AI-Based Detection & Localization of Metal Processing Defects | RapidMiner

【 Background 】

       Quality control is a critical part of industrial product manufacturing, and metal surface defect inspection is one important step within that process. Traditional manual inspection is time-consuming and often inconsistent, especially on complex production lines.
       In this case, the client aimed to build an “identification system for metal surface processing defects”, enabling automatic marking and localization of defects to quickly assess machining conditions and product quality.

Metal processing defect image

Further reading: Product Introduction – Altair RapidMiner

【 Results 】

       Dozens of photos containing metal surface defects were collected in the machining environment, and defects were manually labeled and categorized first. Richin utilized Altair AI to build the defect identification workflow (including converting manually labeled data into Altair AI-compatible formats) to conduct AI training and validation. The final system successfully identifies and localizes defects such as gaps, oil drops, micro oil drops, and water streaks.
       This AI defect identification and localization system provides high efficiency, accuracy, and low human error, significantly improving the productivity of quality assurance teams.
       With defect marking and localization data, manufacturers gain a strong foundation for process improvement. The AI system can also be deployed on production lines for real-time monitoring in the future.

AI Marking System – Defect Labeling and Localization

【 Technical Features 】

• Altair AI integrates with the third-party tool LabelMe, enabling images to be annotated and converted into numerical data usable for AI defect identification model training.

• The AI defect identification and localization system was validated and fine-tuned by adjusting input data, ultimately achieving up to 100% identification accuracy. Even subtle defects can be captured accurately.

• The logic of AI identification technology can also be applied to other fields, such as sound recognition, facial detection, autonomous driving, and more.

【 Metal Processing Defect Identification & Localization 】AI Application Case – 5-Minute Series