Design and analysis of a segmented blade for a 50 MW wind turbine rotor

Abstract

Extreme-size wind turbines face logistical challenges due to their sheer size. A solution, segmentation, is examined for an extreme-scale 50 MW wind turbine with 250 m blades using a systematic approach. Segmentation poses challenges regarding minimizing joint mass, transferring loads between segments and logistics. We investigate the feasibility of segmenting a 250 m blade by developing design methods and analyzing the impact of segmentation on the blade mass and blade frequencies. This investigation considers various variables such as joint types (bolted and bonded), adhesive materials, joint locations, number of joints and taper ratios (ply dropping). Segmentation increases blade mass by 4.1%–62% with bolted joints and by 0.4%–3.6% with bonded joints for taper ratios up to 1:10. Cases with large mass growth significantly reduce blade frequencies potentially challenging the control design. We show that segmentation of an extreme-scale blade is possible but mass reduction is necessary to improve its feasibility.

Keywords

50 MW wind turbine segmentation structural design joint design bolted joint bonded joint

Introduction

The large increase in demand for clean energy and reduction of levelized cost of energy (LCOE) has resulted in the rapid growth of wind turbine blade lengths. However, as blade size has continued to grow, the wind energy industry has encountered design, manufacturing, and transportation challenges. The challenges listed below have led to the development of novel technologies and processes to design, build, assemble, and operate large-scale wind turbines (Johnson et al., 2019; Peeters et al., 2017; Smith and Griffin, 2019; UpWind, 2011):

Maintaining a low mass to reduce loads and facilitate installation;

Assuring quality control in the manufacturing process for correspondence of the as manufactured part to design simulations;

Satisfying land-based transportation constraints on the cargo size and weight.

The multiple constraints on blade size gave birth to the segmentation concept. This involves building and transporting the blade in multiple pieces, as shown in Figure 1, that can be assembled on-site. Segmenting the blade not only can increase the quality of the manufactured pieces, but also generates the possibility of achieving even larger blade lengths to further reduce the LCOE. The large potential of this concept has resulted in a significant amount of research and development with the first segmented blades having been manufactured in the last decades. These designs are either currently undergoing on-site tests or are already in operation (Peeters et al., 2017; Smith and Griffin, 2019; UpWind, 2011). However, despite all these efforts to design and develop segmented blades, there are still many open research questions and challenges regarding the integration of segmentation to current design and manufacturing processes, effective load transfer between segments, and minimization of additional mass to reduce the effect on the turbine performance and cost. Few academic studies have been performed to evaluate the mass effects and technical performance (Bhat et al., 2015), and no prior academic studies have been conducted to determine the cost impacts of various joint concepts. Furthermore, a segmentation study for extreme-scale blades has not yet been undertaken. Thus, it is the focus of this study to provide the research community a first detailed design and comprehensive analysis of various joint configurations for wind turbine blades to address open research questions and challenges.

Figure 1.

Components of capital and operational expenditures (CapEx and OpEx) that are impacted by blade segmentation.

Further LCOE savings can be obtained with offshore wind farms given the larger wind energy sources. The Segmented Ultralight Morphing Rotor (SUMR) project, funded by the Advanced Research Projects Agency-Energy (ARPA-E), has the objective of designing “extreme-scale” rotors with blades over 100 m long and power of 13–50 MW to maximize the U.S. offshore wind power (Chetan et al., 2019a; Griffith and Chetan, 2018; Yao et al., 2021a). For this, an innovative two-bladed downwind turbine design was considered by the SUMR team to reduce tower strike possibilities, mass, loads on the powertrain, and complexity of installation. The SUMR-13i 13.2 MW design with 104.35 m long blades is one example that shows the improvement in performance and reduction in mass compared to conventional three-bladed rotor designs (Pao et al., 2021; Yao et al., 2021a). The SUMR50-S5 50 MW is largest member of the SUMR design family with 250 m long blades that weight over 500 metric tons each (Yao et al., 2021b). However, such extreme blade lengths cannot be manufactured and transported over land with current technology and may require segmenting the blades to make the design feasible with reference to manufacturing and transportation. In this work, we comprehensively study the impact of segmentation on the SUMR50-S5 blade mass and natural frequencies by applying joint concepts available in the literature. A rotor cost analysis, based on the design results shown herein, is presented in Escalera Mendoza et al. (2022). The work presented herein is termed as Level I detailed design and analysis given that the analyses performed are exhaustive and explore multiple variables; however, it is termed as Level I given that segmentation is a complex and large area of research that requires extensive input and effort to address all of its details.

The design and analysis of segmentation for wind turbine blades is performed with a systematic approach. First, a review of the segmentation concepts and applications to wind turbine blades is done. Then, the characteristics of a baseline monolithic one-piece design along with design loads requirements are presented. This information helps understand the design drivers of the existing non-segmented design which are also applicable to the segmented designs. Then, with this information we develop methods to design a segmented blade and perform a Level I detailed structural analysis of blade joints. This is followed by an analysis of the effect of segmentation on the blade mass and the first two parked frequencies of the segmented blade. A qualitative and quantitative cost analysis, based on the results obtained in this work, is presented in Escalera Mendoza et al. (2022) where components of capital expenditures (CapEx) and operational expenditures (OpEx) are examined. Some examples of these are illustrated in Figure 1.

Background review about segmentation of wind turbine blades

Prior studies of segmentation have primarily investigated two concepts for joining segments together: bonded and bolted joints (Peeters et al., 2017). Bonded joints generally include two segments made of fiber reinforced plastics (FRP) that are joined with a structural adhesive (Hart-Smith, 1978; Jørgensen, 2017). This joint type is structurally efficient, inexpensive (material cost) compared to bolted joints, and does not require frequent inspections once cured due to the higher fatigue and corrosion performance of FRP’s compared to metal components. However, bonded joints cannot be disassembled once joined (permanent joint), lack inherent self-alignment and the bond quality is highly sensitive to variations in properties due to environment conditions (temperature, humidity, and surface contaminants) during the joining process (Hart-Smith, 1978; Jørgensen, 2017; Peeters et al., 2017). Also, the complexity, duration, and cost of the installation increases substantially to achieve better control of the bond thickness, material properties, and adherence. To improve the bond quality and facilitate the assembly process, changes to the nominal manufacturing process are required such as adding diverse types of brackets and shims (Peeters et al., 2017; UpWind, 2011). Various authors have studied different types of bonded joints to improve the strength and damage tolerance of the bond. Although, the type of bond is dependent on the geometry and material of the parts to be bonded; for example, the scarf and stepped lap joints are considered to be the most effective for general geometries (Peeters et al., 2017). An example for joining blade segments with a single lap joint is shown below in Figure 2.

Figure 2.

Single lap joint (Peeters et al., 2017) recreated on SUMR50-S5 at 50% blade span (artwork by Daniel Bouzolin).

Bolted or mechanical joints tend to have higher mass and cost, but are easier to assemble, inspect, repair, and can be disassembled contrary to the permanent bonded joints (Peeters et al., 2017). However, bolted joints require frequent inspections and maintenance because of the susceptibility of the metal components to fatigue that can significantly increase the operating costs (Kensche, 2006; Peeters et al., 2017). Nonetheless, mechanical connections such as t-bolts and studs/inserts are already used in all blades to connect the root of the blade to the hub, have been proven to work effectively, and are a basis for the design of mechanical joints to connect blade segments (Peeters et al., 2017; Smith and Griffin, 2019).

Prototypes of segmented wind turbine blades have used both t-bolt and stud/insert joints (Iriarte, 2017; Kensche, 2006; Peeters et al., 2017). T-bolt joints were used in the DEBRA-25 blades (Kensche, 2006) to join two 8.5 m segments through the C-spars, in the JOULEIII project (Dutton et al., 2000) to connect segments of a 23.3 m blade, and in the MEGAWIND project (Vionis et al., 2006) to attach sections of a 30 m blade. The first two prototypes successfully passed static and fatigue testing, but the latter one had premature fatigue failure (Dutton et al., 2000; Kensche, 2006; Vionis et al., 2006). In the UpWind project, it was determined that the most effective joint type to connect two blade segments of a 42.5 m blade was to bond metallic inserts in the spar caps, and then join the two segments together using metallic bolts (Jackson et al., 2005; UpWind, 2011). This led to the development of the 62.5 m INNOBLADE design by Gamesa that was installed in an onshore 4.5 MW G128 turbine (Gardiner, 2013; Pedersen et al., 2009). Until 2013, 27 modular blades had been installed on nine G128 turbines after undergoing extensive testing for validation (Gardiner, 2013; Pedersen et al., 2009).

