NVIDIA Graphics Announces Digital Twin Platform for Scientific Computing

This accelerated digital twin platform for scientific computing consists of the NVIDIA Modulus AI framework for developing machine learning neural network models for physics, and the NVIDIA Omniverse? 3D virtual world simulation platform.

The platform creates interactive AI simulations based on physical information in real time to accurately reflect the real world, enabling simulations such as computational fluid dynamics to be performed 10,000 times faster than traditional engineering simulation and design optimization workflow methods. Researchers are able to model complex systems, such as extreme weather events, with greater speed and accuracy than previous AI models.

NVIDIA demonstrated two examples of the technology's application: NVIDIA's FourCastNet physics machine learning model can simulate global weather patterns and predict extreme weather events such as hurricanes with higher confidence and up to 45,000 times faster than traditional numerical prediction models. In addition, Siemens Gamesa Renewable Energy is using AI to optimize the design of wind turbines.

Ian Buck, vice president of Accelerated Computing at NVIDIA, said, "To address challenges such as climate change, drug discovery and finding new sources of renewable energy, we're accelerating computation using data center-scale AI, which has the potential to deliver a million-fold leap in performance. With NVIDIA's AI Science Digital Twin framework, researchers can discover how to solve these large-scale problems."

NVIDIA Modulus and Omniverse

NVIDIA Modulus takes data and physics into account to train a neural network to create an AI agent model for digital twins. The agent model can reason about new system behaviors in real-time, enabling dynamic, iterative workflows and, when integrated with Omniverse, visualization and real-time interactive exploration.

The latest version of Modulus uses Fourier neural operators for data-driven training, a framework that enables AI to simultaneously solve relevant partial differential equations and also combine machine learning models with weather and climate data, such as the European Center for Medium-Range Weather Forecasts' ERA5 dataset.

NVIDIA Omniverse, a real-time virtual world simulation and 3D design collaboration platform, complements the capabilities of Modulus by enabling real-time visualization and interactive exploration of digital twins using Modulus' output agent models.

NVIDIA FourCastNet

Fourier neural operators and transformers support NVIDIA's FourCastNet physics machine learning model, which was trained using 10TB of Earth system data. As one of the steps towards Earth-2, which NVIDIA CEO Jen-Hsun Huang has announced will be used to create a digital twin of Earth in Omniverse, FourCastNet is able to simulate and predict the development and risk of extreme weather events, such as hurricanes and atmospheric currents, with higher confidence and up to 45,000 times faster.

Karthik Kashinath, Senior Developer Technical Scientist and Engineer at NVIDIA, said, "With digital twins, researchers and decision makers are able to interact with data and quickly explore scenarios, which is nearly impossible with traditional modeling techniques because they are both expensive and time-consuming. As the core of Earth-2, NVIDIA FourCastNet drives the development of Earth's digital twin by modeling the physics and dynamics of global weather faster and more accurately."

Siemens Gamesa Renewable Energy

The digital twin platform has helped wind farms equipped with Siemens Gamesa Renewable Energy wind turbines to significantly accelerate wind farm layout simulation studies, allowing researchers for the first time to use AI to accurately simulate the impact of wind turbine locations on their performance under various weather conditions, thereby optimizing wind farm layouts and increasing power production by 20% compared to previous designs.

Sergio Dominguez, Portfolio Manager for Onshore Digital at Siemens Gamesa Renewable Energy, said, "By partnering with NVIDIA, Siemens Gamesa Renewable Energy is able to dramatically accelerate the speed of computation and the deployment of the latest algorithm development in complex areas such as computational fluid dynamics, and is laying the groundwork for further collaboration in the future. "

To learn more about NVIDIA's digital twin, watch Jen-Hsun Huang's GTC 2022 keynote. Registering for GTC 2022 is free and gives you access to sessions hosted by NVIDIA and industry leaders!