Subcomponent development for sandwich composite wind turbine blade bonded joints analysis

Accepted Manuscript Subcomponent development for sandwich composite wind turbine blade bonded joints analysis Garbiñe Fernandez, Hodei Usabiaga, Dirk Vandepitte PII: DOI: Reference:

S0263-8223(16)32939-7 http://dx.doi.org/10.1016/j.compstruct.2017.07.098 COST 8754

To appear in:

Composite Structures

Received Date: Revised Date: Accepted Date:

22 December 2016 17 July 2017 31 July 2017

Please cite this article as: Fernandez, G., Usabiaga, H., Vandepitte, D., Subcomponent development for sandwich composite wind turbine blade bonded joints analysis, Composite Structures (2017), doi: http://dx.doi.org/10.1016/ j.compstruct.2017.07.098

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Subcomponent development for sandwich composite wind turbine blade bonded joints analysis Garbiñe Fernandeza,*, Hodei Usabiagaa, Dirk Vandepitteb a

Mechanics Department, IK4-Ikerlan, Jose Maria Arizmendiarrieta Ibilbidea 2, Arrasate-Mondragon, 20500 Gipuzkoa, Spain Department of Mechanical Engineering, KU Leuven, Kasteelpark Arenberg 41, Leuven, 3001, Belgium

b

ABSTRACT In the sector of wind energy, the trend to increasing turbine size is ongoing and it will continue to do so. This research focuses explicitly on a particular aspect of structural design which is reported to be very critical in many designs of wind turbine blade wing box structures, namely the connection between the spar web and the spar cap, which are usually built up with sandwich materials. A pyramid structured approach is developed which links local phenomena of stress transfer and failure in the adhesive connection to overall loads on the entire machine. This paper focuses on the subcomponent level of the pyramid scheme and it shows the relevance of investigations and experiments on this level. A specific test structure is designed and manufactured as a C-beam to reproduce load transfer phenomena as they occur in real blades. An experimental test campaign is conducted using different data acquisition principles and sensors to monitor structural behaviour. Results from a finite element model are compared to experimental results and satisfactory results are obtained. Keywords: wind turbine blade; subcomponent; bonded joint; experimental.

*Corresponding author. E-mail address: [email protected] (G. Fernandez).

1. Introduction Fossil fuel dependency reduction has become an important issue all over the world. Renewable energy production, in particular wind energy generation, has seen a dramatic growth. However, wind energy systems face a difficult competition when compared with traditional carbon-based energy sources regarding cost competitiveness. Wind turbine systems have increased in size and power. Fig. 1 shows that the average wind turbine rated power has increased from 0.1 MW in 1985, to current large-scale systems averaging 5 MW. In the future systems are expected to have power levels of 10-20 MW with rotor diameters on the order of 170-240 m.

Fig. 1. Increase in wind turbine scale and power [1].

Wind turbine blades, which are mainly assembled with fibre-reinforced composites components which are glued together, experience a multitude of events that can affect reliability, including variable amplitude cyclic loading, daily and seasonal temperature and humidity changes and, for colder climates, ice impact. For very warm climates, long-term exposure can cause degradation of composite structures and adhesives. For the safe and cost-effective operation of wind turbines, it is necessary to understand how these variables, whether or not in combined effect, can affect reliability. Numerical models should be capable of predicting in a sufficiently accurate way structural deformation of a blade when it is subjected to different loads, in different aerodynamic conditions, and also strength of the blade and its components, including adhesive joints, ply drops, even in the presence of material defects [2, 3]. One of the most critical points in wind turbine blades is the joint between the spar cap and the shear web. This joint is the main focus of this paper. A pyramid structured approach is developed which links local phenomena of stress transfer and failure in the connection to overall loads on the entire machine. For that purpose, a subcomponent structure is designed and tested. Section 2 introduces the pyramid approach and the scope of this paper; the subcomponent structure designed as a C-beam. Section 3 deals with the sizing of this C-beam. Section 4 describes the experimental campaign performed in the coupon level to be able to feed the upper levels correctly. Section 5 presents the finite element model of the subcomponent, and Section 6 gives a detailed description of the experimental test which is conducted on this structure. Section 7 analyses the results obtained in the experimental test and finally, Section 8 compares the experimental and numerical results with a quite fine degree of correlation. 2. Pyramid approach and subcomponent introduction The load carrying capacity of wind turbine blades contains significant uncertainty due to the usage of composite materials and the overall structural behaviour of the blade. In order to reduce these uncertainties, the wind turbine standards IEC 61400-1 [4]

and IEC 61400-23 [5] prescribed that experimental tests be conducted on coupons and on full-scale blades. The main purpose of the coupon level tests is to determine the stiffness and strength of the material in both ultimate and fatigue limit state. Commonly, these tests are performed on small test specimens with basic or bulk material, called material testing. On the other hand, full-scale tests are conducted to verify the strength of the blades in both ultimate and fatigue limit state. The requirements to full-scale testing are given in IEC 61400-23 [5]. Full-scale tests normally contain the following subtests: • Blade properties (weight, elastic properties, natural frequencies, etc.). • Static strength (flap-wise/edge-wise). • Fatigue strength (flap-wise/edge-wise). • Static strength after fatigue tests (flap-wise/edge-wise). A full fatigue test usually takes 3-4 months, whereas a static test takes 1-2 weeks [6]. Normally only one full-scale test is done with each blade type and the tests are normally terminated before failure [7]. Furthermore, the material specimens and the test blade are often manufactured with special care. However, during the service life of wind turbine blades several structural details can cause failure modes and some cannot be detected by coupon testing or in the blade tests, which results in inaccurate life estimation and failure prediction. Failures occur are mainly due to manufacturing or material defects, handling faults or design faults [8]. For the specific case of bonding pastes, according to the Germanischer Lloyd (GL) Guidelines [9], mechanical tests at two different bond line thicknesses (0.5 mm and 3 mm) are required for the material certification. However, adhesive layers with thickness of no less than 10 mm are common practice in the wind turbine blades industry, and it is well known that several materials exhibit overall characteristics which do depend on thickness. Moreover, typical coupon level tests do not take into account geometrical effects like representative thickness or material processing technologies (hand or mechanical mixing), and therefore, their impact on the structural strength. The influence of bonding thickness should not be overlooked. When considering a stress-based failure criterion, and if applying analytical methods developed by Volkersen [10] or Goland and Reissner [11], higher adhesive layer thicknesses always lead to a joint strength increase. However, it has been experimentally demonstrated that a thicker bond line would tend to reduce the strength of both metallic and composite single lap joints [12-14]. Adams and Peppiatt [15] stated that the probability of having porosity and micro-cracks in the joint raise increasing bonding thickness, and therefore, reducing the strength of the joint. Considering the plastic behaviour of the adhesive, Crocombe [12] showed that as adhesive thickness increases, the rate at which the yielding zone develops is faster, so the load bearing capacity of the joint decreases. Gleich et al. [16] have studied the bond line effect, especially for high strength and brittle adhesive. Based on finite element analysis (FEA), they conclude that the maximum strength of adhesively bonded joints seems to occur at an optimum adhesive layer thickness, which depends on the geometry and the materials involved, above and below which strength decreases [17]. Both the stress concentration and singularity at the interface between adhesive and adherend become more intense as bond line thickness increases. Regarding the optimum adhesive layer thickness, a few researchers [13, 16] suggest that an optimum bond line thickness should ideally be kept between 0.1 mm and 0.5 mm. However, this broad observation is not applicable to all cases as there are other variables involved, such as the type of loading (shear, peel, or cleavage), the adherend behaviour (elastic or plastic), and the type of adhesive (ductile or brittle). Furthermore, theoretically, having a range of optimum bond line thickness is not well understood [18]. This is why the effect of adhesive thickness on the strength of adhesively bonded joints still needs investigation. Tests at an intermediate level bridge the gap between elementary material testing and full-scale testing, resulting in a pyramid concept of testing which has ideally a large number of material tests, a lower number of subcomponent tests, and at least one full scale test. Fig. 2 shows the test pyramid applied to wind turbine blade industry. Only full-scale tests (level 3) and the small specimen tests (level 1) are required by certification norms [4, 19, 20].

