Recent developments in the protection of wind turbine blades against leading edge erosion: Materials solutions and predictive modelling

Leading edge erosion of wind turbine blades is the most often observed damage mechanism of wind turbine blades, which causes also additional costs for the maintenance of wind turbines. In this review, recent investigations in the areas of leading edge erosion of blades, anti-erosion coatings, new materials and computa- tional modelling of erosion are discussed. The ideas and results, presented at the annual symposia on erosion of wind turbine blades, organized at DTU Wind since 2020, are reviewed. Recent studies of leading edge erosion, devoted to the computational analysis and materials science aspects of the erosion, are summarized. The application of advanced computational modelling techniques to the analysis of the damage mechanisms in leading edge coatings is demonstrated, including the effect of coating materials properties and structure on the erosion, debonding, humidity and weathering effects, and the analysis of the potential of structured and rein- forced coatings.


Introduction
In order to realize the energy transition to renewables, large expansion of wind energy, in particular, offshore wind energy, is foreseen in next years [1]. The tendencies in the wind energy development include the development of large and extra large wind turbines, located far from coast (e.g., floating wind turbines) and also expansion into new regions (Northern Regions, Northern Sea, but also monsoon regions and other regions with severe weather conditions). The common feature of these tendencies is that it makes the wind turbine maintenance more difficult and expensive, on the one side, and more often required, due to high mechanical and environmental loads (stronger wind on sea, icing, monsoon rains, higher rotation speed).
Leading edge erosion of wind turbine blades is the most often observed damage mechanisms of wind turbine blades [2]. According to the evaluations of frequency of different damage mechanisms, carried out by DTU and Indian National Institute of Wind Energy (on the basis of the analysis of surveys of service teams and wind owners), the surface erosion is the only damage mechanism observed during the first year after the installation, and most critical damage mechanism in the first 5 years of service time of wind turbines in Europe (along with lightning strikes and manufacturing defects). It is also the most expensive degradation mechanism, simply due to the frequency of required repairs [3]. According to the estimations from Ref. [3], minor surface damage costs up to 12 times as much as major structural damage. Fig. 1 shows the schema how the erosion of wind turbine blades occurs due to rain droplet impacts. To overcome the erosion problem, a number of research projects and industrial initiatives were started over last years. Many coatings, leading edge protection systems and shells are available on the market now, including ProBlade Collision Barrier by LM Wind Power, PowerEdge® Care Leading edge protection of Siemens, coatings and tapes by 3M, Bergolin, Duromar, Enercon, Belzona, ELLE (Ever Lasting Leading Edge) soft shell by Poly Tech, Hempablade Edge 171 by Hempel and others. Still, the problem remains actual, especially with view on larger blades, with higher tip speeds and located in the regions of intensive rain and hail loading.
In this paper, recent investigations in the area of wind turbine blade protection against leading edge erosion are reviewed, in particular, developments of new materials and predictive modelling of erosion. The results presented at recent conferences, in particular, the symposia on leading edge erosion of wind turbine blades, organized at the Department of Wind Energy, Technical University of Denmark, in 2020-2022, are analysed [4]. Further, recent investigations at the DTU Wind and IIT Delhi in the area of new coating development, computational design of coatings, and prediction of the erosion and surface degradation are summarized. The promising directions and recommendations to the development of anti-erosion coatings and predictive modelling of the blade erosion are formulated.

Recent investigations in the area of materials for erosion protection: overview
Leading edge erosion of wind turbine blades has been a subject of investigations over several decades. The detailed reviews of investigations of blade erosion and solutions is given elsewhere (see for instance Refs. [5][6][7]). In order to get better overview of lately developed solutions, DTU Wind organized a series of international symposia on leading edge erosion of wind turbine blades in 2020-2023, inviting specialists from research teams and projects active in this area. This section reviews the results, presented at these symposia, as well as some other articles published over last years.

