Recycling Large-Format 3D Printed Polymer Composite Formworks Used for Casting Precast Concrete – Technical Feasibility and Challenges

Two materials were selected for this study: one biobased material and one synthetic/petroleum-based material respectively supplied by Jabil and Techmer Polymer Modifiers. Jabil PLA 3120 WF Pellet was selected as the biobased material and Techmer Polymer Modifiers Electrafil ABS CF20 BK was selected as the synthetic material system. Note that both materials systems comprised 20% filler by weight. Material properties as reported by the suppliers are listed in Table 1.

Fig. 1 depicts the recycling process adopted for the study, where formworks were subject to manufacturing, use, and recycling.

Note material loss was anticipated to occur throughout the recycling iteration. Anticipated losses were anticipated within the manufacturing phase and recycling phase. Manufacturing phase losses resulted from material purge and CNC machining. Recycling phase losses resulted from the shredding/grinding process.

The aforementioned materials were subsequently used to manufacture sets of formwork and in addition, hollow hexagonal prisms. These hexagonal prisms were 3D printed together with the forms. A hollow hexagonal shape with a wall thickness of 3.8 cm was chosen so the printed structure was stable enough to stand on its own. Six 30 cm by 34 cm plates were cut from the side walls of the hollow hexagonal prism. Specimens used for tensile testing were cut out from the plates. The mechanical properties of the 3D printed parts were evaluated based on the tensile test of these test specimens.

The formwork and hexagonal prisms were manufactured using the Ingersoll MasterPrint large-format 3D printer. All manufacturing used single screw extrusion with a screw diameter of 30 mm and a nozzle diameter of 12.7 mm. Fig. 2 depicts the extrusion head during the baseline CF-ABS formwork manufacturing. It is important to note that the as-printed formworks were manufactured as near-net shapes, meaning that the formworks were manufactured slightly larger than the desired dimensions to allow for postprocessing CNC machining of excess material into smooth formwork faces necessary for concrete casting.

Due to the aforementioned anticipated material loss throughout the recycling process, manufacturing parameters for the formworks could be variable. To ensure consistency within experimental material property characterization, the hexagonal prism manufacturing parameters were kept constant for the baseline and recycled forms, as presented in Table 2.

Following the completion of the AM process, subtractive manufacturing methods were used. These methods included the cutting, machining, and drilling of the forms. First, the forms were cut at the corners into four separate panels. These panels then had the interior faces machined to ensure a smooth surface for casting. Finally, each panel had six 12.7 mm holes drilled into them to allow for formwork assembly.

The formworks had a length of 61 cm, a width of 30 cm, and a height of 61 cm. The formworks had a wall thickness of 3.8 cm, which is equivalent to two bead widths, where a bead width refers to the ellipsoid extrudate that is deposited during manufacturing. Fig. 3 details the elevations of the formwork geometry.

Concrete was cast inside of the AM composite formwork to simulate the environment in which the application of this research is intended. For the purpose of this work, concrete was only cast one time within the formworks. The number of castings within formwork is highly dependent on the use case. Some typical wooden formworks are single-use, while other traditional formwork systems, such as steel forms, are used many times. Note that the properties of the concrete were not studied or are of interest for this work.

Before casting concrete, assembly of the formwork sets was necessary. Forms were assembled using 25 mm hex-coil bolts and plate washers. Once assembled, a water-based concrete form-release agent was applied to the interior faces of the forms with a brush. A plastic-covered, expanded polystyrene (EPS) void was then coated in the release oil and placed in the center of the form. The EPS void was used to reduce the amount of waste concrete generated during this work. Concrete was then cast around the EPS voids and was allowed to cure for 24 h. Following this curing period, the formwork was disassembled and stripped from the precast concrete part. A 24-hour curing time was selected for this study as is typical in many civil engineering applications; however, it should be noted that demolding time is application-dependent.

