best congress paper

Training-Free Surface Defect Inspection in Die Casting

Surface defect detection in die casting using computer vision and AI has attracted increasing attention in industrial quality assurance. This study proposes a training-free surface defect inspection method that uses a pretrained AI model to generate feature representations of defect-free products for memory bank construction. During inspection, a test image is processed by the same pretrained model, and its features are compared with those stored in the memory bank. A defect heatmap is then generated using a dissimilarity measure, enabling effective localization and identification of surface defects.

Training-free AI does not require task-specific model training which makes the method particularly suitable for industrial environments where rapid deployment, frequent product changes, limited defect data, and simplified AI implementation are important requirements. The proposed method is evaluated on high-resolution images of diecast products containing small surface defects and challenging imaging artifacts. The results demonstrate that the method can detect small defect regions in large images and provide interpretable fused heatmaps for human verification.

Best paper Authors

  • Bin Chen

    Dr. Bin Chen received his Ph.D. from Duke University. He is a professor of Electrical and Computer Engineering at Purdue University Fort Wayne.

    His research interests include signal and image processing, computer vision, AI/machine learning, and their industrial applications. His work has been supported by DOE, EPA, USDA and industry sponsors.

  • Xiaoming Wang

    Professor Xiaoming Wang is a professor with Purdue University. He started working with Purdue University in 2013 after 5 years working with Hong Kong Productivity Council on diecasting. He has been working closely with diecasting companies combating technical challenges, from conformal cooling dies, new alloy development to AI vision for quality inspection. His research projects are sponsored by Stellantis, NADCA and Defense Logistics. He is working with other universities in the development of a standard testing method for die materials for NADCA. In addition to diecasting, Professor Wang is also doing research on additive manufacture of aluminum alloy matrix composites for high strength and high ductility applications. He is the FEF Key Professor at Purdue University.

  • Eric Kessenich

    Eric Kessenich is a Tooling Manufacturing Engineering Manager at Mercury Marine in Fond du Lac, Wisconsin. He leads engineering activities supporting high pressure die casting tooling and manufacturing operations. With 10 years of experience in the die casting industry, his areas of focus include tooling design, tooling material and coating selection, and advanced technology implementation. Eric holds a Bachelor of Science degree in Manufacturing Technology Management from the University of Wisconsin-Platteville and works closely with internal teams and industry partners to improve casting quality, productivity, and operational performance.

  • Corey Vian

    Dr. Corey Vian has been in the aluminum industry for the past 12 years. He has held multiple engineering related roles, and is currently the manufacturing engineering manager for advanced engineering at Stellantis’ Kokomo Casting Plant. He is active in NADCA, where he is a current Board member, research committee member, and Chapter 25 Vice President; and is also a past recipient of NADCA’s Committee Member of the Year award. He earned his Bachelors, Masters, and PhD from Purdue University and continues to be a part of Purdue’s manufacturing related programs as a member of their Cast Metals Advisory Board and Industrial Advisory Council.

  • Charles Monroe

    Dr. Monroe’s research interests focus on the relationship of part performance, manufacturing process, and material properties with the goal to make robust and more capable components with efficient cost. He believes this is possible through using computer process simulation an understanding the variability and unknowns of the manufacturing process towards achieving the theoretical maximum material limits. Innovative solutions that use multiple manufacturing techniques and extreme conditions to generate improved location specific properties leveraged by designers for application.

    Before joining the MTE Department at UA, Dr. Monroe was an Associate Professor in the department of Materials Science and Engineering at the University of Alabama at Birmingham where he focused on industry and defense research to solve manufacturing problems in cast iron, aluminum high pressure die casting, and cast steel processes.