Thermoplastic and hybrid solutions
Several new materials solutions for anti-erosion coatings were under investigated over last years. A promising direction of the development of new highly resistant coatings is based on using thermoplastic and hybrid thermoplastic based materials [8,9]. British company Armour Edge developed semi-flexible erosion shields, from ultra-tough thermoplastics and with optimal aerodynamic profile. German company SaertexGmbH, in the framework of a collaborative project HyRoS, together with the company Kraiburg and other partners, developed a new thermoplastic-based hybrid material for leading edge protection (LEP) [10]. Further, they developed a coating with embedded heating layer for blade deicing. Zanjani and colleagues [11,12] developed integrated thermoplastic-thermoset hybrid leading edge protection system, based on the co-bonding process (joining of prefabricated parts during the curing process). Thermoplastics show more creep and stress relaxation, as compared to thermosets (due to the restriction of chain motions by cross-linking), and that might be one of reasons for the high potential of thermoplastics for LEP.

Viscoelastic polyurethane coatings
The common assumption is that the viscoelastic properties and damping of coating materials correlate with their erosion protection performance [13][14][15]. While this has not been confirmed in general case, this assumption is widely investigated. Katsivalis and colleagues [16] studied the relationship between exposure time before failure and the storage modulus (characterizing stiffness) and tanδ (ratio of loss modulus to storage modulus of the material, characterizing the energy dissipation potential) for the LEP materials considered. They observed that the correlation between storage modulus and time to failure becomes more significant at higher frequencies. Lower values of storage modulus lead to higher durability. For lower frequencies (100 Hz), time to failure decreases with increasing tanδ, while for higher frequencies (10 6 … 10 9 Hz)time to failure increases with increasing tanδ. Ouachan and colleagues [17] characterised samples of LEP material using the Dynamic Mechanical Thermal Analysis (DMTA) and Time Temperature Superposition methods, and observed that Young modulus of the coating increases with increasing the strain rate.

Structured, reinforced coatings
The particles and other reinforcements in the coatings can cause the dissipation of stress waves from liquid impact, thus, reducing local stresses [13]. It can also increase the strength of the coatings [13]. Ouachan and colleagues [18] added POSS (glycidyl polyhedral oligomeric silsesquioxane) nanoparticles to the polyurethane coatings, and demonstrated that the modification improves damping (at lower temperature). They observed that the particle modified LEP displayed lower tanδ values at all frequencies, as compared to unmodified. The elasticity of the coating increases with frequency of deformation. Pathak and colleagues [19] tested anti-erosion polyurethane coatings with ceramic oxide nanoparticle reinforcements, and observed that nanoparticles reinforced PU coatings show much higher solid particle erosion resistance as compared with pure polyurethane coating.

Structured thick interfaces between LEP and composites
The detachment of coatings is one of the main mechanisms of the blade erosion. In several works, the problem of interface adhesion is tackled by developing gradient, structured interfaces. Cortés and colleagues [20] studied in-mould coatings, and observed that coatings with stiffer and thicker coating-laminate interphase (in coatings with lower degree of curing) lead to better erosion protection performance, than coatings with higher degree of curing and thinner interphases. Erartsin, Salomão and colleagues [21,22] studied the correlation between the cure temperature and interphase thickness and morphology in thermoplastic elastomers, co-bonded to glass fiber reinforced epoxy and polyester composites (integrated leading edge protection, InLEP). They observed that increased cure temperature led to a decrease in the interphase thickness, decrease in the bond strength and more intensive interfacial failure. In the co-bonding process, proposed by Zanjani and colleagues [11,12] for the attachment of pre-fabricated thermoplastic (acrylonitrile butadiene styrene, ABS) coatings to thermoset (polyester) blades, interphase layer is formed, with lower microhardness. The properties of these interphases determine the quality of the LEP attachment. GalvanoPro developed hybrid metallic-polymer LEP with metalized fibrous hybrid metalized carbon layer as a coupling layer which reduces the interfacial stresses and prevents detachment. Pahlevan, from Danish SME Leptek [23], reported the novel adhesion system  for leading-edge erosion shields, using a nanoscale adhesion mechanism, based on polymer brushes technology. He reported 3 times higher adhesion strength as compared to commercially available systems. The conclusion can be drawn that the creation of thick, structured coating/composite interfaces/interphase has the potential to prevent the coating detachment.

