2.1 How often the failure of WT blades takes place?

Among the main causes of WT blade damage, one can list the manufacturing defects, transportation damage, assembly damage, installation damage, lightning strikes, environmental wear, rain, sand, and contaminants caused erosion, bird impacts, thermal cycling, leading and trailing edge erosion, fatigue, moisture intrusion and foreign object impact, egress of moisture through the laminate skin structure, as a result of the surface damage, and mechanical failure.^9-12

With an estimated 700 000 blades in operation globally, there are, on average, 3800 incidents of blade failure each year. Figure 1 shows onsite repair of WT.

According to Carrol et al.,¹³ average failure rate of an offshore WT is 8.3 failures per turbine per year. That includes 6.2 minor repair (costs below 1000 EUR), 1.1 major repair (10³–10⁴ EUR), and 0.43 major replacement and 0.7 failures where no cost data can be categorized. The blades are the fifth biggest contributor to overall failure, with 6.2% (after pitch and hydraulic system, auxialiary components, generator, and gearbox). Blades show 0.456 minor repairs, 0.010 major, and 0.001 major replacement per turbine per year. The minor repair can mean protection tapes or shields,¹⁴ filling and injection, while a major repair is typically a structural repair.¹¹

Also, Dao et al.¹⁵ showed that the failure rate per turbine per year is 4 times higher for offshore than for onshore WTs, for blades and hub. There are various data on annual failure rate of rotors and average downtimes pro failure: 0.11 annual failures with average downtime per failure of 3.2 days based on WMEP database¹⁶ (corresponds to an average downtime of 9 h per year per turbine and probability 0.10% for each turbine¹⁷), 0.23 annual failure rate with an average downtime per failure of 11.4 days based on LWK (corresponds to average downtime of 62 h per year per turbine and to a probability of 0.71% for each turbine).^{16, 17} In Stenberg and Holttinen¹⁸ (VTT database), downtime of a WT due to blade failure is shown to 7% of the total technical downtime, that is, 18 h per year per turbine and 0.21% probability for each turbine.¹⁷ Two percent of turbines per year require blade replacements, mainly because of lightning strikes.¹⁹ The failure rate is rather high in the second year and then again in the 15th and 16th operational years, with spike at year 15.¹⁷ Einarsson,²⁰ referring to Tavner and Branner and Ghadirian,^{16, 17} noted that 75% of failures cause only 5% of downtimes and 25% of failure cause 95% of down times.

A specific question is the estimation of bird impacts. With estimatedly 140 000–328 000 bird impacts in North America,²¹ and of the order of 100 thousands WTs in North America (57.000 WTs across the United States, ~50 wind power plans in Mexico, and ~300 farms in Canada),²² it means roughly one hit per year.

2.2 Costs per WT blade failure

Structural repair of a single wind blade can cost up to $30 000 and a new blade costs, on average, about $200 000.⁵ Preventive maintenance (PM) for one turbine per year costs around 10 000 €, depending on the competence of the technicians and local labor markets (plus the costs of system failures).²³

The estimated operational expenditure (OPEX) for typical offshore WT ranges from 60- … to 87.5-thousand pounds/MW/year⁶ (71 … 10.3 thousands EUR). In Einarsson and IRENA,^{20, 24} the OPEX was given as ranging 11 … 45 $/MWh (in Denmark 15 … 19, in Spain 28, and in Austria 40), with an average 28 $/MWh. Further, wind turbine reinvestments (refurbishment, major overhauls) cost 15 … 20% of the price of WT.²⁵

In the statistics presented in Westberg,²⁶ one scheduled man day costs 750 €, and unscheduled man day costs 1000 €. Costs for replacement of one blade are of the order of 200.000 €. For Finnish statistics based on failure of 92 WTs in Finland in 2000–2004, 70 blades were replaced, leading to largest cost factor in the replacement repair (more than gearbox, generator, and transformer). Blades, together with hub and gearbox, have the highest repair times, repair costs, and number of technicians required for repair, among all other components.¹³

An out-of-service turbine can cost $800–$1600 (USD) per day, with most repairs taking 1–3 days. If a crane is required to repair or replace a blade, the cost can run up to $350 000 per week. An average blade repair (offshore) can cost up to $30 000 (for onshore blades, it can be two times less) and a new blade costs, on average, about $200 000.⁵

In the analysis presented in Tretton et al.,²⁷ component costs of repair are presented. For 1.5- to 2.0-MW WTs, crane costs are $44 000, labor costs for structural repair $23.000, and for nonstructural repair $4.000, and average component costs are for structural repair $88.000 and for nonstructural repair $13.000 (the numbers are for 2010; however, they are well comparable with 30% higher costs reported above for 2019).

