## Abstract

The design of offshore wind turbines is one of the most fascinating challenges in renewable energy. Meeting the objective of increasing power production with reduced installation and maintenance costs requires a multi-disciplinary approach, bringing together expertise in different fields of engineering. The purpose of this theme issue is to offer a broad perspective on some crucial aspects of offshore wind turbines design, discussing the state of the art and presenting recent theoretical and experimental studies.

## 1. Introduction

In past decades, offshore wind energy has attracted a growing interest from scientists and engineers worldwide. After the first offshore wind farm, built in 1991 in relatively shallow waters (2–5 m) off the coast of Denmark at Vindeby, many others have been constructed, especially in the North and Baltic Seas, and new ones are being developed in Europe, USA, China and other countries [1]. Offshore wind energy still represents a small proportion of the total wind energy in the world, about 7 GW out of 318 GW at the end of 2013, but has an enormous potential. The European Wind Energy Association, for instance, estimates that offshore wind energy production in Europe will increase from the 6.5 GW at the end of 2013 to 150 GW in 2030, meeting approximately 14% of Europe's electricity demand [2,3]. Offshore sites offer indeed some considerable advantages over onshore sites, such as generally higher wind speeds with lower turbulence, availability of large areas for large-scale wind farms, lower visual impact at distance from the shore. However, owing to several issues such as more difficult and specialized installation procedures, more expensive support structures, more difficult environmental and working conditions, offshore wind energy may still be more expensive not only than onshore wind energy, but also than conventional power resources. Therefore, closing this gap has become a key step for a future sustainable exploitation of offshore wind energy potential.

Wind energy converters have a very long history that in Europe traces back to the Middle Ages (for more information on the historical development, see Manwell [4]). Nowadays, an offshore wind turbine is a complex ensemble of different components and subsystems: rotor, nacelle with powertrain, control and safety systems [5], and a support structure generally composed of a tower mounted on either a bottom-fixed substructure or a floating device moored to the seabed. Three-bladed, upwind rotors with horizontal axis are the conventional design solution in the industry, although alternative options (two blades, downwind rotor position or vertical axis) are under study. In the 23 years since the Vindeby offshore wind farm was built off the coast of Denmark, turbine size has increased from 450 kW to 3–5 MW and even larger turbines, with rated power up to 10 MW, are being tested to be released in the near future [1].

Most of the existing offshore wind farms are in shallow waters, generally less than 20 m deep. At these sites, wind turbines are mounted on monopiles driven into the seabed or resting on concrete gravity bases [6]. As the water depth increases, however, monopile solutions may require too large diameters to counteract the overturning moment and, for this reason, are not generally considered as economically feasible in waters deeper than 30 m. In transitional water depths, i.e. between 30 and 60 m, wind turbines mounted on space multi-footing substructures such as tripods and jackets are presently considered as a best option [7]. Tripods and jackets have been used at the Alpha Ventus wind farm (Germany) in approximately 30 m of water, and jackets at the Beatrice wind farm (UK) in nearly 45 m of water, for 5 MW turbines. For waters deeper than 60 m, however, current research is investigating wind turbines on floating devices as the most economically sustainable option [8–11]. Configurations under study are generally classified based on the primary physical principles adopted to achieve static stability: the spar-buoy (or ballast stabilized floater), whose stability is provided by a ballast lowering the centre of gravity below the centre of buoyancy; the tension leg platform (TLP) (or mooring stabilized floater), where stability is achieved through mooring lines kept under tension by excess buoyancy in the platform; and the barge (or buoyancy stabilized floater), where stability is achieved through the waterplane area. The spar-buoy and barge are generally moored by catenary lines, but the spar-buoy may be moored also by taut lines. Hybrid concepts using features from the three classes, such as semisubmersible floaters, are also a possibility [9]. The technical feasibility of multi-megawatt floating wind turbines has been already demonstrated by three prototypes: Hywind in the North Sea (2.3 MW turbine on a spar-buoy), WindFloat in the Atlantic (2 MW turbine on a semisubmersible floater) and Fukushima-FORWARD Phase 1 in Japan (2 MW turbine on a semisubmersible floater) [3,12,13]. Several other prototypes are under development and, at this stage, the current effort in industry and research is primarily focused on designing economic floating systems which can compete with bottom-fixed offshore turbines in terms of cost of energy. This has become a crucial goal in the perspective of moving wind turbines further offshore, with the purpose of minimizing visual impact and harnessing the large resources available; it has been estimated, for instance, that only the wind resource potential at 5–50 nautical miles off the US coast could provide the total electrical generating capacity currently installed in the USA (more than 900 GW) [14].

