Determining Optimal Processing Conditions for Fabricating Industrial Moulds with Additive Manufacturing

Abstract

Additive manufacturing has reached a level of reliability and credibility that has already been integrated into specific industries producing final parts or tooling. Among Material Extrusion (ME) techniques, the Fused Granular Fabrication (FGF) method has enabled the development of Large Format Additive Manufacturing (LFAM) using polymeric materials, which has also established its presence in industries working with large prototypes, molds, and tools. This cost-efficient process has proven its applicability and success in manufacturing molds for composites, particularly in short and medium production runs, significantly reducing production times and costs. This paper presents two experiments designed to optimize process parameters when producing molds using the combined FGF and milling approach. These experiments identified optimal extrusion temperatures and extrusion multipliers to minimize defects at both the macro- and microscales for ASA 20 wt.% carbon fiber (CF) material; additionally, a correlation between milling speed, milling strategy, and surface roughness was established. These findings are valuable for industries adopting this innovative production method, as they provide guidance for defining process parameters to achieve the desired surface roughness of a specific part. A case study of the design of an automobile carter mold is presented, concluding that a specific range of milling speeds is required for conventional or climbing milling strategies to achieve a defined surface roughness range.

1. Introduction

Among the Material Extrusion (ME) technologies defined by the ASTM F42 committee [1], Fused Granular Fabrication (FGF) has enabled the production of larger-volume prints and extruded beads, representing a distinct approach to additive manufacturing (AM) using thermoplastic pellets as feedstock [2].

This technique has facilitated the development of Large Format Additive Manufacturing (LFAM), originating from a collaboration between Oak Ridge National Laboratories (ORNL) and Cincinnati Incorporated (Harrison, OH, USA), which resulted in the Big Area Additive Manufacturing (BAAM) system in 2015 [3]. Following this advancement, companies such as Thermwood Corporation (Dale, IN, USA) [4] have introduced industrial 3D printers with printing volumes ranging from 25 m³ to the substantial 450 m³ of the Masterprinter™ developed by Ingersoll Machine Tools (Rockford, IL, USA) [5]. In recent years, additional companies have presented equipment and technologies that fall within the LFAM category, leading to increased research interest in this field [6,7].

LFAM has been applied to produce various large-scale components, including toilet cabinets, excavator frames, boats, houses, and self-driving microbuses [8,9,10]. Notably, FGF combined with machining processes has proven effective and cost-efficient for creating molds across diverse applications. This approach offers a significant advantage over traditional metallic molds (aluminum or steel) by substantially reducing lead times and production costs for short to medium production runs [11].

The number of successful mold developments using FGF is rapidly increasing, particularly in the aerospace, energy, maritime, and automotive industries. For instance, ORNL, in collaboration with companies such as Whirlpool, Boeing, DGT, and Purdue University, has developed molds for expanded foam in refrigeration, compression molding, composite lamination, and metal stamping fixtures. Alliance MG printed a 10 m catamaran mold, BAE Systems produced an aerospace mold using the MasterPrint Ascent system (Macomb, MI, USA), and Thermwood, in partnership with Spirit AeroSystems, created skin molds for aircraft production. This trend extends beyond composite lamination to encompass injection molding, compression molding, and thermoforming.

