AI-Powered 3D Printing Error Detection and Optimization System

Recent advancements and rising popularity of 3d printing technology have increased innovation and advancements in manufacturing vertical. But this has also brought in challenges in waste reduction particularly for materials like PLA, ABS, and PETG. Waste reduction in 3D printing is a crucial challenge due to the huge dependency on thermoplastics like PLA, ABS, and PETG. Each of these materials has distinct characteristics that influence their recyclability, disposal methods, and potential environmental impact. In another parallel, Generative AI is gaining momentum due to its ability in generate complex data based on the provided prompt and model training. This project aims to utilize the benefits and advantages of Generative AI in addressing the challenges of increasing waste in Additive manufacturing processes. Manual inspection of printing defects is inefficient and prone to human errors. AI can automate and improve this process. This paper presents an AI-powered 3D printing error detection and optimization system that leverages computer vision, deep learning, and generative AI to enhance print quality and reduce material waste. The system employs a convolutional neural network (CNN) to classify common 3D printing defects from real-time images and logistic regression model to identify the material type used for 3d printing. This is followed by a Generative AI Model that provides useful insights for the users on the recommendations to use manufacturing processes in more reliable way with a motto to reduce waste reduction during 3D printing processes. For generative AI model, Prompts are provided in such a way that it contains the printing error type and classified material type. Prompt optimization techniques like gradient-descent algorithm and genetic algorithm are considered and both are being used to create different prompt variations and most reliable prompt is sent to Generative AI model. This is to refine the defect descriptions for an LLM-powered recommendation system (Mistral 7B).