In past years, a number of international standards and guidelines have been released for design and assessment of offshore wind turbines [15–17]. In the meanwhile, numerical modelling has progressed towards more sophisticated descriptions of structural components, mechanical and electrical subsystems of modern offshore wind turbines, along with an appropriate treatment of the incident wind and wave fields. Existing numerical models rely on modal representation, multi-body, finite-element (FE) concepts, sometimes combined in mixed approaches, and involve system motion equations to be solved by fully coupled nonlinear time-domain integration, to account for inherent interactions between aerodynamic and hydrodynamic responses. In fact, while aerodynamic loads on the rotor are to be derived from aeroelastic models, considering the complex interaction between air flow and rotor blades, the influence of control systems and support structure dynamics, the latter affects the calculation of the hydrodynamic loads. Hence, although in some cases and especially in the early stages of design, simplified analyses are implemented with responses to wind and wave loads computed separately and next superposed [18,19], only fully integrated time-domain simulations are currently recommended as the basis of final, detailed wind turbine design calculations, owing to the role played by system nonlinearities inherent to rotor aerodynamics, hydrodynamics and control systems [20–22].

For either existing solutions in shallow–transitional waters or future projects in deeper waters, the objective of an efficient and cost-effective design poses engineering challenges and numerical modelling complexities, which can be solved only by an integrated approach combining the expertise in diverse fields such as, among others, aerodynamics, hydrodynamics, structural and foundation engineering; also, vibration control and health monitoring play an important role, in view of ensuring a reliable and continuous power production for sustainable investments [23,24]. Although an exhaustive description of all aspects of offshore wind turbines design is almost prohibitive, this theme issue will attempt to provide a broad perspective on the state of the art and most recent developments, with 16 papers on aerodynamics, hydrodynamics, new design solutions, experimental tests on prototypes, response to environmental loads, vibration control and health monitoring techniques. Attention will be focused primarily on horizontal axis wind turbines, by far the most common configuration adopted in existing installations and future projects, but potential advantages of vertical axis design will be also discussed.

## 2. Aerodynamics and hydrodynamics

Aerodynamic and hydrodynamic models are the basis of any theoretical and experimental study on offshore wind turbines. Therefore, prior to introducing all contributions of the theme issue in detail, a short description of current aerodynamic and hydrodynamic modelling for different types of offshore wind turbines, with a few considerations on applicability, may be appropriate here.

Since the early days of the wind industry, the blade element momentum (BEM) theory has been awarded particular favour as a robust and computationally inexpensive tool to model the rotor aerodynamics and calculate aerodynamic loads. BEM theory couples blade element theory and momentum theory. The blade element theory involves discretizing the blades into a number of independent aerodynamic elements treated as two-dimensional aerofoils, and expressing the global aerodynamic load as the sum of local blade forces obtained from local flow conditions using tabulated aerofoil data. The momentum theory assumes that the loss of momentum across the rotor is caused by the work done on the blades by the air flow, thus providing the global aerodynamic load in terms of the induced velocities that affect the local flow conditions and, consequently, the blade forces built by the blade element theory. Coupling the two theories implies an iterative process to determine the global aerodynamic loads along with the induced velocities near the rotor [25]. The BEM theory relies, however, on some oversimplifying approximations, such as assuming a steady, axial and two-dimensional air flow, and a disc-like modelling of the rotor [26,27]. Most of these limitations have been addressed by incorporating in the original BEM theory appropriate sub-models [27]; for instance, dynamic stall and dynamic inflow models to correct the steady-state assumption, models of yaw and tilt flows to correct the axial flow assumption, correction to two-dimensional aerofoil data to account for three-dimensional effects (e.g. [25] and references therein), and models of tip loss to compensate for rotor disc modelling [27]. With these important modifications, the BEM model is still the most used for rotor aerodynamics in commercial aeroelastic codes.