Enercon has continually grown its number of segmented blades installed (Gardiner, 2016). To overcome transportation challenges due to large chords as the blade grows in length, Enercon developed a cylindrical section to which a trailing edge section (assumed to not be a critical load carrying section) attaches as shown in Figure 3 (Gardiner, 2016). The cylindrical section is built using automated winding of pre-impregnated glass fiber materials that improve the quality of the laminate and reduce fabrication time (Gardiner, 2016). Multiple turbines with segmented blades (101–160 m) that are mechanically joined on site have been installed and are currently operational (Enercon, 2019; Gardiner, 2016; Wind-Turbine-Models, 2011).

Figure 3.

Modular blade design of Enercon. (a) Filament winding. Figure reproduced from Gardiner (2016) with permission from Roth Composite Machinery. (b) Mounted Enercon E-126 7.58 MW. Figure reproduced from Wind-Turbine-Models (2011). (c) Enercon E-126 7.58 MW rotor on ground. Figure reproduced from Wind-Turbine-Models (2011).

GE Renewable Energy has developed a 4.8 MW turbine with a 158 m rotor diameter termed the Cypress Platform, shown in Figure 4, with an onshore demonstration turbine installed in Germany in early 2019 (Noon, 2019). LM Wind Power manufactured the blades that include carbon fiber and are composed of two segments (Noon, 2019). These are joined using a pin joint that includes at least one chord-wise pin that secures a male shear web member to a female shear web member (Yarbrough et al., 2019).

Figure 4.

GE Cypress wind turbine prototype. Figure reproduced from GE Renewable Energy (2021).

Envision Energy installed EN128-3.6 MW shown in Figure 5 two-bladed wind turbine with a 128 m rotor diameter in Denmark and has been in operation since 2012 (de Vries, 2014). The design incorporates a “partial pitch” system with an “extender” (a tubular and fixed-pitch 20 m blade section) and an outer blade section controlled with a pitch bearing system at 20 m station (de Vries, 2014). More information on the joint type has not been disclosed.

Figure 5.

Envision E128-3.6 MW with partial pitch. Figure reproduced from Wind-Turbine-Models (2012).

Nabrawind developed the “Nabrajoint” bonded-mechanical joint with embedded inserts in the lower and bottom spar caps of each of the blade segments as shown in Figure 6 (Callén, 2017; Iriarte, 2017; Lesmes, 2019). These are joined using studs that are pre-tensioned when wedges of the “xpacer” are fastened and the shell pieces are then assembled to the joint flanges to close the open section as shown in Figure 6 (Callén, 2017; Iriarte, 2017; Lesmes, 2019). Nabrawind has performed significant material characterization, subcomponent, and substructure testing since 2016 and is now in the full-scale certification stage (Callén, 2017; Iriarte, 2017; Lesmes, 2019).

Figure 6.

Nabrawind segmented design. Figures reproduced from Lesmes (2019) with permission from Nabrawind Technologies: (a) Xpacer diagram and (b) illustration of Nabrajoint.

Wetzel Engineering investigated the “Space Frame” segmentation concept that consists of fabricating the wind turbine blades in multiple pieces as shown in Figure 7 (Barnhart, 2016; Wetzel, 2014). The modular blade has non-structural skins and the various spar caps are joined together using bonded ribs and are mechanically adjusted by pre-tensioned cables (Wetzel, 2014). This modularization allows large transportation cost savings but increases the number of components to be assembled significantly (Barnhart, 2016; Wetzel, 2014). By changing the conventional design to have non-structural skins the first edgewise frequency and the edge-wise stiffness reduce significantly (Barnhart, 2016).

Figure 7.

Space frame concept. Figures reproduced from Barnhart (2016) with permission from Wetzel Wind Energy Services: (a) modular blade and (b) section components.

SUMR-50 monolithic rotor structural design characteristics

The design characteristics of the monolithic 250 m blade for the 50 MW turbine are presented herein (Yao et al., 2021b). The structural version S5 is summarized in Table 1 and meets international design standards used by the global wind energy industry to ensure the safety and reliability of the blade structure. An analysis of multiple design load cases (DLC’s) by Yao et al. (2021b) shows that the critical case for this blade design is DLC 2.3. The bending moment and shear force distributions of this critical case are shown in Figure 8.

Table 1.

Design scorecard for SUMR50-S5.

Design parameter	Value
Blade length (m)	250
Maximum chord (m)	16.2
Root diameter (m)	12
OpenFAST blade mass (m tons)	502
Center of gravity along span (m)	74.81
Critical design load case	DLC 2.3
Maximum flap-wise root moment (kN m)	9.63E + 05
Maximum edge-wise root moment (kN m)	6.74E + 05
Maximum combined root moment (kN m)	1.18E + 06
Maximum strain ( $μ ε$ )	4910
Allowable tip deflection to avoid tower strike (m)	43
Maximum deflection toward tower (m)	28.5 (DLC 2.3)
Bmodes first flap blade frequency (Hz) (0 rpm)	0.285
Bmodes first edge blade frequency (Hz) (0 rpm)	0.396
Flap-wise fatigue life (years)	>20
Edge-wise fatigue life (years)	>20
Flutter ratio (:)	1.02

Figure 8.

Critical design load case: DLC 2.3: (a) bending moment and (b) shear force.

The extreme blade size of the SUMR50-S5 design significantly challenges the current transportation and manufacturing technology that face non-trivial or impossible tasks unless innovative methods such as segmentation are adopted. Hence, the present study focuses on the segmentation design and analysis to determine the effect of segmentation on the blade mass and the natural frequencies. The effect on rotor cost is analyzed in Escalera Mendoza et al. (2022).

Level I segmentation design procedure and assumptions

A Level I detailed structural sizing is developed and applied for multiple joint configurations. The results of the design study are used to determine joint material usage and the effect of segmented blade mass in parked rotor frequencies. These results are used in Escalera Mendoza et al. (2022) to determine the effect on rotor cost (material, labor, molds, transportation, assembly of segments, installation, and maintenance costs). In this paper we evaluate the feasibility of the segmented designs in relation to the structural performance while benefits/detriments of segmentation compared to the monolithic design regarding logistics and cost are analyzed in Escalera Mendoza et al. (2022). However, segmentation is still the only option to make a 50 MW blade design feasible due to its sheer size. The general procedure to design the blade joints is summarized in four sequential steps that are explained in detail in the following sub-sections:

Select joint locations along the blade span and number of segments;

Select joint type to be designed and analyzed;

Select materials that can be used in the joint for each joint type;

Apply methodologies to design bolted and bonded joints which include strength, manufacturing, geometrical, and load constraints.

Selection of joint locations

Several joint locations were considered for the SUMR50-S5 250 m long blade. As previously mentioned, the design space of segmentation is large. Thus, joints along the span-wise direction are selected to help focus the work on the exploration of joint locations, joint types, materials, number of joints, and taper ratios. The joint locations along the blade span were selected based on the structural results of the one-piece blade and on the current transportation limitations for blade size that are discussed in Escalera Mendoza et al. (2022). A finite element stress analysis of the one-piece or monolithic blade shows that the inboard half of the blade experiences high stresses, thus placing a joint in this region could be a risk. Escalera Mendoza et al. (2022) shows that with land-based transportation only the last 31% of the blade could be transported on highways because the chords of the inboard 69% of the blade exceed the US cargo-size limitations. The inboard section would then need to have joints in the chord-wise direction or require a different method of transportation such as ocean-based transportation which is discussed in detail in Escalera Mendoza et al. (2022). Considering these factors, the locations selected for joint placement are 30%, 40%, 50%, 75%, and 80% of blade span as shown in Figure 9 for segmented blades with two, three, and four pieces.

Figure 9.

Segmentation options for SUMR50-S5.

Selection of joint types and materials

The literature review indicates that the most common joint types are bolted and bonded joints. Detailed design information for both joint types applied to wind turbine blades is limited. However, fundamentals of the joint components, layout, and theory exist to help develop design methods for segmentation of extreme-scale wind turbine blades that are described next.

Three materials are used in the bolted joint design: fiberglass with epoxy resin for the shells, steel for the bolts to join the segments, and adhesive assuming a bolt fitting/insert is bonded to the fiberglass material and then the bolt is fastened to it. Only two materials are used in the bonded joint design: fiberglass with epoxy resin and adhesive. Both joint types share the same fiberglass and adhesive properties.