Metallic and metal-polymer coatings
One of the direction which seems quite promising is the development of erosion shields from metallic materials, nickel or titanium shells [24,25]. The German company Muehlhan has got a patent on metallic coatings [26]. The British company Doncasters proposed to use electroforming nickel cobalt leading edge erosion shields, tested for aerospace applications, helicopters and propeller blades, for wind turbine blade protection. German company GalvanoPro also developed nickel-cobalt leading edge protection with hybrid material as coupling and transition layer. The challenges of this direction include the reliable attachment of metallic coating, fatigue resistance of the coatings, lightning protection, and also the (currently) relatively high price.

Roughness and its effect on the energy production
The critical question asked by many wind park owners and service companiesat which erosion level the energy production is so reduced, that an expensive repair operation is economically justified? There are various estimations on how the erosion reduces the annual energy production, ranging from 1.5% to 7% and more [27]. In a number of studies, the effect of leading edge roughness on the blade aerodynamics was studied [27][28][29], using Computational Fluid Dynamics (CFD) and experimental methods. It can be seen that while many studies are devoted either to the evaluation of the effect of surface roughness on the energy generation in blades, or analysis of blade surface degradation, with prediction of lifetime, there are practically no predictive models of roughness evolution of the blade surface. Kaore and colleagues [30] studied the effect of the tip velocity, surface roughness and the rain intensities on the erosion. They observed that the surface roughness of the coating is the most critical parameter in the rain erosion of coatings.

Moisture, weathering, ultraviolet
The wind turbine blade materials are subject to environmental loading including ultraviolet radiation and high humidity, which can lead to moisture ingress and absorption, swelling, changing properties, In combination with mechanical loading (by rain droplet, wind, etc), the environmental loading speeds up the leading edge degradation. It is of interest that there were almost no presentations on the effect of humidity on the blade erosion at the 1st -3rd symposia on leading edge erosion in 2020-2022 (and practically no publications). However, at the 4th symposium (February 7-9, 2023), several leading groups in this area (including University of Strathclyde/UK and Universidad CEU/SP), report their studies on combined erosion and weathering studies.
In other studies, Gao and colleagues [31] observed that UV-radiation causes photo-oxidation of coatings, free radical excitation, coupling reaction, and can lead to complicated degradation products. Yang et al. [] showed that polyurethane coatings under combined UV, water and oxygen loading, tend to form blisters on the surface, then their chemical composition of the coatings changes (increase of urea and urethane group concentrations; build-up of hydrophilic groups), promoting the water absorption [32]. Godfrey and colleagues [33] studied the effect of cold temperature on solid erosion of polyurethane coatings. They observed that cold temperatures increase significantly the erosion rate leading to plastic erosion behaviour at cold temperatures (and not elastic behavior at ambient temperatures).
This overview of the main activities in the area of new materials for leading edge protection shows the most some interesting and important directions in this area. In the following sections, the application of computational modelling to solving and better understanding of these and other problems of blade erosion is demonstrated. Leading edge erosion mechanisms, materials properties effect and the potential of various materials for erosion coatings are investigated using advanced computational modelling.

Leading edge erosion mechanisms: Computational analysis
In this and next sections, the activity of DTU Wind in the area of predictive modelling of erosion, and the analysis of interrelationships between service conditions, materials structure and anti-erosion performance are presented. The investigations concentrate on several topics shortly discussed above: o Effect of coating materials properties and structure on the erosion, including viscoelastic properties, oDamage mechanisms in leading edge coatings, including debonding, humidity and weathering effects, oЕxploring the potential of structured and reinforced coatings, oMethods of computational analysis and digitization of wind turbine blade degradation.
Using advanced computational modelling methods of the erosion simulation (multiscale and micromechanical modelling, application of machine learning, exploring the potential of digital twins), we seek to analyze he influence of various factors and material parameters on the anti-erosion performance of coatings.

