57 - Forgetti: 3D Printing Spaghetti Made Right - Using Artificial Intelligence to Identify Real Time Mistake During Additive Manufacturing
Description/Abstract/Artist Statement
Fused deposition modeling (FDM) 3D printing is a popular additive manufacturing process, but it often faces challenges such as print failures, which can result in wasted time and materials. Existing failure detection solutions offer limited or proprietary models that hinder open accessibility and customization. This paper introduces Forgetti, an open-source, AI-based software solution designed to detect and respond to 3D print failures. Drawing inspiration from projects such as Pytorch-Wildlife, Forgetti aims to address the lack of public datasets by allowing for the fine-tuning of detection models on diverse failure scenarios. The paper outlines the methodology used to collect, label, and train a model for failure detection, utilizing Ultralytics and Svelte. The software integrates with a web interface to provide real-time failure alerts and system diagnostics, enhancing operational efficiency and security in both hobbyist and industrial environments. This work contributes to the growing need for accessible, customizable, and community-driven solutions in the 3D printing industry.
Faculty Advisor/Mentor
Li, Yaohang
Faculty Advisor/Mentor Department
Computer Science
College Affiliation
College of Sciences
Presentation Type
Poster
Disciplines
Artificial Intelligence and Robotics
57 - Forgetti: 3D Printing Spaghetti Made Right - Using Artificial Intelligence to Identify Real Time Mistake During Additive Manufacturing
Fused deposition modeling (FDM) 3D printing is a popular additive manufacturing process, but it often faces challenges such as print failures, which can result in wasted time and materials. Existing failure detection solutions offer limited or proprietary models that hinder open accessibility and customization. This paper introduces Forgetti, an open-source, AI-based software solution designed to detect and respond to 3D print failures. Drawing inspiration from projects such as Pytorch-Wildlife, Forgetti aims to address the lack of public datasets by allowing for the fine-tuning of detection models on diverse failure scenarios. The paper outlines the methodology used to collect, label, and train a model for failure detection, utilizing Ultralytics and Svelte. The software integrates with a web interface to provide real-time failure alerts and system diagnostics, enhancing operational efficiency and security in both hobbyist and industrial environments. This work contributes to the growing need for accessible, customizable, and community-driven solutions in the 3D printing industry.