Considering, however, that reducing uncertainties and approximations in the calculation of aerodynamic loads is essential for an efficient design, a better understanding of rotor aerodynamics has been sought by alternative methods. Also, more sophisticated descriptions have become of particular interest in view of modelling the aerodynamics of floating wind turbines, where flow conditions may be more complex than in bottom-fixed wind turbines and can hardly be captured by BEM theory, owing to significant low-frequency platform motion (significant pitching motion is encountered at the incident-wave frequency) or severe yaw conditions [28].

Among alternatives to BEM theory, computational fluid dynamics (CFD) methods are certainly the most accurate ones [29]. CFD methods typically solve the Navier–Stokes equations governing the turbulent air flow, assumed to be either compressible or incompressible depending on the ratio between local wind speed and sound speed [25], in conjunction with appropriate turbulence models [29]. Turbulence models are necessary to deal with the wide range of time and length scales involved in a full solution of the Navier–Stokes equations (while in the atmospheric boundary layer the turbulence scales may vary from the order of 1 km to the order of 1 mm, inside the blade boundary layers scales may be even smaller [29]). Turbulence modelling is involved in the large eddy simulation (LES) method, where effects of unresolved small scales are included based on the behaviour of larger scales, in the Reynolds-averaged Navier–Stokes method and in a combination of the two, commonly referred to as the detached eddy simulation method [25,29]. Also, the solution of the Navier–Stokes equations is generally coupled with appropriate models of the wind turbine, such as an ‘exact’ direct modelling of the blades through a body-fitted grid, or the actuator disc (AD), the actuator line (AL) and the actuator surface (AS) models [29]. The first approach may be computationally expensive especially for modelling the boundary layer on the blades, including possible transition, separation and stall, and requires the generation of a high-quality moving mesh, commonly done with different overlapping grids communicating with each other [29]. Mainly to overcome these disadvantages, the AD, AL and AS models have been developed, where the aerodynamic loads are represented in the computational grid by body forces, computed using the wind velocity field obtained from the Navier–Stokes equations and based on tabulated aerofoil data in the same way as in the BEM theory [25]. In the AD model, the body forces are distributed on the entire rotor disc, whereas in the AL/AS models the loads are distributed around lines/planar surfaces along the actual blade positions. Alternative methods to CFD methods solving the Navier–Stokes equations are the so-called free-vortex wake methods, where the motion of fluid particles carrying vorticity is tracked in time and space [25,30]. The advantage over CFD methods is that only part of the space needs to be accounted for, namely the positions of the vorticity elements [26].

As for hydrodynamic loads, models generally depend on the support structure adopted. For bottom-fixed support structures, Morison’s equation is generally used, assuming that wave diffraction and radiation effects are negligible. This assumption is acceptable for slender bodies, i.e. with a small diameter with respect to wavelength of incident waves, and small motion of the support structure, but may become inaccurate for floating supports, where diameters may be large or motion can be significant. Hence, for floating supports alternative methods are adopted to compute hydrodynamic loads, as for instance methods based on the first-order hydrodynamics theory [28], where radiation and diffraction effects are modelled by introducing, in the motion equations of the floating support, frequency-dependent hydrodynamic added-mass and damping terms (radiation), and wave excitation terms depending on frequencies and direction of the incident waves (diffraction), computed in the frequency domain using potential flow theory for regular or irregular sea states [31]. The second-order hydrodynamic theory has been applied in some cases, as second-order loads may excite slow-drift motion in soft-moored platforms, typically those with catenary moorings, or ringing effects in TLPs [32]. Hydrodynamic loads on floating supports can be computed also by CFD methods solving the Navier–Stokes equations [33,34]. Finally, it is worth noting that hybrid approaches coupling LES for wind turbulence and potential flow theory for waves have been proposed, very recently, to assess wave effects on wind farm performances [35,36].