Erosion damage modelling for viscoelastic coatings
The general strategy of computational modelling of leading edge erosion includes several steps: development of random rain scenario, droplet impact modelling, fatigue damage modelling [5]. The main steps of all the computational models of erosion are shown in Fig. 2. The approach is described in several reviews [5,34,35], and implemented in several models [36][37][38].
In order to analyze the potential of the anti-erosion protection improvement by modifying structure and properties of the coating material, computational models of the erosion should include also the complex polymer material behaviour, micro-and nanoscale structures of the coating materials.
In several works, complex viscoelastic laws were implemented in the erosion/liquid impact models, to analyze the effect of material properties In [39], a micromechanical model of viscoelastic polyurethane as a two phase (soft/hard) material is presented. It was demonstrated that the larger mechanical loss coefficient of the material, the less is the stored elastic energy in the material after the impact, and the lower is the likelihood of the damage initiation.
In [36,38], the effect of viscoelasticity of the coating material on the stress field in the coating under liquid impact was studied in numerical experiments. A computational finite element model of liquid drop impact was implemented using the Smoothed Particle Hydrodynamics (SPH) method. Fig. 3 shows a comparison of stress fields in hyper-viscoelastic and elastic coatings: maximum Mises stress plotted versus time (a) and Mises stress distributions (b). In the case of purely elastic coatings, deformation and strain levels are lower and the high stress area is localized near the impact contact area. In the case of viscoelastic coatings, the high stress region is beneath the surface. Thus, the viscoelastic behavior of coatings changes the stress distribution, and the damage mechanisms of coatings.
Comparing these studies, one can observe that viscoelastic coatings have several mechanisms on influencing the erosion damage, including the impact energy dissipation due to internal deformation and changing damage mechanism (highest stress area not on the contact surface but in depth below the droplet impact).

Surface damage and roughness evolution
The damage formation at the blade surface determines the roughening of the blade during the service time.
In order to predict the roughness evolution of wind turbine blade surface during the erosion, a damage evolution model for the leading edge surface was developed [40]. Damage in each given point is calculated, using S-N curve data for the coating material. The damage caused by multiple droplet impacts is calculated at discretization points based on a linear damage accumulation method, and increased after every impact incident.
In the areas where the damage exceeded some given value (e.g., 1) the material was removed, thus, creating roughness on the surface due to the initial damage heterogeneity. In order to simulate the effect of roughness on the further material degradation, the parameter of stress concentration on formed dimples is determined, which depends on the depth of the dimple, and increases the likelihood of the damage formation near the dimple.
Further, simulations of roughness evolution for hyperelastic and viscoelastic coatings was carried out. Fig. 5 shows comparison of roughening of blades with hyperelastic and viscoelastic coatings. It can be seen that he roughnening begins much earlier in elastic coatings, and also the roughness is much larger.
For the validation of this model, roughness measurements of the blade samples eroded in RET (rain erosion tester) were carried out. The comparison of simulated and experimental results showed reasonable agreement.
In the computational studies, the effect of impact velocity on the roughness evolution was investigated. It was observed that increasing the impact velocity leads to the increased roughness of eroded blade, as expected. Fig. 4 shows the calculated roughness as a function of the velocity. This dependence is in fact exponential: with increasing the velocity from 120 m/s to 160 m/s, the roughness height can be increased by 3 or more times. This is an important conclusion: it means that with increasing the wind turbine size, the roughening of blade surface becomes drastically worse.