Following casting, the disassembled formwork panels were washed using a 21 MPa, Troy-Bilt pressure washer (Troy-Bilt, Valley City, OH, USA). The washing procedure consisted of laying the panels on a flat, drainable, surface. High-pressure water was applied to the formwork surfaces until no visible cementitious material remained.

Following cleaning, the formwork panels were reduced to granulates using a two-step shredding and granulation process. Formwork panels were fed into a Cumberland MS 40120 shredder with a mesh size of 51 mm in diameter. This material was then fed through a Cumberland 56U granulator with a mesh size of 6 mm in diameter. The granulated mixture was then used as feedstock for pelletization.

A Leistritz ZSE 27-MAXX twin screw extruder with a throughput of 80 kg/h was used to repelletize the recycled formwork panels. The granulate material was extruded at 247°C and 203°C for the CF-ABS and WF-aPLA, respectively. The granulate mixtures were extruded through a die with six 2.5 mm diameter holes. The formed extruded material was dipped into a water bath and then cut to a length of 6 mm.

A holistic approach was taken for the experimental characterization of material properties. Key thermal, mechanical, and physical properties of both materials were monitored throughout the recycling process to characterize material property degradation and understand where in the recycling process this degradation was occurring. Material degradation was based on a comparison of the experimental results of the recycled material with the baseline. Table 3 presents a summary of measured experimental properties and the corresponding method used to obtain the property.

Using differential scanning calorimetry (DSC), thermogravimetric analysis (TGA), and a rheology study, it was possible to experimentally characterize the glass transition temperature ( $T_g$), the effectiveness of the postcasting washing procedure, material purity, complex viscosity and modulus, and apparent viscosity for both material systems.

All tensile testing was conducted per a modified ASTM D3039 (ASTM 2017) with a deviation in specimen geometry to account for the layered structure of the large-format AM specimen and capture a representative unit cell. Tensile coupon specimen geometry is detailed in Table 6.

An Instron 8801 Servohydraulic test frame with a 100 kN Intron 2527-111 series load cell (Instron Corporation, Norwood, MA, USA) was selected for specimen testing. A constant head-speed test with a head displacement rate of 2 mm/min was used; where specimens were loaded until failure and digital image correlation (DIC) was used to monitor the strain field over the gauge region of the samples. The DIC system used two 12 MP Basler cameras for image capture of a stochastic black and white pattern painted on the gauge region. A combination of Streampix, LabVIEW, and GOM Correlate software packages was used to trigger and capture images during the tensile tests.

Note that the mechanical properties were tested in two principal directions; the 1-direction and the 3-direction, where the 1-direction corresponds to the direction of deposition of the bead, the 2-direction is transverse to the direction of deposition in the plane of the bead, and the 3-direction is through the thickness of the layer. Fig. 4 illustrates the previously defined directional naming convention.

Physical property testing was conducted to experimentally characterize the lengths of the reinforcing fibers and the molecular weight (MW) of the polymers.

Heat flow curves produced from the second heating ramp of the DSC tests were used to establish the glass transition temperature of each material system and can be seen in Fig. 5.

The $T_g$ was determined using the midpoint method of the endothermic peaks. Table 7 presents the midpoint-average $T_g$ values. Note that the CF-ABS samples have two endothermic peaks that both appear to be transition regions: the first at approximately 105°C and the second at approximately 130°C. The literature shows that the first endothermic peak is the true glass transition and that the second endothermic peak is simply a melting point of processed materials used when processing the SAN copolymer. Huseynov et al. (2023) and Billah et al. (2020) both concluded that the second endothermic peak is simply a melting point of an additional processing material. For this work, the first endothermic peak was used to represent the $T_g$ of the CF-ABS material system.

