In a feature published by Usine Nouvelle, Duoverse explains how hybrid artificial intelligence is changing the way engineering simulations are developed and used. 

Traditional simulations often require hours, or even days, of computation to reproduce complex physical behaviour. Although highly accurate, these methods are rarely compatible with operational decision-making. Conversely, conventional AI models can produce rapid predictions but often struggle to generalize beyond the data used during training. 

Duoverse addresses this challenge through a hybrid approach. A first AI model learns directly from thousands of high-fidelity numerical simulations, creating fast surrogate models capable of reproducing physical phenomena. A second AI model continuously learns from operational data, correcting differences between simulation and reality as conditions evolve. 

This combination produces "augmented simulations" that remain physically consistent while adapting to real operating environments. The article also highlights early industrial collaborations, including projects with ArcelorMittal, demonstrating how hybrid digital twins can accelerate engineering analyses and improve industrial processes. 

As industries increasingly seek trustworthy AI solutions, hybrid digital twins represent an important step toward combining computational efficiency with scientific reliability.