Surface finishing is a critical aspect, typically involving machining and applying coatings to achieve the required roughness. The existing literature on this topic is limited. Duty and Springfield provided an initial feasibility assessment of FGF for composite part molds, highlighting the necessity for post-processing or coatings to achieve the desired surface quality [12]. Post et al. investigated the impact of print orientation on machining, finding no significant differences in surface quality or durability, with molds lasting up to five production cycles [13]. Duty et al. also pioneered the use of FGF for Vacuum-Assisted Resin Transfer Molding (VARTM), which requires stringent surface quality, using ABS with 20% carbon fiber (CF) reinforcement [12]. Lind and Lloyd documented the 24 h development of a polished VARTM mold, demonstrating high durability and a geometric precision of ±0.030 mm after milling and 10 production cycles [11]. Love et al. examined the thermal behavior of molds, focusing on the coefficient of thermal expansion (CTE) and thermal conductivity, through the development of a 13 m wind turbine blade mold using ABS with 20% CF [14]. This mold, fabricated in two days, was coated with glass fiber resin and machined, demonstrating the viability of glass fiber coatings for achieving necessary thermal properties, cooling rates, distortion control, and vacuum integrity. While some research has aimed to eliminate coatings, Post et al. successfully produced a seven-part boat hull using only demolding agents on milled surfaces [15]. Billah et al. analyzed the effects of infill orientation by printing three open molds with PPS and 50% CF in different orientations (0°/90° and 90°/0°), revealing that the 0° orientation, aligned with the beads, yielded the highest thermal conductivity [16].

Schniepp et al. evaluated the precision and stability of high-temperature polymer parts before and after thermal cycling [17]. Molds were 3D-scanned to determine geometric deviations and subjected to 120 °C for two hours, followed by cooling to 65 °C for 30, 60, and 90 cycles at 100 psi. The results indicated deformations below 0.5% after 200 h of exposure.

Fiber-reinforced polymers are commonly used in these applications due to their increased stiffness, dimensional stability, and good machinability [18]. Ajijeru et al. have defined the required rheological behavior for the LFAM processing of various materials [19,20]. Fiber reinforcement also enables higher printing temperatures and improved cooling rates [16]. Companies such as Techmer (Clinton, TN, USA), Sabic (Riyadh, Saudi Arabia), Airtech (Huntington Beach, CA, USA), and Matersia™. (Cádiz, Andalucía, Spain) offer commercial materials tailored for mold production, ranging from engineering polymers (ABS, ASA, PC) to technical polymers (PEI, PESU, PPS), all fiber-reinforced, with the latter being suitable for demanding autoclave processes [21].

Regarding polymer machining, Carr et al. identified environmental variables, cutting parameters, and polymer properties as key factors influencing surface quality [22]. The impact of cutting parameters on roughness and precision in injection-molded parts has been extensively studied [23,24,25,26,27], highlighting cutting speed, feed rate, and tool rake angle as critical parameters. Statistical analyses by Chabbi et al. [25] and Madic et al. [26] confirmed that feed rate is the most influential factor, a finding corroborated by Gaitonde et al. using Taguchi methods on various polyamides, including tool material as a variable [27]. Izamshah et al. also found feed rate to be the primary parameter in PEEK milling [28]. Pamarac et al. compared ABS and PLA milling, concluding that ABS requires lower cutting speeds than PLA, which exhibited better surface roughness at higher speeds [29]. Xiao and Zhan emphasized the importance of avoiding viscosity deformation during machining by staying below the glass transition temperature [30]. Yan et al. evaluated the processability of PEEK, PEI, and PMMA at different temperatures, achieving the best surface quality in the fragile regime [31].

This study presents two experiments designed to determine the optimal printing conditions for ASA with 20 wt.% CF parts intended for machining and to identify the optimal process parameters for both additive manufacturing and milling, ultimately optimizing surface quality for mold production.

2. Materials and Methods

The experimental phase was conducted in two stages. The first stage aimed to optimize printing parameters to minimize defects, while the second focused on identifying machining parameters that effectively eliminate micro-defects.

The material used was ASA reinforced with 20 wt.% carbon fiber (CF), supplied by Matersia. This material exhibited the following properties: ultimate tensile strengths (UTSs) of 56 MPa (X-axis) and 10 MPa (Z-axis); tensile moduli of 5800 MPa (X-axis) and 1200 MPa (Z-axis); coefficients of thermal expansion of 12 μm/m·°C (XY-direction) and 13.5 μm/m·°C (XZ-direction); a Vicat softening temperature (VST) of 99 °C; and a melt flow index of 25.62 g/10 min. Prior to printing, the material was dried in a PIOVAN DPA30 industrial dryer (Piovan Group S.p.A., Venezia, Italy) for 4 h at 80 °C, as per the manufacturer’s instructions.