The baseline $T_g$ of both material systems was comparable to values provided by the material compounder and literature. The baseline WF-aPLA Tg of 54.4°C was approximately 11% lower than the value provided by the material compounder and the baseline CF-ABS $T_g$ of 102°C is comparable to values reported in the literature: where Bin et al. (2018) found the $T_g$ of CF-ABS to be approximately 100°C. Since the $T_g$ of the WF-aPLA is relatively low, the dimensional stability of formworks in hot climates could potentially be of concern. If AM WF-aPLA formworks are to be used in hot climates, where formworks are exposed to a temperature near or at the $T_g$, surface warping of the formwork is possible, which could result in imperfect cast concrete parts. If this environmental condition is anticipated, casting in a controlled environment, such as an indoor plant, may be preferable. Additional considerations for loss in structural integrity at temperatures higher than $T_g$ would be necessary for the amorphous polymer typically used in 3D printing; specifically when designing parts for applications with high service temperatures. After the first recycling iteration, there were only minor changes to the mean $T_g$ of each material system, where the percent difference between the baseline, R1 granulate, and R1 pellet is less than 0.5%.

Material purity and effectiveness of the washing procedure were established using TGA. The effectiveness of the washing procedure was determined, through a prewash and postwash comparison. Table 8 is a summary of the prewash and postwash samples, where the effectiveness of the washing procedure is assessed by the percentage of residuals.

The postcasting washing procedure was proven to be effective, with the measured impurities being less than 0.5% for both material systems. The pre- and postwash comparison showed that both material systems had a reduction of impurities of approximately 80% after washing the forms.

Introduced material impurities were determined through a comparison of the residuals of the baseline pellet and the R1 pellet. Mean residuals, for both the baseline and R1 pellet, are presented in Table 9.

Impurities introduced after recycling were determined by comparing the residuals of the R1 pellet with those of the baseline pellet. Following the recycling iteration, it was shown that, of the tested samples, the CF-ABS material system had no introduced impurities; whereas the WF-aPLA system had a mean residual introduced impurity of approximately 0.1%.

Rheological properties of both material systems were characterized for the baseline pellet, recycled granulate, and recycled pellet. Complex viscosity and complex modulus of the materials as a function of angular frequency are depicted in Fig. 6.

Fig. 6 depicts that both the CF-ABS and WF-aPLA material systems, at each stage of the recycling process, are non-Newtonian, shear-thinning fluids. In addition, Fig. 6 reveals that there are varying degrees of material degradation during the material recycling process. Fig. 6(a) presents that there is minor degradation in the CF-ABS complex viscosity between baseline, R1 granulate, and R1 pellet. Fig. 6(c) reveals that there is a noticeable degradation in the WF-aPLA complex viscosity between the baseline, R1 granulate, and R1 pellet. A similar change is recognized in the complex modulus of each material system. It is important to note that granulates were dried in metal drums in a walk-in oven before the pelletization process, which might have resulted in remaining residual moisture. Since aPLA is more sensitive to the presence of moisture at high processing temperatures due to hydrolysis/chain scission reactions, improving the drying procedure can help minimize the degradation of aPLA in future studies. Figs. 6(b and d) depict the entire viscoelastic behavior of both material systems, revealing that both materials become increasingly more flow resistant with increasing frequency. However, the flow resistance of the materials decreases with additional processing. Table 10 displays that there is no statistical significance between the decrease seen in the CF-ABS complex viscosity and complex modulus throughout the recycling iteration. In addition, Table 10 reveals that there is a statistical significance between the changes seen in the WF-aPLA complex viscosity and complex modulus.

The apparent shear rate of the material at the nozzle can be approximated using (Agassant et al. 2017)where $Q=$ the volumetric flow rate; and $r=$ the radius of the nozzle. The shear rate at the nozzle ( $\dot{\gamma }$) of the CF-ABS and WF-aPLA, according to Eq. (1), is approximately 71.7 and $42.7\mathrm{Pa}\cdot \mathrm{s}$, respectively.

The shear stress in the material as a function of shear rate is depicted in Fig. 7; note that these stresses are for low shear rates and are not representative of the shear rates experienced by the material during manufacturing.