Fig. 2. Pyramid scheme for the analysis of wind turbine blades.

When focussing on the joint, a more direct approach to study real wind turbine blade problems is to test adhesive joints at the subcomponent level. A subcomponent represents a selected structural detail of the wind turbine blade and the objective of this procedure is to develop a cost-effective test representation of a blade structure. This type of test bridges the existing gap between coupon and full scale blade testing, simulating complex stress states in the bond line and reproducing blade failure patterns. Examples of structural details in a blade which may be represented in subcomponents are the blade root, trailing and leading edge, spar near root, transition of circular part to aerodynamic shape, sandwiches, repairs, flanges and shear web, spar structure (typically one (or two) I-beam(s) or box beam), bond lines and joints throughout the blade [21]. Some of these structural details are schematically indicated in Fig. 3.

Fig. 3. Examples of structural details in a blade which may be represented in subcomponents. Pictures partially taken from [22, 23].

Some studies investigate the mechanical behaviour of composite I-beam structures. Zhou et al. [24] investigated the static behaviour of a composite sandwich I-beam in a symmetric three point bending test. They showed that for their beam, the failure started at the foam core of the sandwich web. Mandell et al. [25] designed a four point bending test to examine the mechanical behaviour of the shear web material of a beam structure. Hayes [26], Hayes and Lesko [27] and Park et al. [28] investigated the structural behaviour of fibre-reinforced beams for bridges under static and fatigue loading. Park [29] investigated a connection system for such beams. Potter et al. [30] designed and tested an I-beam to investigate an adhesive in a symmetric four point bending test. Samborsky et al. [31] addressed stress concentration effects due to the geometry design and porosity in adhesive joints implemented in the wind industry. They studied numerically and experimentally the performance of several joints with adhesive thickness up to 4 mm. Sayer et al. [32, 33] developed a subcomponent test method in order to investigate the adhesive bond properties. On the other hand, a comprehensive study on the mechanical behaviour of I-beam profile structures was performed in the framework of the work package 3, rotor structures and materials, of the European project UpWind. The outcome of this study was a list of recommendations regarding the design, the modelling and the testing of the subcomponents [34]. And finally, Zarouchas et al. [35] investigated both experimentally and numerically an I-beam subjected to four point bending. Although a quite vast research is performed around wind turbines, a significant degree of uncertainty is still not yet eliminated. The objective of the current study is to understand the bonding joint behaviour in a more comprehensive way. The pyramid approach from Fig. 2 is adapted to the current study leading to the pyramid shown in Fig. 4. This figure shows that the characterisation of a wind turbine blade can be split into several levels. The full wind turbine blade is represented on the top of the pyramid. This is the real case without simplifications or assumptions. This step can be done experimentally or computationally. In this research the full blade characterisation is done computationally using an automated methodology for aero-structural analysis that calculates the global and local stress/strain results on a wind turbine blade. Different wind load cases are studied and the worst case condition is selected to perform subsequent analysis. The developed methodology is used to obtain the stress state in one of the most critical points in a blade. This critical point is the joint between the spar cap and the shear web. This joint is the main focus of the subcomponent level. As the full wind turbine blade model does not include the bonded joint, this subcomponent model, represented as a C-beam, focuses on this particular zone of the blade. Once the stress state both at the spar cap and the shear web is known for a critical load case, the next step is to translate this information into the connection between the spar cap and the shear web. The aim of this step is to improve the understanding of this critical zone of the blade, developing a simplified structure that represents the joint in a realistic way. This is the main focus of this paper. In addition, some coupon level experimental campaign is performed to feed the upper levels with data which are validated at a lower level. The composite material testing is described in this paper while the bulk adhesive and intermediate specimen (bonded joint) experimental campaign and analysis are out of scope of this paper.

Fig. 4. Pyramid concept applied to wind turbine blade.