All specimens were printed using a Discovery 3D Granza Printer (CNC Barcenas, Ciudad Real, Spain), a single-screw FGF printer with a printing volume of 0.5 m³, capable of reaching temperatures up to 450 °C. The printer featured a heated bed capable of reaching 175 °C and a minimum layer thickness of 0.5 mm. A 2 mm diameter nozzle was used for all prints. Constant printing parameters included a 2 mm nozzle diameter, 1 mm layer height, 15 top solid layers, 15 bottom solid layers, 1 contour line, and a bed temperature of 100 °C.

In the first stage, four parallelepiped specimens (98 × 100 × 120 mm, length × width × height) with a 15% infill density and a rectangular pattern were printed to evaluate printing conditions. The studied variables were printing temperature (245 °C and 260 °C) and extrusion multiplier (0.172% and 0.184%). The extrusion multiplier controlled the material flow rate, influenced by software parameters (e.g., virtual filament diameter, path overlap) and printer parameters (e.g., screw spin speed, barrel temperatures). In the second stage, specimens measuring 80 × 80 × 20 mm (length × width × height) were printed, maintaining the constant printing parameters of a 260 °C extrusion temperature and a 0.172% extrusion multiplier.

For milling, a KENDU HMKEN 0200.60 tool (Gipuzkoa, Spain) was used—a two-flute flat end mill with a 12 mm diameter, 30° helix angle, WC-Co (5–10% Co) composition, 22° rake angle, and no coating. Milling tests were conducted on a Kondia Five 400 5-axis machining center (Gipuzkoa, Spain), controlled by a Heidenhain iTNC530 system. The machine’s travel ranges were 660 mm (X-axis), 440 mm (Y-axis), and 510 mm (Z-axis), with a spindle power of 18 kW and a maximum speed of 24,000 rpm.

Cutting parameters were selected based on a literature review to optimize process performance: 150 m/min cutting speed (4777 rpm spindle speed), 0.05 mm/rev feed rate (500 mm/min), and 4 mm cutting depth. Milling was performed in both the up-milling (conventional) and down-milling (climbing) directions and in the direction of material deposition and transverse to it.

Dimensional accuracy, warping, and cracking were measured using a Helios Heta 12™ gauge feeler with a resolution of 0.0001 mm. Micrometrical dimensions, including internal porosity and cracks, were analyzed using images from a Nikon SMZ 800™ stereoscopic optical microscope and ImageJ V1.38 software. Roughness was measured using an Accretech™ portable roughness meter (precision ±0.001 µm), following ISO 21920-3:2021 standards [32].

3. Results and Discussion

The first experiment utilized four specimens printed according to the previously described parameters. Extrusion temperature and flow rate were selected as the primary variables, given their significant influence on both micro- and macro-defects. A 2 × 2 factorial design was employed to analyze the effects of these parameters. The selected extrusion temperatures were 245 °C and 260 °C, and the flow rates were 0.172% and 0.184%.

Prior to milling, a macrogeometric analysis was conducted, beginning with dimensional accuracy measurements.

Table 1. Percentual deviation in general dimensions of ASA 20 wt.% CF specimens at different extrusion temperatures and flow rate multipliers.

Temp/Multiplier	245 °C/0.172%	245 °C/0.184%	260 °C/0.172%	260 °C/0.184%
Height (%)	0.025	0.025	0.025	0.025
Length (%)	0.020	0.030	0.020	0.010
Width (%)	0.020	0.010	0.031	0.020

presents the percentage deviation for each specimen. The results indicate that the low coefficient of thermal expansion (CTE) of fiber-reinforced polymers contributes to improved dimensional accuracy due to their enhanced thermal conductivity. While all specimens exhibited acceptable accuracy, the highest precision was achieved with the combination of a 260 °C extrusion temperature and a 0.184% flow rate.