Although the methods briefly outlined here have allowed significant advances in the understanding of offshore wind turbine aerodynamics and hydrodynamics, an accurate modelling still presents many difficulties. Challenges and innovative solutions are discussed in three papers of the theme issue: the papers by Sorensen *et al.* [37] and Xu *et al.* [38], which present respectively a CFD method and a free-vortex wake method for aerodynamics, and the paper by Viré *et al.* [39], which proposes an integrated CFD approach to aerodynamics and hydrodynamics.

Sorensen *et al.* [37] present a method coupling a CFD solver of the Navier–Stokes equations governing the air flow, EllipSys3D, with an aeroelastic code accounting for the dynamics of the whole support structure under combined wind and wave loadings, FLEX5. In particular, the Navier–Stokes equations solver involves an LES turbulence model and an AL model, while FLEX5 relies on the BEM theory to compute the aerodynamic forces on the blades. Coupling is made with aerodynamic forces and related line coordinates imported from FLEX5 into EllipSys3D, and wind velocities transferred back to FLEX5 at each step of a time domain simulation. The authors discuss the suitability of the AL model for capturing basic features of wind turbine wakes and present, for validation, a comparison with recent experimental data from a model land-based turbine in a wind tunnel. In addition, a simple formula for the near wake length behind a wind turbine is proposed and corroborated through a numerical investigation of the wake structure via proper orthogonal decomposition. The near wake length is found to depend on the number of rotor blades, tip speed ratio, thrust coefficient and turbulence intensity; its importance in the choice of optimal spacing between turbines in offshore wind farms is also discussed.

Xu *et al.* [38] propose a free-vortex wake method to model the aerodynamics of offshore wind turbines on floating platforms. A three-step and third-order predictor–corrector algorithm is developed to solve a finite difference approximation of the wake governing equation. Blade inflow and starting points of vortex filaments are modified considering the motion of the floating platform. Numerical results are presented for assigned floating platform motions, as obtained from experimental data or numerical simulations. The validity of the proposed approach is corroborated by comparison with experimental data and numerical results obtained by alternative approaches.

Within the context of CFD methods, Viré *et al.* [39] propose an innovative method modelling aerodynamic and hydrodynamic fluid–structure interactions on a computational mesh covering fluids (sea water and air) and structures (support structures, rotor blades). In particular, the Navier–Stokes equations are defined over an extended fluid mesh covering fluids and structures, while the effects of the structures on the fluid flows are represented using a thin shell surrounding the structures, which relaxes fluid velocity and structural velocity to one another. The shell is meshed separately, and the region occupied by the shell on the extended fluid mesh is identified by a shell-concentration field. The relaxation between fluid and structural velocities is performed by a penalty force, depending on the fluid–structure relative velocity and on the shell-concentration field. Solutions of the fluid Navier–Stokes equations and dynamic motion equations of the structures are computed in a time domain simulation where, at each time step, coupling is provided by the penalty forces. The method is presented for classical benchmark problems of fluid–structure interactions, including typical ones involved in hydrodynamics and aerodynamics of offshore wind turbines; potential challenges posed by modelling aerodynamic phenomena are also discussed.

## 3. New design solutions

Current strategies to make the offshore wind industry economically attractive focus on the improvement of existing design methods and the introduction of new technologies. In the theme issue, the papers by Kallehave *et al.* [40], Byrne & Houlsby [41], McGugan *et al.* [42], and Borg & Collu [43] are dedicated to these subjects.

