David Loeza, “Proof of Concept of AI-based Corrective Model in Manufacturing”
Mentor: Joseph Hamman, Mechanical Engineering, Engineering & Applied Science (College of)
Poster #71
The future of manufacturing lies in digital innovation. The Connected System Institute (CSI), state-of-the-art engineering lab facility, located at the University of Wisconsin-Milwaukee, serves as an environment for manufacturing industry-academia collaboration and is dedicated to advancing manufacturing through experiential research. CSI is committed to leveraging Artificial Intelligence (AI) to enhance the manufacturing process. Industry 4.0 is the ongoing fourth industrial revolution, which focuses on automation, digitalization, and data exchange in manufacturing and other industries. As Industry 4.0 reshapes the industry, traditional manual processes are being replaced by automated systems and advanced technologies. In traditional manufacturing, AI-based vision models only make detections to determine whether the product is within specification or not. It does nothing after it makes a detection because it requires a human to review the detection. In Industry 4.0, an AI-based corrective vision model would enable manufacturers to detect defects early, identify root causes, and implement corrective action. This would allow industries to optimize quality control processes and reduce product scrap, saving time and money. My role delves into creating a proof of concept for an AI-based corrective model, which would serve as the first steps to leveraging AI-based vision models for the manufacturing process. Approximately 2600 images were used to train the AI-based corrective model using NVIDIA’s AI computing capabilities to perform analysis of image captures to accomplish at least a 90% confidence rate of product failure and root cause of that failure. The AI-based corrective model is also comprised of 2 AI-based model subsets, Python was used to seamlessly integrate them. The demo focuses on demonstrating the decision process of the AI-based corrective vision model if the product is within specification or outside of specification, the identification process of the root cause of failure, and how a feedback mechanism may be applied to the manufacturing process.