Fraunhofer Institute for Material and Beam Technology IWS
Fraunhofer Institute for Material and Beam Technology IWS

LaserKI

Latency-optimized Plant and Sensor Network for Real-time Data Acquisition and Control via AI Models

Motivation and Project Description

In laser material processing, data-driven process models are currently and will continue to be a prerequisite for new products and services. The basis for highly automated, autonomous manufacturing processes is:

  1. The collection and evaluation of relevant process data
  2. The development of AI models for processing the data based on this
  3. The development of specific hardware (FPGA, ASIC) that integrates these models into the plant control and regulation system in a process-oriented manner.

By collecting all relevant process and plant data, it is possible to learn from processes which process parameters need to be adjusted (effects of defects). The necessary understanding of processes and materials is the core competence of Fraunhofer IWS. Building on this, the training of neural networks for the development of AI-based process models uses existing material data in combination with high-resolution, process-oriented, optical sensor technology for the control of complex processing systems. This makes such systems and processes controllable via simple interfaces, even for less experienced operators (one-button process). This development is of outstanding interest both for the adaptation of laser processes to new materials and due to the increasing shortage of skilled workers. However, the basis for this is the real-time connection (< 3 ms latency) of the sensors and the development of AI-based software and hardware for controlling laser processes.

Project Objective

The objective of the LASER-AI research project is therefore to develop latency-optimized laser system and sensor communication for real-time data acquisition and control using AI models for laser powder deposition welding, laser cutting, and laser welding. This allows process data to be used to understand influences on product quality and to develop real-time hardware for the simple control of complex laser processes – from “effects of defects” to the “one button process.”

From “effects of defects” to the “one button process” using the example of laser powder deposition welding. Left: In the first part of the LASER-KI project, the system is to be coupled with a real-time capable OCT sensor. The connection to the Fraunhoer IWS Edge Server enables the training of the models. Right: Based on the trained models, System on Module (SOM)-based FPGAs are developed that control the laser pro-cesses in real time.
© Fraunhofer IWS
From “effects of defects” to the “one button process” using the example of laser powder deposition welding. Left: In the first part of the LASER-KI project, the system is to be coupled with a real-time capable OCT sensor. The connection to the Fraunhoer IWS Edge Server enables the training of the models. Right: Based on the trained models, System on Module (SOM)-based FPGAs are developed that control the laser pro-cesses in real time.