Modular optimized production processes
The Fraunhofer Institute for Material and Beam Technology IWS has developed an innovative solution called SURFinpro to make production processes faster, more precise and more flexible. SURFinpro uses artificial intelligence and optical metrology to detect, classify, visualize and report defects to the production line in real time.
Dr. Christopher Taudt, group leader for surface metrology at the Fraunhofer Application Center for Optical Metrology and Surface Technologies (AZOM), and his team have developed a system that detects surface defects, artifacts and texture changes and evaluates them using artificial intelligence. The system captures surfaces in three dimensions at high resolution, generating continuous information for ongoing production. It classifies defects and provides additional parameters such as defect density, geometric dimensions and defect frequency. This offers significant added value compared to conventional systems.
Higher accuracy at higher speed
The solution has already been successfully used for over a year and analyzes a roll-to-roll process with a width of 70 centimeters. To further increase the optimization potential, the team trains SURFinpro during the ongoing production process. Errors are fed into a neural network using an error catalog to improve anomaly detection. The system must respond dynamically to changes. The researchers are working on developing better neural networks that make do with less data and on implementing new training strategies during ongoing operations.
Fraunhofer AZOM is currently adapting its technology for various application areas, including the continuous manufacturing of fiber composites. The aim is not only to avoid near-surface defects, but also to detect and evaluate components on a multidimensional basis. The semiconductor industry is another target group that will have access to the defect classification algorithms and system, especially for the production of flexible semiconductor material.
Currently, the Fraunhofer AZOM solution uses up to four cameras. The researchers plan to expand the system to include additional camera systems so that it can be used flexibly, regardless of the process. Another focus is on the speed of the solution. High cycle times are required, especially for fiber-reinforced plastics and in textile processing. The team uses machine learning and artificial intelligence techniques to speed up the evaluation and extract information more quickly from less data.
Intelligent and sophisticated modular
An important aspect of the solution is its modularity. Thanks to a sophisticated modular principle with efficient components, SURFinpro is versatile and easily adaptable to different requirements. The individual technologies of the system have been developed as independent building blocks that can also be used effectively in other projects.
Fraunhofer AZOM will present its solution at this year's Laser World of Photonics (June 27-30, 2023) using an application example for the flexible production of solar cells in a roll-to-roll process. Visitors will have the opportunity to experience data recording and evaluation in real time at the Fraunhofer joint booth.
Source: Fraunhofer Institute for Material and Beam Technology IWS. Transparency Note: This text was edited with support from ChatGPT.
