Proposed holistic strategy for mechanical recycling of wind turbine blades for 3D printing and compression molding

There are many solutions being proposed for recycling discarded wind turbine blades, but most solutions are not very ecological, involving either burning or harsh chemicals. In this research, a solution is proposed that has a very low environmental impact and has potential of being scaled to a very large industrial process. The entire wind turbine blade should be recycled and transformed into other products that are recycled and that have economic benefits in the long term. Two of the output products from the current recycling strategy are recycled 3D printing filament and recycled reinforced pellets for compression molding. The research outlines the process, the numerical simulations that help optimize the output and the outline for planning to bring the recycling solution from a small industrial scale to a large scale. This represents ongoing research in this important field of composite recycling.


Turbine blade recycling
The extensive use of composites for wind turbine blades and the accelerating growth of the wind energy industry has led to significant quantities of material waste and undesirable environmental impact [1,2].Although most (around 85%) of wind turbines (like housing and tower parts) are made of recyclable materials like copper and steel [3], non-biodegradable materials such as fiberglass, plastic polymer, and core materials are used to make these high-performance wind turbine blades to enhance their efficiency [4].Turbine blades have a life span of approximately 25 years, and many of the initial generations of turbines are already reaching their end-of-life.Therefore, used wind turbine blades have become a significant source of composite waste generation [2].It is predicted that the annual amount of waste from end-of-life wind turbine blades will increase from over 50,000 tonnes in 2021 to over 200,000 tonnes worldwide by 2034 [5].

Predictive models
The use of predictive models such as numerical models, finite element models, etc., are necessary to facilitate the design process and material assessment in the recycling solution of wind turbine blades.Some of the numerical models are helpful because the methods are well-studied, and there are no restrictions on the composite's geometry, material properties, and fiber angles [16].As the current recycling solution proposed here uses discontinuous fiber recycled composites, the focus of this research is specifically on these types of materials.For discontinuous random fiber reinforcement composites, the material behaviour and performance are affected by several factors: fiber aspect ratio, fiber orientation, reinforcement shape, etc. [17][18][19].Moreover, other factors, such as voids and particles, reduce the accuracy of numerical models.In this sense, using finite element analysis (FEA) can provide good potential property prediction for stiffness and strength, but only if they include all pertinent aspects of the problem.
Explicit modelling of a 3D-printed composite component with all the features mentioned above can be computationally prohibitive.Enormous computational resources and impracticalities are required to obtain a convergent solution when trying to model fibers, particles, and voids in the resulting recycled material.To overcome this hurdle, mechanical analysis methods can be beneficial to obtain the effective mechanical properties of the material by using methods that involve a Representative Volume Element (RVE) [20].An effective RVE must be sufficiently large to include enough variety of reinforcements.In addition, the effective properties of the RVE must statistically represent the actual material properties on a macroscopic scale [21].At the same time, increasing the size of the RVE comes with a higher computational cost.Therefore, the size of RVE should be minimized for computational efficiency but at the same time must guarantee accuracy.

Life cycle analysis
To ensure the sustainable development of wind power, it is essential to better understand the potential life-cycle environmental impact of managing the waste generated during the manufacture, operation and end-of-life of wind turbines.The optimization of potential recycling site locations for blade waste is an integral part.Although this is the least developed aspect of the current research project, it is worth mentioning even at this stage because of its essential nature in comprehensively developing this large industrial recycling operation.

Application to 3D printing and compression molding
Most commercial 3D printing used PLA, ABS, or some similar thermoplastic material as the 3D printing feedstock.Although some 3D printing is done with continuous reinforcing filament, this technique remains a niche operation.The bulk of 3D printing users would greatly benefit from a 3D printing material that has enhanced reinforced properties as well as "predictable" properties.This is the motivation for the 3D printing aspect of this project.However, this entails producing 3D printing filament with an ideal reinforcing fiber average length to maximize the benefits.Fiber lengths that fall outside of the ideal range would be used to make other reinforced pellets that would be used in compression molding of parts.Industrial applications for compression molding would be targeted.In this way, all the fibers produced by the grinding operations would find their way into one of the two categories of manufacturing methods (3D printing and/or compression molding).