The support structure is one of the components where an improved design can allow significant reductions of costs. For simplicity, high degree of standardization and relative ease of installation, support structures on monopile foundations are certainly very attractive. Kallehave *et al.* [40] present current design of monopile foundations and discuss the main drivers for possible optimization. The authors show that introducing full-scale measurements in the design process can allow significant reductions of costs of monopile foundations, with two relevant examples: the West of Duddon Sands offshore wind farm, where remarkable savings of steel tonnage have been achieved upon calibrating pile–soil response based on full-scale measurements; and the Walney offshore wind farm where, for a sample wind turbine, calculated fatigue life expectations have been considerably augmented by using the measured first fundamental frequency. Main drivers for a further optimization are found in an accurate description of pile–soil interaction, development of pertinent S−N curves for fatigue of welded components, damping estimation in operational and parked conditions and less conservative classification of imperfections for shell buckling. The authors suggest that suitable developments in these fields could allow use of extra-large monopiles (10 m diameters) for water depths up to 60 m, with a consequent benefit in terms of cost reduction.

Byrne & Houlsby [41] provide a further insight into foundations design and installation. The authors propose helical piles as an innovative foundation concept for multi-footing substructures in transitional waters, such as tripods and jackets, discussing the advantages over driven piles used in conventional design. Because installing larger wind turbines in increasing water depths will require taller towers and taller submerged substructures, with consequent larger overturning moments at the foundation base, the authors show that helical piles, if appropriately up-scaled with respect to standard dimensions used for onshore applications, can withstand the expected large tension loads at the upwind footing, while some additional advantages exist in terms of installation and decommissioning. Technological challenges are also discussed in the paper.

A new design approach involving damage tolerant structural design and condition monitoring, with specific applications to improved design of blades, is proposed in the paper by McGugan *et al.* [42]. In recognition of the fact that large ‘perfect’ blades are not possible to manufacture, that large blades with defects or damages are too costly to discard, and that manual inspection may be too expensive for offshore wind turbines, the authors propose an original design approach combining damage tolerant materials, built-in sensors to detect evolving damage, non-destructive experimental tests to characterize damage, and detailed FE structural models of damaged blades to estimate residual fatigue life and strength. Examples of levers for obtaining high damage tolerant composite materials for blades are identified in crack bridging provided by fibres connecting crack faces behind a crack tip, and high-energy absorbing interfaces between layers. By the proposed approach the service life of blades may be decided based on their damage state, with a potential for life extension beyond the original planned service life for healthy blades.

Within the search of new technologies, a growing interest is being devoted to vertical axis wind turbines. A comparison between performances of horizontal and vertical axis wind turbines on floating supports is presented in the paper by Borg & Collu [43]. The authors show that vertical axis wind turbines generate substantially different aerodynamic forces on the support structure, resulting in lower tilting moment and higher torque with respect to horizontal axis wind turbines. Further, it is demonstrated that vertical axis wind turbines may benefit not only from lower tilting moments, but also from potentially lower wind turbine mass and lower position of the centre of gravity, with significant advantages in terms of static stability of the system and, consequently, reduced costs of the support structure. Frequency-domain analyses based on the first-order hydrodynamic theory show that a potential disadvantage of vertical axis wind turbines on floating supports is represented by an increased motion in the frequency range including the blade passing frequency (number of blades times rotational speed), with possible overlapping with the range of wave excitations for slow rotational speeds. Comparisons involve the horizontal axis baseline National Renewable Energy Laboratory (NREL) 5 MW wind turbine [44] and the Novel Vertical Axis (NOVA) 5 MW wind turbine [45].

## 4. Experimental analyses

Experimental tests play a fundamental role in the design of offshore wind turbines. *In situ* measurements are obviously highly desirable to assess the accuracy of theoretical predictions on aerodynamics and hydrodynamics, and soil–foundation interaction. However, laboratory tests on scaled models can also provide an insight into the dynamics of the system, and appropriate benchmarking for validation of theoretical models. In this context, experimental tests on scaled floating turbines in wave basins are being carried out in several research centres (see [3] for a summary of ongoing and future scaled tests in wave basins) and are now considered as a fundamental part of preliminary design [46–49]. However, accuracy and reliability of joint aerodynamic–hydrodynamic laboratory tests on scaled wind turbines are major challenges, owing essentially to the difficulties of a simultaneous scaling of wind and wave loadings. The theme issue includes two papers, by Matha *et al.* [50] and Jaksic *et al.* [51], presenting experimental results in wave basins.

