Skip to content

New Approaches in Real-Time Analytics and Computer Vision Powered by Machine Learning in Automated Inspection

  • today
  • access_time 2:00 - 2:30 PM ET
  • location_onSmart Theatre | SME ZONE
  • blur_circularTech Talk

Quality inspection is critical for complex production processes such as welding in the automotive and shipbuilding industries. Due to the intrinsic complexity and many process parameters in welding operations, inspection for defects after production remains the most common approach. Although process automation using robotic systems has enhanced throughput, weld inspection remains mainly manual and prone to the costly discovery of defects at late production stages. The high variabilities during manufacturing and the high number of influencing process parameters are notable barriers to automating quality inspection, especially in welding processes. 

AI and machine learning offer an opportunity to automate inspection during manufacturing with a great potential for reliability and production efficiency enhancements. There is a growing interest in automated defect detection using machine learning and vision software solutions in the food and electronics packaging industries. The adoption of these tools in materials processing operations such as welding is still lagging. In addition to the automation barriers, a high changeover rate complicates the development and scalability of the machine-learning-powered solutions for these processes. 

In this presentation, we review new approaches in real-time analytics and computer vision powered by machine learning. We discuss the issues around common machine learning methods in industrial applications, followed by a discussion on adapting these tools to enable automated inspection in welding operations and provide manufacturing operations with actionable insight regarding process quality in real-time. 

With the established nature of the manufacturing sector in mind, the adoption of advanced technologies in this industry requires special attention with regard to integration with production lines and workflow. We present a case study related to the adoption of automated inspection at a metal fabrication plant and discuss how the intrinsic nature of machine learning in conjunction with advanced IoT infrastructure can facilitate seamless integration causing minimal interruptions in manufacturing operations. We further discuss how a collaborative approach can benefit both manufacturers and AI solution provides with regards to continuous solution improvements and access to data.