Machine Learning Enhances Precision and Efficiency in Metal Laser Processing

Metal laser processing is widely used in sectors such as automotive, aerospace, and healthcare for tasks like precision welding and metal 3D printing, thanks to its high accuracy and flexibility. However, the technique is extremely sensitive to material properties and parameter settings—minor deviations can result in production defects. Traditionally, this has required extensive trial-and-error and expert tuning, leading to high costs.

Researchers at the Swiss Federal Laboratories for Materials Science and Technology (Empa) have optimized this process using 1machine learning. In metal 3D printing—specifically powder bed fusion (PBF)—their algorithm analyzes data from optical sensors in laser equipment to identify processing modes (conduction or keyhole) in real time. It then automatically adjusts parameters, reducing the number of required experiments by two-thirds while maintaining high quality. This advancement could significantly lower the operational threshold of PBF systems, facilitating broader adoption.

The team also developed a real-time control system based on field-programmable gate arrays (FPGAs), integrated with PC-based machine learning algorithms to enhance laser welding processes. The FPGA ensures fast and precise control, while the PC algorithm continuously learns from data to improve system intelligence over time. This approach allows the system to adapt to unexpected issues—such as surface defects—resulting in more stable and reliable production.