The fiberglass-epoxy material properties presented in Table 2 correspond to the shell skin material used in the SUMR50-S5 design which are the same as the properties used of the design of the SNL100-00 100 m glass blade (Griffith and Ashwill, 2011). For consistency, only one fiberglass-epoxy material is evaluated in this paper. Steel bolts are assumed to have the highest proof strength (1215 MPa) of bolts currently available, which corresponds to the BUMAX ultra bolt (BUMAX, 2019) due to the sheer size of the blade and the extreme bending moments it experiences. The elastic constants of this material are not publicly available, thus general steel elastic constants are used for analysis (Matweb, 2019) as shown in Table 2 where the “t” column refers to the cured ply thickness.

Table 2.

Fiberglass and steel material properties.

Material	$E_{L}$ (GPa)	$G_{LT}$ (GPa)	$ν_{LT}$ (:)	t (mm)	$ρ$ (kg/m³)
SNL triax fiberglass	27.7	7.2	0.39	1	1850
Steel	205	80	0.29	:	7850

There is a large variety and number of adhesive materials applicable to bonding composite sections. Therefore, additional attention is put in the selection of adhesive materials to be used for both joint design types. Seven candidate adhesive materials applicable to bonding composite parts and that have available data-sheets are presented in Table 3. In this, the material properties used in structural and mass calculations are shown (modulus of elasticity, mass density, and single lap shear strength) as well as the fabrication characteristics (maximum work time, cure time, and recommended bond thickness) (3M, 2012a, 2012b; Hexion, 2019a, 2019b; Plexus, 2018a, 2018b, 2018c): (Williams, 2019, personal communication). All seven adhesive materials are evaluated for both joint design types due to the large variation in properties.

Table 3.

Adhesive material properties.

No.	Material name	$E_{t}$ (GPa)	$ρ$ (kg/m³)	LSS(MPa)	Maximum work time (hours)	Minimumt (mm)	Maximumt (mm)
1	3M-W1101	3.7	1330	23	2	0.5	10
2	3M-W1115	1.93	1130	27.6	0.33	0.1	1
3	HEX-535-537	3	1126	33	2.17	1	10
4	HEX-135G-137GF	5.5	1250	15	3.5	1	10
5	PLEXUS-MA300	1.137	970	26.2	0.1	0.3	3.2
6	PLEXUS-MA530	0.8274	950	17.2	0.67	0.75	18
7	PLEXUS-MA560	0.827	950	17.2	1.17	0.75	25

Et: tension modulus of elasticity; LSS: single lap shear strength; maximum work time: time until adhesive starts to gel; minimum t: minimum adhesive thickness recommended; maximum t: maximum adhesive thickness recommended.

The Poisson’s ratio or shear modulus is not given in the data-sheets of any of tabulated adhesive materials. We believe this to be due to the high variability of mixing and curing conditions (Samborsky, 2019, personal communication). The lap shear strength of adhesives is highly dependent on the shear properties of the adherends as shown in the corresponding data sheets (3M, 2012a, 2012b). The adhesive properties are also highly dependent on the environment conditions during adhesive mixing and placement, on the bonding surface preparation and the cure cycle control (3M, 2012a, 2012b; Hexion, 2019a, 2019b; Plexus, 2018a, 2018b, 2018c). Knowing the adhesive properties more accurately for a particular application requires testing samples with the adherends intended to be used in the blade design as recommended in the data-sheets of the adhesive materials presented herein. Since no material characterization has been performed in our study, we assume that the adhesives of Table 3 have a Poisson’s ratio equal to the fiberglass adherend (0.39) (Griffith and Ashwill, 2011).

The next sub-sections describe in detail the joint design methodology. Screenings on the adhesive materials presented in Table 3 are performed during the structural sizing process based on the following criteria. If a material meets the criteria, then it forms part of the feasible group of adhesive materials and it is used for further calculations. Else, it is excluded from further calculations and from the final results.

Criterion 1: The time to place adhesive on the arc-length of the cross-section to be bonded does not exceed the maximum work time defined by the manufacturer. The NREL cost model (Bortolotti et al., 2019) indicates that 60 m/hour of adhesive can be placed on the blade shell to bond a shear web. This is used to calculate the time required to place adhesive on the bonding surface of the joint. This component of the NREL cost model is applied because it uses a bottoms-up approach (Bortolotti et al., 2019), thus it is assumed to still be applicable to blades longer than 100 m. But more research is warranted to further improve cost models for larger turbines;

Criterion 2: The bond thickness meets the minimum and does not exceed the maximum values recommended by the manufacturer.

Bolted joint design methodology

The Level I detailed bolted joint design is performed with the following steps:

Determine the required bolt area using a combination of axial, shear, and torque loads in a von Mises bolt stress calculation, and using fatigue analyses with DLC 1.2 loads to meet a 20 year life;

Based on the bolt area results from step 1, calculate the bolt diameter, number of bolts, and composite shell thickness required to fit the bolts;

Screen the adhesive materials by comparing the time required to place adhesive on the bonding surfaces and the work time specified by the manufacturers;

Calculate the span-wise bolt length and adhesive thickness assuming a double lap shear joint and using DLC 2.3 loads. A second adhesive screening is performed in this step to select the designs that do not exceed the adhesive thickness limit and the lap shear strength specified by the manufacturer;

Determine the mass of the segmented blade and the effect on natural frequencies.

Figure 10 shows the geometry of the bolted joint used for analysis. The wall thickness of the fitting and the gap between inner and outer sections are ignored to simplify calculations.

Figure 10.

Span-wise cross-section view of real versus assumed bolted joint geometry.

First, the total bolt area required to withstand the shear and axial forces at the joint are determined with equation (1) (shear force on the bolt) and equation (2) (von Mises stress on the bolt) which were used to analyze and size the root bolts for SUMR-D (Bay et al., 2019; Chetan et al., 2021; Kaminski et al., 2021; Yao et al., 2019, 2020). We chose to use DLC 1.2 loads in this step because the bolts are made of metal that has a much lower fatigue Wöhler exponent (m) than fiberglass and carbon fiber. But results with DLC 2.3 loads are checked, and used to size the bolt length and the adhesive thickness in subsequent steps.

V_{bol t_{i}} (N) = \sqrt{F_{x_{i}}^{2} + F_{y_{i}}^{2}} + \frac{| M_{z_{i}} |}{y_{i}}

(1)

σ_{v m_{i}} (Pa) = \frac{1}{A_{bol t_{i}}} \sqrt{F_{z_{i}}^{2} + 3 * V_{bol t_{i}}^{2}}

(2)

The term $F_{x}$ is the out of plane shear force, $F_{y}$ is the in plane shear force, $F_{z}$ is the axial load, and $M_{z}$ is the bending moment about the $z$ axis for the ith joint location. The $y_{i}$ and $A_{bolt}$ terms refer to half of the cross-section depth and the area of all bolts at the ith joint location respectively. The equations are used with MLife (Hayman, 2012; Hayman and Buhl, 2012) to obtain the bolt area that results in a 20-year fatigue life for Wöhler exponent $m$ equal to three (value used for metals). A factor of safety (FS) of 1.8, that comes from the multiplication of 1.5 (generally used for metals) and 1.2 (hole fitting factor), is applied to the von Mises stress equation.

Second, the required bolt diameter and number of bolts are determined by equating the number of required bolts to the number of bolts that fit along the ith cross-section arc-length (number of possible bolts) with equations (3) and (4) where $D$ is the diameter of the bolt.

No . of Required Bolts = \frac{Required Total Bolt Area}{Area for One Bolt} = \frac{A_{bolt}}{0.25 π D^{2}}

(3)

No . of Possible Bolts = \frac{Arc - length}{5 D}

(4)

The number of possible bolts is determined by assuming a bolt pitch distance (p) of five times the bolt diameter (p × D = 5 × D) based on historical practice in the aerospace industry to avoid stress concentrations due to proximity of holes to each other. Similarly an edge distance of three times the bolt diameter (e × D = 3D) is assumed for bolts on composite materials to avoid failure due to bearing, tear-out and shear-out on the laminate (DOD, 2002). However, failure in the bolted joint is more likely to occur due to bending and cracks that grow with blade deflection due to the thrust and gravity loading. Hence, an edge distance of one and a half the bolt diameter (e × D = 1.5 × D) is used instead based on the measurements from the SUMR-D built blades (Bay et al., 2019; Chetan et al., 2021; Kaminski et al., 2021; Yao et al., 2019, 2020). The minimum shell thickness at the joint is then two times the minimum edge distance (2 × e × D = 3 × D) as illustrated in Figure 11. In some cross-sections the nominal pitch distance may not be met due to the shorter cross-section depth near the trailing edge. This is not considered as a setback in Level I design but would require of more attention in a more rigorous Level II structural evaluation.