Matha *et al.* [50] outline an integrated approach to the design of wind turbines on floating supports, combining in a multi-step optimization procedure simple formulae for a preliminary selection of relevant design parameters, numerical simulations accounting for coupled dynamic behaviour with a different accuracy level, and scaled testing methodologies in wave basins to validate numerical simulation results, adjust the hydrodynamic coefficients used in the simulations, check potentially critical cases and quantify model uncertainties. Challenges involved in a fully coupled aero-hydro-servo-elastic numerical modelling, and in a simultaneous aerodynamic and hydrodynamic scaling, are discussed thoroughly. The proposed integrated design approach is applied for a wind turbine on a post-tensioned concrete spar, conceived within the European KIC AFOSP project as a low cost and durable design solution. Upon illustrating a preliminary design based on simple formulae relating to the spar basic geometry and some key design parameters such as the hydrostatic restoring stiffness in pitch direction and still water pitch eigenperiod, the authors describe the numerical results obtained by two coupled aero-hydro-servo-elastic simulation codes with different levels of complexity. Next, the results of experimental tests carried out in a wave basin and a wave flume on a scaled prototype, with the wind loading simulated by a steady force applied at the tower top using a suitable mechanical system, are presented in conjunction with pertinent numerical results obtained from a dedicated simplified nonlinear model of the tested scaled prototype. The authors show that experimental and numerical results agree very well in terms of pitch and yaw free decay responses, pitch and surge response amplitude operators for regular waves and no wind, pitch and surge responses for irregular waves and maximum operational thrust force at the rotor, allowing adjustment and validation of important system parameters for hydrodynamic analyses. Data collected by the integrated experimental and numerical investigation on the scaled prototype provide a general concept viability check on the selected design solution, for refined numerical and experimental analyses to be performed with more accurate fully coupled numerical simulations and experimental models at larger scale.

Jaksic *et al.* [51] present an experimental methodology to assess nonlinear signatures of the dynamic response of floating wind turbines. Tests are carried out on a scaled TLP for wind turbines in a wave basin, under wave loading only. The delay vector variance (DVV) method is used to capture nonlinear signatures of the dynamic response by processing, in particular, force measurements in the mooring lines measured by load cells and displacements and velocities measured by a laser Doppler vibrometer (LDV), under different wave conditions, i.e. scaled regular monochromatic waves and waves corresponding to various sea spectra. LDV measurements of displacements and velocities show that degrees of nonlinearity differ depending on the input, thus allowing excitation force identification. Load cell measurements of mooring line forces validate previous findings that operational loads on the tower are dominated by pitch and roll motions of the floating device. The proposed methodology may serve as a basis to develop monitoring and control strategies against undesirable levels or types of dynamic response, and to help in estimating changes in system characteristics over the life cycle of the structure. It is readily applicable to full-scale devices under real environmental loads and has the significant advantage of requiring output only without any numerical modelling of the device.

## 5. Design load cases

The search for an improved design and new technologies is to be performed in conjunction with accurate investigations on the response to environmental loads, covering normal and particularly unfavourable conditions [52–54], as prescribed by standards and guidelines [15–17]. In the theme issue, the papers by Muskulus [55] and by Alati *et al.* [56] present numerical studies on offshore wind turbines under normal and extreme loading conditions of particular interest.