Objective
The aim of this research includes but is not limited to three phases: I. Develop an optimum process for recycling glass fibers from scrap wind turbine blades as reinforcements in 3D printing and compression molding based on low environmental impact of the process and high consistent improvements in the properties of the recycled materials.II.Establish novel algorithms and predictive models to facilitate the design process and material assessment based on Phase I. III.Propose life cycle assessments for the above recycling solutions (Phase I) and optimize a plan for recycling site locations in Canada.The distinct project phases are defined and summarized below.Previous work and relevant results can be found in [22][23][24][25][26][27].

I.
Develop an optimum process for recycling glass fibers from end-of-life wind turbine blades as reinforcements in 3D printing and compression molding.a) Raw material acquisition: Grind turbine blade pieces to obtain recyclate fibers with a cost-effective and efficient method.b) Sieving: Sieve the recyclate fibers to obtain fibers featuring a suitable length range for both the 3D printing process and compression molding.Optimize the sieving process to obtain the most fibers within a short sieving time.c) Pelletizing: Pelletize fibers for the use of 3D printing and compression molding.d) 3D printing: Develop and validate the optimum extrusion process parameters to produce filaments with a consistent diameter for use in 3D printing.l Evaluate the compatibility of the recycled glass fibers with the base polymer.l Eliminate air bubbles in the preform by dehydrating the polymer.l Optimize the extrusion process parameters to produce filaments with a tolerance of ±0.05 mm in diameter.l Conduct Differential Scanning Calorimetry (DSC) and rheology tests on the pure and reinforced thermoplastic filaments to identify the processing and operation window.l Compare the mechanical properties of the pure and reinforced thermoplastic filaments with the available commercial filament.l Optimize the design and manufacturing parameters of 3D printing.l Print tensile test specimens in compliance with ASTM standards to determine the tensile properties of the parts made of pure and reinforced thermoplastics.l Measure other important properties through bending tests, impact tests, fracture toughness tests, etc. e) Compression molding: Develop and validate the optimum parameters of the compression molding process to produce parts with enhanced mechanical properties.
l Evaluate the compatibility of the recycled glass fibers with the base polymer.l Eliminate air bubbles by dehydrating the polymer.l Conduct Differential Scanning Calorimetry (DSC) and rheology tests on the pure and reinforced thermoplastic preforms to identify the processing and operation window.l Compare the mechanical properties of the pure and reinforced thermoplastic parts with the available commercial parts.l Optimize the design and manufacturing parameters of compression molding.l Prepare tensile test specimens in compliance with ASTM D638 standards to determine the tensile properties of the parts made of pure and reinforced thermoplastics.l Measure other important properties through bending tests, impact tests and fracture toughness tests.
II. Establish novel algorithms and predictive models to facilitate the design process and material assessment of the above recycling solutions.a) Novel algorithms: Propose new modified random sequential adsorption (MRSA) methods for the efficient generation of the RVEs with hybrid and arbitrary-geometry reinforcements while satisfying geometric periodicities and no intersections among entities without tedious looping or Boolean calculations.b) Predictive models: Establish representative volume element (RVE) models and perform stress and strength analysis of recycled fibre composites fabricated by 3D printing and compression molding, making use of PYTHON, ABAQUS, Moldflow and PAM-RTM.c) Boundary conditions: Improve the PYTHON code from EasyPBC to enable the ease of enforcing periodic boundary conditions (PBCs) without periodic meshes.d) Structural characteristics: Fibers, resin particles, voids, etc. will be characterized by using Microscopy, Micro-CT, ImageJ and Avizo.The proposed methodology for the recycling of wind turbine blades can be found in Figure 1.A threestage grinding methodology will be applied to obtain the fibers for the filament extrusion.Of the three known methods for recovering fibers (grinding, pyrolysis, and chemical methods like solvolysis), grinding is the method that uses up the least energy by far and thus has the lowest environmental impact on the fiber-recovery process.This is according to a study by Hani et al. [28] published in 2022 and the resulting energy results are shown below in Figure 2. Figure 2. Energy consumption for three different methods for recovering fibers from composite materials [28].