Fatigue debonding of multilayered coating
Another damage mechanism of leading edge coatings is the debonding from the substrate composites. Stress waves due to repeated liquid impacts, reflected from interfaces, can lead to fatigue damage of interfaces, leading to the debonding. In order to simulate the fatigue caused debonding of coatings, a complex of programs for finite element  code Abaqus was developed [41]. The program complex included a cohesive zone model for fatigue delamination rate simulation, a program for multiple impact loading modelling (Abaqus VLOAD subroutine), and FE model of multilayered coating with cohesive contact between layers. Fig. 6 shows stress field along the cross section of the multilayered coating. Fig. 7 shows the surface stress distribution due to multiple liquid impacts, obtained using the developed Abaqus subroutine VDLOAD.
This approach is based on the interface degradation model, in which fatigue damage in the cohesive law envelope is accumulated according to a simple two-parameter law [42]. The constitutive fatigue damage model was implemented as a user-written subroutine UMAT for ABA-QUS using Turon's mixed-mode cohesive law formulation [43]. The proposed cohesive fatigue model can predict interfacial failure in multilayered coatings attached to blade composites. Fig. 8 shows examples of the results, obtained with the model, damage initiation on interfaces between coating and putty, and putty and composite. The coating layer of thickness 100μ was modelled on the top of the putty layer of thickness 200μ, and an adhesive layer of thickness 10μ is placed between the coating and putty with tie constraints to both layers. Further, both layers are modelled on the 2 mm composite substrate, with an adhesive layer placed between the putty and composite with tie constraints. Penetration among the layers is prevented by defining the contact between the interfaces of the layers.   Three cases are analysed: elastic, viscoelastic, and hyperelastic combined with viscoelastic coating properties. The multilayer coating model is analysed for cohesive fatigue delamination rate subjected to repetitive droplet impact loading at arbitrary locations. An accelerated analysis is performed by providing the smaller damage threshold value to predict the results in less time. The delamination rate results are compared for both the cases mentioned above. It is observed that the delamination rate for the coating with elastic and viscoelastic properties is faster than that with hyperelastic combined with viscoelastic properties. It is of interest that the likelihood of debonding between the putty and  composite seem to be higher than between putty and coatings.
The hyper-viscoelastic coating has been observed to provide excellent damping performance by reducing the shockwave behaviour during raindrop impact events. Therefore, impact stresses are lower, which causes less damage in the adhesive layer (interface region) between coating and putty. Whereas in the case of elastic coatings, poor damping causes higher stresses in the coating, leading to more damage in the adhesive layer. Furthermore, the adhesive layer between the putty and coating was damaged more than the adhesive layer between the coating and putty. It is learnt that the stiff composite substrate reflects the stress waves produced due to the generation of shockwave behaviour during impact. The reflected stress waves were approximate twice the initial magnitude, causing more deformation and damage to the adhesive layer between putty and composite, as shown in Fig. 8(b). Additionally, the moisture ingress in the coating system causes hygroscopic swelling, and residual stresses cause faster degradation of the adhesive layer.

Weathering and stresses due to ingress of moisture
Wind turbine staying offshore are subject to high environmental loading, e.g. high humidity, additionally to all other loads. The combination of mechanical loads, high humidity and temperature variations can lead to residual strain in coatings, hygroscopic swelling in polymers, and ultimately to higher local strain concentration, and debonding/ delamination of the coating. The moisture ingress in polymers promotes the relaxation of the polymeric chain, consequently degrading the mesoscale modulus of the material [44,45].
In order to investigate the moisture diffusion in the multilayered coating system and stress evolution, a computational finite element model was developed (see Fig. 9). A coupled moisture-displacement problem was solved by adopting a thermal-moisture direct analogy in ABAQUS/Standard [46,47]. As shown in Fig. 9, saturation weight gain by the coating system at 90% relative humidity and 60 • C is approximately 2.5% of the initial weight. Fig. 10(a) shows moisture diffusion in the coating layers at 600 h.
The hygroscopic stress field developed due to moisture diffusion in the coating sample is shown in Fig. 10(a). It can be seen that stress variation is high between the layers' interfaces. The stress changes between the polyurethane and putty layers cause additional shear stresses along interfaces caused by the moisture ingress. This create additional mechanism, leading to coating debonding.
It can be seen that the humidity has potentially a strong effect on the stress distribution in the coating, causing earlier degradation, especially in multilayered coatings.
In this section, computational models of various mechanisms of leading edge erosion of wind tyrbine blades are presented. Some of the models are still in development, or validated, but they allow reproducing all the acting degradation mechanisms, and allow the prediction or erosion, and analysis of the factors influencing the erosion process.
Among others, some interesting observations were made: • Surface roughening of blade surface due to erosion (the main cause of energy generation reduction) increases drastically with increasing the rotation speed (example:it can increase by 3 times when the speed increases by 30%). It means that the erosion problem can become even more critical for extra-large wind turbines. • The roughness is however sufficiently lower for viscoelastic, coatings than for elastic. Viscoelastic coatings have also different damage mechanism, from elastic coatings, with highest stress area not on the contact surface but in depth below the droplet impact. These conclusions echo the observations about high potential of highly damping, thermoplastic or soft polyurethane phase containing coatings, discussed in section 2. • High humidity and moisture ingress can lead to additional stresses in the coatings, thus, reducing the coating lifetime.

