PaddlePaddle/PaddleScience

What it solves

PaddleScience is a scientific computing suite designed to solve complex problems in physics, chemistry, and meteorology. It addresses the challenge of solving differential equations and simulating physical systems by combining deep learning with scientific laws.

How it works

Built on the PaddlePaddle framework, it leverages automatic (high-order) differentiation and deep neural networks. The suite supports three primary solving methodologies:

• Physics-driven: Using physical laws to guide the model (e.g., PINNs).
• Data-driven: Learning patterns directly from datasets.
• Physics-data fusion: Combining both physical mechanisms and data for improved accuracy.

It also integrates with the sympy library for symbolic equation representation and supports complex STL geometry sampling and various boundary conditions (Dirichlet, Neumann, Robin).

Who it’s for

It is intended for researchers and engineers in scientific computing, fluid dynamics, structural analysis, materials science, and meteorology who want to use AI to accelerate simulations or solve inverse problems.

Highlights

• Broad Domain Coverage: Includes 20+ cases across fluids, structures, and weather forecasting.
• Scientific Tooling: Supports symbolic computation via sympy and complex STL geometry operations.
• Flexible Solving: Offers a choice between purely data-driven, purely physics-driven, or hybrid approaches.
• Research Efficiency: Features experiment source code tracking and one-click parallel experiment launching.