III. Propose life cycle assessments for the above recycling solutions and optimizing a plan for recycling site locations in
IOP Publishing doi:10.1088/1757-899X/1293/1/0120046 a) Grinding: The scrap turbine blades will be first cut down into smaller pieces (20 cm !20 cm) and placed in a grinding machine (ECOWOLF, INC.) which contains a hammer mill system and a metal classifier with a hole diameter of 3 mm.b) Double sieving: A double-sieving mechanism will be demonstrated to classify the recycled materials (a mixture of fibers and resin powder) from the grinder, and two grades of granulated materials will be obtained.The granulated materials obtained from the initial grinding process will be sieved through a stainless-steel screen with a hole diameter of 0.1 mm.The larger-sized recycled material will be then re-fed into the sieve shaker for the second sieving operation to extract more fine fibers.Details on the exact range of ideal fiber sizes and the factors which determine the optimum results is proprietary information (patent pending) which cannot be revealed in the current research paper.c) Visual inspection: Microscopy will be carried out for the recycled materials after sieving to guarantee the fibers are in the desired length range before the filament extrusion process.
In the end, over 60% of the ground recycled material that comes out of the sieve should be usable for the filament extrusion process with a nozzle diameter of 0.7 mm.The remaining sieved residue could be either reground and reused or processed with larger nozzle diameters.
Fibers obtained from turbine blade waste and PLA pellets (Ingeo 4043D, Natureworks LLC, Blair, Nebraska) will be fed into a dehydrator machine for a drying process of 4 hours at 60 o C to reduce the moisture content of the pellets to below 250 ppm.The materials will be then placed in a twin-screw extruder (Leistritz ZSE18HP-40D, Nuremberg, Germany) with 8 subzones and a pelletizer to produce glass fiber-reinforced pellets.The pellets will be re-dried and fed into a single screw extruder (FilaFab, D3D Innovations Limited, Bristol, UK) to produce glass fiber-reinforced filaments with a diameter of 1.75 ± 0.05 mm.A laser micrometer with an accuracy of ±2 µm will be used to monitor the filament diameter along the extrusion.The single and twin-screw extruder parameters are summarized in Table 1.
Table 1.The single and twin-screw extruders parameters (recommended for the specific extruder that was used in the test program) Once filaments are prepared, glass fiber-reinforced specimens with a total thickness of 3.36 mm and a [0] !" 3D printing direction sequence will be prepared using a 3D printer Prusa i3 Mk2S.The manufacturing process and design parameters are summarized in Table 2.These are typical values only, that were used to get good printing results with the specific 3D printing equipment used in this project.Fibers within an undesired length range (longer fibers) will be used in the compression molding process.As stated before, it is essential to try to recycle all the fibers resulting from the grinding process.Those fibers that are not optimal for 3D printing filament will nevertheless still be useful for making compression molding pellets.Corresponding simulation work will be carried out via Moldflow and PAM-RTM.Plan details will be enriched as the project progresses.In the end, after careful sieving operations, around 60% of the generated recyclates are short fibers, which will be used for filament extrusion.About 30% are medium fibers which will be used for compression molding while the remaining 10% will be either disposed or used as concrete filler (see Figure 3).

