Building a Navier-Stokes Solver in Python from Scratch: Simulating Airflow

2026-03-22 · Source: Towards Data Science · Field: Technology & Digital — Software Development & Engineering, Data Science & Analytics · Depth: Intermediate, medium

Summary

This guide details the derivation and Python implementation of a simple incompressible Navier-Stokes (NS) solver for Computational Fluid Dynamics (CFD). It targets data scientists and engineers seeking to understand numerical simulations by translating partial differential equations into discretized Python code, emphasizing vectorized operations with NumPy. The solver simulates airflow around a bird's wing profile, demonstrating how pressure and velocity evolve in a fluid. Key steps include solving the Pressure-Poisson equation using Jacobi iteration and applying finite difference schemes for time, advection, diffusion, and pressure terms. The implementation uses NumPy arrays for initialization, wing geometry definition via an immersed boundary mask, and a main solver loop that builds source terms, solves pressure, updates velocity, and applies boundary conditions. Qualitative results show expected pressure and velocity distributions, with pressure forces dominating viscous friction.

Key takeaway

For data scientists or engineers looking to deepen your understanding of numerical simulations beyond high-level libraries, implementing a basic CFD solver in Python is highly beneficial. You will gain practical experience with partial differential equations, discretization techniques, and vectorized computing with NumPy. This foundational knowledge is crucial for developing more complex scientific computing applications and understanding machine learning architectures that rely on similar principles.

Key insights

Implementing a Navier-Stokes solver from scratch clarifies fluid physics and numerical simulation mechanics.

Principles

• Pressure and velocity are coupled in fluid dynamics.
• Finite difference schemes discretize PDEs for computation.
• Pressure forces dominate viscous forces in airflows.

Method

The solver uses explicit Euler for time, upwind for advection, and central differences for diffusion/pressure. It iteratively solves a Pressure-Poisson equation via Jacobi iteration to maintain incompressibility.

In practice

• Use NumPy for vectorized operations in scientific computing.
• Implement immersed boundaries with Boolean masks for complex geometries.
• Integrate pressure fields to calculate lift and drag forces.

Topics

• Computational Fluid Dynamics
• Navier-Stokes Equations
• Finite Difference Method
• Numerical Simulation
• Python Scientific Computing

Best for: Data Scientist, Software Engineer, Machine Learning Engineer

Related on AIssential

• FLAME 2023: Diving into the Future of Fluid Dynamics and Machine Learning — AI techniques are transforming simulation by integrating physics-based modeling with data-driven machine learning.
• AeroJAX: JAX-native CFD, differentiable end-to-end. ~560 FPS at 128x128 on CPU [P] — AeroJAX enables differentiable CFD simulations for seamless integration into ML optimization and learning pipelines.
• The Bug That Ruined Game Physics For Decades — A new fluid simulation method prevents volume loss and handles complex dynamics by design, making advanced theory practical.
• From Euler to Dormand-Prince: ODE Solvers for Flow Matching Generative Models — Higher-order ODE solvers significantly improve Flow Matching generative model sampling efficiency and quality, especially for undertrained models.
• Operator learning for the 2D incompressible Navier-Stokes equations: a conformal prediction approach in the data-scarce regime — The perturbation-based conformal prediction framework quantifies uncertainty in neural operators efficiently, especially in data-scarce regimes, by comparing noisy and original label predictions.
• Learning universal approximations for partial differential equations with Physics-Informed Broad Learning System — PIBLS offers a backpropagation-free, least-squares optimization approach for PDEs, achieving faster, more accurate solutions than PINNs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards Data Science.