Structure of coatings and its effect on the erosion
Apart from viscoelastic and highly damping coatings, structured, engineered, reinforced coatings represent a promising direction for new generation of blade protections.
In this section, some previous studies on the influence of structures and properties of coatings are summarized.
In the computational studies in Ref. [48], the effect of multilayer coatings on the stress distribution in the coatings was studied. Various bilayer coating structures were tested in computational simulations, among them: stiff upper coating/soft lower coating ("up-stiff"), soft  upper coating/stiff lower coating ("up-soft") and homogeneous coatings. Fig. 11 shows the schema of the model and maximum von Mises stress plotted versus the time for each of three models. It was shown that the highest stresses are observed for the stiff upper coating case and inversely, lowest stresses are observed for soft upper coatings.
In [36], it was also demonstrated that we surface can drastically improve the erosion resistance. Even thinnest liquid film on the blade surface can reduce the maximum local stress by about 30%, thus, extending the coating lifetime. Fig. 12 shows droplet splash after impact on a dry surface, and on a wet surface and maximum Mises stress depending on the time, for dry and wet surface. While reinforcing particles in coatings tend to strengthen coatings,  and potentially dissipate and scatter the impact energy [13], it was observed both in computational and experimental studies in Ref. [22], that the weakly bonded, weakly adhered particles can trigger the erosion and damage in coatings. Thus, good attachment of reinforcing particles in the anti-erosion coatings is extremely important for the coating quality.
The very important factor influencing the performance of antierosion coatings is the availability of defects, voids, microcracks. According to Refs. [36,48], the erosion degradation of coatings starts from voids/defects. The availability of voids reduces the coating lifetime drastically.

Multiscale modelling of structured nanoreinforced anti-erosion coatings
In [13], it was demonstrated that stress wave scattering on reinforcements, inclusions embedded in the coatings causes energy dissipation after droplet impact, thus, diverting the energy from the material damage and new surface formation. In this section, the recent investigations in the area of reinforced (fiber or particle) coatings are reviewed.
In order to analyze the effect of nanoscale coating material structure on the coating damage resistance, a multiscale micromechanical model of structured coating degradation was developed [49]. The model is based on the submodelling techniques and unit cell approach (i.e., a representative volume with inclusions, which reflects the material behavior at nanoscale).
Unit cells with disc shaped randomly oriented nanoparticles, long "snake shaped" or zigzagged fibers and/or round particle or voids were generated, and tested in the multiscale model subject to multiple liquid impact. Figs. 13 and 14 show schemes of the multiscale models, with the upper (macro)scale model of liquid droplet impact, and submodel (unit cell model of material) with internal structure. Several specific cases were studied: polyurethane reinforced by Kevlar fibers pulp; polyurethane reinforced by graphene nanoparticles; bio-based polymer reinforced by nanocellulose. Also, the porosity of the material was taken into account.
The main observation was that fiber pulp reinforcement leads to lower stress concentration in the vicinity of voids. The local stress was reduced by 10% … 15%, with 4% content of Kevlar fibers, for instance. Given that the erosion damage is initiated near the voids in the coatings, the reduced stresses near the voids lead to the better anti-erosion performance of the coatings. Therefore, the availability of fibers in the coatings allows shielding the voids. It was estimated that the reduction of stress on voids by 10% can lead to 80% increase in lifetime, while a reduction of stress by 15%, leads even more than 350 times higher lifetime. Thus, the fiber pulp reinforced coatings have a great potential for the anti-erosion protection.
Further studies were directed to the potential of particle reinforcement, in particular, graphene nanosheets. Unit cell computational models of polymer reinforced with graphene sheets were developed for various orientations, density and sizes of graphene nanosheets (Fig. 14). Computational simulations of stress wave propagation and damage evolution in graphene reinforced polymers were carried out. Fig. 14 shows the schema of multiscale model of droplet impact on structured graphene reinforced and porous coating, and finite element model of the submodel unit cell. Several cases were investigated, among them, the cases of dilute and concentrated distributions of randomly arranged and oriented graphene nanosheets in polymers. In dilute case, the average  distance between nanoplatelets (5 μm diameter) was 16 μm, while in concentrated case, the distance was 6 μm. It was observed that the dilute graphene reinforcements cause a negligible shielding of stresses on voids, while high density nanoparticle reinforcement leads to the strong reduction of stresses on voids, 40-50% reduction in the stress than the value without graphene reinforcement.

