The research, published in Frontiers in Manufacturing Technology, targets extrusion-based 3D printing, known in the standards as material extrusion (MEX) and in its commercial forms as fused filament fabrication or fused granulate fabrication. The technique is cheap, adaptable, and cleanroom-compatible because it needs neither powders nor photoreactive liquids, which makes it a favorite for biomedical polymers such as polycaprolactone, polylactide, and PLGA. But according to the authors, no holistically automated production infrastructure for this process has been established anywhere in academic production technology, and industrial lines still lean heavily on manual steps for material handling and quality control. That gap is the central obstacle to scaling personalized medical devices.

To ground the concept in a concrete use case, the team defined a narrow but demanding product: cell-free, porous tissue-reconstruction scaffolds made of polycaprolactone granulate, printed by fused granulate fabrication. The scaffolds weigh between 10 and 100 grams and measure no more than 100 by 100 by 50 millimeters. Polycaprolactone has a long clinical track record in sutures and implants, and granulate feeding avoids the extra step of producing filament. The choice of a cell-free product matters for process control, because the critical quality attributes of such scaffolds are geometric and mechanical rather than biological: dimensional accuracy, porosity, and the linkage between printed layers.

From a systematic literature review that filtered more than 1,800 papers down to 30 relevant studies, the researchers mapped which process parameters actually matter. Deposition speed, extrusion temperature, and extrusion pressure or rate emerged as the most frequently studied variables. For their use case, the team classified just two as critical process parameters requiring continuous monitoring: the extrusion rate and the extrusion temperature. Other variables, such as nozzle diameter, platform temperature, and cool-down time, are fixed once and validated; ambient temperature and humidity are handled by the cleanroom itself. On the quality side, dimensional accuracy, porosity, and layer linkage were designated critical quality attributes, while biocompatibility is determined by the material choice and biodegradability cannot be measured during production at all.

With requirements in hand—regulatory compliance for cleanroom production under Good Manufacturing Practice, autonomy during production, flexibility and scalability for batch-size-one manufacturing, and integrability with existing equipment—the team used the morphological box method to collect and compare automation solutions for every function in the process chain. A value-benefit analysis then selected the best-rated approach for each function. The result is a coherent architecture in which a magnetic planar drive plays two roles at once: it is the intralogistics transport system for the whole line, and it forms the horizontal X and Y axes of the 3D printer itself.

The planar drive works through magnetic levitation. Movers carrying arrays of permanent magnets float frictionlessly above a stator surface and can be positioned in the plane with a tolerance of 50 micrometers. Each mover carries an exchangeable, sterile, disposable substrate plate on which the scaffold is printed. Because there is no mechanical contact, particle abrasion is minimized—a decisive advantage in a cleanroom, where every shed particle is a potential contaminant. The concept for using a planar drive in 3D printing was first patented in 2022 and validated in the German BMBF-funded research project BellaFactum; the new paper now presents it as the backbone of a complete production system.

The decoupling of print bed and extruder is what unlocks parallelism. While one printed layer cools on one mover, another mover can slide under the extruder and receive the next print job. Movers can also detour to a camera station after every five printed layers, where high-resolution reflected-light images serve three purposes at once: detecting over- or under-extrusion by measuring filament position and thickness, judging layer linkage from the shading at filament intersections, and calculating porosity from the distances between filament lines. If a print fails a threshold, the mover carries the defective substrate to a disposal storage location, reloads, and the same job restarts automatically—no human intervention required.

Final inspection is handled by laser distance measurement, which records a point cloud of the finished scaffold’s outer dimensions and compares it against the planned CAD geometry to catch deformations such as warping. Material logistics are equally automated: a cooled central reservoir stores polycaprolactone granulate under the low-humidity, low-temperature conditions the hydrophilic polymer demands, a gravimetric dosing unit meters precise amounts into transport containers, and an autonomous mobile robot ferries them to the extruder hopper, triggered by an ultrasonic sensor when material runs low. An overarching control software called COPE, developed at Fraunhofer IPT, orchestrates the entire system in a manner the authors describe as compliant with FDA CFR Part 11 requirements for electronic records.

The concept was physically realized as a demonstrator the team calls DIAMS, short for Dynamically Interconnected Additive Manufacturing System. It combines a Beckhoff XPlanar planar drive of twelve tiles, each 240 millimeters square, levitating four movers of 155 by 155 millimeters at a height of 2 millimeters, a screw-based granulate extruder on a vertical linear axis, and a microscope camera on its own axis. A multi-agent path-finding algorithm written in Python, based on model-predictive control with conflict-based search, coordinates the movers collision-free, while a TwinCAT function reads standard G-code and translates it into synchronized motion of the planar and linear axes.