Using the data presented in Fig. 7 and the Herschel–Bulkley (H-B) model, it was possible to obtain the apparent viscosity at the desired shear rates. The H-B model iswhere $k=$ the consistency index; ${\sigma }_0=$ the yield stress; and $n=$ the flow behavior index.

The shear stress values and fitting parameters were obtained through a least squares curve fit; calculated values from this fit are listed in Table 11.

The apparent viscosity of each material system at the nozzle was calculated using Eq. (3). The calculated values are presented in Table 12.

The CF-ABS had less degradation of apparent viscosity between the baseline and R1 results when compared with the WF-aPLA material system. The CF-ABS baseline to R1 granulate has a percent difference of approximately 17% and an approximate difference of 49% between the baseline and R1 pellet. The WF-aPLA baseline to R1 granulate has a percent difference of approximately 22% and an approximate reduction of 75% between the baseline and R1 pellet. These reductions in rheological properties, for each material system, can be correlated to the respective degradation of molecular weight (Bremner et al. 1990; Bernreitner et al. 1992; Chadwick et al. 2004; Bourmaud and Baley 2007; Ares et al. 2010; Yuan et al. 2011). Fig. 8 depicts the correlation between the material apparent viscosity and the MW, showing that the CF-ABS material system is more sensitive to changes in MW compared with the WF-aPLA. Two recent studies, evaluating the recyclability of natural fiber–reinforced composites, explored the effects of multiple recycling iterations on the composite rheological properties, concluding that reductions in material viscosity are products of reductions in polymer molecular weight and reduction in reinforcing fiber length (Rosenstock Völtz et al. 2020; Zhao et al. 2022). The dependency of viscosity on the composite constituent materials indicates that the degradation in viscosity is of concern due to the nocent nature of the thermomechanical recycling process on the reinforcing fibers and polymer matrix, where reductions in material viscosity could pose as a limiting factor. Continued reductions in viscosity at melt temperature would affect the ability of the extruded material to hold its shape (Chu et al. 2022; Bhandari and Lopez-Anido 2023). However, additional studies have been conducted to evaluate the reclamation of material properties after recycling, through the addition of both virgin material and additional additives. One such study focused on the thermomechanical properties of a 3D printed virgin/recycled PLA blend with a chain extender additive (CE), concluding that the incorporation of the CE improved rheological properties including complex viscosity (Cisneros-López et al. 2020). Another study explored the use of mechanical recycling of fused filament fabrication (FFF) 3D printing filaments as a method of waste reduction, focusing on the chemical, thermal, and thermomechanical characterization of both virgin and virgin/recycled PLA blends; finding that viscosity increases with an increased ratio of virgin material to recycled material (Agbakoba et al. 2023).

In addition, Fig. 9 provides a physical presentation of alterations in extruded material appearance, texture, and shape following one recycling iteration.

The tensile properties of the material systems investigated are presented in Table 13. The corresponding stress–strain relationships of the materials are depicted in Fig. 10.

The measured baseline tensile properties revealed that both the CF-ABS and WF-aPLA materials were stronger in the 1-direction than the 3-direction; this holds for the recycled material systems. The materials are expected to have a higher strength and modulus in the 1-direction when compared with the 3-direction, due to the preferential alignment of fibers. In addition, the 3-direction contains “welding lines” formed by incomplete polymer diffusion along the deposited layers.

Table 13 reveals that the mean tensile strength and mean tensile modulus for the CF-ABS material system, in the 1-direction, decrease after the material is recycled. Following recycling, the mean tensile modulus decreased by approximately 1.4% and the mean tensile strength decreased by approximately 18%. The decrease in mechanical performance of the CF-ABS, in the 1-direction, can be attributed to the fiber attrition of the carbon fibers. Hwang and Cho (2019) concluded that the mechanical performance of CF-ABS was highly dependent on the aspect ratio of the fiber reinforcement. In addition, Table 13 also reveals that the mean tensile strength and mean tensile modulus of the WF-aPLA, in the 1-direction, increases by 2.7% and 4.3%, respectively. Increased performance of the WF-aPLA in the 1-direction can be attributed to the increased aspect ratio of the wood flour fibers; resulting from further fibrillation during the recycling process. A study conducted by Robertson et al. (2013) found that thermoplastic composites reinforced with natural fibers displayed higher strength when reinforced with large aspect ratio fibers.