Figure 11.

Pitch (p) and edge (e) distance definitions for bolted joint design (artwork by Daniel Bouzolin).

Third, the adhesive materials in Table 3 are screened using the maximum work time recommended by the manufacturer (first screening). For the bolted joint, the adhesive work time in equation (5) is the time required to place adhesive inside one cavity/bolt hole in the shell. An adhesive material is only used for further calculations if the calculated work time does not exceed the manufacturer’s maximum work time.

Adhesive work time required (hr) = 2 π * radiu s_{bolt} (m) * 0.0167 hr / m

(5)

Fourth, the bolt length and adhesive thickness are sized to avoid lap shear failure in the adhesive. For this, a double lap shear joint is selected assuming there exists adhesive on the top and on the bottom of the bolt as shown in Figure 12. The shear stress in the adhesive as part of a double lap shear joint is calculated with equations (6) and (7) (Budynas et al., 2010), and compared to the allowable lap shear strength of the adhesive using a SF of 2.45 (assuming post-cured bonds and applicable for the abnormal load case DLC 2.3) based on guidelines from the Germanischer Lloyd (GL) standard for offshore wind turbines (GL, 2012). Thermal effects are not included for simplicity, but these can be non-trivial and should be taken into account in Level II design.

τ (x) = \frac{ω P}{4 b} [\frac{\cosh (ω x)}{\sinh (ω c)} + \frac{2 E_{o} t_{o} - E_{i} t_{i}}{2 E_{o} t_{o} + E_{i} t_{i}} * \frac{\sinh (ω x)}{\cosh (ω c)}]

(6)

ω^{2} = \frac{G_{a}}{t_{a}} [\frac{1}{E_{o} t_{o}} + \frac{2}{E_{i} t_{i}}]

(7)

Figure 12.

Double lap shear joint diagram showing the adherends (fiberglass shell and steel bolt) and the adhesive when a tension load “P” is applied.

In equations (6) and (7) (Budynas et al., 2010), the bond lap shear strength is $τ$ , the width of the bond is $b$ , the axial tension load is $P$ , the thickness of the material is $t$ , the modulus of elasticity is $E$ , the shear modulus of elasticity is $G$ , and the length of the bond is along the span of the blade. Based on equations (6) and (7), the largest shear stress occurs at the edges of the bond ( $x = c = L / 2$ ). This equation is used to calculate the required bolt length on one segment or side of the joint; hence, the total length of the bolt is twice this value. The tension load $P$ is assumed to be equal to $F_{z}$ (axial blade load at the corresponding blade span location) and the width of the bond $b$ is equal the arc-length of the bolt cross-section.

A rapid optimization is performed with the loads from the maximum combined loading case output from MExtreme (Hayman, 2015) for DLC 2.3 to determine the bond length and adhesive thickness that result in the minimum joint mass while meeting manufacturing and adhesive strength constraints (second screening of adhesive materials). The optimization uses the gradient based and nonlinear programing solver “fmincon” with the sequential quadratic programing “sqp” algorithm of MATLAB because analytical equations and a sufficient number of constraints are defined (MATLAB, 2019).

The axial force due to the bending moment loads were not included in the design of the bolt because to obtain the axial force from blade bending moments it is required to have prior knowledge about the number, location, and size of the bolts. Thus, the joint is designed with the methods described above first and afterward a structural performance evaluation is performed on the design by including the bending moments. The methodology for including the bending moment is described next.

Structural evaluation including bending moment methodology

The objective of this evaluation is to determine the feasibility of the design in regard to fatigue for each bolt when all blade beam loads are included in a von Mises stress (GL, 2012) calculation using equations (8) to (21). A pre-tension load is added to the bolt to exert an equivalent compression load on the surrounding joint structure that helps keep the joint segments together (Budynas et al., 2010). The step-by-step methodology is described next:

Determine the distribution of axial loads at each joint cross-section between the bolts and the surrounding fiberglass (Budynas et al., 2010). The stiffness factor $K_{s}$ is the ratio of the axial stiffness of the bolts versus the total axial stiffness of the cross-section. $A$ is area and $E$ is the elastic modulus of elasticity.

(a) Total cross-section area

A_{total} = A_{fiberglass} + A_{bolts}

(8)

(b) Stiffness factor

K_{s} = \frac{E_{steel} A_{bolts}}{E_{steel} A_{bolts} + E_{fiberglass} A_{fiberglass}}

(9)

2. Determine the axial force taken by each bolt and the axial force taken by the fiberglass surrounding the bolt. The pre-tension factor (PTF) is assumed to be 65% of the steel yield stress. The force is $P$ , $M$ is bending moment, $n_{bolts}$ is number of bolts, $I$ is the second moment of area of the full joint cross-section, $x$ and $y$ are the coordinates of the bolt on the cross-section in accordance to Figure 13, and $σ_{y}$ is yield stress.

(a) Axial/normal force on bolt due to:

(i) Axial force on cross-section

P_{Z} = \frac{F_{Z}}{n_{bolts}}

(10)

(ii) Force from local flap-wise bending moment

P_{MY} = \frac{M_{y} * x_{bolts}}{I_{flap - wise cross - section}} * A_{bolt}

(11)

(iii) Force from local edge-wise bending moment

P_{Mx} = \frac{M_{x} * y_{bolts}}{I_{edge - wise cross - section}} * A_{bolt}

(12)

(iv) Bolt pre-tension

P_{PT} = σ_{y_{steel}} * PTF * A_{bolt}

(13)

(b) Total local axial force on:

(i) Bolt

P_{total axia l_{B}} = K_{s} * (P_{z} + P_{MY} + P_{MX}) + P_{PT}

(14)

(ii) Fiberglass

P_{total axia l_{F}} = (1 - K_{s}) * (P_{z} + P_{MY} + P_{MX}) - P_{PT}

(15)

3. Determine the shear force taken by each bolt. The angle to the bolt from the reference axis is $ϕ$ and the radius of gyration to the bolt is $ρ_{g}$ in accordance to Figure 13 where the flap-wise and edge-wise directions are indicated. The polar moment of inertia of the full joint cross-section is $J$ .

(a) Shear force on bolt due to:

(i) Force along the flap-wise direction on cross-section

P_{X} = \frac{F_{X}}{n_{bolts}}

(16)

(ii) Force along the edge-wise direction on cross-section

P_{Y} = \frac{F_{Y}}{n_{bolts}}

(17)

(iii) Local force along the flap-wise direction due to torque on cross-section

P_{M z_{X}} = - \cos ϕ * \frac{M_{Z} * ρ_{g}}{J_{cross - section}} * A_{bolt}

(18)

(iv) Local force along the edge-wise direction due to torque on cross-section

P_{M z_{Y}} = \sin ϕ * \frac{M_{Z} * ρ_{g}}{J_{cross - section}} * A_{bolt}

(19)

(b) Total local shear force on bolt:

P_{Total Shea r_{B}} = \sqrt{{(P_{X} + P_{M z_{X}})}^{2} + {(P_{Y} + P_{M z_{Y}})}^{2}}

(20)

4. Calculate the von Mises stress $σ_{vM}$ on each bolt:

σ_{v M_{bolt}} = F S_{Strength} \frac{\sqrt{P_{total axia l_{B}}^{2} + 3 P_{total shea r_{B}}^{2}}}{A_{bolt}}

(21)

Figure 13.

Orientation of axes and positive angles. Y+ points toward the trailing edge of the airfoil and X+ to the low pressure surface of the airfoil.

A Wöhler exponent $m$ equal to five, also applicable to metals, is also evaluated based on further review of the GL (2012) standard. Because the bending moment and pre-tension are included in this evaluation, we use a lower safety factor of 1.15 to evaluate fatigue. This factor is selected from Tables 5.3.1 and 5.3.2 of the GL (2012) standard assuming that the bolted connections could be periodically monitored and maintained, that they would have poor accessibility, and if failure were to occur then the result would be either wind turbine failure or consequential damage.