3.1 Contribution of the failure, downtime, and repair costs to the energy costs

The LCOE of a wind farm can be estimated as²⁸

(1)

where AEP—annual energy production, r—discount rate, and n—year. Maintenance costs include corrective maintenance (fixing the turbine after the failure occurs) and PM.¹⁵ OPEX is determined as follows:

(2)

where OPEX_fixed—fixed cost elements, not depending on failure and technologies, for example, rental, administration, insurance; C_CM—costs of random failures (unplanned maintenance); C_PM—cost related to regular, planned maintenance activities (inspection, monitoring, and control).¹⁵

As noted in Newman,⁶ unscheduled activity (either reactive or proactive) constitutes 65% of O&M costs for each of the WTs considered, which is extraordinary high, and would not be acceptable in other, more mature industries. Thus, the problem of unscheduled repair deserves a special attention. The annual costs of unscheduled maintenance/failures can be calculated as the cost of repair and downtimes, summarized over amount of repairs:

(3)

where c_D—costs of downtime per hour, T_D—duration of downtime, N_F—amount of failures per year, and C_rep1—costs of single repair. The single repair costs can be calculated:

(4)

where C_trans—costs of transportation, C_equip—equipment and materials, and C_labor—costs of field labor. C_labor = Mt_repc, where M—size of team, c—hourly costs, and t_rep—time of repair.

Thus, the costs of unscheduled maintenance activities, repair of failures, can be estimated as follows:

(5)

The main parameters related to repair technology are “input parameters”—duration of work t_w, size of team M, equipment and material costs C_equip, and the “output parameter” amount of repairs N_F (or inversely, time between repairs, 1/N_F).

3.2 Technological cost reduction possibilities

The composite repair includes the following steps¹²: getting information about damage (e.g., from regular monitoring or control), transportation of repair team, identification of damage, design the repair scheme, fabricating repairing parts (patch), removing damage, grinding, surface preparation, attaching repairing parts, curing, post-repair inspection, documenting the repair, and monitoring repair zone.

The formulas above allow the evaluation of saving potential of different technologies. For instance, using ultraviolet (UV)-cured adhesives has some saving potential. Taking the example of seven layers of laminate 120 × 50 cm, one can estimate the saving potential as follows: with time distribution for the steps of repair sanding/surface preparation 120 min, surface preparation 30 min, lamination 45 min, heating with heating blanket for 4 h, surface sanding, painting (5 h), the duration of work t_rep ~ 12 h 45 min. With the size of team three persons (major repair)¹³ and the average hourly costs per technician 70 €,²⁷ the costs for such a repair will be at the level of 3500 €. By replacing the heating curing by UV curing (i.e., 20 min instead of 4 h) would allow to reduce the costs by ~9 … 10 times. It should be noted that the UV curing repair requires also additional equipment, typically UV gun, which requires additional investment.

The necessary duration of repair has an effect also on the so called “repair window,” that is, season where a repair (for instance, offshore) can be carried out. The repair of WT blades is possible only during relatively low wind, no storm time. Consider first the rain even as stopping event for blade repair. The time to next rain/storm can be considered as a random value distributed by exponential function²⁹:f(t) = 1 − exp(−γt), where γ—season-dependent rain frequency. The probability that the time between two consecutive rain or storm events exceeds the required time for the repair is therefore

(6)

where t_rep—time for repair (which was estimated above around 12 h, without transportation and waiting time). Assuming that the repair team decides to go for repair, if this probability exceeds 0.5, one can estimate the critical rain/storm frequency allowing for repair:

(7)

In other words, the repair is possible if the expected time between rains/storms is 44% more than the necessary repair time (including all the necessary steps). Assuming that the use of quickly curing resins allows shortening of necessary repair time from 12 to 7 h, the requirements toward rain frequency are much less. Another important aspect is the influence of the ambient temperature on the repair conditions. Apparently, variations of temperature have stronger influence on the quick 20-mincuring than on longer curing (however, the thermal curing is realized under better thermal isolation).