3. C-beam sizing

Before proceeding to the C-beam development, a load case has to be defined which is relevant for a wind turbine blade. The employed wind turbine blade as a reference generates a full turbine power of 5 MW. The selected extreme wind load case [36] occurs when the wind turbine is working at rated conditions and an unexpected severe gust takes place. The gust occurs so suddenly that the blade does not have the time to be pitched. Therefore, the pitch angle is the same as rated conditions, but the speed of the wind is higher than the rated value. To calculate stress/strain map on the blade, an automated procedure developed for the calculation of loads on wind turbine blades is used. This iterative methodology consists on building up a full 3D model of the wind turbine blade in several cross-sections along its length and calculating the aerodynamic properties taking into account the 3D characteristics. Once the pressure coefficients are estimated the wind loads can be introduced as a smooth pressure distribution to the 3D model. With this methodology [37], the wind loads depend on blade deformations, that is, wind loads are coupled with blade deformations. This iterative procedure is performed until convergence of results is reached. The major advantage of the proposed procedure is that as the wind loads are introduced as a smooth pressure distribution, undesirable and unphysical stress concentrations are avoided which is critical when software like Bladed or FAST is used. The discrete forces and moments obtained in this type of codes generate unrealistic stress concentrations on the 3D blade model. Furthermore, while a CFD calculation can take hours or days, the proposed procedure needs a much shorter computation time in the iterative procedure. A validation is done comparing the results with those obtained from Bladed and FAST and good correlation is obtained. The blade section where stress reaches a maximum value both at the spar cap and in the shear web is taken as a reference, and just to assess that the extreme cases that wind turbine blades have to withstand are all fully covered, a multiplication factor of 2 is applied. This value is selected to compensate for possible higher stresses in blades with a different bond line design. The spar cap is loaded with tensile and compressive components of stress which occur simultaneously. Normal stress in the axial direction is caused by the bending moment in the blade, while the shear web is loaded by shear stress due to shear force. The bonding paste that joins both parts is thus subjected to a multi-axial stress state which depends on all of them. Once the stress state is identified, the next step is the development of a model of the assembly of the spar cap to the shear web with the aid of a subcomponent model. The C-beam geometry is suitable for that purpose as the beam flanges work as spar cap, while the beam core acts like the shear web. The flanges and core are bonded to each other in this critical joint. Fig. 5 shows the part of the blade that it is studied together with the geometry and material distribution of the subcomponent. The material distribution is as close as possible to the original structure. The flanges of the C-beam are composed in majority of uni-directional composite material and some tri-axial composite material as skins. The sandwich material of the shear webs has a foam core and bi-axial composite skins. The different composite materials are composed by glass fibre and epoxy. These components are bonded with an adhesive; for this analysis a bi-component epoxy based bonding paste from Momentive/Hexion is selected. This is an adhesive approved by Germanischer Lloyd (GL) standard for use in wind turbine blades.

Fig. 5. Schematic description of the subcomponent structure.

The next step is the definition of the test configuration. Both the C-beam and the test itself have to be defined to cause failure in the beam through the bonded joint and not in the region with boundary conditions as peak stresses as the corresponding failure would be misleading. The concepts which are traditionally available are 2-point bending (clamped boundary condition), 3-point bending and 4-point-bending, with the trade-off here between the two former options. Fig. 6 shows the distributions of the shear force and the bending moment in a clamped beam and in a 3-point bending configuration. They are similar with the exception of the clamping moment which is replaced by two finite forces with a finite lever. A theoretically clamped configuration of the beam involves a constant shear force over its entire length and a maximum bending moment at the clamped edge, which meets the requirements of the current problem. However, it is highly impractical to actually build a set-up with a clamped boundary condition. The clamping device inevitably requires a finite length of the beam, which necessarily turns it into a 3-point bending set-up. Bending moment is maximum in the cross-section where the load is applied and shear force is maximum in the shorter part of the span.

Fig. 6. 2 and 3 point bending tests; shear force and bending moment distributions over beam length.

In theory, the most realistic configuration is the 2-point bending test, because the stress state is the closest to what is required. However, to avoid the impractical clamping condition, the structure is converted to an unsymmetrical 3-point bending configuration. To circumvent the practical difficulty of a clamping system, a third support point is added (3-point bending) at a finite distance from the middle support. In addition, the section between both supports is strengthened with a steel

reinforcement to the web of the beam. Thick steel strips are attached to both sides of the web of the composite beam to avoid failure due to high shear forces in the section between both support points. The next step is the sizing of the C-beam to reproduce a realistic stress state with the right components and the correct magnitude. The target stress state in the flanges and the core is known from the full blade finite element analysis. On the other hand, an adhesive thickness of 10 mm is prescribed as real wind turbine blades have such bonding paste thicknesses. At this point, some C-beam parameters are fixed while others are variable in order to define the subcomponent as realistic as possible: • Fixed parameters: o overall length of the beam o material distribution that resemble spar cap – shear web configuration (material information in Section 4) o test configuration (unsymmetrical 3-point bending test) • Variable parameters: o applied force (or displacement) o width and thickness of the flanges o width of the core and adhesive layer o height of the core Fig. 7 shows schematically the pre-defined parameters and variables.

Fig. 7. Pre-defined parameters and variables which need a definition for the C-beam.

To fix the design parameters, standard equations from beam theory are applicable and sufficiently accurate with the prismatic geometry of the beam:

ߪ௕ =

‫ܯ‬. ‫ݕ‬ ‫ܫ‬

(1)

߬=

ܸ. ܳ ‫ܫ‬. ܾ

(2)

where σb is the normal stress, M is the bending moment, y is the distance from the neutral axis, I is the moment of inertia, τ is the shear stress, V is the shear force, Q is the first moment of area and b is the width of the section at the position where shear stress is calculated. For simplification, this study is performed assuming that the subcomponent has an I-beam geometry. As the beam is composed of more than one material, it is necessary to take this into account in the calculations. In the classical beam theory the strain is continuous across the beam cross section but bending stress is discontinuous. A method to analyse beams with more than one material is to first select a reference material and then use equivalent areas to represent the increased (or decreased) stiffness of the other materials. The new equivalent cross section is assumed to be made completely from the reference material. In order to use the material strength effectively over the volume of the beam, the uni-directional composite material in the flanges of the beam is designed in a tapered geometry. Therefore, thickness of the flanges is not constant over the beam length. As illustrated in Fig. 8, flange thickness is maximum at the clamped side and minimum at the load application beam extreme, decreasing in uni-directional (UD) layers in a smooth way.

Fig. 8. Uni-directional layers ply-drop along the beam length.

Fig. 9 shows the stress state in the adhesive joint parts of the beam. Both upper and lower joints are represented in this figure. On the one hand, the dotted pink curve refers to the adhesive joint at the lower flange of the beam, where compressive and shear stresses are expected to build up. On the other hand, the blue solid line refers to the adhesive joint at the upper flange of the beam, where tensile and shear stresses are expected to build up. The left side graph shows the bending stress along the length of the beam. The compressive stresses are plotted with absolute values. Due to the ply-drop in the uni-directional material, maximum bending stresses occur rather to the right of the middle support point. The right hand side graph represents the shear stress evolution over the blade length, which remains nearly constant for both joints.

Fig. 9. Bending and shear stress components in the bonded joints as calculated with beam theory equations.

