Manufacturing Quality Inspection Using AI and Edge Computing

C. Ouyang, T. Cook, C. Lu
United States

Keywords: AI vision machine learning manufacturing inspection


High standard of quality, inspection cycle time and accuracy, management of large number of devices, are just a few common issues in manufacturing quality inspection. Artificial Intelligence (AI) assisted quality inspection in an actual manufacturing environment along with distributed edge devices can significantly improve inspection accuracy and time. Edge computing is moving compute to where the data is. Edge Computing for Devices provides users with a new architecture for node management. It is designed specifically to minimize the risks that are inherent in the deployment edge nodes. Also used to manage the service software lifecycle on edge nodes fully autonomously and supports model management through sync service to facilitate the storage, delivery, and security of models and metadata packages. Models trained by AI software such as Power AI Vision are used for production-level quality inspection, resulting in improvements in both efficiency and quality assurance. Edge devices and cameras are used both as input to train the vision models as well as continuous retrain and deployment.  Models are deployed on devices that are small yet powerful for inference coupled with real time photos using a camera or mobile phone. A model management system allows for fast and accurate model training and deployment.  An example use case is interface bent pin inspection. For the Power and Z systems, I/O slots line the back of the cage where cards are inserted during assembly. If the pins on the slots are not perfect, then the cards cannot be installed to functional properly. Detection of bent pins is important pre installation, as testing, locating and removing cards afterwards is very time consuming. There are often dozens of slots each with hundreds of tiny pins which are nearly impossible for humans to detect with the naked eye. Key benefits to using AI and edge devices for manufacturing include: Deep learning and vision models are continuously trained and improved. Multiple models and algorithms are used to test and create the best fit. The speed at which the machines move and inferencing that take place is unmatched by humans. Both models and inferencing run at scale with embedding AI and edge services into the manufacturing process.  By combining AI with Edge computing, enterprises can rapidly deploy a fully optimized and supported platform with blazing performance.