The University of Bath has filed a patent for automated material calibration on FFF systems.
The patent document describes a 3D printing system that could automatically identify an unknown material, predict its properties and select suitable printing parameters. That could remove a mountain of trial and error work when working with a new material.
The application, WO2025068712A1, is simply titled “3D Printing”.
The document is unusually broad, covering several related ideas involving machine learning, spectroscopy, automated extrusion measurement and even multimodal printing. However, the most interesting portion is the attempt to make the 3D printer determine what material it has been given and work out how to print it.
Let The Printer Figure It Out
The proposed workflow begins by obtaining a vibrational spectrum of the material, potentially using infrared spectroscopy. That spectral data is fed into a machine learning model trained to predict at least the identity of the material, and potentially properties such as glass transition temperature, melting temperature or density.
A second step then uses those predicted characteristics to determine appropriate process parameters.
For FFF, those could include extrusion temperature, bed temperature, feed rate, printing speed, cooling fan speed, layer height, raster angle and other settings. The patent specifically contemplates automatically adjusting parameters when the material loaded does not match what the machine expected.
In other words, instead of selecting “PLA” from a menu and assuming the spool really is PLA with known characteristics, the printer could examine the actual material and make its own assessment. Incredible!
That becomes even more interesting when you must deal with recycled materials, experimental blends, unlabeled spools or materials where the chemical formulation changes from batch to batch.
The concept could also help manufacturers developing open material systems. One of the biggest complications of accepting arbitrary materials is creating dependable profiles for them. To do so, many major manufacturers have set up complex and growing libraries of print profiles for operators. Automated characterization with this approach could reduce the need for such libraries.
Closing The Loop
The application goes further by describing automated extrusion width measurement.
A camera observes extruded material during printing. Machine vision and instance segmentation identify the extrusion, and the system compares its measured width against an expected value. If the difference exceeds a tolerance, printing parameters can be adjusted automatically.
That is potentially quite useful because extrusion width is a direct result of several interacting factors: material flow, temperature, speed, nozzle geometry and material behavior. Measuring the actual extrusion gives the controller information that commanded extrusion values alone cannot provide.
The patent even discusses monitoring consecutive extrusions rather than relying on occasional measurements, suggesting a genuinely closed loop process rather than a simple calibration routine.
There are other concepts packed into the filing, including a printer capable of switching between FFF and electrohydrodynamic printing modes, special electrically insulated printing beds and nozzle arrangements.
This sounds very promising, but it remains a patent application rather than a product announcement. The University of Bath is not a 3D printer manufacturer, so we might expect them to partner with a major 3D printer manufacturer that can actually implement these concepts.
Spectroscopy hardware and machine vision would also add cost and complexity to a printer.
Desktop and industrial 3D printer manufacturers have already automated bed leveling, calibration and process monitoring. Automatically identifying materials and generating appropriate process settings could be another step toward machines that require far less material specific knowledge from 3D printer operators.
I wish this feature had been implemented long ago, as it would have save me from quite a number of operational mistakes.
Via Espacenet