After initial subcomponent dimensions are defined, a small improvement is added to the structure. The 10 mm thickness of the adhesive is kept in the middle zone of the beam where failure is expected to occur, while the remainder of the beam has a thickness with a mean value of 3 mm. Having this common adhesive thickness of 10 mm encountered in real wind turbine blades, similar manufacturing defects as seen in rotor blades are expected to occur. The transition of bond line thickness is done in a smooth way to avoid stress concentrations. The component is manufactured by a company in the Basque Country (Spain) which manufactures wind turbine blades among other composite pieces [38]. Fig. 10 (a) shows the subcomponent.

Fig. 10. (a) C-beam subcomponent with its main characteristics, (b) C-type geometry of the core.

In order for this beam to resemble a real wind turbine blade, the sandwich core is manufactured as illustrated in Fig. 10 (b). The core is not centred and it has the typical C-type geometry that is common in real blades. Moreover, the bonding paste is rounded. 4. Coupon level testing The subcomponent is composed by uni-directional, bi-axial and tri-axial composite material, adhesive and a foam. The properties of the selected foam [39] and adhesive [40] are taken from their datasheets. On the other hand, an experimental campaign is performed to assess composite properties. The selected fibres are from Saertex [41]. Saertex fabrics, also known as NCFs (non-crimp fabrics), are characterised by stretched fibres within the individual layers. In addition, an epoxy based resin is used as the matrix to manufacture these composite coupons. Employing the same family products both in the composite and in the adhesive avoids compatibility problems between them. In resume, quasi-static tests in uni-axial tension and in shear are performed. 4.1. Quasi-static tests on composite material coupons in uni-axial tension To perform quasi-static tensile tests in the different composite materials, ASTM D 3039 [42] standard is followed. The different coupon geometries are defined following this standard and a total number of 5 coupons of each type are manufactured. Fig. 11 shows an image with all the coupons broken after quasi-static tensile tests. Most of them have the fracture line outside the clamping system. These tests are used to determine values of Young’s modulus, tensile strength and extrapolated maximum strain as an extensometer is used. Fig. 12 shows the stress/strain results for the different composite coupons. The results show repeatability and they are in accordance with the bibliographic data.

Fig. 11. Uni-directional, bi-axial and tri-axial composite coupons after quasi-static tensile tests.

Fig. 12. Stress/Strain results for the different composite coupons.

4.2. Quasi-static tests on composite material coupons in shear In addition to quasi-static tests in uni-axial tension, materials are tested to shear in order to obtain material properties in that plane. The Iosipescu test is conducted for that purpose. The ASTM D 5379 [43] standard is followed and 5 coupons of each material are tested. In parallel, 2D Digital Image Correlation (DIC) is used in the beforehand painted coupons to obtain strain values. The test set-up is described in Fig. 13 with one coupon inserted in the device and the camera ready to record the strain values. Fig. 14 shows the failure modes obtained for the different composite materials. Only the uni-directional coupons are broken in the notched part. In the case of bi-axial and tri-axial composites, failure occurs outside the notched area; therefore the ultimate results obtained are not feasible. Nevertheless, the shear modulus is derived. Again, comparison of these results with bibliographic data shows that results are consistent.

Fig. 13. Set-up for Iosipescu shear test method.

Fig. 14. Failure modes (marked by the pale areas on the coupons) for the different composite materials.

5. FE model of the subcomponent structure The C-beam finite element (FE) model is built in Ansys v16.2 [44]. The beam is modelled with solid elements (solid185, 3D 8node structural solid elements) and all the geometrical aspects are taken into account, such as the asymmetric geometry of the beam, adhesive thickness changes and rounded shape of the adhesive end. In addition, a sub-modelling procedure [45] is used to obtain a detailed strain map in the mid-section of the beam where failure is expected to occur. Sub-modelling is an FEA modelling technique used to get more efficient results in a local region of the model. When the whole structure cannot be meshed with the desired mesh refinement because the model can be too large and costly of solving, the sub-modelling approach can be used. In this case, a coarse mesh is used for the whole beam structure (see Fig. 15 (a)), while the sub-model has a finer mesh (see Fig. 15 (b)). For the boundary conditions application in the coarse model, half a roller is modelled to apply displacement, and for the clamping system the combination of the cylindrical supports (fixed in the radial direction) and the steel plates is considered as shown in Fig. 16. The analysis is performed in 3 main steps to avoid problems with the displacement application. First, a zero displacement is imposed, then, the 10 % of the total displacement is imposed, and on the last step the total displacement is applied. The material properties are defined in Table 1. It should be mentioned that due to the brittle behaviour of the different materials, the inelastic strains are discarded facilitating the computation time. Although an error is introduced, in general, it is assumed that is negligible.

Fig. 15. (a) Beam finite element model with a coarse mesh, (b) Sub-model of the middle part of the beam with a finer mesh.

Fig. 16. Beam finite element model with boundary conditions. Table 1 List of employed materials and their sources. Material Adhesive Uni-directional composite Tri-axial composite Bi-axial composite Foam Steel

Location Union between core and flanges Flange core Flange skins Sandwich skins Sandwich core Roller & clamp reinforcement

Properties source Datasheet [40] Experimental data & datasheet [41] Experimental data & datasheet [41] Experimental data & datasheet [41] Datasheet [39] Datasheet [44]

Quasi-static non-linear (large displacement) analysis is performed with a total of 28 sub-steps. Once the coarse model is calculated, results are transferred to the sub-model by defining boundary conditions to be able to calculate results in the fine model. The finer mesh model of the middle part of the beam is shown in Fig. 15 (b). All the components have at least 2 elements through the thickness. The whole structure with the coarse mesh has 597217 nodes and 407039 elements, with a total length of 1500 mm. On the other hand, the sub-model has 762617 nodes and 177600 elements with a length of 300 mm. 6. C-beam experimental set-up definition The set-up concept for this experimental test is shown in Fig. 17. This structure is designed to withstand more than 80 kN, much more than the load expected for beam failure (below 17 kN). On the clamped side, the C-beam is reinforced by glueing steel plates to each side of the web. The cylindrical inserts allow for rotation of the beam. On the other end of the beam, a set of rolling elements is used to apply the load. The system is displacement controlled. A load cell is mounted in this device to measure force. In addition, a measurement laser is used to determine beam tip displacement. Both ends of the beam are attached to a firm foundation.

Fig. 17. Test set-up design to perform the experimental test on the C-beam.