Muskulus [55] carries out a comparative analysis of some simplified analytical models of rotor thrust, with special focus on their suitability for fatigue life assessment of the baseline NREL 5 MW wind turbine [44] mounted on a jacket. Simplified rotor load models provide loads to be applied to a standard FE model of the support structure, in approximation of the actual rotor loads that would be obtained from a fully coupled time-domain simulation. Therefore, reliable models may be of great interest for design and structural optimization applications, to build/compare diverse solutions without requiring the time-consuming full modelling of rotor aerodynamics. The author considers rotor thrust models currently available in the literature: a spectral model based on rotor load spectra, and a model involving a thrust coefficient computed for steady wind or by a statistical analysis using two different mathematical forms; in addition, the author introduces a stochastic model consisting of a trend, a quasi-periodic component and noise. Fatigue analyses are carried out on an FE model of the whole structure, where the rotor thrust is applied as a top horizontal force, and hydrodynamic loads are generated by Morison’s equations. The comparison shows that the rotor thrust models provide substantially different predictions of fatigue life at different sections of the jacket. The important conclusion is that more effort is required to develop rotor thrust models that can reliably replace full modelling of rotor aerodynamics.

Alati *et al.* [56] investigate the seismic response of the baseline NREL 5 MW wind turbine [44], mounted on two different bottom-fixed substructures for transitional waters, a tripod and a jacket. The study is carried out by fully coupled nonlinear time-domain simulations on full system models (i.e. including support structure, rotor blades and nacelle components, with fixed and flexible foundation models), implemented in BLADED, a combined modal/multi-body dynamics code with BEM modelling of rotor aerodynamics and Morison’s wave forces [57]. A large earthquake set and different load cases are considered, i.e. earthquake striking in the operational and parked states, and earthquake triggering an emergency stop. The analysis shows that stress resultant demands are significantly increased by earthquake loads even for moderate peak ground accelerations, and that fully coupled nonlinear time-domain simulations on full system models are essential to capture relevant information on the response of the rotor blades, which cannot be predicted by analyses on simplified models allowed by existing standards. The analysis also shows that stress resultant demands due to earthquake loads may be higher than demands from some typical design loads prescribed by IEC 61400-3 [15], in general for the highest levels of peak ground acceleration. Although a definitive answer as to whether earthquake loads are design driving can be given only considering site-specific conditions, the results shown in the paper substantiate the need for an accurate seismic assessment when installing offshore wind turbines in seismically active areas.

## 6. Vibration control

As multi-megawatt machines with higher power production and larger rotor diameters are being deployed, new challenges arise such as an increased flexibility of blades and support structures, whose vibrations can cause not only higher risk of fatigue damage with attendant increase of operations and maintenance costs, but also compromise the power output of the turbines. Passive, active and semi-active vibration control strategies exist to mitigate increased vibrations [58–60]. In the theme issue, the papers by Basu & Staino [61] and by Jaksic *et al.* [62] focus on vibration control.

Basu & Staino [61] present a comprehensive review of well-established vibration control strategies. Focusing on blades, the authors discuss the disadvantages of using active pitch control to alleviate aerodynamic loads on blades (pitch control means that the blade is rotated about its axis by means of hydraulic or electric actuators located inside the wind turbine structure). The authors note that active pitch control, originally introduced with the primary purpose of tracking the optimum tip speed ratio to maximize power generation and regulating rotor speed at the rated wind speed, cannot provide, when used for control purposes, a major reduction of aerodynamic loads on the blades without compromising on the mechanical power generated by the rotor. As alternative and more efficient approach, the authors propose an innovative dual control strategy combining passive pitch control and active tendons inside the hollow structure of the blades, where the design variables, i.e. the control force in the tendons and the pitch angle, are obtained from a pareto-optimal formulation involving an objective function formulated by assigning weights to the two conflicting objectives of vibration reduction and power generation. While reduction of vibrations is achieved by the active tendons, passive pitch control is activated to either maximize or restore power to rated power when needed. Benefits are that passive pitch control avoids the actuator dynamics of the active pitch control and the generation of dynamic stresses, while the force demand on the active tendons is reduced as the pitch control reduces the aerodynamic loads.