Proof-of-principle tests printed a cuboid scaffold of 50 by 50 by 20 millimeters with a layer height of 200 micrometers five times, using a 400-micrometer nozzle at 130 degrees Celsius and a deposition speed of 15 millimeters per second. The measured base dimensions averaged 49.8 millimeters with a deviation of plus or minus 0.2 millimeters, and the height averaged 19.9 millimeters with a deviation of plus or minus 0.1 millimeters—a maximum shape deviation of 0.4 millimeters from the CAD model. The team observed increased stringing, indicating that print parameters still need optimization, but the general printability of the levitating system was demonstrated. The authors are candid about the limits: the concept covers only polycaprolactone scaffolds in a defined size range, validation used uniform cuboids rather than complex personalized geometries, and economic evaluations remain qualitative. Their next steps include discrete-event simulation of the production line to identify the throughput scenarios where this levitating factory genuinely outperforms conventional printer farms, and deeper research into adaptive, resource-efficient path planning for multi-agent manufacturing systems. If those studies succeed, the frictionless factory floor may move from laboratory demonstrator to the production halls where personalized medicine is actually made.

Subject of Research: Automated extrusion-based 3D printing of personalized medical scaffolds using magnetic planar drive technology

Article Title: Development of an end-to-end automated production concept for extrusion-based additive manufacturing of personalized medical scaffolds

Article References: Janning, K., König, S., Herbst, L., Nießing, B., & Schmitt, R. H. (2025). Development of an end-to-end automated production concept for extrusion-based additive manufacturing of personalized medical scaffolds. Frontiers in Manufacturing Technology, 5, Article 1572842. https://doi.org/10.3389/fmtec.2025.1572842

Image Credits: AI Generated

DOI: 10.3389/fmtec.2025.1572842

Keywords: 3D printing, additive manufacturing, material extrusion, medical scaffolds, personalized medicine, magnetic levitation, planar drive, polycaprolactone, cleanroom production, quality control, automation, biomedical engineering

The study, led by Virginia Uralde and Fernando Veiga with colleagues Alfredo Suarez and Tomas Ballesteros, addresses a problem that has grown acute as metal 3D printing has matured. Unlike conventional subtractive manufacturing, in which milling, turning, grinding and drilling remove material from a solid workpiece, additive manufacturing builds parts layer by layer, unlocking geometries that would otherwise be impossible: hollow sections, intricate internal lattices and topology-optimised structures that slash weight. That design freedom, however, comes at the price of complexity. With each process offering a different balance of feedstock, energy source, deposition rate, accuracy, build volume and post-processing burden, the question of which technology to use no longer has an obvious answer.

The researchers’ first move is deceptively simple: rule out what cannot work. Before any ranking takes place, three design-phase criteria act as exclusion filters: the geometric complexity of the part, the material it must be made from, and its overall dimensions. A technology that cannot print a particular alloy, or whose build volume cannot accommodate a large component, is eliminated before the analysis begins. This pre-screening stage, the authors argue, is the study’s key novelty, because it tailors the subsequent comparison to the specific application rather than producing a generic, one-size-fits-all ranking. The underlying hypothesis is blunt: there is no universally optimal additive manufacturing technology for all materials, sizes and shapes.

Once the field has been narrowed, the framework deploys the Analytical Hierarchical Process, or AHP, to rank the surviving candidates. AHP decomposes a decision into a hierarchy: the overall objective at the top, criteria in the middle, and alternatives at the bottom. Decision-makers compare elements in pairs using Saaty’s famous 1-to-9 scale, where 1 means two items are equally preferable and 9 means one is extremely preferable to the other. These subjective judgments are converted into numerical weights through eigenvector calculations, and a consistency ratio, which must fall below 0.1, verifies that the comparisons are logically coherent rather than arbitrary. The method’s strength, the authors note, is precisely its ability to transform qualitative expert judgment into transparent mathematics, a valuable asset in a field where quantitative data can be scarce, unreliable or quickly outdated.

The team considered five technological alternatives spanning the main families of metal additive manufacturing. Powder Bed Fusion, or PBF, which fuses fine metal powders layer by layer, offers high geometric complexity, tight tolerances and smooth surface finishes, with typical layer thicknesses of just 0.02 to 0.06 millimetres, but limited productivity and build size. Material Extrusion, or MEX, uses metal-filled filaments followed by debinding and sintering, offering accessible equipment and low costs at the expense of mechanical performance. Directed Energy Deposition, or DED, which deposits material directly into a melt pool, was subdivided by energy source into laser-beam, electron-beam and arc-based variants, the last of which stands out for very high deposition rates of roughly 10 kilograms per hour and suitability for large components and repair work.

