Accurate blind welds: lasers meet AI
Expertise in laser welding - the BIAS institute in Bremen, Photo: ECOMAT/Raveling
Using lasers to weld nickel-based materials in industrial processes is a real challenge. The Institut für angewandte Strahltechnik GmbH (institute for applied (laser) beam technology, known as "BIAS"), in Bremen, has come up with an innovative automated manufacturing process that makes it possible to weld together nickel-based components safely and securely, even in hidden T-joints.
In many sectors, whether it be power-generation, transport or mechanical engineering, the transition to sustainable, renewable sources of energy will require new technologies. Manufacturing processes that are modern and more efficient are part and parcel of these new technologies. The ability to use lasers to create welded seams in nickel is a critical application for the future.
The BIAS Bremer Institut für angewandte Strahltechnik GmbH is working closely with a number of industrial partners to find solutions to this challenge. The reason: the automated welding processes previously used in industry were slow and, to some extent, not precise enough. There were also very few scientific descriptions and approaches for the use of lasers to weld pure nickel components, until now.
The problem:
In industrial manufacturing processes, bars have to be attached to a nickel sheet. The process itself requires the bars to lie underneath the sheet, but they have to be welded from above. These T-shaped welded seams must be created without being seen or, in other words, "blind".
Two nickel sheets, welded into a T-shape. The T-joint is welded together from above, i.e. "blind"., Photo: ECOMAT/Raveling
If the bar is less than 1.5 mm wide and more than one meter long, these thin metal sheets can move fractionally during the production process. As a result, the weld created by the pre-programmed laser beam deviates very slightly from the actual edges of the bar, resulting in flaws in the welded seam.
Research project: AI-based detection of deviations in welded seams in a T-joint
This problem – referred to as welding concealed T-joints in technical jargon – is the subject of a research project at BIAS.
In the last three years, the team lead by BIAS Project Leader, Marcel Möbus, has found a solution which is based on one of this Bremen Institute's key areas of expertise: laser welding using sensors for real-time monitoring.
The team's approach uses what is known as an OCT measuring beam. OCT stands for "optical coherence tomography". The beam is used for OCT measurement processes. This measurement method uses light to discover the depth of the welding process. It is similar to ultrasound, but it is light-based, and much more precise (resolution in the micrometre range).
The weak but precise OCT beam is aligned on the same path as the laser beam. During the welding process, the beam measures the depth of the "keyhole", a very small vapour capillary, which the laser creates in the material.
If the bar position shifts during welding, the OCT measurement tool identifies this on the basis of the depth information. "To handle this, we have developed an event-dependent position-adjustment system which detects the edges before the bar is even left and adjusts the position of the laser beam during welding, so that it is always placed on the bar", said Möbus.
Top: Extreme example in which the actual linear feed direction is from left to right (seen from above) and the laser beam path has been corrected by the position-adjustment system. Below: Covered web plate, marked in red, and real-life result of correcting the laser beam path when the feed direction is linear. Photo: BIAS
Control algorithm identifies edges and prevents defects in the weld seam
What makes the new technology special: The control system intervenes before an error even occurs. The system identifies that the laser is getting close to an edge, before it even does so, and adjusts the welding position accordingly. This effectively prevents flaws being created in the welded seam.
The team achieved this by using a precise evaluation of the OCT measured values together with a sophisticated control algorithm. As Möbus explained, "We then also used this data to train an AI system, which was then implemented in the control system. This increased the accuracy of the control system even more, because the AI can identify when an adjustment is needed up to 200 milliseconds earlier."
Working towards achieving validation for the process
As Möbus said, the technology used here is based on standard industrial products, to which the team has added the appropriate evaluation and control system algorithms. In addition to the OCT measuring technology, acoustic or temperature sensors are also used to detect welding flaws, in real time. The team uses an FPGA controller (a hard-coded micro controller), which could also be used at an industrial scale, to integrate the AI. Up to now, the setup has been working well in the laboratory. The next task for the project team is to get the system validated for use in industrial applications and then conclude the project by the end of 2025.
The BIAS project team with Timon Ahlers, Ronald Pordzig and Marcel Möbus, Photo: ECOMAT/Raveling
First scientific publications about using lasers to weld pure nickel
In addition to exploring the use of OCT and neural networks to adjust the positioning of welding lasers, the BIAS is also breaking new ground by investigating the use of lasers to weld pure nickel. As Möbus explained, "There have been very few scientific publications about using lasers to weld pure nickel, until now. We are the first team to apply a scientific approach to the challenging techniques involved in welding concealed T-joints, with the involvement of AI, and publish the outcomes. It's actually something we're very proud of."
According to Möbus, who is the Project Leader, was one of the greatest challenges was to develop these technologies to handle ever faster welding speeds, with minimal tolerance for errors. As he went on to say, "Thanks to innovative laser beam sources and the way we developed our processes, we can now achieve welding speeds of over 20 m/min. This is more than twice the speed that was possible previously. It represents a huge step forward for the technology, whilst at the same time improving the reliability of the process."
All this means that the Institute, which is also a partner with ECOMAT, Bremen's Research and Technology Centre, is clearly demonstrating the expertise in materials and process development that is to be found in Bremen, a city in which many other institutes and industrial partners are working together to find new solutions for the world of technology.
Publications:
Krämer, A; Pordzik, R; Mattulat, T: Influence of the airborne sound sensor position on the detectability of acoustic emissions during deep penetration laser welding; conference transcript 13. Mittweidaer Laser Conference, Scientific Reports, Journal of the University of Applied Sciences Mittweida, No. 3 (2023) 57-62 (ISSN 1437-7624)
Ahlers, T; Pordzik, R; Mattulat, T: Approaches for automatic detection of mispositioning during laser welding in hidden T-joints using optical coherence tomography; conference transcript 13. Mittweidaer Laser Conference, Scientific Reports, Journal of the University of Applied Sciences Mittweida, No. 3 (2023) ISSN 1437-7624
Pordzik, R; Seefeld, T: Identifying the direction of weld path deviations in laser deep penetration welding of hidden T-joints by means of OCT; Journal of Laser Applications (ICALEO 2024) paper no. 0877-1306-000092
Möbus, M; Pordzik, R; Krämer, A; Mattulat, T: Process comparison of laser deep penetration welding in pure nickel using blue and infrared wavelengths; Welding in the World (2024) https://doi.org/10.1007/s40194-024-01713-9
Möbus, M; Pordzik, R; Seefeld, T: Influence of intensity distributions on the process dynamics during laser deep penetration welding of pure nickel with a flexible ring mode laser source; J. Laser Appl. 36, 042073 (2024) https://doi.org/10.2351/7.0001529
Mattulat, T: Understanding the coaxial optical coherence tomography signal during the laser welding of hidden T-joints; Journal of Laser Applications 36 (2024) 012003, doi: 10.2351/7.0001157
Krämer, A; Henze, I; Pordzik, R; Radel, T: Inline detection of process anomalies during laser deep penetration welding of hidden T-joints; 13th CIRP Conference on Photonic Technologies (LANE 2024), eds.: M. Schmidt, C.B. Arnold, K. Wudy. Elsevier B.V. [Volume 124] (2024) 526 - 529, https://doi.org/10.1016/j.procir.2024.08.167
Information about the funding project: https://www.enargus.de/detail/?id=7518736
The new technologies needed to combat the changing climate will require new materials such as materials derived from nickel. The Bremen-based BIAS developed the first automated laser welding process in the world that can do something quite unique.