Table 13 also reveals that the mean tensile strength and mean tensile modulus of both material systems, in the 3-direction, increases after recycling, where the CF-ABS was approximately 2% stronger after recycling and the WF-aPLA was approximately 27% stronger, in the 3-direction, after recycling. The observed increase in 3-direction tensile strength can be attributed to the increased interlaminar adhesion between the layers, as a result of increased molecular diffusion through the layers (Bhandari et al. 2019).

A matrix dissolution was used to isolate the reinforcing fibers from the thermoplastic matrices to be measured. Fiber length distributions and corresponding Weibull fits are depicted in Fig. 11.

Figs. 11(a–c) depict the fiber length distributions for the carbon fibers (CFs) for increasing levels of processing. Figs. 11(a–c) depict that additional processing of the CFs creates a more normal distribution of fiber lengths, with increasing density of the shorter lengths. Figs. 11(d–f) depict that the wood flour fibers (WFs) distribution does not increase in normality and the distribution with a right skew remains. In addition, Fig. 11 depicts the Weibull fit for each material system’s corresponding level of processing. Table 14 lists two fitting parameters for the Weibull distribution.

In addition to fiber length determination, aspect ratios were also measured throughout the recycling iteration. The significance of the change in aspect ratio between the baseline and the recycled material was determined using analysis of variance (ANOVA) $t$-tests.

Table 15 reveals that the aspect ratio of the CFs decreases with additional processing, which was to be expected due to the chopping and shredding nature of the recycling process. Table 15 also presents that there is no statistical significance between the difference in the CF baseline and R1 granulate. However, there is a statistical significance between the difference in the CF baseline and the R1 pellet, which is revealed by the $P$-value being less in value than the level of significance. This means that the CFs maintain their lengths during the purely mechanical portion of the recycling process, with only minor degradation; and the majority of fiber attrition is happening in the additional processing step of material extrusion and pelletization.

Interestingly, the WF aspect ratio increased with material recycling. Table 15 reveals that the increase in WF aspect ratio between the baseline and the R1 granulate, as well as the baseline and R1 pellet, is statistically significant since the $P$-values are significantly smaller than the significance level. This means that the WFs are fibrillating and becoming more fiberlike, rather than being particulates.

GPC was used to quantify the molecular weight of the compounded material at multiple stages within the recycling iteration. Results from the GPC analysis are presented in Table 16.

The ABS saw a reduction in average mean molecular weight of approximately 0.5%, between the baseline and the R1 granulate, and a reduction of approximately 5% between the baseline and R1 pellet. The polydispersity (PD) of the ABS remained unchanged after the shredding/granulation process. However, the PD of the ABS increased by approximately 3% after the repelletization of the recycled granulate material; meaning that the molecular weight distribution of the ABS becomes broader after recycling. The amorphous PLA saw a reduction in average mean molecular weight of approximately 8% between the baseline and R1 granulate, and a reduction of approximately 31% between the baseline and the R1 pellet. In addition, the PD of the aPLA increased by approximately 5% following the recycling process. If the observed negative trend of MW continues throughout successive recycling interactions, the composite material will no longer be printable.

As anticipated, material loss was noted between the baseline print and the R1 print. The CF-ABS and WF-aPLA systems experienced reductions in material weight of approximately 50 kg/system. In addition to the material loss within the recycling process, there was an allocation of material for the hexagonal prisms. The total reduction in material resulted in a reduction in the number of forms that could be printed per material system. The baseline had a total of ten 3D printed forms, five forms per material system. The R1 cycle had a total of six 3D printed forms, three forms per material system.