6.1. Strain measurements on the beam In order to obtain the strain map on the beam, strain gauges are used in conjunction to a Digital Image Correlation (DIC) system. DIC is used on one side of the beam and as the beam is not symmetrical, some strain gauges are mounted at the other side of the structure. In addition, some strain gauges are glued at the tri-axial material to measure axial strain. 6.1.1. Strain gauges Three linear strain gauges are glued to the top flange. In addition, two other linear strain gauges are glued to the bottom flange. Linear gauges are chosen for that purpose as the flanges in general transfer tensile and compression loads. On the other hand, as the middle part of the beam is subjected mostly to shear, two shear strain gauges are glued to the web. Finally, some strain gauges are placed in the adhesive itself. Rosette type strain gauges are used as it is expected that the adhesive is subjected to shear and axial loads. Fig. 18 shows a schematic overview of the different strain gauge locations. The right part of the beam is clamped, while on the left side of the beam load is applied through prescribed displacement. The gauges are located in the central part of the beam, where failure is expected to occur. Linear gauges for tensile and compressive strain measurements are numbered from 1 to 5. Shear gauges to measure strain at the biaxial skin are defined as 6 and 7, and finally, gauges in the adhesive material are numbered from 8 to 10. For future research, placing some strain gauges outside the expected failure zone, more specifically around the transition between the reinforced and unreinforced part would be interesting as it gives more information about the whole system.

Fig. 18. Schematic overview of the location of the different strain gauges.

6.1.2. Digital Image Correlation (DIC) Digital Image Correlation (DIC) is employed to obtain the strain map on the whole selected section of the beam. 3D DIC is used just to assess that all the information is retained. Fig. 19 shows the full set-up of the beam experiment. The zone of the beam measured with the DIC, which is delimited with two red arrows, is the central part of it, as it is expected that the failure is going to happen there. The advantage of DIC is that gives full-field measurement in a non-intrusive way. However, no out of plane displacement is recorded. As in the transition from flanges to the bonding paste there is a geometrical discontinuity (all the beam height is not in the same plane), the information in this transition (flange to bonding paste) is lost.

Fig. 19. Full set-up of the beam experiment (red markers show the zone in which strain is recorded by the DIC system).

6.2. Acoustic Emission (AE) measurements on the beam Acoustic emission (AE) is defined in ASTM E1316, Standard Terminology for Non-destructive Examinations [46], as “the class of phenomena whereby transient elastic waves are generated by the rapid release of energy from localised sources within a material”. For the C-beam test, 4 AE sensors are available. These sensors are mounted along the length of the upper flange as shown in Fig. 20.

Fig. 20. Acoustic emission sensors placed on the top flange of the beam.

6.3. Other data acquisition devices Two dial gauges are placed around the clamping system of the beam as shown in Fig. 18. The function of the dial gauges is to verify if the beam is fully clamped. 6.4. Experimental test procedure The displacement load application is done in different steps. First, load is increased until 4 kN, once this value is reached, the system is maintained in that situation for 4 minutes. When this period is finished, the load is increased with another 4 kN, and again, this load is maintained for 4 minutes. This procedure is repeated during the test until beam failure. This process is done to take into account AE requirements. AE procedures prescribe that when a structure is kept at a certain load level, during the first 2 minutes are recorded some hits, but the next 2 minutes hits should no longer be recorded unless the structure is permanently damaged. 7. Experimental test results analysis 7.1. Force-displacement graph and failure mode analysis Fig. 21 (a) shows the force-displacement graph obtained from the load cell and the laser. For the subsequent analysis, 8 load levels are represented as the loading history. This figure shows that force increases linearly with displacement up to a maximum level of 15 kN. Then an event takes place, but the beam still withstands a high load. After further increasing the prescribed displacement, the beam fails completely. The overall picture is that the beam has a critical and a post-critical behaviour.

Fig. 21. (a) Force-displacement graph with 8 load levels for stress and strain analysis, (b) Final failure of the beam structure, (c) Force-displacement comparison between experimental and simulation data.

Fig. 21 (b) shows the ultimate stage of the test after load has dropped to zero. The upper flange has separated from the web in the vicinity of the clamping system. The flange separation was not clean, because there are some composite fibre layers in the web itself. Therefore it seems that the strength of the bonded joint was sufficient to give rise to failure in the adherend. Finally, Fig. 21 (c) shows the comparison between force-displacement data between experimental and simulation. It is observed that the FE model is stiffer than the experimental data. However, a load drop from approximately 15 kN to approximately 12 kN takes place before final failure (see Fig. 21 (a)). Fig. 22 reveals failure mechanisms at about 45° on both sides of the beam. These failures are in the upper adhesive bonding joint and both are located almost at the same position along the beam length. Possibly failure initiated at this point and then propagated in both directions.

Fig. 22. (a) Failure zone on the beam, (b) Bonding joint failure at one side of the beam, (c) Bonding joint failure at the side of the beam.

7.2. Information obtained from strain gauges Fig. 23 shows the linear strain gauge measurement during the test. The scheme of the number references of the different gauges can be found in Fig. 18. This picture also shows that the absolute values of strains decrease in the last minute of the recorded time signals. This is due to the load drop that takes place before final failure (see Fig. 21 (a)). Another remarkable point is that when load has been applied, the strain values increase, and when the load build-up is interrupted for 4 minutes, the strains keep constant. Finally, comparing gauge values, it seems that all of them have similar values, therefore, it seems that strain values are quite constant over the full length of that section of the beam.

Fig. 23. Linear strain gauges measurements through test duration.

Fig. 24 shows the major e1 and minor e2 strain measurements in gauge numbers 6 and 7. These gauges are placed in the middle height of the web (more details in Fig. 18). Focusing on gauge 7 results, the same characteristics as in the linear strain gauges are repeated. On the one hand, there is a decrease in the values of strains at the same time as load drops (see Fig. 21 (a)). On the other hand, when the load build-up is interrupted for 4 minutes, strain values keep constant too. Gauge 6 records an unrealistic increase of strain when the first failure mode occurs (approximately at time 1020 s).

Fig. 24. Shear strain measurements in the beam core.

Finally, the strain measurements for the upper bonded joint are shown in Fig. 25 (a) and Fig. 25 (b) shows the major and minor strain measurements results on the lower bonded joint of the beam. Gauge number references are summarised in Fig. 18. As the bonded joint in the bottom part of the beam is subjected to shear and compressive loads, absolute maximum strain values are the minor strains, e2. For the upper case, the bonded joint is subjected to shear and tensile loads. This time again, a reduction of absolute values of strain is evident when load amplitude decreases. In addition, gauge 9, which is close to the clamping system records higher strains.

Fig. 25. Major and minor strain measurements (a) in the upper bonded joint, (b) in the bottom bonded joint.