Bonded joint design methodology

The Level I detailed bonded joint design is described in this section. Different types of bonded joints have been developed and studied. Some of these are illustrated in Figure 14. Double lap shear joints have larger bonded joint strength and are not subject to failure due to eccentric load paths compared to single lap shear joints (Chamis and Murthy, 1991; Smith, 1973), but are also not as complex to model as other higher strength joints like the stepped lap joints. Hence, the double lap shear joint is considered appropriate to do a Level I study.

Figure 14.

Bonded joint geometry versus strength: recreated from DOD (2002) (artwork by Daniel Bouzolin).

The simplified double lap shear joint layout shown in Figure 15 is used for Level I design in this paper. The double lap shear joint type includes one inboard (closer to the blade root) section “A” that has an empty space to receive the flange of the outboard (closer to the tip) “B” section. The two parts are joined using adhesive. The additional adhesive between the vertical boundaries of side A and side B in Figure 15 is not considered in the present analysis. Details of higher complexity are part of a more rigorous Level II design, thus are outside of the scope of this paper.

Figure 15.

Double lap shear joint diagram.

The main steps in the design process for bonded joints are listed below:

Screen the adhesive materials based on the work time required to bond the blade segments;

Determine span-wise length and thickness of joint components using DLC 2.3 loads;

Calculate the segmented blade mass and determine the effect on natural frequencies.

The first screening of adhesive materials is performed by comparing the work time required to bond the blade segments, that is calculated with equation (22) and depends on the full arc-length of the blade cross-section, and the maximum allowed work time defined by the manufacturer.

Adhesive Work Tim e_{bonded} (hr) = Arc - lengt h_{joint} (m) * 0.0167 hr / m

(22)

A gradient based optimization is performed using the down-selected adhesive materials from the first step. The “fmincon” solver with the “sqp” algorithm of MATLAB is used with equation (6) (MATLAB, 2019), in which both outer and inner adherends are made of fiberglass, to find a feasible design with minimum joint mass. The joint mass is a function of the thickness of the adherends, the thickness of the adhesive and the span-wise bond length. The span-wise bond length and the thickness of the adherends and adhesive are calculated with a SF of 2.45 (GL, 2012). The optimization has constraints on the thickness of the adherends, the double lap bond shear stress and adhesive thickness (allowable values are specified in Table 3 for the last two). Thus, this optimization also is a second adhesive screening.

The thickness of all adherends are set to be equal and are initially defined to be one third of the maximum shell thickness of the monolithic blade at the joint span-wise location. However, the thickness of the adherends can be increased to satisfy strength requirements and adhesive constraints because the shell thickness is not defined by the 1.5D edge distance of the bolted joint.

Segmentation design results

Bolted joint design results

The fatigue life of the metal bolts at each joint location are shown in Figure 16 as a function of the total bolt area determined with equations (1) and (2). The total bolt area is smaller for outboard locations because the loads are lower. There is a significant decrease in bolt area at the 75% station compared to the 50% station due to a threefold reduction in loads. The methods used to determine the bolt diameter and number of bolts with Figure 16 results are illustrated in Figure 17 for the 50% blade span.

Figure 16.

Fatigue life versus total bolt area for bolted joints in different blade span locations.

Figure 17.

Number of bolts and shell thickness versus bolt diameter for joint at 50% blade span.

The number of bolts that can fit along the arc-length of the cross-section is determined by the intersection of the blue and red curves in Figure 19 (bolts required vs bolts possible for a pitch distance of 5D) that are given by equations (3) and (4). The shell thickness limit is half of the depth of the cross-section which is equivalent to a solid joint. Any point below the thickness limit is considered feasible. The required shell thickness is calculated with a 1.5D edge distance assumption. The resultant bolt diameter and number of bolts using this method are tabulated in Table 4.

Table 4.

Diameter and quantity of bolts for all joint locations.

Joint location % of blade span	Bolt diameter (m)	Number of bolts (:)	Total bolt area (m²)
30	0.2	29	0.92
40	0.22	21	0.82
50	0.23	16	0.66
75	0.12	19	0.21
80	0.09	23	0.16

Table 5.

Bonded joint—screening of adhesive materials.

Joint location % of blade span	First screening
	Work time (hour)	Feasible materials
30	0.493	All except 2 and 5
40	0.395	All except 2 and 5
50	0.314	All except 5
75	0.196	All except 5
80	0.186	All except 5

The first screening of the adhesive materials is performed using the largest bolt diameter (0.23 m at 50% blade span) and equation (5) that results in 43 seconds required to place adhesive on each bolt cavity. This is eight times less than the smallest maximum work time in Table 3, but may be possible to achieve with automated machines or robotic arms. Based on this result we proceed to the next step using all the seven adhesives. Figure 18 shows the mean and standard error with 95% confidence of the bolt lengths obtained through the optimization as well as the individual results for each of the seven adhesive materials denoted by “M.”

Figure 18.

Span-wise bolt length versus bolted joint location.

In general, the bolt length reduces closer to the blade tip due to the decrease in the load magnitude and blade chord as shown in Figure 8. The results for materials six and seven overlap because these adhesives have similar structural properties. Figure 19 shows the results of the fatigue analyses including the bending moment and bolt pre-tension for two different Wöhler exponent values and three safety factors (1, 1.15, and 1.35).

Figure 19.

Time to failure of critical bolts for various combinations of Wöhler and safety factors: (a) top (low pressure side) bolt and (b) bottom (high pressure side) bolt.

In the fatigue analysis, a safety factor of one means that no safety factor is applied. A safety factor of 1.15 is analyzed based on the revision of the GL (2012) standard as explained in the previous section. This safety factor is considered appropriate despite being lower than the previously used 1.8 value because in this fatigue evaluation the bending moment and pre-tension are included. A safety factor of 1.35 is typically applied to the wind loads (GL, 2012) and is analyzed here to help identify the load limits of the bolted joint designs. The top bolt (low pressure side) experiences the largest stresses and is subject to failure earlier than other bolts due to the downwind configuration of the wind turbine combined with the pre-tension load. Based on these results, the bolted joint designs are feasible in terms of stress (DLC 1.2 and DLC 2.3) and fatigue life (DLC 1.2). The latter using an adjusted Wöhler exponent of five and a fatigue safety factor of 1.15.

Bonded joint design results

Table 5 shows the results of the first adhesive screening based on the work time required to place adhesive along the arc-length of the joint cross-section calculated using equation (22). Then, an optimization is performed with remaining feasible materials to find the thickness of adherends, thickness of adhesive and length of the bond. The latter is shown in Figure 20.

Figure 20.

Initial bond length versus bonded joint location.

The length of the span-wise bond reduces when the joint is closer to the blade tip because the loads become smaller closer to the blade span. However, for bonded joints, unrealistically small bond lengths are obtained because the width of the bond, which is in the denominator of equation (6), is assumed to be equal to the arc-length of the large blade cross-section and this reduces the shear stress significantly. The bond length may be higher if the local effect of the blade bending moments were included. However, including the bending moments is a complex task that would require higher fidelity models and these are outside of the scope of this paper. Instead, we keep the optimized adherend and adhesive thicknesses, but investigate the effect of making the length of the span-wise bond longer. Example results for the 50% blade span are shown in Figure 21.

Figure 21.

Adjusted margin of safety (a value of zero indicates an optimal design) versus bond length factor at 50% blade span.

As the bond length grows, the adjusted margin of safety levels off to a specific value because of three reasons: the presence of the hyperbolic functions in equation (6), the evaluation is performed at the ends of the bond, and the outer adherends have the same material properties and thickness. Thus, equation (6) simplifies to equation (23) (with $x = c = L / 2$ which is at the ends of the bond where the stress is highest):

τ (L / 2) = \frac{ω P}{4 b} [\frac{\cosh ω * L / 2}{\sinh ω * L / 2} + \frac{1}{3} \frac{\sinh ω * L / 2}{\cosh ω * L / 2}]

(23)

With larger bond lengths the cosh and sinh terms increase rapidly as shown in Figure 22. By the point where the bond length reaches about 5–16 times the original length (depending on the blade span station and the adhesive material) the cosh/sinh and sinh/cosh terms approximate the value of one such that the equation (6) further simplifies to equation (24):

τ (L / 2) = \frac{ω P}{3 b}

(24)

Figure 22.

Value of equation term versus bond length factor at 50% blade span.

New and much longer bond lengths are selected by taking the first point that results in the maximum adjusted margin of safety per material and per joint span-wise location. The results of this procedure, that lengthens the bond and uses equation (6) (method 1), are shown in Figure 23.

