Over the last fifteen years, GE Research has built an advanced and robust portfolio of platforms that use Artificial Intelligence and Computer Vision systems to improve the quality and speed of parts and material inspection. GRC uses multiple light spectra (IR to visible to UV) and multi-geometry models (2D, 3D) to find defects such as cracks, pits, dents, nicks in high-value gas and wind turbine parts. Technology first developed to characterize cellular structures is now leveraged to inspect the integrity of Ceramic Matrix Composites (CMC Panels) production for military jets. GE Additive uses advanced metal sintering to build complex parts that are extremely difficult, if not impossible to build using current standard manufacturing techniques. Automated inspection technology from GRC can monitor the build process via advanced cameras or contact sensors to detect minute variations from plan and predict defects even before the part has been built, resulting in reduced waste for the customers. GRC continues to enhance the accuracy of defect detection by building Digital Twins of these parts to track how they perform not only at a point in time but over their entire lifetime.
The ability to perform automated inspection on newly made parts or parts that have come in for service have fundamentally changed the approach to quality control in advanced manufacturing. Multiple businesses have changed their strategic direction and investment plan to incorporate this technology.
Automated inspection creates a significant impact on three fronts. First, it increases the speed of inspection without sacrificing quality – they are demonstrated to be as good if not better than current methods. Faster inspection means customers can get serviced parts and components back sooner. Second, it reduces cost of operation. With automated inspection, the business can improve their bottomline performance. GE Businesses have clear line of sight to realizing over $100M of cumulative bottom line impact by employing automated inspection technologies. Third, automated inspections do not suffer from the variability of human inspections leading to a strong quantitative approach towards more reliable systems.
At GE Research, we can mesh our knowledge of physical systems, our expertise in computer vision and AI, our software and robotics insight to create an end-to-end solution that works for the customer on the shop floor, not only in the lab.
Zhang, S., Wen, L., Bian, X., Lei, Z. and Li, S.Z., 2018. Single-shot refinement neural network for object detection. In IEEE CVPR.
Sankaranarayanan, Swami, Yogesh Balaji, Arpit Jain, Ser Nam Lim, and Rama Chellappa. "Unsupervised domain adaptation for semantic segmentation with gans." arXiv preprint arXiv:1711.06969 2 (2017).