Process parameter optimization for alumina ceramic parts manufactured by fused deposition modelling

Abstract

Ceramic materials are gaining relevance in Additive Manufacturing (AM), although the lack of standardized process parameters limits the repeatability and comparability of printed parts. This study proposes an optimization procedure for the fabrication of alumina (Al₂O₃) components using Fused Deposition Modelling (FDM) with a ceramic–polymer composite filament. The methodology was structured in two phases. First, the printing parameters related to extrusion conditions, speed profiles, layer configuration, and infill strategies were iteratively optimized using green parts. Then, the influence of chemical debinding and sintering was considered to refine the geometric parameters and scale factors necessary to ensure dimensional stability in final ceramic parts. The optimized parameter set was evaluated by fabricating standardized test artifacts according to ISO 52902:2023 to assess dimensional accuracy, resolution, and surface finish, and results were benchmarked against polylactic acid (PLA) printed under optimal conditions. The findings show that the proposed procedure enables reliable fabrication of alumina parts and provides objective performance data; however, dimensional deviations and surface artifacts increase after sintering due to material shrinkage and thermal deformation. The study recommends the adoption of structured parameter optimization workflows to support industrial integration of ceramic FDM and highlights the need for future work on optimizing debinding and sintering profiles to improve final part stability.

Introduction

AM has undergone continuous growth since its inception in rapid prototyping, evolving into a technology with the potential to become a cornerstone of the industrial sector¹. The ability to fabricate complex geometries has enabled design and manufacturing solutions to meet the increasing demands of industry^2,3. AM has demonstrated its capacity to address key challenges such as integration into production lines, large-scale manufacturing, and component weight reduction. However, its widespread adoption is often constrained by the high costs associated with both machinery and raw materials.

The impact of AM is particularly significant in sectors such as aerospace and automotive manufacturing^4,5, where it facilitates the production of lightweight structures. In the medical field⁶, AM enables the development of patient-specific interventions with enhanced precision and efficacy. Furthermore, its adoption is expanding into other domains, including the energy sector, electronics, and architecture⁷, where it is streamlining workflows and optimizing design processes for professionals^8,9. The continuous advancement of AM technologies, ranging from novel exchange formats to the development of advanced materials and the integration of numerical simulation, underscores its potential to play a central role in the future of industrial manufacturing. Nevertheless, despite the rapid progress of AM, several limitations must be addressed before it can be fully established as a standard technology in manufacturing, particularly concerning metallic and ceramic materials¹⁰. The precision and printing speed of current AM systems still require significant improvements to minimize post-processing steps. Moreover, certain materials that are widely used in conventional manufacturing processes—such as technical ceramics—remain only partially compatible with AM technologies^11–13.

There are various approaches for producing ceramic components using additive manufacturing processes. One such method is Stereolithography (SLA or VPP-UVL/C), a technology that employs a ceramic paste instead of the plastic resin used in conventional stereolithography for polymers. This paste consists of ceramic particles suspended in a liquid photosensitive resin, which acts as a binder during printing and is subsequently removed through thermal post-processing¹⁴. Another technique is Selective Laser Sintering (SLS or PBF-LB/C) using ceramic powder. In this process, fine layers of preheated ceramic powder and binder are fused together to form the final part. Unlike other additive manufacturing techniques, SLS does not require extensive post-processing, as the ceramic particles can achieve the desired solid-state form without the need for a polymeric base that must later be removed. However, processes such as isostatic pressing or pressure infiltration can enhance the green density of the fabricated parts¹⁵. A different category includes Binder Jetting (BJ or BJT/PC), a technology that enables the production of composite parts made of polymeric and ceramic materials. Successive layers of ceramic powder are selectively bonded using a polymeric adhesive, which acts as a binder and is subsequently eliminated through post-processing polymerization at moderate temperatures¹⁶. This study focuses on another additive manufacturing technology: fused deposition modelling of Al₂O₃, commonly referred to as FDM. According to ISO 52900:2021, this technique is classified as MEX-TRB/C/Al₂O₃¹⁷.

The FDM process utilizes solid-state materials, such as filaments or pellets, which are melted and extruded through a heated nozzle, depositing successive layers to form a three-dimensional object. In the case of technical ceramic 3D printing, layers of polymer-based material containing a ceramic load are deposited and fused together to create a solid component. In certain applications, the characteristics of the homogenized material may be suitable for the intended use of the part. However, in many cases, additional thermal and/or chemical post-processing steps are required to remove the polymer content, as is the case in this study. Technical ceramic 3D printing facilitates the manufacturing of complex and precise components, overcoming the limitations of traditional manufacturing methods. This advantage allows ceramic materials to be leveraged in high-performance sectors such as the automotive and aerospace industries. However, it is important to acknowledge that FDM-based ceramic part designs are inherently constrained by the operating principles of the process, the materials employed, and the technological limitations of the system¹⁸. Furthermore, due to the necessity of post-printing processing for ceramic materials^19,20, issues such as fractures, crack formation, and deformations may occur²¹.