Phase II
The generation of an RVE cuboid, containing fibers and other entities (inner voids and resin particles) is essential to FE analysis of 3D printed filament material and compression molding pellet material.
The RVE cubic volume will be considered first, which will be assigned a size obtained through a sensitivity analysis and comparison with experimental data.
The fibers are modelled as straight cylinders with circular cross-sections of constant radius.This is due to small aspect ratios, low volume fraction [29], and a simplification because non-constant radius fibers would have an insignificant effect on the elastic properties [30].The constant radius is set as the mean value of the actual fiber radius measured from experiments, and the length range of fibers is obtained following the classification operation.The orientation, distribution, and content of fibers from experiments are also verified.Specifically, Scanning Electron Microscopy (Hitachi UHR Cold-Emission FE-SEM SU8000) and optical microscopy (Nikon, Tokyo, Japan) is used to characterize the fiber orientation and fiber distribution in the filament.The filament is sectioned longitudinally and transversely, where the former is used to check the fiber orientation and the latter reveals the fiber distribution.The samples sectioned longitudinally are immersed in an epoxy resin, ground, and polished in preparation for the microstructural analysis.Grinding is done using 120 grit, 220 grit and 600 grit sandpaper.Polishing is performed using a 10 µm diamond slurry, followed by a 5 µm diamond slurry, and finished with a 0.3 µm alumina suspension.To verify the fiber volume content and fiber orientation in 3D space, a micro-CT scan is performed, and the sample entities are reconstructed in Avizo 3D software.The generation of the fibers is realized through different modified random sequential adsorption (MRSA) algorithms by using ABAQUS and PYTHON scripts.
Inner voids and resin particles emerge during the filament extrusion and 3D printing processes.A micro-CT scan is performed on sample materials, and the sample entities are reconstructed in Avizo 3D software to characterize the inner voids and resin particles.The diameters of the voids and resin particles in the RVE model are assumed to be the average values obtained from the software.
A homogenization procedure is usually applied to the RVE to obtain the overall elastic properties.To apply the periodic boundary conditions on the RVE planes, an identical mesh pattern needs to be created on the opposite faces of the RVE through a copy mesh scheme.However, due to the complex structure of the model, a periodic mesh cannot always be guaranteed.Instead, an interpolation technique is implanted into the plugin tool EasyPBC [31] to impose PBC with non-periodic meshes.
FE analysis of 3D-printed recycled composite parts made from PLA and recycled glass fibers will be performed and the effects of fiber content, fiber aspect ratio, fiber orientation, local fiber density, inner voids, resin particles, etc., will be discussed [27].
An example of the RVE hierarchical structure of recycled glass fiber-reinforced specimens is shown below (Figure 4) [27].All entities with different geometries are generated in the RVE frame while maintaining geometric periodicity.Fibers with different orientations (UD: unidirectional, RD: randomly directional), voids and resin particles are considered in this RVE model.

Phase III
The initial step before applying the above recycling solutions is the concentration and management of end-of-life blades from different wind farms within the interested area, which can be formulated as a facility location problem (FLP) of recycling sites.
By optimizing the site locations of recycling centres, life cycle assessment with aspects of the dynamic amount of waste resources, transportation among wind farms and recycling sites, relative energy cost and carbon footprint can then be studied.
Facility location plays a crucial role in the strategic planning of various public and private organizations [32].When addressing the issue of blade waste, it is essential to take into account multiple factors, including distance and cost from the corresponding demand points.In this specific context, the demand points are represented by wind farm locations in Canada.
Among the various models for facility location, the covering problem stands out as a popular and compelling choice.Typically, in covering problems, the services provided by facilities are determined based on the distance between the customer and the facilities.Specifically, a customer can receive service from any facility located within a predetermined coverage distance or coverage radius [33].Numerous scenarios can be considered as covering problems, such as determining the optimal number and locations of public schools, police stations, libraries, hospitals, public buildings, post offices, parks, military bases, radar installations, branch banks, shopping centers, and waste-disposal facilities [34].Similarly, the problem of establishing recycling points for blade waste can also be viewed as a covering problem.
The dynamics of facility location problems encompass the potential modification of existing facilities or the development of new ones over time.These dynamic aspects must be taken into consideration to effectively manage changing parameters.To ensure the practical applicability and implementation readiness of relevant models, it is crucial to incorporate the uncertainty of parameters.This allows for a balanced assessment between the benefits brought about by facility location changes and the associated costs of potential modifications.
In the case of recycling sites for blade waste, a notable challenge arises from the inconsistent supplies that fluctuate over decades.Given this scenario, it is common for organizations to contemplate relocating their facilities within a specific time horizon, without causing disruptions to ongoing activities.Thus, the possibility of facility relocation should be factored into the decision-making process, allowing for careful evaluation of associated implications and trade-offs.
Clustering plays an important role in data analysis that is used as a common method in modern scientific research.Several clustering approaches include Centroid-based clustering, Density-based clustering, Distribution-based clustering, Hierarchical clustering, etc.
The proper approach for the blade waste problem could be considered as a density-based clustering (Figure 5) since the local density of wind farms and corresponding waste in different regions varies in Canada.In the example shown in Figure 5, yellow and blue circles can be considered as wind farms in different regions, while grey spots can be assumed as possible recycling sites in this given optimization problem.Note that the location of the recycling sites is subject to change (relocation) as the waste supply changes.d.The annual processing volume of each wind farm should be balanced as much as possible.
The analysis steps for plastic waste management from wind turbine blades in Canada are summarized in Table 3.The table outlines a digital infrastructure for the management of turbine blade waste and design for the recycling site locations.

