BEAM_IDL – Multiple laser BEAM-shaping monitoring and IDentification boosted by deep-Learning algorithms

Idea and relation to VEDLIoT

Increasing demand for the use of Aluminum (Al) in the manufacture of Electric Vehicles (EV) –especially in battery closures and chassis attachments– requires optimized laser welding schemes. Intrinsic Aluminum-associated welding issues are being overcome with new dynamic and adaptable laser beam approaches, mainly to increase laser-material interaction. Process control for these new laser manufacturing scenarios requires novel in-process monitoring strategies demanding pattern and process event recognition. BEAM-IDL proposes the application of optimized DL algorithms for visible and infrared (IR) laser welding process image recognition. This novel approach will allow to, in the future, link feature extraction with process parameters, opening the gate to novel process optimization.

BEAM-IDL will contribute to the VEDLIoT ecosystem extension with mainly 3 aspects:

  1. The proposed laser manufacturing-based use case could benefit VEDLIoT as the use of a 4-channel dataset could help in testing the robustness, viability and needs for adaptation of the different middleware tools.
  2. The proposed use case contributes to enlarging the set of VEDLIoT use cases, as it demonstrates the application of the technologies in a relevant industry like laser welding where Software defined Laser Beam Shaping (SdLBS) is required, especially focused on light weight manufacturing.
  3. The application field of BEAM-IDL’s solution can be extended to any kind of real-time process monitoring application with high processing requirements and image/data fusion needs.


  1. State of the art of monitoring and laser welding equipment for aluminum materials, definition of process requirements and market analysis.
  2. Definition of industrial constraints, hardware interfaces and software packages.
  3. Hardware and software architecture design for the development of image identification procedures for laser-based manufacturing processes with embedded DL algorithms.
  4. Development of robust machine vision, AI and DL solutions, data processing and welding parameters.
  5. Integration of proof-of-concept at operational prototype manufacturing system.
  6. Definition of the final specifications of the developed technology and business model.
  7. Industrial component manufacturing case in end-user’s operational environment. Several trials will be performed at LORTEK’s facilities for industrial performance evaluation.


BEAM-IDL pretends to develop an infrared (IR) and visible image recognition tool for different laser beam shape identification by means of the application of a dedicated classification and segmentation Deep-Learning (DL) algorithm. The use of different laser beam shapes is a state-of-the-art tendency for the optimization of welding processes of materials with high conductivity and fluidity like Aluminum (Al). This start approach is especially interesting for the multiple Al-based Electric Vehicle (EV) parts and unions and in general for applications in light weight manufacturing. Various beam geometries increase process parameter configuration complexity, thus the application of automatic procedures to monitor and control the process is the path to follow. For that, the first step will necessarily be the application of in-process monitoring strategies, which will allow determining what is arising during the welding process.

BEAM-IDL proposes as an initial stage for this fully automatic laser processing strategy, the use of process cameras for laser beam shape and energy distribution recognition. Laser welding is an intrinsically and extremely fast procedure, so the application of complex automatic image recognition algorithms is not trivial. Ad-hoc and accelerated DL-based algorithms will allow recognizing in-process the shapes of the laser beams being applied.

Expected Impact

The expected short-term impact is the validation of the proposed approach for laser welding processes, heavily focusing on machine vision and artificial intelligence algorithms and their implementation in embedded computing devices for quality assurance and process control applications. Based on the actual set-up, an embedded heterogeneous multiprocessing system is planned to use the VEDLIoT framework to better exploit the edge-computing capacities of the laser solution. The objective of EXOM Engineering in the project is the further development of laser technology (both the method and the system/equipment) which allows a higher digitalization of the process, supporting the engineers using the solution during the definition of process parameters windows as well as identifying quality issues in manufacturing lines.

The exploitation of the project results will turn around the commercialization of products (algorithms, laser control card and scanning optics) which shall support the robustness and adoption of dynamic beam shaping technologies for aluminum welding in different industrial applications. Shortening the development and manufacturing cycle, reducing the consumption of time and material resources, and facilitating the use of aluminum laser welding for a wider range of applications also in other industries like aeronautics or energy production. After the project runtime, a strong commitment to exploit and commercialize the resulting product is ensured thanks to the strong alignment of the proposed solution with the actual offer of EXOM Engineering in the field of high-power laser applications.

Further info/links