Numerous studies highlight the critical role of process parameters in AM, particularly in FDM^22–26. The effectiveness and efficiency of the process are significantly influenced by these parameters. While their impact has been extensively analysed in the context of polymeric materials, there remains a substantial knowledge gap regarding the use of FDM for technical ceramics. Consequently, the identification and application of optimal parameter values are essential to fully exploit the potential of the process and to establish an optimized parameter profile independent of external factors that may affect the final outcome²⁷.

In this study, based on the selected optimal parameter profile, demonstrator artifacts are employed to assess the relative capabilities of the technology in terms of geometric accuracy and minimum achievable dimensional features. The evaluation follows ISO 52902:2023²⁸, which defines a set of standardized artifacts designed to assess the performance of additive manufacturing technologies concerning precision, resolution, and surface quality. The geometries of these artifacts are provided in STEP format by the standard itself, ensuring dimensional accuracy by minimizing potential distortions introduced during digital model conversion. This approach guarantees a dimensional precision of at least 2.5 μm.

The following sections provide a detailed description of the materials and equipment used throughout the various stages of the manufacturing process. Additionally, the methodology employed to establish the optimal parameter profile is outlined, followed by a presentation and discussion of the results obtained at different phases of the optimization procedure. Finally, the study’s key conclusions are summarized, and potential future research directions are proposed to further advance this field.

Materials and methodology

The study was structured into two stages (Fig. 1). During the first phase, starting from a universal parameter profile, the printer’s threshold and safety temperatures were adjusted, along with the motor startup power and the pressure exerted by the drive wheels on the filament. Subsequently, the optimization of the printing speed profile and filament parameters was carried out using “open artifacts.” This was followed by the optimization of printing parameters for the fabrication of solid infill structures, including the production of samples to determine a micro-step in the Z-axis that would ensure a smooth surface finish and uniformity in the lower layers.

Fig. 1.

Methodology for process optimization.

Once the printing process was optimized using the measured dimensions of the printed parts without applying any scale corrections, the second stage focused on optimizing the parameters for parts that had already undergone chemical and thermal treatments. In this phase, the number of bottom layers, perimeters, infill parameters, and top layers were optimized, resulting in a refined parameter profile for post-processing operations. Finally, using these optimized parameters, standardized test artifacts were fabricated according to ISO 52902:2023, which defines the geometries of the parts to be manufactured for the evaluation of both precision and resolution.

For the generation of 3D files of the components under study, SolidWorks (v.2023) was used to create the digital models, which were then exported in a format compatible with slicing software. The digital information was processed using PrusaSlicer v.2.7.1, where key production parameters were defined. This included specifications such as layer height, perimeters, infill density, skirt, support material, and printing speeds, among others (Fig. 2).

Fig. 2.

PrusaSlicer software interface. A screenshot of process parameters configuration.

Additionally, the software allowed for the selection of the working material and provided a simulation of the 3D printing process, including the estimated material deposition times.

This study was conducted using a ceramic material, Zetamix Alumina (A1 100-BA), supplied by the company Nanoe in filament form with a diameter of Ø1.75 mm. The material composition consists of 55% alumina (Al₂O₃) and 45% binder system by volume, which translates to 83% and 17% by weight, respectively. The remaining material properties are presented in Table 1.

Table 1.

Zetamix alumina properties.

Printing temperature	150 °C
Solvant debinding	Acetone
Sintering temperature	1550 °C under air
Shrinkage	x, y = 20.8% ±1%/z = 23.2% ±1%
Density	98–99%
Specific Gravity [g.cm-3]	2,5
Melt Flow Rate [g/10(min)]	200
Melt Volume Rate [cm3/10(min)]	80
Moisture Absorption [%] 24 h/7 days	< 0,1%/<0,3%

The printing of the components was carried out using the Prusa i3 MK3S + printer (Fig. 3a). For the chemical post-processing (chemical debinding process), the SonoSwiss SW3H ultrasonic equipment was used (Fig. 3b). This system, through temperature and agitation, accelerates the removal of the plastic binder from the parts. The device was equipped with a gas extraction hood to eliminate gaseous residues generated by the chemicals during the operation (95% acetone concentration) as they evaporate. For thermal debinding and sintering, a Nabertherm LHTCT furnace was used, capable of reaching temperatures up to 1,600 °C (Fig. 3c).

Fig. 3.

Equipment used for the artifacts manufacture. (a) 3D printer. (b) Ultrasonic tank. (c) Furnace for heat treatment.

Experimental procedure and results

The following describes the optimization procedure followed in the different phases, along with the results obtained throughout the process.

Phase 1: optimization of the printing speed profile and filament parameters

The initial adjustments were made using a control test specimen, following the parameters listed in Table 2. However, certain values had to be modified in the initial startup code of the printer to address issues encountered at the beginning of the printing process. First, it was necessary to limit the safety threshold temperature to 110 °C by modifying the coded line “M302 [S]”, as the default value was set to 150 °C. This adjustment ensured that the material could be extruded at lower temperatures than those typically used for polymeric filaments.