Figure 23.

Comparison of bond length versus bonded joint location using various methods.

Results from a second method (method 2), that is formulated in equation (25) with an additional safety factor of 2.45, are also added (Chamis and Murthy, 1991; Osnes et al., 2011). This indicates that the length of the bond can be designed such that two times the axial force per unit width does not exceed the allowable lap shear strength.

bond length (m) = \frac{2 * FS * F_{z} / (1 m)}{τ_{allowable}}

(25)

The bond lengths resultant of the initial sizing obtained with method 1 are not used in further analyses because they are in unrealistically short. Instead, we use the extended bond length that results in the maximum adjusted margin of safety, illustrated in Figure 21 for the 50% blade span, because its calculation still includes both the adhesive and adherend properties. The results of method 2 are illustrative and helpful to support the lengthened bond results obtained with equation (6).

Blade mass results

Three taper ratios (ply dropping of fiberglass layers to linearly build up to the joint shell thickness) are evaluated to study the effect of segmentation on the blade mass: one-to-one (1:1), one-to-ten (1:10), and one-to-twenty (1:20). The last two are typically used as a general guideline in the design of composite structures in the aerospace industry. The guideline recommends using a 1:20 taper ratio in the direction of the main load and 1:10 along the transverse direction to the main load. The joint shell thickness is the summation of the thickness of adherends and adhesives at the location of the joint on the blade span. A bolted joint has one inner adherend of thickness equal to the bolt diameter, two outer adherends that are fiberglass laminates and are defined by the edge distance rule of 1.5D, and two adhesive layers. A bonded joint instead has all adherends (two outer and one inner) made of fiberglass and two adhesive layers. These are illustrated in Figures 10, 12, and 15.

The segmented SUMR50-S5 blade options were modeled using AutoNuMAD which is a semiautomated tool developed at the University of Texas at Dallas and is based on the NuMAD framework (Berg and Resor, 2012). AutoNuMAD was introduced by Chetan et al. (2019a, 2019b), and further capabilities were added in Yao et al. (2021b, 2021a) and Escalera Mendoza et al. (2021). The internal blade structure was modified only at the region spanned by the joint structure using the results from the previous sub-sections while the rest of the blade was left unmodified. This means that materials that are not part of the joint, such as spar cap carbon and shell core, were removed at the region of the joint to include the joint material instead. Twenty-four models were created to represent the segmented options that vary in number of segments, location of segments, taper ratio, and joint type. Only the average results for bond length, bolt length, adherend thickness, and adhesive thickness were used to reduce the number of models. Then, a custom tool was created to convert the AutoNuMAD blade models to a format used by WISDEM (Bortolotti et al., 2019; NREL, 2019). WISDEM is a tool created by NREL that as part of its suite of capabilities can provide detailed blade material and direct labor costs, which are of interest for Escalera Mendoza et al. (2022). For consistency, the blade mass results in this section were calculated using WISDEM.

The blade mass includes the mass of all materials that form part of the blade structure except for root bolts, lighting protection system (LPS), leading edge, and trailing edge adhesive, and overlays. WISDEM automatically makes cost estimates of these components based on a procedure described by Bortolotti et al. (2019), but these items were not structurally sized and consequently are not present in the blade mass results. The monolithic blade mass obtained with WISDEM is 526 metric tons and Figure 24 shows the percent difference of the mass of the segmented options relative to the monolithic result.

Figure 24.

Change in blade mass due to presence of joints for various segmentation options and taper ratios: (a) bolted and (b) bonded.

The blade mass increases when joints are present mainly due to the addition of fiberglass despite of the removal of spar cap carbon and core mass at the joint. A greater amount of segments, larger taper ratios, and joints located at blade spans stations with larger chord result in larger mass growths. Controlling the mass growth is a challenge for bolted joints because many fiberglass layers are added to fit the bolt, driven by the 1.5D edge distance assumption, along the full arc-length of the joint cross-sections and the chords for a 250 m blade are of extreme size. Fiberglass is the largest contributor to the joint mass (76%–99% depending on joint location and taper ratio) because the volume of the fiberglass is much larger than of the adhesive (which makes the adhesive be less than 1% of the joint mass in all segmentation options) and bolts combined, the latter being much denser than fiberglass.

For bonded joints, the fiberglass is the largest contributor to mass but the joints are lighter than bolted joints because no steel bolts are present (one less material) and because the shell thickness at the joint is smaller (not defined by edge distance rule) that makes bonded joints less sensitive to changes in the taper ratio. The adhesive mass is less than 2% in bonded joints.

Comparison of parked frequency results

The mass and stiffness effect of the joint structure on the first flap-wise and edge-wise frequencies is evaluated using AutoNuMAD (Chetan et al., 2019a, 2019b; Escalera Mendoza et al., 2021; Yao et al., 2021a, 2021b) and Ansys Mechanical APDL (AnsysÒ, 2018) at zero revolutions per minute (rpm). Bolted joints modeled in AutoNuMAD include: mass and stiffness effect of the additional fiberglass, only mass effect of bolts and adhesive, and mass and stiffness effect of removing material present at the joint location that corresponds to the monolithic blade but not to the segmented blade (e.g. tapering out the shell core while tapering in the fiberglass of the joint). The bolts and adhesive are discrete at the joint location (they do not form a continuous layer along the cross-section arc-length), thus it is not possible to accurately capture the stiffness effect of bolts and adhesive using AutoNuMAD. The mass effect of the bolts and adhesive can be captured by modeling these as layers and adjusting the corresponding densities to match the mass that corresponds to the average joint geometry results presented in the previous sub-sections. An inaccurate stiffness effect of the bolts and adhesive is avoided by setting the Young’s modulus and the shear modulus to very small values. Table 6 shows the frequency change due to bolted joints in reference to the frequencies obtained with Ansys for the monolithic one-piece blade (first flap-wise of 0.261 Hz and first edge-wise of 0.351 Hz). For example, the largest frequency reduction for the first flap-wise mode is for option 3-S-B where the frequency is reduced by 21.7% to 0.205 Hz.

Table 6.

Percentage change on parked blade frequencies with bolted joints.

Segmentation option	First flap			First edge
	1:1 (%)	1:10 (%)	1:20 (%)	1:1 (%)	1:10 (%)	1:20 (%)
2-S	−3.85	−3.70	−7.28	−0.39	−9.45	−16.6
3-S-A	−5.14	−7.38	−12.2	−0.12	−9.19	−15.5
3-S-B	−5.02	−14.1	−21.7	−0.58	−17.4	−26.6
4-S	−9.22	−21.5	−20.3	−1.74	−18.4	−26.4

A segmented blade with bolted joints has lower first flap-wise and first edge-wise frequencies. In general, the decrease is larger with higher taper ratios because the blade mass grows more significantly than the increase in stiffness due to the joint material. However, the change in frequencies is also dependent on the location of the joint. Joints closer to the root have a smaller effect on frequencies despite being the heaviest. The frequencies are more sensitive to changes in mass near the blade tip, but near the blade tip the joint mass is smaller compared to near the blade root. Large reductions in frequencies are not desirable because small frequencies are more difficult to control.

Table 7 shows the change in the first two monolithic blade frequencies due to the bonded joints. Bonded joints modeled in AutoNuMAD include: mass and stiffness effect of the additional fiberglass and adhesive (since both form continuous layers), and mass and stiffness effect of removing material at the joint location that belongs to the monolithic blade but not to the segmented blade.

Table 7.

Percentage change on parked blade frequencies with bonded joints.

Segmentation option	First flap			First edge
	1:1 (%)	1:10 (%)	1:20 (%)	1:1 (%)	1:10 (%)	1:20 (%)
2-S	1.20	0.53	−0.42	1.51	1.37	0.91
3-S-A	0.95	0.09	−1.08	1.37	1.17	0.65
3-S-B	0.72	−0.56	−2.48	1.12	0.51	−0.76
4-S	0.20	−1.59	−3.78	0.94	0.17	−1.07

For all bonded segmentation options, the change in frequencies shown in Table 7 is either negligible or small compared to the results with bolted joints that could have a smaller impact in the control design. Higher taper ratios reduce the frequency values due to the larger joint mass added and the longer extent of the joint along the blade span. An increase in frequency indicates that the increase in stiffness is more significant than the change in mass. Segmentation options with frequency reductions show that the increase in mass is more considerable than the change in stiffness. The results in this section show that some bolted joint options and all bonded joint options could be structurally feasible in terms of strength, mass, fatigue, and frequencies.

