Empa, SUPSI and Bystronic Develop Self-Guiding Laser Cutting Method Using Cameras and Microphones

To assess the cut quality itself, the laser is able to use cameras and microphones. Image: Roland Richter, Empa

(IN BRIEF) Empa, SUPSI and Swiss machine manufacturer Bystronic have developed a self-guiding laser cutting method that uses cameras, microphones and machine learning to assess cut quality and adjust machine settings automatically. The project, supported by Innosuisse, addresses the challenge that laser cutting machines often need time-consuming test series and repeated recalibration for different alloys and material thicknesses, especially when cutting thick metal workpieces. Empa is developing acoustic methods that use sound recordings from microphones inside the machine, while SUPSI is working on optical camera-based evaluation; researchers found that acoustic assessment can evaluate cut quality almost as well as cameras, while requiring cheaper equipment. The team trained models using metal thicknesses of 6 mm, 10 mm, 15 mm and 25 mm, focusing on cut-edge roughness and burr formation, and is now linking acoustic and optical models to a central control system that could allow machines to self-correct instantly. The two-year project is expected to conclude in autumn 2026, and the resulting system could be retrofitted to existing machines using a PC, microphones and cameras.

(PRESS RELEASE) DÜBENDORF, 14-Aug-2026 — /EuropaWire/ — Researchers at Empa, the University of Applied Sciences and Arts of Southern Switzerland, known as SUPSI, and Swiss machine manufacturer Bystronic have developed a method that enables laser cutting machines to assess cut quality using cameras and microphones and adjust their settings autonomously when needed.

To train their models, the researchers conducted a series of tests using different material thicknesses. From left: Bystronic engineer Konstantinos Skovolas, Bystronic Global manager innovation Andreas Lüdi and Empa researcher Naval Mehta. Image: Roland Richter, Empa

The work is being carried out as part of an Innosuisse project aimed at improving laser cutting of metals, particularly when processing materials with different thicknesses and properties.

Laser cutting is a powerful and flexible process that allows metals to be cut quickly and precisely, including workpieces with complex shapes and small batch sizes.

Using near-infrared wavelengths and high power, laser beam cutting can melt metal in fractions of a second and can even penetrate steel plates up to 2.5 centimetres thick.

The process is increasingly in demand in industrial applications, including the automotive sector.

However, laser cutting can be volatile, especially when working with thick materials.

To achieve the required cut quality, machines often need to be readjusted for each alloy and material thickness.

Manufacturers may need to run targeted test series to configure a laser cutting machine correctly.

Even small changes in the material being cut can require the process to be repeated, making adjustment time-consuming.

The project partners aim to address this challenge by giving the laser cutting machine “eyes” and “ears”.

Existing machines could be easily retrofitted with the new system. Image: Roland Richter, Empa

SUPSI is working on camera-based evaluation of cut quality, while Empa is developing acoustic methods.

Roland Richter, an Empa researcher in the Multifunctional Materials and Interfaces laboratory and a member of Elia Iseli’s team, said the goal is to allow the machine to continuously assess the cut and adjust itself.

Assessing cut quality through sound

Although laser cutting is carried out with light, the process produces significant sound.

Where the laser hits steel, molten and vaporised metal and plasma are created.

A process gas also flows continuously over the cut to remove ablated material.

Empa researchers are using data from numerous experiments and machine learning models to identify signals about cut quality within this complex background noise.

Richter said the team was able to show that cut quality can be assessed acoustically almost as effectively as with cameras, while the required equipment is significantly cheaper.

Cutting quality is mainly determined by two factors: the roughness of the cut edge and the formation of burrs.

A smoother cut edge indicates better quality.

Burrs form when material blown out of the cut settles on the back of the workpiece.

Thick workpieces are especially prone to burr formation because they require high laser power during cutting.

Reducing burrs means less post-processing and material waste, which can save both time and costs.

To train their models, Empa researchers conducted tests using four different material thicknesses: 6 millimetres, 10 millimetres, 15 millimetres and 25 millimetres.

In each test, the researchers assessed cut quality manually while simultaneously recording ambient noise.

The team used nine microphones placed at different positions inside the machine.

The resulting models were then refined using smaller test series involving different alloys.

At the same time, SUPSI followed a similar approach using optical cameras.

Bystronic engineer Konstantinos Skovolas, Bystronic Global Manager Innovation Andreas Lüdi and Empa researcher Naval Mehta were involved in the testing work.

Toward instant self-correction

The two-year project is scheduled to conclude in autumn 2026.

In the final phase, Empa researchers are working to connect the acoustic and optical models to a central control system that acts as the machine’s “brain”.

This system is designed to allow the laser cutting machine to evaluate cut quality and immediately optimise its own settings.

Richter said the model could eliminate the need for time-consuming test series by enabling the laser to automatically adjust to the best parameters for each material.

The system could also be retrofitted to existing machines.

According to Empa, the retrofit would require a PC with the control algorithm, along with several microphones and cameras.

The project partners said the collaboration benefits both industry and research.

Richter noted that the project has allowed researchers to test and further develop models for simulating and controlling laser cutting processes using powerful industrial lasers that would otherwise not be available to them.

Through the project, Empa, SUPSI and Bystronic are advancing a self-guiding laser cutting approach that combines acoustics, optical monitoring and machine learning to improve process efficiency, reduce setup times and support more reliable industrial metal cutting.

Media Contacts:

Dr. Roland Richter
Empa, Multifunctional Materials and Interfaces
Phone +41 58 765 63 04
roland.richter@empa.ch

Dr. Elia Iseli
Empa, Multifunctional Materials and Interfaces
Phone +41 58 765 63 28
elia.iseli@empa.ch

Dr. Andreas Lüdi
Bystronic Laser AG, Global R&D Innovation
Phone +41 62 956 36 46
andreas.luedi@bystronic.com

Anna Ettlin
Communications
Phone +41 58 765 47 33
redaktion@empa.ch

SOURCE: Empa