7.3. DIC results Fig. 26 and Fig. 27 show respectively the longitudinal direction and the shear strain maps which are obtained with DIC. In both figures, the first (or upper) picture is taken when maximum load is applied (15 kN), and the second (or lower) picture is taken just

after sudden load decrease takes place (approximately 12 kN). The time gap between the upper and the lower pictures is about one second. The load decrease which is evident Fig. 21 (a) (load-displacement) and also in Fig. 23-25 (strain gauges), is observed in DIC results (Fig. 26 and 27) too. Fig. 26 shows the longitudinal strain values. It is evident that the upper part of the beam is subjected to tensile load, leading to the maximum values in the upper flange, where uni-directional (UD) composite material is present. The bottom part of the beam transfers compressive load, and this time again, the UD composite is the material with highest strain values. Finally, in the area around the neutral axis of the beam, longitudinal strains are around zero. Fig. 27 shows the shear strain values. The bi-axial composite skin of the sandwich core has the maximum strain values. Once again, in the load drop from 15 kN to 12 kN, the strain values decrease.

Fig. 26. Longitudinal direction strain exx decreases when first load drop occurs.

Fig. 27. Shear strain exy decreases when first load drop occurs.

In order to be able to analyse variations in strain values along the beam length, vertical lines are drawn from the top flange to the bottom one at different beam sections. A total of 7 lines are drawn (L0 to L6) as shown in Fig. 28. In addition, 2 red arrows are added to the figure to correctly position this picture in the whole structure (see Fig. 19). The right hand side arrow, where line 1 (L1) is located, is around at 600 mm from the clamped end of the beam. The left hand side arrow, where L0 is located, is around 855 mm starting from the clamped end.

Fig. 28. Vertical lines for beam length strain variation analysis.

Fig. 29 shows the strain evolution in the selected line L1 at different load levels; major and minor strains are plotted. The corresponding 8 points at selected instants are shown (see Fig. 21 (a)). The y axis from Fig. 29 refers to the beam height, with Y = 0 mm referring to the neutral axis. The x axis refers to major and minor strain values. As explained in Section 6.1.2, DIC does not give reliable results in the areas where the specimen is not planar, this is why in Fig. 29, small zones of transition from flange to bonding joint are lost. The left hand graph from Fig. 29 shows the major strain evolution during the test. The maximum load corresponds to the maximum strain values. As expected, major strain values are maximum in the top flange where tensile load occurs. Highest values of minor stress are located in the bottom flange.

Fig. 29. Major e1 and minor e2 strains evolution at line 1 (L1) during the test.

Fig. 30 illustrates the major e1 and minor e2 strain values when maximum load is reached (15 kN) for the 8 different lines selected in Fig. 28. Therefore, these are the maximum strain values which occur on the beam. The aim of Fig. 30 is to analyse if there are differences in strain values over the blade length where DIC is used. Although some differences exist, it seems that the values are more or less constant over the beam length.

Fig. 30. Major e1 and minor e2 strains at maximum load (15 kN) at different blade lengths represented by 8 lines.

Fig. 31 shows the major and minor strain values taken from both top and bottom flanges. The purpose of this picture is to see if there are strain variations at the same material over the beam length where DIC is used. The x axis represents beam length. The right end corresponds to around 600 mm starting from the clamped end of the beam (where L1 is located in Fig. 28), while the left end corresponds to around 855 mm starting from the clamped end (where L0 is located in Fig. 28). There are 2 elements of evidences which confirm the previous analysis. On the one hand, this picture clearly shows that maximum absolute strain values are found at maximum load level (15 kN). On the other hand, major strains e1 occur mainly at the upper flange, while minor strains e2 are mostly in the bottom flange.

Fig. 31. Major e1 and minor e2 strain values at top and bottom flanges.

For the adhesive joint, a similar analysis is shown in Fig. 32. Load steps are the same as in previous analysis (see Fig. 21 (a)). The general tendency for the adhesive joint seems to be similar to the UD material of the flanges.

Fig. 32. Major e1 and minor e2 strain values at top and bottom bonded joints.

Finally, Fig. 33 shows strain evolution at the neutral axis of the beam, in the biaxial skin, where bending strain is close to 0 and shear is maximum. This time again, the maximum strain values are obtained just before the first load drop.

Fig. 33. Major e1 and minor e2 strain values at the biaxial skin.

Fig. 31-33 show quite large fluctuations in the strain values over the analysed blade length. These fluctuations are more pronounced in minor e2 strains at the upper flange (UD material) and adhesive joint, in the major e1 strains at the bottom flange (UD material) and adhesive joint, and finally in the minor strain values in the biaxial skin. All these fluctuations are observed at positions of minimum strain values for those materials. In principle, they should be smoother. However, taking one example, such as the one given in Fig. 34, the strain map shows some fluctuations in the strain values. In conclusion, overall strain results are found to be reliable in the regions where maximum values are observed.

Fig. 34. Major e1 strain values in the beam when maximum load is reached.

7.4. Acoustic Emission results Fig. 35 shows the evolution of the number of AE hits during the test. The left y axis shows the number of hits, the right y axis gives the applied force and the x axis shows time. Channel 1 is the sensor that was placed closest to the clamping system, while channel 4 is the sensor that was placed closest to the load application point (see Fig. 20). The first main conclusion is that almost all the damage is observed near channel 1, as this is the sensor with highest number of hits. On the other hand, one important aspect of AE can be clearly seen: if a pre-loaded structure is healthy and without any damage, no hits are registered until the load value exceeds the previous load level. The lack of structural damage is evident as Fig. 35 does not show any hits until 650 s approximately.

Fig. 35. Number of AE hits appearance during the test.

Fig. 36 is quite similar to the previous one, but this graph shows something different. Instead of the number of hits, this graph plots the energy of hits. This time again, channel 1 is the sensor with highest energy values, but a special comment applies to channel 3 where an increase in energy values at maximum load application is observed. When this first failure mode occurs, significant damage is accumulating around this point, as the number of hits is low (Fig. 35), but the energy of them is quite high.

Fig. 36. Energy values of the different channels during the test.

The AE peak amplitude is associated with the magnitude of fracture [47], and the b-value is defined as a slope of the amplitude distribution. Therefore an effective index related to the states of fracture can be defined with the b-value. In the improved b-value, the range of AE amplitude is determined based on such statistical values as the mean and standard deviation. This is a method for determining the amplitude range independent of the range of amplitude, this way, b-values are obtained quantitatively [47]. Fig. 37 shows the improved b-value on the left y axis, the energy of hits on the right y axis and force in x axis. It is clear that when the improved b-value (ib) is minimum, what happens around 13-14 kN, there is a high energy increase. This means that in that period, the beam suffers significant damage.

