NVIDIA has introduced an accelerated digital twins platform for scientific computing comprising the NVIDIA Modulus AI framework for developing physics-ML neural network models, and the NVIDIA Omniverse 3D virtual world simulation platform.
The platform can create interactive AI simulations in real time that are physics-informed to accurately reflect the real world. Simulations such as computational fluid dynamics are accelerated up to 10,000x faster than traditional methods for engineering simulation and design optimisation workflows.
With the platform, researchers can model complex systems, such as extreme weather events, with higher speed and accuracy when compared to previous AI models.
“Accelerated computing with AI at data centre scale has the potential to deliver million-fold increases in performance to tackle challenges, such as mitigating climate change, discovering drugs and finding new sources of renewable energy. NVIDIA’s AI-enabled framework for scientific digital twins equips researchers to pursue solutions to these massive problems,” said Ian Buck, Vice President of Accelerated Computing at NVIDIA.
NVIDIA Modulus takes both data and the governing physics into account to train a neural network that creates an AI surrogate model for digital twins. The surrogate can then infer new system behaviour in real time, enabling dynamic and iterative workflows. Integration with Omniverse brings visualisation and real-time interactive exploration.
The new release of Modulus allows data-driven training using the Fourier neural operator, a framework enabling AI to solve related partial differential equations simultaneously. It also integrates ML models with weather and climate data, such as the ERA5 dataset from the European Centre for Medium-Range Weather Forecasts.
Complementing Modulus, NVIDIA Omniverse is a real-time virtual world simulation and 3D design collaboration platform. It enables the real-time visualisation and interactive exploration of digital twins using the output surrogate model from Modulus.
Fourier neural operators and transformers enable the NVIDIA FourCastNet physics-ML model, trained on 10TB of Earth system data. As a step toward Earth-2, FourCastNet emulates and predicts the behaviour and risks of extreme weather events such as hurricanes and atmospheric rivers with greater confidence and up to 45,000x faster.
“Digital twins allow researchers and decision-makers to interact with data and rapidly explore what-if scenarios, which are nearly impossible with traditional modeling techniques because they’re expensive and time consuming. Central to Earth-2, NVIDIA’s FourCastNet enables the development of Earth’s digital twin by emulating the physics and dynamics of global weather faster and more accurately,” said Karthik Kashinath, Senior Developer Technology Scientist and Engineer of NVIDIA.