i); %Wave Drag Component F1iner(i)=CinertiaWave*SeaWaterDens*3.14*wb^2/8*A1Wave(i)*h(i);%Wave inertia Component F1(i)=((F1drag(i))^2+(F1iner(i))^2)^.5; %Wave Induced Force F(1,i)=Modal(1,1)*F1(i)+Modal(2,1)*F2(i); %Modal Transformation 1st DOF F(2,i)=Modal(1,2)*F1(i)+Modal(2,2)*F2(i); %Modal Transformation 2nd DOF end %%Force Simluation compeleted

%Numerical Inetgration Using Average Acceleration Method %Computing the correspodnig values dt=0.1; k_hat1=kk(1,1)+2/dt*cc(1,1)+4/dt^2*mm(1,1); a1=4/dt*mm(1,1)+2*cc(1,1); b1=2*mm(1,1); k_hat2=kk(2,2)+2/dt*cc(2,2)+4/dt^2*mm(2,2); a2=4/dt*mm(2,2)+2*cc(2,2); b2=2*mm(2,2); %k_hat1=w1^2+2/dt*2*zeta1*w1+4/dt^2*1; %k_hat2=w2^2+2/dt*2*zeta2*w2+4/dt^2*1; %a1=4/dt*1+2*2*zeta1*w1; %a2=4/dt*1+2*2*zeta2*w2; %b1=2; %b2=2;

[126]

%1st DOF Analysis %Initializing Dynamic Propertis u1(1)=0; uD1(1)=0; uDD1(1)=0;

p1=F(1,:); delta_p1(1)=p1(2)-p1(1); delta_p_hat1(1)=delta_p1(1)+a1*uD1(1)+b1*uDD1(1); delta_u1(1)=delta_p_hat1(1)/k_hat1; delta_uD1(1)=2/dt*delta_u1(1)-2*uD1(1); delta_uDD1(1)=4/dt^2*(delta_u1(1)-dt*uD1(1))-2*uDD1(1);

for i=2:1000 uD1(i)=uD1(i-1)+delta_uD1(i-1); uDD1(i)=uDD1(i-1)+delta_uDD1(i-1); delta_p1(i)=p1(i+1)-p1(i); delta_p_hat1(i)=delta_p1(i)+a1*uD1(i)+b1*uDD1(i); delta_u1(i)=delta_p_hat1(i)/k_hat1; delta_uD1(i)=2/dt*delta_u1(i)-2*uD1(i); delta_uDD1(i)=4/dt^2*(delta_u1(i)-dt*uD1(i))-2*uDD1(i); u1(i)=u1(i-1)+delta_u1(i-1);

end %2nd DOF Analysis %Initializing the Dynamic Propertis u2(1)=0; uD2(1)=0; uDD2(1)=0;

p2=F(2,:); delta_p2(1)=p2(2)-p2(1); delta_p_hat2(1)=delta_p2(1)+a2*uD2(1)+b2*uDD2(1); delta_u2(1)=delta_p_hat2(1)/k_hat2; delta_uD2(1)=2/dt*delta_u2(1)-2*uD2(1); delta_uDD2(1)=4/dt^2*(delta_u2(1)-dt*uD2(1))-2*uDD2(1);

for i=2:1000 uD2(i)=uD2(i-1)+delta_uD2(i-1); uDD2(i)=uDD2(i-1)+delta_uDD2(i-1); delta_p2(i)=p2(i+1)-p2(i); delta_p_hat2(i)=delta_p2(i)+a2*uD2(i)+b2*uDD2(i); delta_u2(i)=delta_p_hat2(i)/k_hat2; delta_uD2(i)=2/dt*delta_u2(i)-2*uD2(i); delta_uDD2(i)=4/dt^2*(delta_u2(i)-dt*uD2(i))-2*uDD2(i);

[127]

u2(i)=u2(i-1)+delta_u2(i-1); end U1st=Modal(:,1)*u1; U2nd=Modal(:,2)*u2; Utot=U1st+U2nd; StorageMat(j+1,:)=Utot(1,:); StorageMat(j+2,:)=Utot(2,:); %stores the 1st DOF displacements %stores the 2nd DOF displacements

Disp(j+1,:)=StorageMat(j+1,:); %Stores the the diplcements for 1st DOF Disp(j+2,:)=StorageMat(j+2,:)-StorageMat(j+1,:); %Stores the the differential diplcements between 1st and 2nd DOF end AbsDisp=abs(Disp); for i=1:NoLoops/2 %Finding the max displacement for each DOF and place thenm inside a vector MaxDisp1(:,i)=max(AbsDisp(2*i-1,:)); MaxDisp21(:,i)=max(AbsDisp(2*i,:)); end

[128]

6.4 SIMULATION OF WIND AND WAVE FORCES RESULTS

6.4.1 WAVE FORCE VS. TIME

[129]

6.4.2 WIND FORCE VS. TIME

[130]

6.4.3 DISPLACEMENTS VS. TIME

[131]

[132]

6.4.4 ABSOLUTE DISPLACEMENTS VS. TIME

[133]

[134]

6.4.5 PDF AND CDF FOR 2 DAYS HOURLY MAXIMUM DISPLACEMENTS

U1:

[135]

[136]

U21:

[137]

[138]

You might also like