Thermomechanical stresses are known to induce distortions in printed parts, particularly at the base. This phenomenon, termed warping, is common in thermoplastic materials processed via AM, resulting from contractions during the cooling phase. Minimizing warping is crucial for achieving improved surfaces, facilitating the precise positioning and clamping of specimens, and ensuring stability during milling. Figure 1 illustrates the measurement locations, with the sample’s center serving as the reference point.

The measurements of the printed specimens revealed the highest deviations at corners 1, 5, 21, and 25 (Figure 1), reaching up to 0.3 mm when measured with a caliper.

An analysis of variance (ANOVA) was conducted to determine the most influential variable. The results indicated that temperature significantly impacts geometrical distortions, yielding a p-value of 0.0079 (below 0.05 for a 95% confidence level). This effect may be attributed to temperature’s influence on the melt flow index (MFI), which can affect a material’s adhesion to a build platform. As demonstrated in Table 1, a 15% increase in temperature resulted in a reduction in thermomechanical deformations.

To analyze microscale geometric defects, such as pores and cracks, across varying orientations and distances from the perimeter, all specimens underwent milling with a stepped profile, as depicted in Figure 2. Measurements were conducted using an optical stereoscopic microscope, and the resulting images were subsequently analyzed with ImageJ™ software. Following image calibration, the Threshold Colour tool within ImageJ was used for data processing. This tool segmented the images into color-coded regions based on predefined thresholds for pores or cracks, enabling the measurement of surface areas in square millimeters. These defect analyses were performed for both the perpendicular and aligned milling directions relative to the material deposition direction.

Internal porosity measurements were conducted at depths of 4, 8, and 12 mm across various layers and at different orientations relative to the deposited beads. This analysis aimed to distinguish between defects aligned with the beads (cracks) and those perpendicular to them (pores). The results presented in

Table 2. Porosity values in mm² of ASA 20 wt.% CF specimens.

Temp/Multiplier	245 °C/0.172%	245 °C/0.184%	260 °C/0.172%	260 °C/0.184%
Porosity (%)	0.86%	0.92%	0.86%	0.92%
Pore areas at layer 4 (mm2)	0.0148	0.0047	0.0102	0.0023
Pore areas at layer 8 (mm2)	0.0163	0.0053	0.0119	0.0031
Pore areas at layer 12 (mm2)	0.0112	0.0066	0.0092	0.0046

indicate a reduction in porosity with increasing flow rate, from 0.172% to 0.184% at 245 °C. At 260 °C, the benefit of carbon fiber reinforcement, specifically its enhanced heat dissipation, minimizes inclusions resulting from material degradation and gas emissions [33].

To quantify porosity and assess the influence of various parameters on pore formation, an ANOVA was performed. This analysis examined the effects of perimeter distance, flow rate, and temperature. The results indicated that flow rate, or the extrusion multiplier, was the most significant factor determining internal pore content, yielding a p-value of 0.000. This is consistent with the expectation that increasing the flow rate results in higher internal material density, effectively filling the internal pores formed during inter- and intralayer material deposition.

Similarly, an analysis was conducted to quantify the crack length and porosity observed during milling in the direction aligned with the deposited beads, using ImageJ™ v1.38.

Table 3. Surface defect area values of ASA 20 wt.% CF specimens.