Table 2.

Basic printing parameters.

Printing temperatures
Nozzle temperature	150 °C
1 st layer nozzle temperature	150 °C
Bed temperature	40 °C
1 st layer bed temperature	40 °C

Perimeters and layers
Extrusion nozzle diameter	0.4 mm
Layer high	0.15 mm
1 st layer high	0.2 mm

On the other hand, the layer thickness of 0.15 mm was chosen as an optimal balance between dimensional precision, surface quality, and process time. Several layer heights (0.10, 0.15, 0.20 mm) were tested during the optimization stage. A smaller layer height improved surface finish but led to higher internal stresses and longer print times, while thicker layers degraded the resolution of fine details required by ISO 52,902 artifacts. The selected value of 0.15 mm minimized the “stair-stepping” effect while maintaining acceptable mechanical integrity of the ceramic-loaded material. This layer thickness is also in line with the manufacturer’s recommendations for Zetamix alumina filaments, which suggest a range of 0.10–0.20 mm for accurate and stable deposition.

A wall thickness equivalent to three perimeters was deliberately selected to ensure sufficient mechanical strength and to compensate for possible dimensional contraction during debinding and sintering. Preliminary trials showed that fewer than three perimeters resulted in fragile “green” parts prone to delamination and geometric distortion during post-processing. Choosing three perimeters provided the best compromise between dimensional stability, printing time, and material consumption, while preserving the integrity of small geometric features defined by ISO 52,902. This value also aligns with typical design guidelines for FDM ceramics, where multi-perimeter walls help counteract the brittle nature of ceramic-binder composites prior to sintering.

Additionally, the current of the digital potentiometer controlling the stepper motor had to be reconfigured. The corresponding line “M907 [E]” was modified to set the extruder stepper motor current to a value of 260, preventing the supplied power from reaching levels that would cause material clogging due to filament compression between the drive gears. Finally, the tension applied by the drive wheels on the filament was manually adjusted. Excessive tension at the entry point of the drive system led to filament breakage as it was being pulled by the toothed wheels.

The reduction in printing speed (approximately 30–35% lower than standard PLA) is directly related to the ceramic particle content within the filament. The presence of alumina particles significantly increases the viscosity and abrasiveness of the material, which affects the dynamics of extrusion. Lowering the print speed ensures a more homogeneous flow through the nozzle, prevents under-extrusion, and enhances interlayer adhesion. Moreover, reduced speed minimizes the mechanical stress exerted on the feed mechanism and mitigates the risk of filament breakage, which is common in highly loaded composites.

Once the initial conditions were defined, the experiment was carried out using a geometry that allowed for the observation of possible defects derived from the parameters studied. The fabrication of 20 × 20 × 10 mm prisms was chosen, applying the “vase mode” available in PrusaSlicer software, without applying any scale correction coefficient. The conditions observed for the specimens were as follows:

• The specimens must not present deformations in either the walls or the base.

• The minimum possible shrinkage should be observed on the edges of the specimen.

• The printed layers must be uniform.

• The rounding of the specimen’s corners must be avoided.

• The walls (P1 – P4) must have a thickness of 0.4 mm.

Based on these criteria, the optimization process was initiated, carrying out several iterations that allowed for the updating of printing parameters (Table 3). As a result of the first print, in which the parameter profile shown in the first column of the table (iteration 1º) was used, the specimen exhibited a lack of uniformity in the base layers, discontinuities in the external perimeters, rounded corners, and material shrinkage (Fig. 4a). Additionally, the wall thicknesses deviated from the target, with differences ranging between 0.2 mm and 0.3 mm.

Table 3.

Print speed profiles and parameters.

Speed (mm/s)		Iteration
		1º	2º	3º	4º	5º
External perimeters		20	17.5	15	12.5	10
Solid filling		30	30	25	20	20
Solid top filling		35	30	25	20	20
Support		35	30	25	20	20
1 st layer		20	17.5	15	12.5	10
Filament
Extrusion multiplier		1	1	1.1	1.05	1.05
Geometric features
Number of solid bottom layers		2	3	3	3	3
Wall thickness	P1	0.38	0.37	0.41	0.40	0.40
	P2	0.38	0.38	0.42	0.40	0.40
	P3	0.37	0.37	0.41	0.39	0.40
	P4	0.38	0.38	0.41	0.40	0.40

Fig. 4.

Printed test specimens. (a) First iteration (b) Fifth iteration.

By modifying the printing parameters in the second iteration, shrinkage was reduced, but the edges remained deformed and exhibited slight rounding. In this iteration, the number of bottom layers was increased, reducing fragility and facilitating the removal of the specimen from the machine’s build platform. However, the wall thicknesses remained unchanged.

The third iteration achieved greater uniformity, with fewer discontinuities in the layers, but the selected parameters failed to eliminate shrinkage and rounded corners. On the other hand, wall thicknesses increased to 0.4 mm compared to the values of the previous iteration, as reflected in the column corresponding to the 3rd iteration.