Fig. 37. Improved b-value (ib) graph for damage characterisation.

8. FE model and experimental results comparison The results described below are for a load application which corresponds to a displacement of 68.4 mm or 17.1 kN. The model is stiffer than the real beam. This is because the numerical model does not include a detailed model of the clamping system, as the main objective of the model is to compare strain maps. Furthermore, displacement gauges show that the beam is not perfectly clamped and this information is not taken into account in the FE model. 8.1. Comparison between strain gauges and FE model Maximum values of the different strain gauges at first failure (15 kN, see Fig. 21 (a)) are compared with FE sub-model results. Fig. 38 shows this comparison. The y axis represents strain values and the x axis the position along the beam length. In the FE model, mean strain values are taken from the same geometrical positions as the real strain gauges. The number references of the strain gauges are listed in Fig. 18. Triangle shape points are calculated with the sub-model, and the diamond shape points are the strain gauges results. Both major e1 and minor e2 strain values are compared in this graph. The correlation between them is quite good, except for the case of biaxial skin core (gauges 6 and 7). In this part of the beam the results differ between the experimental and the numerical ones. The FE sub-model predicts higher strain values at the shear web than is obtained experimentally. At this moment, no further attention is paid to this discrepancy as the neutral axis of the beam is not the critical area of the structure.

Fig. 38. Comparison between FE sub-model results and the different strain gauges output.

8.2. Comparison between DIC and FE model At the same beam position but at its other side, an analysis is performed using DIC. Strains are compared along several lines along the length of the beam. These lines are around 255 mm long. There is one line on the upper flange (UD material), another one on the bottom flange (UD material), two lines each on both adhesive joints and a last one in the middle of the core. Fig. 39 illustrates the comparison between DIC results and FE results. Major e1 and minor e2 strain values are plotted. The solid lines refer to DIC results while the dotted lines are results from the FE sub-model. Conclusions are similar to previous analysis with strain gauges. Strain values are quite similar except on the biaxial skin.

Fig. 39. Comparison between FE sub-model results and DIC data.

9. Conclusions As one common failure site in wind turbine blades is the joint between the spar cap and the shear web, a subcomponent which is based on this joint is studied. A similar geometrical structure can be represented using a C-beam, with the flanges representing the spar cap and the beam web representing the shear web. Commonly, the FE model of the whole blade does not take the bonded joints into account, therefore this C-beam is an interesting structure to study the adhesive bonded joint behaviour for this specific case. This subcomponent has some of the particular characteristics of adhesive joints in wind turbine applications. On the one hand there is the multi-axial stress state that has to be sustained, and on the other hand, high joint thicknesses up to 10 mm are built-in in this structure. Therefore, common rules cannot be applied to this specific case. The C-beam is designed and manufactured to be as close as possible to real components. This is why commonly used materials are used. The bonding paste is an epoxy based bi-component adhesive from Momentive/Hexion. Unfortunately only one beam is manufactured and statistical data are not available. For this reason, the structure is designed for minimum uncertainty due to geometrical properties. The C-beam is tested by application of a prescribed displacement on one side while the other side is clamped. The failure seems to occur on the bonded joint itself with 45° cracks appearing on both sides of the upper bonded joint where the adhesive is subjected to tensile and shear stresses. Different devices are used to extract as much information as possible for maximum validation of measurement data. Load cell and laser give load information, while strain gauges and DIC give strain map information. Acoustic emission gives an idea of where and when damage starts and dial gauges are used to evaluate the clamping system suitability. Finally, the strain values are compared to those obtained with a numerical model and quite good agreement between them is obtained. For future research, more C-beams should be manufactured and tested in order to obtain statistical results. This way more reliable results can be achieved. Although really valuable information has been extracted from this test, there is room for improvement for follow-up tests. Maximum information is extracted from the test if a reliable finite element model with sufficient detail is available. It is recommended to conduct a linear elastic FE analysis to prepare for the experimental set-up and the design of the subcomponent. Indeed, the FE model should be improved by using contact model in order to simulate the interactions between cylindrical inserts and holes. Moreover, taking the non-linearity of the different materials may improve the quality of the numerical predictions. Regarding the experimental set-up, more sensors should be used in order to monitor not only the unreinforced part, but possibly also in the reinforced part. This may give a full structural behaviour understanding of the subcomponent structure. Therefore, due to the complexity of the test structure, the test data and the simulation data should be complementary to be able to update the model for maximum extraction of validated data. In addition, it would be interesting to test this subcomponent in fatigue. Wind turbine blades are subjected into high fatigue cycles and the importance of this type of load should not be underestimated. Acknowledgements This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References 1.

2. 3. 4. 5. 6. 7. 8.

Loth E, Steele A, Ichter B, Selig M, Moriarty P. Segmented ultralight pre-aligned rotor for extreme-scale wind turbines. 50th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, 09 - 12 January 2012, Nashville, Tennessee. Vassilopoulos AP. Fatigue and Fracture of Adhesively-bonded Composite Joints. Behaviour, Simulation and Modelling. Woodhead Publishing Series in Composites Science and Engineering: Number 52, 2015. da Silva LFM, Öchsner A, Adams RD. Handbook of Adhesion Technology. Springer, 2011. International Electrotechnical Commission (IEC). IEC 61400-1. Wind turbines -Part 1: Design Requirements. International Standard, 2005. International Electrotechnical Commission (IEC). IEC 61400-23. Wind turbine generator systems -Part 23: Full-scale structural testing of rotor blades. International Standard, 2001. Williamson K. Push and pull – testing wind turbine blades, 2012. Available at: http://www.renewableenergyfocus.com/view/22799/push-and-pull-testing-windturbine-blades/ (Accessed: 26 March 2016). Toft HS, Sørensen JD. Reliability-based design of wind turbine blades. Structural Safety 33 (2011) 333–342. Rossmanith HP. Structural Failure: Technical, Legal and Insurance Aspects. E & FN Spon, an imprint of Chapman & Hall, London, UK, 1996.

9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47.