245 °C	Start 4 mm	Middle 8 mm	End 12 mm	Start 4 mm	Middle 8 mm	End 12 mm
Multiplier	0.172%	0.172%	0.172%	0.184%	0.184%	0.184%
Layer 4 (mm2)	0.0005	0.0389	0.0326	0.0249	0.0328	0.0412
Layer 8 (mm2)	0.0002	0.0001	0.0092	0.0001	0.0012	0.0138
Layer 12 (mm2)	0.0009	0.0011	0.0276	0.0000	0.0082	0.0129
260 °C	Start 4 mm	Middle 8 mm	End 12 mm	Start 4 mm	Middle 8 mm	End 12 mm
Layer 4 (mm2)	0.0006	0.0006	0.0007	0.0005	0.0005	0.0005
Layer 8 (mm2)	0.0002	0.0002	0.0001	0.0079	0.0113	0.0128
Layer 12 (mm2)	0.0004	0.0005	0.0005	0.0082	0.0148	0.0256

presents the measured defect area values at different layers and perimeter distances.

These results were subsequently used in an ANOVA to determine the influence of perimeter distance, material flow rate, distance to the perimeter, and temperature on crack formation. The analysis revealed that all parameters, except the milling distance to the borders, significantly affected crack formation, with temperature exhibiting the strongest influence (p-value = 0.0037).

Figure 3 illustrates the types of defects quantified in Table 2 and Table 3. Figure 3a depicts the defects observed during milling perpendicular to the deposition direction, representing an inherent characteristic of this process. Figure 3b shows cracks resulting from interlayer cohesion failures.

The values for all specimens were plotted against temperature and flow rate and fitted with potential regression models, as shown in Figure 4, due to their good fit and physical relevance. The results indicate that crack formation increases with distance from the perimeters. Furthermore, at 260 °C, an increase in flow rate led to increased crack formation and reduced adhesion, likely due to material overflow. Conversely, the optimal results were obtained at 260 °C with an extrusion multiplier of 0.172%.

To validate the developed models, the percentage error between experimental values and model-predicted values was calculated. A maximum acceptable error threshold of 10% was established. Applying this criterion, specimens printed at 245 °C with an extrusion multiplier of 0.175% were excluded from further analysis.

Based on the remaining results, the optimal printing parameters were determined to be an extrusion temperature of 260 °C and an extrusion multiplier of 0.172%. While this combination resulted in slightly higher internal porosity compared to the 0.184% multiplier, it yielded greater material stability and printing regularity, with reduced warping and cracking and improved dimensional stability. Additionally, milling at a distance of 4 mm from the perimeter effectively reduced crack formation across all configurations. Utilizing these optimized printing parameters, a subsequent experiment was conducted to determine the optimal milling conditions.

The specimens for the second experiment (Figure 5) were printed and milled according to the parameters outlined in the Materials and Methods Section. The objective was to identify the milling conditions that yield the lowest arithmetic mean roughness values on the milled surfaces. Consequently, the variables studied in this stage were feed rate and milling strategy (conventional milling/up milling or climbing milling/down milling).

Consistent with the literature findings [25,26,27,28], feed rate was considered the most influential parameter in polymer milling. The feed rates examined were 0.025 mm/rev, 0.050 mm/rev, 0.075 mm/rev, and 0.1 mm/rev, corresponding to 238.85 mm/min, 477.7 mm/min, 716.55 mm/min, and 955.4 mm/min, respectively.

Once the CNC code was generated and specimens milled, the arithmetic mean roughness (Ra) values were measured on every milled surface at every defined feed rate, as is presented in

Table 4. Arithmetic mean roughness (Ra) values of feed rates.

	Conventional Milling		Climbing Milling
Feed Rate (mm/min)	Parallel	Perpendicular	Parallel	Perpendicular
238.85	0.475 µm	0.495 µm	0.550 µm	0.540 µm
477.7	0.540 µm	18.300 µm	0.967 µm	0.790 µm
716.55	1.460 µm	1.060 µm	2.133 µm	2.467 µm
955.4	2.433 µm	2.689 µm	2.830 µm	3.520 µm

. The measurements were conducted considering the direction of the deposited filament, as shown in Figure 5.