Bio-based sustainable coatings
The potential of fiber and particle reinforcements in coatings to improve the erosion protection performance opens also a path to the development of sustainable, bio-based anti-erosion coatings. Usually, bio-based polymers show strength and stiffness much lower than those of polymers used anti-erosion coatings. In Ref. [49], it was suggested to develop new coatings using cellulose fiber pulp reinforcerment in the bio-based polymers, which increases the damage resistance and strength of coatings. Nanocellulose fibers show chemical inertness, high strength, and outstanding stiffness, low density, low coefficient of thermal expansion, dimensional stability. Considering the above properties, these can be used as a reinforcing material in the anti-erosion coatings of wind turbine blades.
In order to explore the potential of using nanocellulose reinforcement for shielding stresses on voids (as observed above), computational studies of polymer with nanocellulose reinforcement under rain droplet impact were carried out, using the approach similar to that described in section 4.2. A unit cell model of nanocellulose fiber-reinforced coating with a single void is developed, and used as a submodel in the rain drop impact simulations. Since nanocellulose fibrils are much thinner than   [28], with kind permission from Elsevier). Kevlar fibers considered above, the fibrils of radius 6 nm and length of 1800 nm were modelled here as 3D beam elements. Fig. 15 shows a finite element model of unit cell with a number of nanocellulose fibrils, and a micrograph of real nanocelluose fibrils. It was, again, observed that in the case of dilute reinforcement, no stress reduction near voids due to the reinforcement is observe, while at high content of nanocellulose reinforcement, reduction of local stress by 10% was observed. The effect is similar to that observed in section 4.2, namely, nano reinforcements do reduce the stresses on voids, but only if their content is relatively high, and not in dilute distribution. The observation of the potential of nanocellulose to delay the degradation of coatings opens the way to creating protective, bio-based and sustainable anti-erosion coatings for wind turbine blades. In [50], the experimental validation for the development of nanocellulose reinforced coatings for wind turbine blade anti-erosion protection is presented. It is demonstrated that the cellulose reinforcements allow the improvement of the coating performance up to 70%.

New graphene and hybrid nanoparticle reinforced coatings
In order to explore the high potential of nanoengineered, particle reinforced coatings, samples of new coatings were manufactured, with polyurethane (PU) matrix, incorporating functionalised graphene nanoplatelets (f-GNP) and silica-based sol-gel [51,52]. Mechanical tests of various compositions of the coatings (Young's modulus, elongation at break, tensile strength and modulus of toughness) were carried out. The novel nanoreinforced polyurethane coatings with graphene and hybrid (graphene/silica) reinforcement have been further tested using a Single Point Impact Fatigue Tester (SPIFT) to evaluate their potential erosion resistance. Scanning electron microscopy has been used for analysis of damage after SPIFT testing. It has been demonstrated that the nanoreinforced coatings have significantly greater resistance to erosion [52]. Polyurethane with hybrid GNP + SG nanoparticle reinforced coatings exhibited lifetimes up to 13 time greater than pure polyurethane coatings. Fig. 16 shows the SPIFT tests results, amount of impact hits up to first failure, for different coatings. This study proved the high potential of nanoengineered coatings. The further upscaling and utilization of the nanoparticle engineered coatings for the practical exploitation depends on the practical feasibility of technologies of the coatings attachment and manufacturing.
In this section, the potential of structured, nanoreinforced coatings was investigated, using advanced computational models. Some interesting conclusions can be drawn. Coatings with fiber pulp reinforcement (for instance, Kevlar fiber pulp or nanocellulose fibrils), or high density nanoparticlereinforcement allow reducing the stresses on eventual defects in the coatings, thus, extending the coating life. This opens the path to the development of new, also bio-based, coatings.