Based on the obtained data, the extrusion multiplier was reduced and print speeds were further decreased. This adjustment yielded more acceptable results, which became optimal by the fifth iteration when only external parameters and the first-layer speed were reduced (by 50% compared to the 1 st iteration). Under the working conditions shown in the column corresponding to the 5th iteration, the results presented in Fig. 4b were obtained. As seen, this iteration resulted in a specimen without deformations, shrinkage, or rounded edges. Additionally, the uniformity of the bottom layers remained consistent with the previous stage but showed improvement compared to the first print layer.

After this final iteration, an optimized speed profile was established for printing parts with this ceramic filament. Furthermore, the expected wall thickness of 0.4 mm was achieved, confirming that the assigned extrusion multiplier value was correct and that variability in the results for this characteristic was minimal.

As a summary, Table 4 presents the initial and optimized speed profiles. The print speeds were reduced for all parameters, reaching a reduction of up to 50% of the original value. Additionally, the extrusion multiplier was originally set to 1, whereas in the optimized profile, this factor was increased by 5%.

Table 4.

Comparison of initial and optimised speed profiles.

	Initial speed (mm/s)	Optimised speed (mm/s)
Perimeters	25	15
Small perimeters	20	10
External perimeters	20	10
Filling	30	15
Solid Filling	35	20
Solid top filling	35	20
Support	35	20
Bridges	35	20
Gap filling	35	20
1 st layer	20	10

With the fully optimized profile, the optimization process for the solid infill printing parameters was carried out. In this phase, adjustments were made to the number of external perimeters, the number of bottom and top layers, and the infill percentage of the part. It is important to note that these parts will be analyzed without undergoing the necessary chemical and thermal post-processing treatments required for final components, meaning their robustness is not yet equivalent to what it will be after those treatments.

Following the recommendations of the ceramic filament manufacturer, a two-dimensional rectilinear infill pattern with a 5% infill percentage was used for the first iteration. The specimen design employed in this stage was very similar to the one used in the previous phase, aiming to continue observing possible imperfections that might occur. The dimensions of the prism remained 20 × 20 × 10 mm, and no scale correction coefficient was applied, as the specimens were not intended to undergo any chemical or thermal post-processing. To draw conclusions regarding the quality of the infill and the top layers, the fabrication process was analysed not only upon completion of printing but also at two intermediate points, at 65% and 85% of the process.

Under these working conditions, the results shown in Fig. 5a were obtained. As observed, when the fabrication reached 65% completion, the infill was not deposited uniformly and homogeneously, leading to many layers where the material lines turned into overhangs and solidified in incorrect positions, leaving large gaps in the infill mesh. Moreover, at 85% of the printing process, the low infill density of the specimen resulted in poor deposition of the first top layer, causing it to warp and retain these deformations upon solidification. This also led to the upper layers of the specimen lacking a stable base for deposition, ultimately reducing the quality of the final surface finish. The final result of this iteration exhibited a top layer with noticeable discontinuities and a lack of uniformity. This was due to an insufficient number of top layers in the piece. Additionally, the number of perimeters and bottom layers was also inadequate, as the specimen’s robustness and resistance were too low, making it highly susceptible to minor compressive forces. Two more iterations were conducted, adopting the values presented in Table 5. In the third iteration, the size and proportion of gaps in the infill mesh were significantly reduced. Most of the overhangs were located at the upper part of the grid, meaning that the gaps left during printing did not compromise the structural integrity, as the printhead passed over these areas during the deposition of the first top layers, filling in and levelling out these defects. With the increased infill density, greater uniformity and precision in the formation of the initial top layers were observed (Fig. 5b). Finally, the robustness of the specimen in this iteration improved considerably, providing sufficient resistance to ensure safe handling.

Fig. 5.

Printed test specimens. (a) With initial geometric profile (b) With optimized geometric profile.

Table 5.

Geometrical and filling parameters.

	Iteration 1	Iteration 2	Iteration 3
Geometrical parameters
Number of perimeters	1	2	2
Number of upper layers	3	4	4
Number of lower layers	3	4	4
Filling parameters
Filling percentage (straight pattern)	5	7.5	10
	65%	85%	100%

The final experimental study conducted in this first phase was the analysis of the micro-step in the Z-axis, a parameter that ensures a good surface finish and uniformity in the bottom layers. This characteristic largely depends on how the extrusion head is assembled, as it determines the initial distance between the extruder nozzle and the build platform. Additionally, this parameter varies depending on the printer used, making it impossible to establish a single configuration value.

The objective of this experimental process was to determine an approximate range of values where a machine with a standard assembly (with the extrusion head mounted approximately one millimetre away from the heater for the printing material) ensures a good surface finish in the bottom layers of printed parts. To achieve this, six specimens were fabricated using the optimized parameter profile from the previous experimental stages (speed and geometric parameters). These specimens, in the form of prismatic boxes and printed in “vase mode” with dimensions of 20 × 20 × 3 mm, were also manufactured without scale correction.

To draw conclusions regarding this study parameter, the following conditions were observed:

• Uniform deposition of the first layer.

