Project Subtitle
Quality Control of Vehicle Assembly using an Ultrasonic Imaging Sensor with Embedded Artificial Intelligence
Companies and Partners
Tessonics Inc., University of Waterloo, NRC’s Aerospace Research Centre, Ford-Werke (Ford Company) GmbH, RWTH Aachen University
Problem
One of the essential production steps in industrial production is the automated joining of metallic structures. One of the most important joining methods is welding. In automobile production industries, Resistance Spot welding (RSW) and Laser Brazing welding (LBW) are primarily used with the help of the robot-guided system. An enormous high-quality assurance effort is still required for this automated joining process which in turn is very expensive.
Goal
An AI-based welding monitoring is developed that predicts the quality of the weld, thereby reducing the costs of quality inspection. With the help of ultrasound data and the ground truth annotation, an AI-based image segmentation algorithm is trained to correctly predict segmentation for weld nuggets for both RSW and LBW. The results obtained from the trained model are used for meaningful characterization of the welding process based on the company’s guidelines. The trained model is integrated into the production environment to provide real-time analysis of quality criteria.
Outcome
Real time intelligent AI-based welding monitoring systems for Resistance Spot welding and Laser Brazing welding. Significantly reduced manufacturing and quality inspection costs.