Abstract:Scientific discovery and engineering design are currently limited by the time and cost of physical experiments, selected mostly through trial-and-error and intuition that require deep domain expertise. Numerical simulations present an alternative to physical experiments but are usually infeasible for complex real-world domains due to the computational requirements of existing numerical methods. Artificial intelligence (AI) presents a potential paradigm shift by developing fast data-driven surrogate models. In particular, an AI framework, known as Neural Operators, presents a principled framework for learning mappings between functions defined on continuous domains, e.g., spatiotemporal processes and partial differential equations (PDE). They can extrapolate and predict solutions at new locations unseen during training, i.e., perform zero-shot super-resolution. Neural Operators can augment or even replace existing simulators in many applications, such as computational fluid dynamics, weather forecasting, and material modeling, while being 4-5 orders of magnitude faster. Further, Neural Operators can be integrated with physics and other domain constraints enforced at finer resolutions to obtain high-fidelity solutions and good generalization. Since Neural Operators are differentiable, they can directly optimize parameters for inverse design and other inverse problems. We believe that Neural Operators present a transformative approach to simulation and design, enabling rapid research and development.

What problem does this paper attempt to address?

The paper primarily explores the application of Neural Operators in accelerating simulation and design in scientific research. Specifically, the paper attempts to address the following core issues: 1. **Limitations of Physical Experiments**: Current scientific research and engineering design are constrained by the high cost, time consumption, and reliance on trial and error and intuition of physical experiments, which require deep domain knowledge. 2. **Limitations of Numerical Simulations**: Although numerical simulations can serve as an alternative to physical experiments, existing numerical methods often have computational demands that are too high to be feasible for complex real-world scenarios. 3. **Artificial Intelligence (AI) as a Solution**: The paper proposes the Neural Operator framework, an AI-based approach that can learn mappings between functions over continuous domains, such as spatiotemporal processes and partial differential equations (PDEs). Neural Operators have the following characteristics: - They can perform zero-shot super-resolution predictions at unseen new locations, inferring high-resolution results from low-resolution data. - Compared to traditional numerical methods, Neural Operators can achieve speedups of 4 to 5 orders of magnitude in many applications. - Neural Operators can integrate constraints from physics and other domains to obtain high-fidelity solutions, and because they are differentiable, they can be directly used for solving inverse problems and optimizing designs. 4. **Limitations of Neural Networks and Advantages of Neural Operators**: Standard neural networks have fundamental limitations in scientific modeling, such as the inability to learn mappings between functions over continuous domains. In contrast, Neural Operators overcome this limitation by approximating the underlying operators (i.e., mappings between input and output function spaces). 5. **Specific Application Scenarios**: The paper showcases application cases of Neural Operators in various fields, including computational fluid dynamics, weather forecasting, and materials modeling, with particularly notable speed improvements in weather forecasting. In summary, the core objective of this paper is to demonstrate how Neural Operators can serve as a transformative method to accelerate simulation and design work in scientific research and engineering design, thereby significantly enhancing the speed and efficiency of research and development.

Neural Operators for Accelerating Scientific Simulations and Design

Physics-Guided, Physics-Informed, and Physics-Encoded Neural Networks and Operators in Scientific Computing: Fluid and Solid Mechanics

Neural Operator: Learning Maps Between Function Spaces

An Advanced Physics-Informed Neural Operator for Comprehensive Design Optimization of Highly-Nonlinear Systems: An Aerospace Composites Processing Case Study

Multiscale Neural Operator: Learning Fast and Grid-independent PDE Solvers

Neural Operator: Is data all you need to model the world? An insight into the impact of Physics Informed Machine Learning

Magnetohydrodynamics with Physics Informed Neural Operators

Physics-Informed Deep Neural Operator Networks

Learning nonlinear operators in latent spaces for real-time predictions of complex dynamics in physical systems

Neural-Parareal: Dynamically Training Neural Operators as Coarse Solvers for Time-Parallelisation of Fusion MHD Simulations

Finite Operator Learning: Bridging Neural Operators and Numerical Methods for Efficient Parametric Solution and Optimization of PDEs

Physics-enhanced Neural Operator for Simulating Turbulent Transport

Physics-Informed Geometry-Aware Neural Operator

Neural Dynamical Operator: Continuous Spatial-Temporal Model with Gradient-Based and Derivative-Free Optimization Methods

Neural Operator: Graph Kernel Network for Partial Differential Equations

Towards Stability of Autoregressive Neural Operators

NeurOLight: A Physics-Agnostic Neural Operator Enabling Parametric Photonic Device Simulation

Improved generalization with deep neural operators for engineering systems: Path towards digital twin

Neural Inverse Operators for Solving PDE Inverse Problems