• Strong adhesion between the deposited filaments in the first layer.

• Good surface finish on the bottom of the specimen.

• Avoiding deformations at the edges and corners of the base.

Unlike previous cases, this process was not divided into experimental iterations where each set of working parameters was based on the results of the previous iteration. Instead, the procedure involved fabricating multiple specimens with different micro-step values in the printer’s Z-axis (Z-offset, or the working distance between the extruder and the heated bed).

By comparing the results, conclusions were drawn about the valid parameter range. It is important to note that the smaller this distance, the better the fusion between deposited layers. However, this also increases the risk of over-extrusion, affecting wall thickness and increasing the likelihood of nozzle clogging.

As shown in Fig. 6, specimens 1 and 2 exhibited significant discontinuities in the first layer, very low uniformity in the bottom layers, and poor surface finish. On the other hand, specimens 5 and 6 displayed some deformations in the corners of the base, with specimen 6 also showing geometric defects on the edges of the bottom surface. Regarding the specimens with intermediate values (specimens 3 and 4), good uniformity and adhesion were observed in their first layers, without deformations at the edges or corners, and without discontinuities in the base.

Fig. 6.

Test specimens printed with different Z-axis micro-step.

During the fabrication of specimen 7, the extrusion head nozzle became clogged, leading to the cancellation of its print. Consequently, the corresponding parameter value was deemed invalid. The micro-step values used in the printing of each specimen are presented in Table 6.

Table 6.

Z-axis micro-step values.

Test specimen	1	2	3	4	5	6	7
Micro-step (mm)	1.10	1.20	1.30	1.35	1.40	1.50	1.60

Phase 2: optimization of geometric and infill parameters

During this phase, the printed parts were analyzed after undergoing the debinding and sintering processes, meaning that the designs already included scale factors for each of the three axes. The steps carried out in this order allowed for the optimization of the number of bottom and top layers, the number of perimeters, and the infill parameters. All specimens manufactured in this phase underwent the final processes necessary to obtain functional parts.

First, chemical debinding was performed. This process involved a chemical bath in acetone with a concentration of 99.5%, using an ultrasonic device at a constant temperature of 40 °C. The bath duration was, in all cases, over two hours, following the manufacturer’s recommendations. The parts were then dried at room temperature for two hours. To ensure complete removal of any residual solvent, the specimens were placed on an absorbent surface.

Second, thermal debinding was carried out according to the stages shown in Fig. 7a. This treatment, which removes any remaining plastic material not extracted during chemical debinding, consisted of a heating phase lasting 22 h and 40 min, followed by a cooling phase at room temperature, which took approximately 24 h.

Fig. 7.

Thermal processes. (a) Debinding. (b) Sintering.

Finally, sintering was performed to reduce the porosity caused by the removal of plastic material during debinding. The heating, temperature maintenance, and cooling stages of this process are detailed in Fig. 7b.

The initial scale corrections applied to compensate for the volumetric reductions occurring during the thermal stages were 124.5% for the X and Y axes and 124% for the Z axis.

To optimize the number of bottom layers, five types of test artifacts (“vase mode”) were fabricated, each with a different number of bottom layers (ranging from 4 to 8). Before undergoing thermal treatments, it was verified that the weight reduction caused by chemical debinding exceeded 5%, in accordance with the manufacturer’s recommendations. This condition was met in all models, with a reduction exceeding 13% in all cases (Table 7).

Table 7.

Configuration of test artefacts for the optimization of the number of bottom layers.

Artifact Test Code	Number of bottom layers	Thickness (mm)	Chemical debinding time (min)	Initial weight (g)	Weight reduction
1.1	4	0.65	150	2.940	14.01%
1.2	5	0.80	150	3.093	13.63%
1.3	6	0.95	180	3.274	13.65%
1.4	7	1.10	180	3.466	13.73%
1.5	8	1.25	180	3.588	13.53%

It was confirmed that specimens 1.1 and 1.2 were too thin, giving them a fragile appearance. Conversely, specimen 1.5 exhibited a plastic-like texture in its matrix, indicating that it had not fully hardened. Considering the manufacturer’s recommendation of a minimum wall thickness of 1 mm for printing, specimen 1.4 (with 7 bottom layers) was selected as the optimal solution.

The next step was to determine the optimal number of perimeters. In this case, applying all the parameters optimized up to this point, three types of artifacts were fabricated, also in “vase mode,” each with a different number of perimeters (2, 3, and 4). The configuration of the specimens and the results after chemical debinding are shown in Table 8.

Table 8.

Configuration of test artefacts for the optimization of the number of perimeters.

Artifact Test Code	Number of perimeters	Thickness (mm)	Chemical debinding time (min)	Initial weight (g)	Weight reduction
2.1	2	0.80	180	3.466	13.73%
2.2	3	1.20	180	4.286	12.78%
2.3	4	1.60	210	5.106	13.38%

In this case, specimen 2.2 (bottom thickness of 1.1 mm and lateral wall thickness of 1.2 mm) stood out from the others, exhibiting high hardness and robustness across all surfaces, with homogeneous characteristics and properties throughout its geometry.