Application of machine learning to the analysis of the blade erosion
Blade surface erosion is a complex process, controlled by multiple physical effects and interactions, and influenced by a number of stochastic processes (rain distribution, material defects, properties variations). Deterministic mechanical models, like presented in sections 3 and 4, have therefore some constraints when applied to rain erosion of real materials. Therefore, the artificial intelligence methods represent an attractive option for the modelling of erosion, overcoming the limitations of continuum models. Machine learning techniques can be used to complement the finite element models, and to provide predictive estimation of different parameters. This can be realized by using surrogate models with machine learning algorithms which are trained with data from finite element analyses [53].
In the following analysis, the application of machine learning methods to the analysis of erosion is demonstrated for the analysis of the effect of droplet size on the erosion process. After a range of expected droplet diameters is defined and finite element impact simulations are performed for the selected diameter values within that range, the machine learning algorithm is trained to fit the output of interest of these simulations, (here, stress values). To demonstrate this method, a twolayer neural network (NN) was used to fit the stress output of finite element models for impacts of droplets with diameters between 1 mm and 4 mm. The inputs of the NN were the droplet diameter and the position relative to the point of impact while the output was a maximum principal stress amplitude (Sa) value. The training and validation data were used to find the optimal number of units of each of the two layers by considering the mean square (mse) and average (mae) errors. The NN parameters obtained from the training dataset are used for predictions for the validation set. The predicted and simulated Sa fields are illustrated in Fig. 17. Graphs of Sa values versus the distance from the point of impact are shown in Fig. 18, at different depths below the impact surface. While there were some differences for the maximum stress values between the predictions and the simulation, the surrogate model was able to produce accurate results, even with a simple neural network model of 2 layers.
The purpose of this technique is to make fast predictions for stress values that will occur during the droplet impact of any diameter within the expected range of values. The predictions can then be used in coating lifetime estimation models and allow them to take into account the stochastic nature of droplet sizes. The benefit is that the surrogate model can provide similar results to the finite element simulations but in a much shorter time. Surrogate modelling techniques could also be used to fit the output of other numerical tools such as the coating lifetime prediction tool or the roughness prediction tool.
Another application of AI on the field of leading edge erosion is the damage detection via image processing. Images are frequently collected from blades of operating turbines by drones or humans and are processed by applying AI algorithms.

Digital twin of the eroding wind turbine blade
The erosion is microscale process, which can make it more difficult to link direct experimental observation of the process with the computational models [54]. In this section, the suggested approach to the development of digital twin of eroding blade is formulated. The idea is that the surface roughness can be automatically photographed, and the digitized micrographs are introduced in the computational model of the blade surface degradation as input roughness. Using the computational model of surface damage and roughness evolution (e.g, described in section 3) or machine learning approach (as demonstrated in section 5), one can estimate and predict the roughness evolution (for given coating properties and rain conditions). The roughness determined in the simulations is validated and corrected by comparison with new micrographs. The surrogate models, as described in section 5, allows predicting the output of the physics-based model within seconds. Fast computations enable us to quickly test different parameters and environmental conditions for their effect on blade erosion. It will allow to make fast decisions for the operation of turbines and schedule optimal repair intervals. Fig. 19 shows the schema of such design of the digital twin of eroding wind turbine blade. The model development is underway now.
For reproducing the real 3D microgeometry of the blade surface, photogrammetry approach is used [55]. This involves taking multiple images of an object which overlap each other and then an image stitching software is used to create a 3D reconstruction. During reconstruction, the software extracts common points between images and calculates the distance between points belonging to the object and the camera positions. A camera can be mounted on an inspection drone which will take images of the leading edge on a stationary blade. Fig. 20 shows the leading edge and the reconstructed 3D model. This 3D reconstruction was obtained, using a handheld smartphoneto capture 24 images of an eroded leading edge on a decommissioned blade, and commercial software 3DF Zephyr.

Conclusions
Recent developments in the wind turbine blade protection against leading edge erosion, are reviewed, on the basis of last year publications, works presented on the annual DTU symposia on leading edge erosion over last four years, as well as studies carried out at DTU Wind.
The most common directions of development of new blade coating materials were grouped into thermoplastic and hybrid thermoplastic coatings, highly viscoelasic coatings, structured interfaces to enhance the coating attachments (using hybrid fibrous layers, polymer brushes, or interphase layers), electroforming metallic leading edge erosion shields and structured nanoreinforced coatings. Other important topics are the roughness evolution on the blade surface, and effect of humidity and other environmental factors.
Further, the application of advanced computational modelling techniques to the analysis of the damage mechanisms in leading edge coatings, effect of coating materials properties and structure on the erosion, debonding, humidity and weathering effects, potential of structured and reinforced coatings.