Skip to main content

AI algorithm optimises parameters for laser marking


Researchers at the Max Planck Institute for Informatics have developed an artificial intelligence-based method for determining parameters for laser marking applications.

The work is being done in collaboration with laser manufacturer Trumpf, and has now been funded €800,000 via the German Federal Ministry of Economics and Climate Action (BMWK) EXIST startup programme.

Lasers can be used to colour metallic surfaces with high-resolution, robust marks such as serial numbers, data matrix codes or even artwork. However, such heat-induced colours can be extremely difficult to predict due to the number of parameters in play.

For example, the laser speed, power and pulse duration, along with the type of metal being processed, will all influence the resultant mark colour.

According to the researchers, up until now, in order to obtain the required colour tone of mark, laser engineers have had to manually edit the process parameters in a trial-and-error fashion until the desired result was achieved. 

“This very inaccurate and inefficient approach has hindered industrial adoption of colour laser marking, and also similar laser production methods that depend on optimising the interplay of a large set of laser parameters," said Vahid Babaei head of the Artificial Intelligence aided Design and Manufacturing research group at the Saarbrücken Max Planck Institute for Informatics.

To address this, Babaei and his colleagues have developed an evolutionary algorithm, which when combined with a customised sorting method is able to automate and streamline this ‘parameter exploration’ process.

The exploration algorithm repeats the same steps until it finds the best possible result: First, it starts with the laser making markings based on randomly selected parameters. The algorithm then measures the properties (colour and resolution) of these markings and then, based on these, calculates the next set of parameters for the laser to use to mark. This is where the customised sorting algorithm comes in: it sorts the best-performing process parameters of the previously marked colours according to various metrics, such as resolution or colour saturation. The evolutionary algorithm, in turn, uses these sorting results to generate a new generation of colours that contains the best characteristics of the "parent generation." This iterative process runs until there is no significant improvement in the result and thus the best possible outcome has been found.

According to the researchers, the algorithm is highly scalable to different types of lasers and substrates, and can therefore be used to optimise laser system parameters for a wide range of applications in addition to colour marking. The work will consequently help make a "whole set of laser material processing methods feasible for industrial use", the scientists shared in an announcement of their work. 


Lasers can be used to mark metallic surfaces in high-resolution and in a wide range of colours. (Image: AIDAM)

"With this, we have developed the first computational, automated solution for a whole range of highly complex problems of the laser material processing industry that were until now still solved 'manually' by trial and error," confirmed Babaei. "We strongly believe that colour marking is just the tip of the iceberg and our algorithm can accelerate many different processes dealing with surface activation through lasers, like for example changing the haptics of a material."

To bring their algorithm to market maturity, Babaei’s team is now being funded by the EXIST start-up programme of the German Federal Ministry of Economic Affairs and Climate Action (BMWK), to the sum of €800,000. The researchers have also brought on Trumpf as an industrial partner.  

Paul Stumpf, business development manager at Trumpf, remarked: "We believe this technology can have a major impact on the laser marking industry and beyond. The technology treats the physical laser-material interaction side of the process as a black box and takes advantage of the power of data and artificial intelligence algorithms."

Media Partners