The change in mass and stiffness due to the addition of joints also has an effect on the flutter performance of the blade. The effect on flutter was evaluated using a flutter tool (Griffith and Chetan, 2018) that indicates a tendency for flutter speeds to increase with the addition of joints due to the increase in stiffness at the joint locations. However, further research is warranted to investigate the effects of segmentation on flutter.

Conclusions

A detailed Level I design and analysis of a segmented rotor blade for an extreme-scale 50 MW wind turbine is presented herein. Detailed methodologies were developed to investigate the impact of segmentation on the blade mass and blade frequencies while evaluating its structural feasibility. A matrix of options are studied that include number of joints, joint materials, joint types (bolted and bonded) and taper ratios. The results obtained are useful for the design of increasingly large blades, to further reduce LCOE, because these provide guidance on segmentation design, answers to foreseen challenges of increasing blade lengths and of implementing segmentation on blades, and point attention to research topics that are highly valuable to continued growth of the wind energy industry.

It is shown that segmenting the blade increases the blade mass with larger mass increase with bolted joints, joints closer to the root due to the extreme size of chords and with higher number of joints. For taper ratios up to 1:10, the blade mass increases by 4.1%–62% with bolted joints and by 0.4%–3.6% with bonded joints. The taper ratio (ply dropping) has a non-trivial effect on blade mass for options with bolted joints. Challenges on controlling mass growth for the bolted joint type are observed and alternative design methods that reduce this are needed.

A modal analysis is performed to evaluate the effect of segmentation on the first two natural frequencies of the blade. Segmented blades with bolted joints show significant reductions in the natural frequencies due to the large blade mass, that in most cases exceeds the additional stiffness obtained by adding large amounts of fiberglass to fit the bolts (driven by the 1.5D edge distance rule). Large reductions on the natural frequencies are not preferred because these increase the difficulty of control design for wind turbines. On the other hand, segmented blades with bonded joints have trivial to small changes in the natural frequencies because bonded joints are more lightweight and have a smaller impact on the local stiffness.

The Level I structural research studies show that segmenting a blade can be structurally feasible and creates awareness about the effect on mass and frequencies of an extreme-scale blade design. A cost analysis based on the design results obtained herein is presented in Escalera Mendoza et al. (2022) that includes direct labor, material, manufacturing, assembly, installation, and transportation as depicted in Figure 25.

Figure 25.

Ocean based transportation of segmented two-piece 50 MW blades (artwork by Daniel Bouzolin).

Footnotes

Acknowledgements

The research presented herein was funded by the US Department of Energy Advanced Research Projects Agency-Energy (ARPA-E) under the Segmented Ultralight Morphing Rotor project (award number DE-AR0000667). The authors are grateful for the support of the ARPA-E program and staff. The authors also acknowledge the support of the entire SUMR team. The authors also appreciate the help of Daniel Bouzolin from the University of Texas at Dallas. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of ARPA-E.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the U.S. Department of Energy Advanced Research Projects Agency-Energy (ARPA-E) (award number DE-AR0000667).

ORCID iDs

Alejandra S Escalera Mendoza

Shulong Yao

Mayank Chetan

D Todd Griffith

Data availability statement

The data that support the findings of this study are not openly shared due to confidentiality. Limited data may be shared from the corresponding author upon reasonable request.

References

3M (2012a) Wind blade bonding adhesive W1101 technical data sheet. Available at: https://multimedia.3m.com/mws/media/723392O/3m-wind-blade-bonding-adhesive-w1101.pdf (accessed 24 July 2019).

3M (2012b) Wind epoxy structural adhesive W1115 technical data sheet. Available at: https://multimedia.3m.com/mws/media/642707O/3m-wind-epoxy-structural-adhesive-w1115-technical-data-sheet.pdf (accessed 24 July 2019).

AnsysÒ (2018) Academic research mechanical, Release 19.2. https://www.ansys.com/academic/terms-and-conditions (accessed 21 December 2021).

Barnhart

(2016) Space frame blade technology. Available at: https://www.slideshare.net/sandiaecis/ryan-barnhart-space-frame-blade-technology (accessed 25 July 2019).

Bay

Damiani

Fingersh

, et al. (2019) Design and testing of a scaled demonstrator turbine at the national wind technology center. In: AIAA Scitech 2019 forum, p.13. San Diego, CA: American Institute of Aeronautics and Astronautics. DOI: 10.2514/6.2019-1068.

Berg

Resor

(2012) Numerical manufacturing and design tool (NuMAD v2.0)for wind turbine blades: User’s guide. Technical report SAND2012-7028, Sandia National Laboratories (SNL). DOI: 10.2172/1051715.

Bhat

Noronha

Saldanha

(2015) Structural performance evaluation of segmented wind turbine blade through finite element simulation. International Journal of Mechanical Engineering and Mechatronics 9(6): 996–1005.

Bortolotti

Berry

Murray

, et al. (2019) A detailed wind turbine blade cost model. Technical report NREL/TP-5000-73585, National Renewable Energy Laboratory (NREL), Golden, CO. DOI: 10.2172/1529217.

Budynas

Nisbett

Tangchaichit

(2010) Shigley’s Mechanical Engineering Design, 9 edn. New York: McGraw Hill.

10.

BUMAX (2019) BUMAX ultra fastener. Available at: https://www.bumax-fasteners.com/bumax-grades/bumax-ultra/ (accessed 25 July 2019).

11.

Callén

(2017) Nabrajoint: The solution for XXL blades segmentation. Technical report PPT-0398-RevA, Nabrawind Advanced Wind Technologies, Bremen, Germany.

12.

Chamis

Murthy

(1991) Simplified procedures for designing adhesively bonded composite joints. Journal of Reinforced Plastics and Composites 10(1): 29–41.

13.

Chetan

Griffith

Yao

(2019a) Flutter predictions in the design of extreme-scale segmented ultralight morphing rotor blades. In: AIAA Scitech 2019 forum, pp.3–6. San Diego, CA: American Institute of Aeronautics and Astronautics. DOI: 10.2514/6.2019-1298.

14.

Chetan

Sakib

Griffith

, et al. (2019b) Aero-structural design study of extreme-scale segmented ultralight morphing rotor blades. In: AIAA aviation 2019 forum, p.4. Dallas, TX: American Institute of Aeronautics and Astronautics. DOI: 10.2514/6.2019-3347.

15.

Chetan

Yao

Griffith

(2021) Multi-fidelity digital twin structural model for a sub-scale downwind wind turbine rotor blade. Wind Energy 24: 1368–1387.

16.

de Vries

(2014) Close-up: Envision’s EN128/3.6MW direct drive turbine. Available at: https://www.windpoweroffshore.com/article/1227512/close-up-envisions-en128-36mw-direct-drive-turbine (accessed 25 July 2019).

17.

DOD (2002) Polymer matrix composites guidelines for characterization of structural materials. In: Document Automation and Production Service (DAPS), MIL-HDBK-17-1F: Composite Materials Handbook, vol. 1. Philadelphia, Pennsylvania, United States Department of Defense, pp.7.14–7.34.

18.

Dutton

Kildegaard

Kensche

, et al. (2000) Design, structural testing, and cost effectiveness of sectional wind turbine blades. Technical report JOR3-CT97-0167, Rutherford Appleton Laboratory, Oxfordshire, England.

19.

Enercon (2019) Enercon E-160 EP5. Available at: https://www.enercon.de/en/products/ep-5/e-160-ep5/ (accessed 25 July 2019).

20.

Escalera Mendoza

Chetan

Griffith

(2021) Quantification of extreme-scale wind turbine performance parameters due to variations in beam properties. In: AIAA Scitech 2021 forum, pp.4–6. Virtual: American Institute of Aeronautics and Astronautics. DOI: 10.2514/6.2021-1603.

21.

Escalera Mendoza

Yao

Chetan

, et al. (2022) Techno-economic analysis of blade segmentation for an extreme-scale 50 MW wind turbine rotor. Journal In Preparation.

22.

Gardiner

(2013) Modular design eases big wind blade build. Composites World. Available at: https://www.compositesworld.com/articles/modular-design-eases-big-wind-blade-build (accessed 25 July 2019).

23.

Gardiner

(2016) Automated filament winding aids segmented blade production. Composites World. Available at: https://www.compositesworld.com/blog/post/automated-filament-winding-aids-segmented-blade-production (accessed 25 July 2019).

24.