To weight the criteria, the researchers surveyed 260 respondents: 130 additive manufacturing experts, 70 designers and 60 customers of metal parts. The questionnaire posed 17 paired comparisons on a 1-to-9 scale, asking participants to trade off criteria against one another. Nine sub-criteria were organised into two phases of the additive workflow. The print phase covered machine price, manufacturing cost, productivity, reliability and part finish after printing; the final phase covered final part properties, post-processing complexity, process sustainability and user preferences. The results were striking: the print phase received a weight of 0.683, more than double the final phase’s 0.317, and reliability emerged as the single most important criterion overall, with a global weight of 0.304. Final part properties followed at 0.186, then manufacturing cost at 0.147, while sustainability, machine price and user preferences ranked lowest.

Interesting differences emerged between respondent profiles. Experts and designers both weighted the print phase heavily, drawn to factors such as productivity and process reliability, with the experts placing particular emphasis on reliability. Customers, by contrast, cared more about the finishing phase, because the final properties of the delivered part matter most to them. For criteria that could not be answered from the literature, such as sustainability perceptions and user preferences, a second survey of purely technical interviewees supplied the paired comparisons needed to score each alternative.

When all five technologies were evaluated against the full set of criteria, arc-based Directed Energy Deposition, or DED-Arc, came out on top with a final score of 0.249, narrowly ahead of Powder Bed Fusion at 0.231, with DED-EB, DED-LB and MEX trailing at 0.179, 0.172 and 0.170 respectively. The result tracks the underlying physics and economics. DED-Arc machines cost roughly 110,000 euros, the lowest of the group, and its arc-welding deposition achieves productivity around 10 kilograms per hour, roughly fifty to one hundred times that of PBF, which manages only 0.1 to 0.2 kilograms per hour. PBF, however, dominates on quality: surface roughness of Ra 6 to 12 micrometres, tensile strengths of 800 to 1100 megapascals in Ti6Al4V, and the highest survey scores for reliability and as-printed finish, reflecting the fine powder particles that yield smoother surfaces. Its weakness is cost, with the highest manufacturing cost of any alternative, driven by material waste and energy consumption.

The framework’s real power shows itself when the pre-selection filter is applied. In a worked example, the researchers considered a part of medium-high geometric complexity, medium size, made from aluminium. The filter excluded DED-EB and DED-Arc, leaving PBF, DED-LB and MEX as viable candidates. Re-running the AHP with only these three alternatives reshuffled the ranking dramatically: PBF surged to first place with a score of 0.415, followed by MEX at 0.306 and DED-LB at 0.279. The lesson is that the ‘best’ technology is not an intrinsic property of a process but a function of the part, the material and the constraints, and that a ranking computed without feasibility screening can be actively misleading.

The authors are candid about the method’s limits. The criterion weights reflect the preferences of the surveyed group rather than universal priorities, and different expert populations could produce different rankings. Technical characteristics were drawn from literature and expert assessment and vary with machine, material and process parameters, so the output should be read as decision support rather than a definitive verdict. AHP also assumes criteria are independent and uses deterministic comparisons, leaving uncertainty and interactions unmodelled; the researchers suggest fuzzy or probabilistic extensions as future work. And because metal additive manufacturing is evolving rapidly, the technical data underpinning the framework need periodic updating, though the two-stage structure itself can absorb new alternatives, screening conditions and weights without modification. For manufacturers weighing six-figure machine investments against part quality, throughput and sustainability, the study offers something rarer than a winner: a defensible, customisable way to find the right tool for each job.

Subject of Research: Multi-criteria decision-making for selecting optimal metal additive manufacturing technologies using the Analytical Hierarchical Process

Article Title: Method for selecting the optimal technology in metal additive manufacturing using an analytical hierarchical process

Article References: Uralde, V., Veiga, F., Suarez, A., & Ballesteros, T. (2026). Method for selecting the optimal technology in metal additive manufacturing using an analytical hierarchical process. Heliyon, 12(15), Article e45519. https://doi.org/10.1016/j.heliyon.2026.e45519

Image Credits: AI Generated

DOI: 10.1016/j.heliyon.2026.e45519

Keywords: metal additive manufacturing, analytical hierarchical process, multi-criteria decision making, powder bed fusion, directed energy deposition, material extrusion, DED-Arc, reliability, manufacturing cost, productivity, technology selection, 3D printing