Automated Weld Seam Inspection has been requested by automotive leaders. Since 2019, Mapvision has been developing Weld Seam Inspection based on AI technology. The introduction of the WSI 2024 marked a significant milestone for Mapvision, enabling anomaly detection for weld seams. This technology utilizes the same images that are already used for dimensional measurement and presence check. Weld Seam Inspection is completely integrated into the Mapvision Q-Series, the in-line inspection system.
To address the requirements for detailed defect data, Mapvision has further developed Weld Seam Inspection with its WSI 2025. This latest version meets elevated quality standards and requirements set by automotive OEMs and integrates advanced AI technology to improve quality assurance and overall production efficiency.
The WSI 2025 utilizes generative AI models to improve its capabilities, transitioning from Anomaly Detection technology to Defect Detection technology. This advancement allows not only for the recognition of defects but also for the identification of their specific types. This capability offers more statistics. Detailed defect information allows for specific adjustments for different defect types. It also facilitates root cause analysis of defects, and the implementation of corrective actions based on data-driven insights.
Fig 1: The Key Differences between the WSI 2024 and the WSI 2025
The WSI 2025 employs a universal semantic segmentation model that analyzes images and segments them into various layers. This allows the system to recognize and classify defects accurately. Additionally, the WSI 2025 incorporates a second AI model that specializes in analyzing segmentation outputs, providing comprehensive information about detected defects. One of the key features of the WSI 2025 is its ability to measure the dimensions of weld seams. By recognizing the outlines of the weld seams, the system can perform dimensional analysis, providing valuable insights into the quality of welds.
The WSI 2025 reduces the time and effort required for seam specific training. Unlike the WSI 2024, which required a large amount of ‘OK’ training samples, the 2025 version comes pre-trained and ready for immediate use. This streamlined implementation process ensures faster deployment and quicker results.
Fig 2: How the WSI 2025 works
One of the primary challenges in defect detection is acquiring enough ‘NOK’ samples to train AI models. To address this issue, the WSI 2025 utilizes generative models to create artificial defect samples. This approach enables Mapvision to extensively test the system without relying solely on real defect samples, which are fewer and more difficult to obtain. By generating artificial defects, the WSI 2025 offers comprehensive datasets for training and testing. As a result, the system achieves high accuracy in defect detection, ensuring reliable inspection results.
■ Weld seam presence
■ Skip & gap
■ Burn through
■ Porosity
■ Craters
■ Dimensional defects >5 mm
The Mapvision WSI 2025 marks a significant breakthrough in automated Weld Seam Inspection technology. Utilizing AI-powered defect detection algorithms, this system can inspect 150 weld seams in just 40 seconds, achieving defect accuracy rates between 97% and 100% for supported defect types. Detailed defect analysis and data-based corrective actions help maintain quality control standards and improve production efficiency.
If you're interested to learn more, book a product demo or watch our webinar here.