2026 ASEE Annual Conference & Exposition

Manufacturing plays a key role in the U.S. economy, and its impact continues to grow. With the expanding presence of Industry 4.0 and Industry 5.0, manufacturing jobs require a broader set of skills, including comprehension and management of human-centric manufacturing systems. Specifically, for next-generation smart manufacturing processes (e.g., 3D printing), skills such as process features understanding (e.g., physical machine behaviors, anomaly detection, and troubleshooting protocols) are highly demanded. Therefore, engineering students must acquire this knowledge within the limited number of course hours to fulfill the demands of a highly competitive job market, which seeks to reduce training costs and time. The effective learning expected in manufacturing courses is highly correlated with student-machine engagement and interaction. Nevertheless, traditional instructional methods (e.g., static lectures and observational labs) often lack the interactivity necessary to guarantee a comprehensive system-level understanding. Large language models (LLMs) can serve as a promising solution to generate direct interactivity with students, resolving emergent questions during the learning experience. Standard LLMs do not have access to real-time manufacturing data to produce specific insights about the manufacturing system’s status; thus, this often leads to generalized responses or hallucinations. To address this, an LLM-powered educational platform integrated with a 3D printing system is developed. (1) The platform captures real-time sensor data embedded in the 3D printer (e.g. nozzle and bed temperature, XYZ acceleration, motor current). (2) A Python-based graphical user interface (GUI) plots the process parameter signals. (3) The platform incorporates a chatbot powered by a domain-specific (DS) LLM, which takes sensor data for real-time interaction. For (3), a Llama 3 model is fined-tuned using relevant DS academic literature, user queries, and targeted responses. Here, the real-time manufacturing data is used to enrich the answers with specific machine status, facilitating real-time and adaptive interaction. Finally, the developed LLM-powered platform is expected to enhance the traditional learning process by providing a mechanism for human-machine adaptive learning in real-time.