In a Circular Economy, high-value metal products, such as those in the heavy-duty industry, are returned back into the system and are ripe for remanufacturing. These returned products are called “cores” by the industry. Gauging the quality of these cores is critical to providing viable materials for any remanufacturing program.

Today, visual inspection of a core’s damaged components is not reliable enough. The potential for errors is significant. A cost-effective method for recognizing common failure modes for damaged, high-value components, such as a cylinder head, is needed. Accurately recognizing damage and failure modes, such as porosity variations and cracks, requires a much higher order of reasoning than typical machine vision methods provide.

In this project webinar, Iowa State’s Paul Kremer, the project lead, and the University of Dayton’s Gül Kremer will discuss their research, which seeks to address this issue. More specifically, automated visual defect inspection systems that can be trained and integrated using commercial off-the-shelf technologies offer the promise of rapid, reliable detection of defects. Developing system software architectures that use updateable, deep learning models coupled with other established image acquisition and processing techniques  is part of the core task in putting together a workable system. However, calibration and validation of each hardware/software component, and of the overall system, through multiple calibration and benchmarking processes with clear definition of ground truth, is of critical importance.

In this webinar, a case study will be presented describing the integration of a software and hardware system for defect detection in cylinder heads undergoing remanufacturing.