GE Renewable Energy (2021) Cypress onshore wind turbine platform. Available at: https://www.ge.com/renewableenergy/wind-energy/onshore-wind/cypress-platform (accessed 20 June 2021).

25.

GL (2012) IV Rules and guideline industrial services. In: GL Renewables Certification, Guideline for the Certification of Wind Turbines. Hamburg, Germany: Germanischer Lloyd, pp.4.1–6.51.

26.

Griffith

Ashwill

(2011) The Sandia 100-meter all-glass baseline wind turbine blade: SNL100-00. Technical report SAND2011-3779, Albuquerque, New Mexico, Sandia National Laboratories.

27.

Griffith

Chetan

(2018) Assessment of flutter prediction and trends in the design of large-scale wind turbine rotor blades. Journal of Physics Conference Series 1037: 042008. DOI: 10.1088/1742-6596/1037/4/042008

28.

Hayman

(2012) MLife theory manual for version 1.00. Technical report NREL/TP-XXXXX, National Renewable Energy Laboratory (NREL), Golden, CO.

29.

Hayman

(2015) MExtremes manual version 1.00. Technical report NREL/TP-XXXXX, National Renewable Energy Laboratory (NREL), Golden, CO.

30.

Hayman

Buhl

Jr (2012) Mlife users guide for version 1.00. Technical report NREL/TP-XXXXX, National Renewable Energy Laboratory (NREL), Golden, CO.

31.

Hexion (2019a) Epikote-TM resin MGS BPR 135G-series and EPIKURE-TM curing agent MGS BPH 134G-137GF technical data sheet. Available at: https://www.hexion.com/CustomServices/PDFDownloader.aspx?type=tds&pid=45668d44-5814-6fe3-ae8a-ff0300fcd525 (accessed 23 July 2019).

32.

Hexion (2019b) Epikote-TM resin MGS BPR 535 and Epikure curing agent MGS BPH 537 technical data sheet. Available at: https://www.hexion.com/CustomServices/PDFDownloader.aspx?type=tds&pid=c4ce8f44-5814-6fe3-ae8a-ff0300fcd525 (accessed 23 July 2019).

33.

Iriarte

(2017) Logistic cost model: Identifying modular blades market. Technical report PPT-0944 Rev. A, Nabrawind Advanced Wind Technologies, Bremen, Germany.

34.

Jackson

Zuteck

van Dam

, et al. (2005) Innovative design approaches for large wind turbine blades. Wind Energy 8(2): 141–171.

35.

Johnson

Bortolotti

Dykes

, et al. (2019) Investigation of innovative rotor concepts for the big adaptive rotor project. Technical report NREL/TP-5000-73605, National Renewable Energy Laboratory (NREL), Golden, CO. DOI: 10.2172/1563139.

36.

Jørgensen

(2017) Adhesive joints in wind turbine blades. PhD Thesis, Technical University of Denmark (DTU), Denmark. DOI: 10.11581/DTU:00000027.

37.

Kaminski

Noyes

Loth

, et al. (2021) Gravo-aeroelastic scaling of a 13-MW downwind rotor for 20% scale blades. Wind Energy 24: 229–245.

38.

Kensche

(2006) Fatigue of composites for wind turbines☆. International Journal of Fatigue 28(10): 1363–1374.

39.

Lesmes

(2019) Advanced validation strategies for segmented blades. Technical report PPT-1358, Nabrawind Advanced Wind Technologies, Bremen, Germany.

40.

MATLAB (2019) Version r2019a. Natick, MA: The MathWorks Inc.

41.

Matweb (2019) ASTM A709 steel grade 50 properties technical data sheet. Available at: http://www.matweb.com/search/datasheet_print.aspx?matguid=9ccae9a8aeca4739b63ca27f2dc24970 (accessed 30 July 2020).

42.

Noon

(2019) Green giant: Cypress, GE’s huge new onshore wind turbine, comes to life. Available at: https://www.ge.com/reports/green-giant-cypress-ges-huge-new-onshore-wind-turbine-comes-life/ (accessed 25 July 2019).

43.

NREL (2019) WISDEM Version TBD. Golden, CO: National Renewable Energy Laboratory.

44.

Osnes

Guthu

McGeorge

(2011) Chapter 13: Strength of bonded overlap composite joints in marine applications. In: Camanho

Tong

(eds) Composite Joints and Connections, Woodhead Publishing Series in Composites Science and Engineering. Philadelphia, Pennsylvania, Woodhead Publishing, pp.399–422.

45.

Pao

Zalkind

Griffith

, et al. (2021) Control co-design of 13 MW downwind two-bladed rotors to achieve 25% reduction in levelized cost of wind energy. Annual Reviews in Control 51: 331–343.

46.

Pedersen

Rua

IADL

Sola

, et al. (2009) Sensorised blade joint. Technical report US20090116962A1, Gamesa Innovation and Technology SL. Available at: https://patents.google.com/patent/US20090116962A1/en (accessed 7 May 2021).

47.

Peeters

Santo

Degroote

, et al. (2017) The concept of segmented wind turbine blades: A review. Energies 10(8): 1112.

48.

Plexus (2018a) MA 300-revision 7 technical data sheet. Available at: https://itwperformancepolymers.com/media/1145/ma300-data-sheet_rev07.pdf (accessed 20 July 2019).

49.

Plexus (2018b) MA 530-revision 14 technical data sheet. Available at: https://itwperformancepolymers.com/media/10759/ma530-data-sheet_rev14.pdf (accessed 26 July 2019).

50.

Plexus (2018c) MA 560-1-revision 10 technical data sheet. Available at: https://itwperformancepolymers.com/media/1156/ma560-1-data-sheet_rev10.pdf (accessed 28 July 2019).

51.

Smith

Griffin

(2019) Supersized wind turbine blade study: R&D pathways for supersized wind turbine blades. Technical report 10080081-HOU-R-01, Lawrence Berkeley National Lab, CA. DOI: 10.2172/1498695.

52.

Smith

(1973) Adhesive-bonded single-lap joints. Technical report NASA-CR-112236, National Aeronautics and Space Administration (NASA), Hampton, VA.

53.

UpWind (2011) Upwind: Design limits and solutions for very large wind turbines—A 20 MW turbine is feasible. Technical report, Risø Technical University of Denmark (DTU) National Laboratory, Denmark. Available at: http://www.ewea.org/fileadmin/files/library/publications/reports/UpWind_Report.pdf (accessed 25 July 2019).

54.

Vionis

Costales

Mieres

, et al. (2006) Development of a MW scale wind turbine for high wind complex terrain sites; the MEGAWIND project. In: Proceedings of the European wind energy conference, Athens, Greece, p.1.

55.

Wetzel

(2014) Modular blade design & manufacturing. Available at: https://www.slideshare.net/sandiaecis/2014-wind-turbine-blade-workshop-wetzel (accessed 24 July 2019).

56.

Wind-Turbine-Models (2011) Enercon E-126-7.58MW. Available at: https://en.wind-turbine-models.com/turbines/14-enercon-e-126-7.580 (accessed 25 July 2019).

57.

Wind-Turbine-Models (2012) Envision E-128-3.6MW. Available at: https://en.wind-turbine-models.com/turbines/284-envision-e128-3.6mw-pp-2b#datasheet (accessed 25 July 2019).

58.

Yao

Chetan

Griffith

(2021a) Structural design and optimization of a series of 13.2 MW downwind rotors. Wind Engineering 45: 1459–1478.

59.

Yao

Chetan

Griffith

, et al. (2021b) Aero-structural design and optimization of 50 MW wind turbine with over 250-m blades. Wind Engineering. Epub ahead of print 24 July 2021. DOI: 10.1177/0309524X211027355.

60.

Yao

Griffith

Chetan

, et al. (2019) Structural design of a 1/5th scale gravo-aeroelastically scaled wind turbine demonstrator blade for field testing. In: AIAA Scitech 2019 forum, p.1. San Diego, CA: American Institute of Aeronautics and Astronautics. DOI: 10.2514/6.2019-1067.

61.

Yao

Griffith

Chetan

, et al. (2020) A gravo-aeroelastically scaled wind turbine rotor at field-prototype scale with strict structural requirements. Renewable Energy 156: 535–547.

62.

Yarbrough

Caruso

Huth

, et al. (2019) Joint assembly for a wind turbine rotor blade. Technical report US20190040842A1, General Electric Co. Available at: https://patents.google.com/patent/US20190040842A1/en (accessed 22 November 2021).