Germanischer Lloyd (GL). Guideline for the Certification of Wind Turbines, Edition 2010. Germanischer Lloyd Industrial Services GmbH, Hamburg, Germany. Volkersen O. Die nietkraftoerteilung in zubeanspruchten nietverbindungen mit konstanten loschonquerschnitten. Luftfahrtforschung, 1938, 15:41. Goland M, Reissner E. The stresses in cemented joints. J Appl Mech 11: 1944, A17–A27. Crocombe AD. Global yielding as a failure criterion for bonded joints. Int J Adhes Adhes 1989; 9(3):145–53. da Silva LFM, Rodrigues TNSS, Figueiredo MAV, de Moura MFSF, Chousal JAG. Effect of adhesive type and thickness on the lap shear strength. J Adhes 2006; 82(11):1091–115. Kim KS, Yoo JS, Yi YM, Kim CG. Failure mode and strength of uni-directional composite single lap bonded joints with different bonding methods. Compos. Struct., 2006, 72, 477–485. Adams RD, Peppiatt NA. Stress analysis of adhesively-bonded tubular lap joints. J Adhes 9:1, 1997. Gleich DM, van Tooren MJL, Beukers A. Analysis and evaluation of bondline thickness effects on failure load in adhesively bonded structures. J Adhes Sci Technol 2001; 15(9):1091–101. Vallée T, Correia JR, Keller T. Probabilistic strength prediction for double lap joints composed of pultruded GFRP profiles part I: Experimental and numerical investigations. Composites Science and Technology 66 (2006) 1903–1914. Tang JH, Sridhar I, Srikanth N. Static and fatigue failure analysis of adhesively bonded thick composite single lap joints. Composites Science and Technology 86 (2013) 18–25. Guidelines for Design of Wind Turbines. Second edition. Det Norske Veritas, Copenhagen and Wind Energy Department, Risø National Laboratory, Denmark 2002. Jackson JK, Zuteck MD, Van Dam CP, Standish KJ, Berry D. Innovative Design Approaches for Large Wind Turbine Blades. Wind Energ. 2005; 8:141–171. Nijssen R, Westphal T, Stammes E, Lekou D, Brøndsted P. Rotor structures and materials – strength and fatigue experiments and phenomenological modelling. Adhesives, Sealants & Surface Treatments for Wind Turbines. Available at: http://showcase.designnews.com/sites/showcase.designnews.com/files/blade%20image%20for%20OEM%20release.jpg (Accessed: 30 May 2016). Polyurethane adhesives bring wind of change. Available at: http://www.henkel-adhesives.com/industrial/2014-48361polyurethane-adhesives-bring-wind-of-change-47443.htm (Accessed: 30 May 2016). Zhou G, Forbes AJ, Foster N. Static behaviour of preconditioned glass-fibre reinforced epoxy sandwich I-beams. Composites Science and Technology 59, 1999, 963-973. Mandell JF, Samborsky DD, Comps DW, Scott ME, Cairns DS. Fatigue of composite material beam elements representative of wind turbine blade substructure. NREL/SR 500-24379 Report, November 1998. Hayes MH. Structural Analysis of a pultruded composite beam: Shear Stiffness determination and strength and fatigue life prediction (Dissertation). Virginia Polytechnic Institute and State University, 2003. Hayes MH, Lesko JJ. Failure analysis of a hybrid composite structural beam. Composites: Part A 38 (2007) 691-698. Park K-T, Kim SH, Lee Y-H, Hwang Y-K. Pilot test on a developed GFRP bridge deck. Composite Structures 70 (2005) 48-59. Park K-T. Performance verification of a new pultruded GFRP bridge deck-to girder connection system. Composite Structures 81 (2007) 114-124. Potter KD, Davies R, Barett M, Godbehere A, Bateup L, Wisnom M, Mills A. Heavely loaded bonded composite structure: design, manufacture and test of I-beam specimens. Composite Structures 51 (2001) 389-399. Samborsky DD, Sears AT, Mandell JF, Kils O. Static and fatigue testing of thick adhesive joints for wind turbine blades. ASME wind energy symposium, 2009. Sayer F, Post N, van Wingerde A, Busmann HG, Kleiner F, Fleischmann W, Henkel M. Testing of adhesive joints in the wind industry. In: Proceedings European Wind Energy Conference; 2009, Marseille. Sayer F, van Wingerde A, Busmann HG. Development of a new test methodology for adhesive joints in the wind industry. European Wind Energy Conference, Warsaw; 2010. Nijssen RPL. Recommendations and procedures for qualification tests of materials and subcomponents. UpWind document D.3.1.3, February 10, 2011. Zarouchas DS, Makris AA, Sayer F, Van Hemelrijck D, Van Wingerde AM. Investigations on the mechanical behavior of a wind rotor blade subcomponent. Composites: Part B 43 (2012) 647–654. Burton T, Sharpe D, Jenkins N, Bossanyi E. Wind energy. Handbook. John Wiley & Sons, Ltd, 2001. Fernandez G, Usabiaga H, Vandepitte D. Weak coupling aeroelastic analysis for the structural characterisation of wind turbine blades. International Conference on Noise and Vibration Engineering (ISMA 2014), Leuven, Belgium, 15-17 November 2014. Mandiola Composites, S.L. Available at: http://www.mandiola-composites.com. AIREX C 70. Available at: http://www.airexbaltekbanova.com/airex-c70-pvc-foam.html (Accessed: 10 March 2016). EPIKOTE™ Resin MGS™ BPR 135G-Series and EPIKURE™ Curing Agent MGS™ BPH 134G-137GF. Available at: http://www.hexion.com/Products/TechnicalDataSheet.aspx?id=8230 (Accessed: 25 March 2016). Saertex glass fibres (uni-directional, bi-axial, tri-axial). Available at: https://www.saertex.com/en/products/datasheet-glass (Accessed: 8 March 2016). ASTM Standard D 3039. Standard Test Method for Tensile Properties of Polymer Matrix Composite Materials. ASTM International, West Conshohocken, PA 19428-2959, United States, 2000. ASTM Standard D 5379. Standard Test Method for Shear Properties of Composite Materials by the V-Notched Beam Method. ASTM International, West Conshohocken, PA 19428-2959, United States, 1998. ANSYS® Academic Research, Release 16.2, Help System, ANSYS Mechanical APDL Structural Analysis Guide, ANSYS, Inc. Miller E. Submodeling in ANSYS Mechanical: Easy, Efficient, and Accurate, 2013. Available at: http://www.padtinc.com/blog/the-focus/submodeling_ansys_mechanical (Accessed: 21 June 2015). ASTM Standard E 1316. Standard Terminology for Nondestructive Examinations. ASTM International, West Conshohocken, PA 19428-2959, United States, 2006. Shiotani T, Yuyama S, Li ZW, Ohtsu M. Application of AE improved b-value to quantitative evaluations of fracture process in concrete materials. Journal of Acoustic Emission, 2001, 19, 118-132.