This table of values shows an atypical value at a feed rate of 477.7 with conventional machining on the face perpendicular to the beads. This is because the specimen presented a crack on this surface, generating an excessively high roughness value, that should not be considered. An ANOVA was conducted to assess the influence of milling strategy and feed rate on surface roughness. The results revealed p-values of 0.0128 for milling strategy and 0.000 for feed rate. This indicates that feed rate has a significantly greater impact on surface roughness than milling strategy, with a 95% confidence level.

Figure 6 presents two potential regression models that predict average roughness as a function of feed rate, stratified by milling strategy. The blue line represents the fitted regression line, the green lines illustrate data dispersion, and the gray lines indicate the model error.

Both regression models were validated by calculating the percentage error, with a maximum acceptable error of 10%. All accepted values were within this limit. Based on these models and the values presented in Table 4, a feed rate of 238.85 mm/min resulted in excessively low roughness levels, falling below the desired range. Conversely, feed rates of 716.55 mm/min and 955.4 mm/min produced arithmetic mean roughness (Ra) values exceeding the established range. A feed rate of 477.7 mm/min yielded Ra values within the required range, excluding the outlier attributed to an internal crack. Notably, both models demonstrated an increasing trend in Ra with an increasing feed rate, consistent with the findings in the reviewed literature [28,29]. This phenomenon occurs because a lower feed rate results in a greater volume of material removal per unit surface area, leading to a reduced milling footprint.

To illustrate the practical application of these findings, a case study involving the design of a composite material mold for a car engine carter is briefly presented. The carter, a critical engine component, stores lubricating oil. Through extensive research, the carter’s model specifications and the mold requirements for manufacturing it with an epoxy resin and 70% carbon fiber fabrics were determined. While detailed product specifications are beyond the scope of this study, the surface roughness requirement was established at 0.762–1.524 µm, as specified in Original Equipment Manufacturer (OEM) manuals [34,35].

Considering these requirements, a mold was designed, printed in two parts, and subsequently assembled. The mold’s dimensions were adjusted to account for the differential contraction coefficients of the composite materials used for the part and the mold. This design, illustrated in Figure 7, demonstrates the practical utility of this research. The optimized printing parameters minimize micro- and macro-defects, and the milling parameters can be precisely adjusted to achieve the required surface quality, both in terms of time and roughness.

4. Conclusions

This study addresses a gap in the literature related to the specific technological approach of manufacturing molds using Large Format Additive Manufacturing (LFAM) combined with milling, where surface quality or roughness is crucial for successful product development. As previously mentioned, there is limited literature in this specific field. Therefore, the authors provide a comprehensive review of existing studies, covering various aspects such as surface treatments, coatings, durability, dimensional accuracy, and material selection. This review, along with examples of industrial applications, highlights the reliability of this manufacturing method for industries such as the energy, aerospace, and maritime industries.

Through two distinct experiments, optimal printing conditions for ASA 20 wt.% carbon fiber (CF) material were identified, focusing on key variables: temperature and the extrusion multiplier. The results of this research demonstrate that 260 °C and an extrusion multiplier of 0.172% should be chosen to reduce macro-defects such as warping, cracking, and dimensional inaccuracies and to minimize internal micro-defects like pores and cracks. These defects were quantified to facilitate further decision-making. Additionally, a 4 mm distance to the perimeter was identified as a key parameter, establishing 4 mm as the optimal distance for the analyzed conditions. After selecting the appropriate temperature and extrusion multiplier, a second experiment was conducted to identify key process parameters influencing the lowest arithmetic mean roughness values for the milled surfaces. The results indicate that a milling speed of 477.70 mm/min, combined with both conventional and climbing milling strategies, is the optimal condition evaluated.

To conclude, a specific case study of mold development was presented. In this case, the mold was developed using the conditions defined in this research. This mold achieved a specified range of surface roughness of 0.762–1.524 µm as defined in the product specifications. According to this research, these Ra values can be achieved with milling speeds ranging from 319 to 545 mm/min for up milling and 375 to 700 mm/min for down milling, process parameters that can significantly accelerate the milling process, leading to a reduction in production times and costs.