Digital twins and LLMs integrated to enhance manufacturing and design

Researchers are exploring the integration of large language models (LLMs) and knowledge graphs to enhance the utility of digital twins in manufacturing and mechanical design.

One framework integrates structured knowledge graphs with LLMs to interact with in-process digital twins of 3D printers. By retrieving live telemetry and historical data, the system provides context-aware responses and information models, aiming to reduce the manual effort and multidisciplinary expertise typically required to identify anomalies or defects. This approach has shown potential to outperform standard retrieval-augmented generation methods.

In the field of mechanical design, particularly for complex transportation systems, digital agent-based frameworks are being developed to navigate multi-objective design tradespaces. These frameworks utilize machine learning predictors trained on datasets from digital twins to map performance targets directly to feasible design configurations, seeking to automate portions of the embodiment design process and reduce reliance on traditional trial-and-error methods.