NDT Plenary Paper: NDE 4.0: realising zero-defect mass production of bonded joints by integrating AI into the advanced real-time ultrasonic quality monitoring process
R Maev
Canada
This presentation will demonstrate recent results in developing advanced non-destructive evaluation (NDE) technology, which allow for the realisation of zero-defect mass production of bonded joints by integrating artificial intelligence (AI) into the real-time ultrasonic quality monitoring process. Some successful results will be demonstrated, including implementation of the real-time automated spot weld quality evaluation of ultrasonic B-scans using deep learning original algorithms. Using a proprietary deep learning algorithm allows the welds to be classified as good, acceptable or bad in real time during the welding process. As a result, this solution enables the NDE specialists to modify currently used automotive adaptive systems, which control the welding process and produce high-quality welds that match the production-level requirements. Additionally, there is another important aspect of this problem – and this is crucial for manufacturing cycle times – that such recognition is realised in real time. The developed high-speed proprietary algorithms allow the system to make real-time decisions and send an online request to the welder to modify the welding parameters to make sure that at the end of each spot welding cycle the quality of each weld corresponds to a good weld, which is a great example of an advanced NDE 4.0 solution.
As an alternative, an in-line integrated advanced ultrasonic monitoring system with real-time AI-driven weld process characterisation is envisioned to create actionable feedback to the weld controller. The resultant models developed were highly efficient, with extremely fast inference speeds, well within the computational time constraints. The study shows that adaptive welding using ultrasonic process monitoring backed by AI-based data interpretation is possible and is a perfect example of an NDE 4.0 contribution towards zero-defect manufacturing (ZDM), which has been a dream for decades.
The author is confident that when the various aspects of this problem are resolved and the persistence of the results demonstrated, the developed approach to zero-defect quality will revolutionise the modern manufacturing process, especially the automotive assembly plants globally. This technology has the potential to bring big savings in production, reduce labour costs and eliminate unnecessary destructive tests, which are still part of today’s quality inspection process.
As an alternative, an in-line integrated advanced ultrasonic monitoring system with real-time AI-driven weld process characterisation is envisioned to create actionable feedback to the weld controller. The resultant models developed were highly efficient, with extremely fast inference speeds, well within the computational time constraints. The study shows that adaptive welding using ultrasonic process monitoring backed by AI-based data interpretation is possible and is a perfect example of an NDE 4.0 contribution towards zero-defect manufacturing (ZDM), which has been a dream for decades.
The author is confident that when the various aspects of this problem are resolved and the persistence of the results demonstrated, the developed approach to zero-defect quality will revolutionise the modern manufacturing process, especially the automotive assembly plants globally. This technology has the potential to bring big savings in production, reduce labour costs and eliminate unnecessary destructive tests, which are still part of today’s quality inspection process.