With the geometric data from all the specimens fabricated thus far, the scale factor was further adjusted, resulting in a value of 126.52% for the X and Y axes and 129.30% for the Z axis.

The next step, optimizing the infill parameters, was carried out in two stages. First, the infill percentage was optimized, followed by the pattern selection. Initially, three specimens were fabricated with infill percentages of 10%, 15%, and 20%, all using a rectilinear pattern. They underwent chemical debinding for 180 min, and it was confirmed that all of them met the mass reduction requirement of over 5%.

After the thermal treatments, it was observed that specimen 3.1 exhibited some fragility and discontinuities in the infill due to a lack of support points. This issue arose from the applied percentage, which increased the travel distance of the print head before reaching the next anchoring point. On the other hand, specimen 3.3 underwent significant torsion around the Z-axis of the machine after the thermal stages, despite maintaining good strength and hardness. The optimal value was thus found in specimen 3.2.

Using this information, four test artifacts with different infill patterns were analyzed: the previously fabricated rectilinear pattern, triangular pattern, 2D honeycomb pattern, and gyroid pattern. Following the same validation procedure as in the previous steps, it was observed that the resolution, strength, and stiffness obtained were very good across all four infill patterns. However, the triangular and 2D honeycomb patterns resulted in an approximate 20% increase in print time compared to the rectilinear and gyroid patterns. The gyroid pattern was discarded, as the ceramic filament manufacturer does not recommend using this type of infill for 3D-printed parts with complex geometries. Therefore, the rectilinear infill pattern was determined to be the most suitable option for fabricating parts with this material.

With all the data gathered throughout the complete optimization process, a final specimen was fabricated (Fig. 8), confirming that the optimized parameters allowed for the production of functional and structurally sound parts, free of visible defects.

Fig. 8.

Test artifact manufactured with all optimized parameters.

This second phase concluded with the fabrication of several test artifacts, incorporating all optimized parameters. In accordance with ISO 52902:2023, these artifacts allow for assessing the technical capabilities of this additive manufacturing process. A total of 16 artifacts were fabricated: 8 in PLA and 8 in Al₂O₃. These sets of 8 pieces correspond to those defined in the standard: Three artifacts for precision evaluation, four artifacts for resolution assessment and one additional artifact to evaluate the achievable surface finish.

Regarding precision, two linear artifacts (Fig. 9a and b) were fabricated to evaluate linear positioning accuracy along a single machine direction. Depending on the specimen’s orientation (aligned with the X-axis or Y-axis), measurement errors due to tool positioning failures can be identified. This provides the necessary data to adjust the system accordingly.

Fig. 9.

Artifacts for accuracy evaluation. (a) Linear PLA. (b) Linear Al₂O₃. (c) Z-axis PLA. (d) Z-axis Al₂O₃.

Additionally, an artifact for the Z-axis (Fig. 9c and d) was fabricated, which does not require support structures for assessment along the third machine axis. This setup allows for distinguishing measurement errors due to failures in the Z-axis positioning system, enabling recalibration of the drive system when necessary.

Regarding resolution, several test artifacts were fabricated

A pin resolution artifact (Fig. 10a and b) to determine the thinnest pin the machine is capable of producing. A hole resolution artifact (Fig. 10c and d) to qualitatively evaluate the minimum producible internal cylindrical features. This artifact also assesses how easily the system can remove unsolidified material from small-diameter holes. A rib resolution artifact (Fig. 10e and f) to determine the minimum wall thickness achievable using this additive manufacturing system. A slot resolution artifact (Fig. 10g and h) to evaluate the minimum slot thickness the system can produce or the minimum spacing between two geometric features achievable with this process. In all four cases, measurements of pin diameters, hole diameters, rib widths, and slot widths also provide valuable data on the precision of the system.

Fig. 10.

Artifacts for resolution evaluation. (a) PLA pins. (b) Al₂O₃ pins. (c) PLA holes. (d) Al₂O₃ holes. (e) PLA ribs. (f) Al₂O₃ ribs. (g) PLA slots. (h) Al₂O₃ slots.

Finally, an artifact was fabricated to evaluate the surface finish achievable by the system. In this case, a piece was designed to provide the same conclusions as the artifact proposed in the standard, with a simpler geometry and less complexity regarding the required printing paths and material to be used (Fig. 11a and b). This allowed the evaluation of the surface finish of an inclined surface, observing the effects of overhangs. In addition to this factor, others can also be considered, such as the diameter of the extruder nozzle, gravity, heat dissipation, and the operating parameter profile used by the machine during modeling. All of these factors cause the texture and finish of the pieces to vary across their surface.

Fig. 11.

Artifact for surface finish evaluation. (a) PLA. (b) Al₂O₃.