Summary and discussion
This research offers a comprehensive framework for the recycling of glass fibers from end-of-life wind turbine blades, including experiments (Phase I), simulations (Phase II) and LCA (Phase III).In Phase I, fibers will be recycled through FFF 3D printing and compression molding process.Corresponding simulation work will be carried out in II.Impact of recycling solutions from Phase I will be evaluated in Phase III.Based on the data obtained from Phases I and III, recycling site selection and optimization will be conducted.The recycling process and associated projects are summarized in Figure 6.Furthermore, this method uses techniques that can be scaled up to large industrial process.Mechanical grinding machines come in various sizes, thus with larger and multiple grinders, large amounts of fiberglass waste can be decomposed.The sieving process can also be scaled up because the technology for large sieving and sorting operations already existing in the technology of mineral sorting for the mining industry.These sorting machines from the mining industry could be adapted for use in the current fiberglass sorting operation.Finally, in the production of reinforced material for 3D printing, large-scale twin-screw extruders already exist in the polymer processing industry, thus large-scale operations for producing reinforced pellets can be easily developed.These pellets, in turn, can be either used in compression molding industries or fed into industrial-grade filament makers for producing reinforced filament for the 3D printing industry.
It is clear that in this on-going research, not all aspects have been fully developed.The science behind the grinding, sieving, sorting, and production of recycled 3D printing materials is well-advanced and ready for industrial scale automation.However, the first industrial process will be used to recycle a very specific source of similar wind-turbine materials.The industrial process will have to be flexible enough to handle variations in the type of input material (fiber weight percentage, resin type, additives, etc).Life cycle assessment methods will have to take account of these variations and to date, the development of the LCA analysis is not as well defined.This is an area where the research will have to catch up.

Expected contributions
The proposed research is expected to result in the following contributions: l Develop an optimized methodology for transforming the wind turbine blade recyclates into processable pieces for filament extrusion and compression molding processes.l Develop the optimum process for the extrusion of reinforced thermoplastic filaments as feedstock for 3D printing.l Develop reinforced FFF 3D printed and compression molded parts through the proper selection of the design and manufacturing process parameters.l Characterization of the structural properties of 3D printed and compression molded parts made of thermoplastics reinforced with recycled fiberglass.l Develop novel algorithms and predictive models to facilitate the design process and material assessment of the above recycling solutions.l Propose life cycle assessments for the above recycling solutions and optimize recycling site locations in Canada.l Boost wind turbine blade recycling from lab to market.

Figure 3 .
Figure 3. Recycling process showing the three resulting waste streams.These results are with the currently optimized process and are subject to changes (improvements) in the future.

Figure 5 .
Figure 5. Example of density-based clustering for wind farms.

Figure 6 .
Figure 6.The process of recycling and corresponding projects.

Table 2 .
Manufacturing and design parameters for specimen 3D printing.

Table 3 .
Analysis steps for waste management from wind turbine blades in Canada.