Instead of time-consuming machining, large-format 3D printing is taking over, a process that monitors and continuously adjusts itself during production. This is the foundation of the Czech-German AI4MultiDirectAM project, whose ambition is to transform the production of molds for composite parts into a faster, more precise, and more cost-effective process.
Molds for composite parts, such as car bodies or technical covers, are still often manufactured using outdated methods. This process works, but it is slow, expensive, and material-intensive. It all starts with a block of material that must be milled, repaired, and repeatedly modified. Weeks pass before the final mold is created. Costs rise, and so does the amount of waste.
When AI Takes Over 3D Printing
In the AI4MultiDirectAM project, mold production works completely differently. Instead of slow machining, large-format 3D printing controlled by artificial intelligence takes over. During production, the system continuously monitors what is happening, evaluates changes, and immediately adapts to them.
The project is led by Entry Engineering, CXI TUL, Fraunhofer IWU, RTT Automation, and LAKOWA. Their shared goal is to transform large-format 3D printing into a practical and reliable tool for real-world industrial production. The project is primarily aimed at the production of large-scale molds for composite parts in the automotive and aerospace industries, where it could replace costly machining of composite wood or aluminum.
The fundamental difference from conventional 3D printing lies in the very principle of production. It does not simply print layer by layer in an upward direction. The technology relies on six-axis robots capable of printing in multiple directions—along curves, at an angle, and on curved surfaces. This opens the door to producing more complex shapes, greater design freedom, and less reliance on support structures. But just as production gains greater freedom, the demands on quality control also increase. And that’s where artificial intelligence comes into play.
Eyes and Senses in Robotic Printing Inspection
For the system to recognize that something is starting to go wrong during printing, it must first see what is actually happening during production. That is why the project is building a literal network of eyes and senses that monitors the entire process. It monitors the extruder’s behavior, the robot’s movement, the nozzle temperature, and how the part itself is formed layer by layer. Thermal cameras and LiDAR sensors positioned around the printing workstation assist with this.
But simply collecting data isn’t enough. It must be properly synchronized in time, stored, and prepared so that artificial intelligence can learn from it. The project is therefore also building a smart data infrastructure. This infrastructure can handle both standard data and complex data, such as videos, thermal recordings, or 3D point clouds from LiDAR. In other words: it’s not enough to have a smart machine. You also need smart memory.
This will make it possible to create a digital twin of the entire process. This will be a detailed production record showing what happened during printing, where an error occurred, and how the system responded to it.
So, in other words, data is used to monitor the printing process. If something is off—such as layer width, temperature, or deformation—the AI will suggest a change. And that change will be immediately applied to the next layer. No waiting for an operator, no overheating, no re-prints.
Entry will put 3D printing to the test
The first year of the project revealed where large-format 3D printing most frequently encounters issues. The team identified three key problems: uneven heat distribution within the printed object, warping during cooling, and dimensional inaccuracies caused by robot movement or material properties. In other words: the part may warp, layers may not bond properly, and the final shape may deviate from the design. In development, this is a complication; in industry, it is often the reason a product fails. This is precisely where AI-assisted inspection is intended to help.
At Entry in Ohrazenice near Turnov, a new LSAM workstation is being set up with a KUKA Quantec KR 250 robot, a CEAD S25 extrusion head with a capacity of 25 kilograms per hour, and a newly designed print chamber. For the company, this is a practical step forward. “It’s a great opportunity to collaborate on the development of a very interesting technology that is directly applicable to our production of tools for manufacturing composite parts,” says Michal Amrich. However, as he adds, the potential applications are much broader, ranging from printing large-format structural parts and mockup models to large-format visual and decorative elements.
Next step: sensors, data, and online monitoring
The entire system is being designed to handle real-time process interventions in the future based on sensor data and AI recommendations. Work continues on sensor integration, generating data for AI training, and optimizing data flows. The first planned deliverable is a functional prototype of a sensor head for online monitoring of large-format 3D printing.
It is precisely online monitoring and software corrections that are intended to help stabilize the operation of LSAM technology. As Michal Amrich explains, temperature control and print speed are particularly challenging. “The mass of material you create during printing is significantly greater than with conventional FDM prints. It cools much more slowly, so you can’t layer the print uncontrollably. The print could collapse or become misshapen.” And this is where AI comes into play. If the shape is not maintained, the program adjusts subsequent layers based on sensor data and its evaluation. If surface defects or stringing appear, the AI adjusts the print parameters, such as ambient temperature, melt temperature, or print speed. According to Amrich, this reduces the risks of producing imperfect prints, economic losses, and time delays.
Where large-format printing makes sense
However, he notes that large-format printing isn’t for everyone. “If you’re mass-producing large plastic parts, you’ll likely continue to manufacture them using injection molding, thermoforming, or another production technology,” he says. It’s particularly suitable for prototyping or single-unit and small-batch production. In such cases, LSAM can save both the material and time required to produce prototype tooling or allow the prototype to be printed and finished directly.
According to Michal Amrich, there are many potential applications. At the same time, however, he points out that engineers, developers, and designers often do not yet take this production method into account when designing their products. “They are unaware of the technological limitations and do not know how to properly design parts for LSAM.” This is one reason why Entry is open to sharing this know-how and welcomes interested parties.
Award for Czech-German Cooperation
Moreover, in its very first year, the project won the main prize from the Czech-German Chamber of Commerce and Industry for cross-border cooperation. And rightly so, because AI4MultiDirectAM promises manufacturing that monitors itself during the process. And when it detects that something is wrong, it intervenes before a defect occurs.