Table 9 shows the nominal values and measurements taken for each characteristic to be evaluated in the precision artifacts manufactured with both materials. According to the standard, the values are identified by a letter (A and B) corresponding to the artifact being evaluated and a number (1–9) corresponding to the characteristic being measured on the artifact. Figure 12a represents these characteristics for the linear artifacts, while Fig. 12b shows them for the Z-axis artifact.

Table 9.

PLA and Al₂O₃ precision artifacts measurements.

Precision artifacts
Code	Linear X [mm]			Linear Y [mm]			Z-Axis [mm]
	A			A			B
	Nominal	PLA	Al2O3	Nominal	PLA	Al2O3	Nominal	PLA	Al2O3
1	2.50	2.48	2.56	2.50	2.45	2.54	4.50	4.61	4.42
2	5.00	4.87	5.00	5.00	5.03	5.01	8.25	8.20	8.17
3	5.00	4.88	5.02	5.00	5.04	5.03	13.88	13.81	13.68
4	5.00	4.90	5.01	5.00	5.03	5.10	19.50	19.60	19.28
5	2.50	2.44	2.57	2.50	2.55	2.57	23.25	23.22	22.88
6	12.50	12.50	12.01	12.50	12.57	11.89	25.75	25.81	25.35
7	10.00	10.00	9.63	10.00	10.01	9.59	28.55	28.22	27.76
8	7.50	7.47	7.22	7.50	7.52	7.10	32.00	32.05	31.5
9	5.00	4.90	4.70	5.00	5.03	4.76	37.63	37.68	36.84

Fig. 12.

Diagrams of the characteristics to be evaluated. (a) Linear. (b) Z-axis.

Based on the data, it is observed that with PLA, the largest deviations in the measurements of the linear artifacts occur along the X-axis, with values always lower than the nominal. In contrast, on the Y-axis, the measurements are higher than the nominal (except for measurement A1). Additionally, on the Y-axis, the deviations are less spread out (min. A7 = 0.01 mm and max. A6 = 0.07 mm), with the average measurement deviation being 0.03 mm. On the X-axis, the maximum deviation of 0.13 mm corresponds to measurement A2, and the minimum deviation is found in measurements A6 and A7, where the recorded value coincides with the nominal. In this case, the average deviation is 0.06 mm. Regarding the Z-axis artifact, there is greater dispersion and larger deviations in the measurements, with values both above and below the nominal.

When we look at the data for alumina (Al₂O₃), we first notice that, in both the X and Y linear artifacts, the values corresponding to the ribs (A1 – A5) are equal to or greater than the nominal, while the values for the gaps (A6 – A9) are lower. Moreover, the deviations for these gaps are significantly higher than those for PLA, reaching a deviation of 0.61 mm for measurement A6 in the Y linear artifact. However, it is important to note that both artifacts (linear X and linear Y) have experienced slight twisting in their respective axes, causing the distances between each of the ribs to be affected. In the case of the Z-axis, all values are lower than the nominal, with a maximum deviation observed in measurements A8 and A9.

On the other hand, Table 10 shows the nominal values and the measurements taken for each of the characteristics to be evaluated in the resolution artifacts manufactured with both materials. In this case, and according to the standard, the values are identified with the letters C, D, E, and F depending on the artifact being evaluated, and with a number (1–5 for pins and holes and 1–6 for ribs and slots) corresponding to the characteristic being measured on the artifact. Figure 13 shows these characteristics for the four artifacts manufactured to evaluate resolution.

Table 10.

PLA and Al₂O₃ resolution artifacts measurements.

Resolution artifacts
Artifact	Pins [mm]			Holes [mm]			Ribs [mm]			Slots [mm]
Code	C			D			E			F
	Nominal	PLA	Al 2 O 3	Nominal	PLA	Al 2 O 3	Nominal	PLA	Al 2 O 3	Nominal	PLA	Al 2 O 3
1	4.00	3.81	4.03	4.00	3.88	3.89	6.00	5.91	5.95	6.00	6.04	5.76
2	3.00	2.87	3.02	3.00	2.81	2.94	5.00	4.92	5.00	5.00	5.07	4.62
3	2.00	1.89	-	2.00	1.80	1.92	4.00	3.92	3.95	4.00	4.07	3.67
4	1.00	1.08	-	1.00	-	-	3.00	2.90	3.02	3.00	3.06	2.74
5	0.50	-	-	0.50	-	-	2.00	1.88	2.03	2.00	2.09	2.03
6							1.00	0.96	1.10	1.00	1.07	0.92

Fig. 13.

Diagrams of the characteristics to be evaluated. (a) Pins. (b) Holes. (c) Ribs. (d) Slots.

It can be observed in Table 10 that, for both PLA and Al₂O₃ artifacts, not all characteristics to be measured are present, as it was impossible to reproduce all the pins (see Fig. 10a and b). In the case of Al₂O₃, the process was only able to produce the Ø4 mm and Ø3 mm pins. The Ø2 mm pin detached from the base of the artifact during the sintering process, the Ø1 mm pin dissolved during the chemical debinding stage, and the system was unable to model the Ø0.5 mm pin. As for the holes, the Ø1 mm and Ø0.5 mm holes were not materialized with either material. Specifically, the Ø1 mm hole in the Al₂O₃ artifact closed up due to the reduction the piece underwent during the sintering process. However, unlike the precision artifacts, the deviations in the measurements (C1, C2, C3, D1, D2, and D3) of these two artifacts are smaller with Al₂O₃ than with PLA.

