April 2021

28 n MACHINE VISION April 2021 www.drivesncontrols.com Linescan cameras team up with AI to inspect carbon fires H elicopters, planes and racing cars rely on carbon fibre parts to give them strength without excessive weight. The global carbon fibre market is expected to reach $6.3bn by 2025. With the increasing demand for carbon fibre, its manufacturers are confronted with serious challenges: n enhancing properties such as stiffness and strength, while maintaining quality; n ensuring that the fibres can adapt to different manufacturing processes, from braiding directly a roving spool, to fabrics scrims, which are draped manually or using robots; and n complying with stringent safety requirements. Carbon fibres – also known as graphite fibres – are composed mostly of carbon atoms and are 5–10μm in diameter, or thinner than a strand of human hair. The fibres get their strength by being twisted tightly together like yarn. Carbon fibre is five times stronger than steel and twice as stiff, yet much lighter, making it ideal for applications such as aviation. The materials are also making inroads into the automotive sector. Carbon fibre composites could halve passenger car weights, improving fuel efficiency by almost 35% without compromising the performance of the car or the safety of its passengers. SGL Carbon, a manufacturer of graphite and composite materials, develops high- performance carbon fibre materials at its plant in Meitingen, Germany. It is leading a project, funded by the German Federal Ministry for Economic Affairs and Energy, aimed at developing new high-strength carbon fibres. As part of the AirCarbon project, SGL has been working with the Fraunhofer Institute for Casting, Composite and Processing Technology IGCV to develop an optical inspection system that uses AI-enhanced image processing to control carbon fibre material quality. “The automatic detection of very small fibre defects in carbon fibre production is still not completely resolved,” explains Kristina Klatt, SGL’s head of carbon fibre development. “The uniform colouring of fibres and fibre defects makes it particularly difficult for optics and software to achieve good results.” Previous test systems have been limited, for example, to inspecting fabrics that have discernible contrast resulting from the knitted threads, making deviations easier to recognise. Fraunhofer IGCV took another approach. Using a wave linescan camera with a true- colour RGB sensor, it came up with a system that provides seamless, continuous surface monitoring, online during operation. The system combines a Chromasens allPixa wave camera with an adapted neural network, resulting in high image-processing flexibility. It can monitor both upstream and downstream processes. The pre-trained neural network adapts to new conditions and learns to find errors. It can recognise whether a component is good or bad, with a claimed accuracy of better than 90%. However, this only works for patterns that are particularly heterogeneous. “In the case of fibres, there are very different errors that are also subject to varying environmental conditions,” explains Andreas Margraf, a project manager who works on online process monitoring at Fraunhofer IGCV. “The researchers and SGL Carbon were interested in how often and where the defects occur in the individual sections.” The researchers have found this can be achieved by expanding the neural networks, so that each pixel of an image is classified as either “good”(not defective) or“bad”(defective). Anomalies in the fibre can be identified by classifying larger groups of pixels as bad. The technology allows defective fibres to be identified reliably from the images recorded by the Chromasens camera, whose 15,000-pixel resolution is high enough to capture tiny defects. “The level of detail the system offers online could only be exceeded in the laboratory under a microscope,” says Margraf. Because deep black materials absorb significant amounts of light, carbon fibres are difficult to make visible in image-processing systems. They therefore need special lighting to generate high-contrast images. The German researchers are using Chromasens Corona II LED lights that provide up to 3,500,000 lux of illumination, along with a patented technology that that uses elliptical reflectors instead of lenses to focus the LEDs. Defects can be made visible to the filament level (10μm), allowing even the smallest anomalies to be identified. Defect-detecting software developed by Fraunhofer is now being used in real carbon fibre production at SGL Carbon’s manufacturing plant. The aim is to test the software with various parameters. n German researchers have developed an optical system for inspecting carbon fibre materials for tiny defects, thus avoiding potential problems such as misaligned filaments. The system provides data that could help to improve carbon fibre production “significantly”. The technologies developed as part of the AirCarbon project can identify tiny defects in carbon fibre materials