Application of control devices is currently under investigation not only to mitigate blade vibrations, but also floating support motion [60]. In the theme issue, Jaksic *et al.* [62] investigate the use of multiple tuned liquid column dampers (MTLCDs) for dynamic response mitigation in floating wind turbines. Results of experimental tests on a scaled TLP in an ocean wave basin (the same studied in [51]), under waves corresponding to standard sea spectra and a constant top load simulating wind thrust, show that MTLCDs have the potential to reduce tension demand in mooring lines, as well as surge motion. The effectiveness of MTLCDs is corroborated by numerical simulations. Finally, the authors show that the DVV method, discussed in [51], may provide output only statistical markers to monitor potential changes in the dynamic behaviour of the system equipped with MTLCDs.

## 7. Health monitoring

Ensuring a reliable power production is essential for an economically attractive development of offshore wind energy. In this context, structural health monitoring plays a crucial role, especially in view of deploying wind farms in remote areas, under challenging conditions. Health monitoring should generally concern all system components, with particular attention to key components such as gearbox and blades. Challenges and innovative approaches in health monitoring of offshore wind turbines are discussed in three papers of the theme issue, by Antoniadou *et al.* [63], Tippmann *et al.* [64] and Myrent *et al.* [65].

Antoniadou *et al.* [63] provide an interesting review of vibration-based advanced signal processing, machine learning methods and sensing technologies for damage detection in wind turbine components. The authors emphasize that joint time–frequency and cointegration analyses are the most suitable signal processing methods for dealing with the non-stationarity of the wind turbine response, especially with the difficulties posed by separating the inherent non-stationary characteristics due to operational and environmental conditions (wind turbulence, for instance) from time-varying features due to a potentially evolving damage. In this context, an example of an empirical mode decomposition method applied to gearbox damage detection is reported. Several sensing technologies and machine learning methods are discussed, and an experimental study using auto-associative and radial basis neural networks is presented for damage detection in a blade under fatigue loading. Finally, the authors illustrate the potential of a population-based approach to damage detection, where the condition of a single wind turbine could be determined based on the measurements obtained from other wind turbines in the farm, using machine learning methods and data available from supervisory control and data acquisition systems. An example involving power curves (power production versus wind speed) acquired from the Lillgrund wind farm shows that power curves are generally robust with respect to individual differences in the turbines (locations, different sensors and generators) and that, therefore, deviation from expected power curves could potentially be used to detect damage in a wind farm.

Tippmann *et al.* [64] present two innovative damage detection methods for blades based on impulse response functions (Green’s functions) passively reconstructed using ambient noise and sensors. Passive reconstruction is particularly attractive since many sources of diffuse acoustic noise exist in the operating conditions of offshore wind turbines. The first method identifies damage as a source of nonlinearity breaking down the reciprocity of the impulse response function; a principal component analysis, based on a set of features built on impulse response function forward and backward components, is performed to compare damaged and undamaged states. The second method is a match-field processing approach, matching the passively reconstructed impulse response function in the damaged state to an experimentally derived replica in the undamaged state. Experimental results on a blade show that both methods can identify damage location. Potential issues in full-scale applications of the proposed methods are discussed.

Myrent *et al.* [65] propose a method for detecting shear web disbond in composite blades. The authors carry out aeroelastic simulations on the baseline NREL 5 MW wind turbine mounted on a monopile foundation using FAST, a combined modal/multi-body dynamics code with BEM modelling of rotor aerodynamics [66], and identify non-blade and blade measurements that processed by standard data analysis methods, and combined in a suitable identification algorithm, lead to the detection of the presence of a disbond and its severity in terms of length along the blade. In the numerical simulations, blades are modelled by Timoshenko beam elements, with stiffness matrices built starting from a suitable three-dimensional model of the composite layers in the blade, where shear web disbond is accounted for by separating shear web and blade shell along a varying length. Excellent probabilities of detection and classification by severity are found, for most common operating wind conditions.

## Footnotes

One contribution of 17 to a theme issue ‘New perspectives in offshore wind energy’.

- © 2015 The Author(s) Published by the Royal Society. All rights reserved.