On the other hand, when we look at the ribs and slots, there are also smaller deviations with Al₂O₃ than with PLA. However, the measurements taken in the Al₂O₃ artifact are affected by the twisting process that the pieces experienced in both cases, altering their original geometry. This deformation is more noticeable the smaller the thickness of the rib or the walls that defined the slots.

Nevertheless, while it is important to consider the comments regarding materialization issues with some of the characteristics of the different artifacts, Fig. 14 shows, graphically, the average values of the measurements recorded in both PLA and Al₂O₃.

Fig. 14.

Average values of the deviations measured in the different artifacts.

Finally, regarding the artifact manufactured to assess the surface quality, the piece made of PLA shows no significant defects on the top surface, while on the bottom surface, defects start appearing at 45° and become more pronounced as the angle reaches 60°, with a significant defect at 70° inclination. In the case of Al₂O₃, the ceramic piece exhibits many defects along both the top and bottom surfaces, and, in fact, it did not manage to complete the printing of the piece. The bottom surface shows significant defects from 30° to 60°. Additionally, starting at 65°, the piece loses all definition, and the overhangs of the piece cause the printing to stop, leading to a blockage in the print head.

Conclusions

Considering that in order to observe the true potential of a technology like AM, it is essential to produce parts with an optimized set of parameters²⁵, this article develops a procedure for optimizing parameters for the manufacturing of ceramic parts (Al₂O₃) through a process based on FDM. As an illustration of the technological capability of this process, the article also includes a comparison with a widely studied and known polymeric material, PLA.

The optimization procedure is structured in two phases. The first phase only addresses the printing stage, thus preventing the debinding and sintering processes, which are necessary to obtain the final part, from interfering with the desired objectives. These objectives are achieved in this phase through iterative processes that adjust various parameters, primarily those related to printing speeds, and secondly, geometric and infill aspects. It is important to note that during this stage, no scaling coefficient is needed. The second phase of the procedure involves working with final parts and, therefore, applying operations that remove the binder material and provide consistency and strength to the parts. In this phase, iteration of the manufactured parts is not necessary, and optimized parameters can be obtained through the manufacture of artifacts with different values applied to these parameters. In this way, the time for the procedure is reduced, as the operations that take more time (chemical and thermal) are carried out simultaneously for the different parts.

Although the optimization procedure is designed for ceramic materials, it would be transferable to the fused of metal parts using fused deposition modeling processes. It would also be applicable to other additive manufacturing technologies that, for these materials — ceramic and metallic — require debinding and sintering operations to achieve final geometries and mechanical properties. However, the suitability of the procedure should be assessed beforehand from the perspective of the time these technologies require for the initial printing of parts.

On the other hand, once the profile of optimized parameters has been obtained through the developed procedure, the comparison made with PLA and Al₂O₃ allowed the following conclusions to be drawn:

• The data obtained from both the linear artifacts and the Z-axis artifacts made of Al₂O₃ do not achieve the precision reached in those corresponding to the PLA parts.

• Regarding achievable resolution, the additive manufacturing system using FDM performs better for PLA-manufactured artifacts than for Al₂O₃. PLA allows the creation of smaller features than the filament used in this study, whereas, in the case of alumina, the dimensional reduction that occurs during the sintering operation negatively affects features that both the slicer software and the printer can process but ultimately disappear in the post-process.

• As a result of the post-printing operations (chemical and thermal), all the Al₂O₃ parts have undergone some kind of warping, which has altered the original geometry of the parts.

• Surface finishes are substantially better in the PLA-manufactured artifact, with the geometry of the Al₂O₃ piece not materializing when the part reaches an overhang with an inclination greater than 65°.

Finally, it should be noted that this work does not evaluate the mechanical properties achieved by adopting the optimized parameter procedure in the manufacturing of the specimens. Given that thermal treatments are decisive in this aspect, it would be advisable to consider optimizing the parameters governing these processes (time-temperature) based on the configuration recommended by the manufacturer.

Author contributions

Conceptualization, V.M. and E.C.; methodology, V.M. and L.M.; formal analysis: L.M., A.G. and F.P.; validation: E.C., A.G. and F.P.; writing—original draft preparation, V.M. and L.M.; writing—review and editing, E.C., A.G. and F.P.; visualization, L.M. and A.G.; supervision, V.M., L.M. and E.C.; project administration, V.M. and E.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was part of a project funded by University Institute of Industrial Technology of Asturias (IUTA), grant number SV-23-GIJON-1-01, also by MICIU/AEI/10.13039/501100011033 and, as appropriate, by “ERDF A way of making Europe”, by “ERDF/EU”, by the European Union; grant number PID2021–125992OB-I00.

Data availability

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.