5.1 Internal defects/voids as triggering factor for destruction of adhesives and coating elements

The X-ray tomography analysis of the WT blade erosion and scanning electron microscopy (SEM) analyzes presented in Mishnaevsky et al.,³² demonstrated that the surface damage of blade coatings is triggered by the local defects (see Figure 2). It was observed in SEM investigations^32-35 that the initial surface cracks form from the heterogeneities (voids, bubbles, or reinforcing particles) in the coatings and not in the sites of highest stress as estimated for homogenized contact models. Similarly, as noted in previous studies,^{33, 34, 36} defects and voids in repair adhesives trigger post-repair degradation of WT blade (see Figure 3). Katnam et al.³⁷ observed that “micro-voids are frequently seen in adhesive bonds cured at elevated temperatures.” Void formation is connected to air entrapment during mixing or application, vaporization of the moisture, problems of substrate surface preparation, and so on. Using X-ray microtomography scans, Katnam and colleagues analyzed the microvoid content on dispenser-mix and hand-mix specimens and concluded that the strength depends on the mixing technique and microvoids.

Thus, the working mechanism of all the “accumulated” failure of WT blades (surface erosion, delamination, and matrix cracking) is assumed to be that interaction of external cyclic loading (rain impact on surface, cyclic loading on interfaces) and internal, manufacturing defects. The failure of WT blades is triggered by the defects available in the material. This is apparently not the case for force majeure impacts, like bird impact or transportation damage; however, this mechanism is assumed to work for most other “routine” damage.

5.2 Estimation of effect of defects in adhesives on the post-repair lifetime of blades

In order to estimate the effect of the voids on the fatigue behavior of adhesives, let us consider a unit volume of polymer with voids, located in the bonding layer between the patch and main structure (Figure 4). The material is subject to repeated mechanical load (due to the wind blade rotation and deformation at bending). Due to repeated cyclic loading, irreversible deformations are accumulated in the polymer chains, and at some Nth cycle, a crack starts from the void. The void in this case acts as notch, creating stress concentration.

The damage accumulation due to cyclic loading can be described using the concept of continuum damage mechanics.³⁸ Two damage parameters are introduced: volume damage D_v (void density in adhesive layer) and D_s (interface defects density). Voids are formed in the adhesive layers due to imperfect curing and suboptimal curing regimes.^{37, 39, 40} Thus, the value D_v in this context characterizes the quality of curing regimes. The interface defects are formed due to bad adhesion, and thus, D_s characterizes the adhesive properties of the glue. Assuming first D_s = 0 (ideal adhesion between the adhesive and composite), the initial volume damage of the voided volume can be defined as D_v = s/lw, where S ~ Σd²—sum of surface areas of the voids in the section, d—diameter of each void, lw—thickness of the adhesive layer multiplied by the width w. In more general case, the damage value D is determined as tensor from all the available defects.

Following the continuum damage mechanics approach, the local stress is defined as σ₀/(1 − D).^{38, 41} The Young modulus E of the considered material is reduced under cyclic loading due to accumulated irreversible deformations and microscale defect accumulation. If the S–N curve of the polymer material is given as N = aσ^−m, where N—lifetime (amount of cycles to failure) a and m—parameters of the material, then the lifetime of the layer with given voids is

(8)

Near a spherical void, the stress reaches 3 times of remote applied stress. Assuming that the damage in voided material starts not in arbitrary place, but in the vicinity of void, in the point of highest stress, the lifetime of a layer with void is reduced by

(9)

The thickness of adhesive can be of the order up to 30 mm for the blade of 70 m and above.^{42, 43} Assuming m = 0.5, one can estimate that the post-repair lifetime (time until next repair) decreases by 7%, if the total section area of voids is increased from 40 to 4 mm and by 25% if m = 2.

For instance, in Bourchak and Aid,⁴⁴ the values for high density polyethylene were determined as follows: σ = AN^b, A = 33.5, b = −0.0728. It gives m = 1/b = 13.7. In Wang et al.,⁴⁵ fatigue behavior of plasticized epoxy resin was studied. The S–N curve was determined as N = 1.857e17σ^−10.752. Thus, the sensitivity of repaired structures to the voids/defects in adhesives is rather high: with increasing the total void section area from 40 to 400 μm, the lifetime is reduced by 13% for epoxy and by 15% for high density polyethylene.