CFD · OpenFOAM · LLM · UN Youth Delegate
Talking to the Wind
Using computational fluid dynamics and large language models to decode how atmospheric stability governs urban pollution — from Shanghai street canyons to the UN climate stage.
When Physics Meets Policy
This project bridges high-fidelity numerical simulation, miniature physical experiments, and AI-driven environmental tools to answer a question with real urban planning consequences: does the atmosphere trap or dilute pollution?
// The Question
Trapping or Dilution?
Growing up in Chongqing, I wore dust masks daily. But meeting Uncle Deng in a Pingxiang coal mine — his lungs scarred by pneumoconiosis — taught me air pollution is a matter of life and death.
I noticed a puzzle: foggy mornings persist in low-traffic Qingpu, while Lujiazui’s rush-hour air clears by noon. The difference is atmospheric stability — the thermodynamic force that either traps pollutants in canyons or dilutes them through mixing. No study had systematically compared its effects on street-to-residential transport. I built that missing picture.
OpenFOAM Simulation
High-resolution RANS simulations of a Shanghai street canyon (H/W = 2) and adjacent residential neighborhood under three stability regimes: unstable, neutral, and stable.
Physical Validation
A miniature experimental platform with 3D-printed buildings, ice blocks for temperature stratification, smoke tracers, and PM2.5/CO/CO2 sensors qualitatively confirmed CFD predictions.
LLM-Powered App
A smartphone prototype fuses CFD physical rules, real-time sensor data, and urban maps through a multimodal LLM (Qwen-plus via LangChain) for natural-language air quality queries.
Global Advocacy
Presented at UN Climate Change Conference SB62 in Bonn (June 2025). The research was also submitted to the Shing-Tung Yau Science Awards and Korea Science & Engineering Fair.
Project Gallery
Key Results
3
Stability Regimes
H/W=2
Street Canyon Ratio
SB62
UN Conference
KSEF
Science Fair Entry
From Simulation to Global Advocacy
At SB62, I argued that urban planning must incorporate stability-specific strategies — widening ventilation corridors, optimizing building setbacks, deploying vegetation barriers — rather than relying solely on emission caps. Policymakers responded with interest in piloting these concepts in new development zones.
The most important lesson was methodological humility. CFD gives beautiful contours, but without physical validation and real-world boundary conditions, it is just colored pixels. By combining simulation, miniature experiment, and LLM inference, I learned that robust environmental science requires triangulation. My next step is a field campaign using portable sensors and drone platforms to capture the mesoscale gap between fixed monitoring stations and building-scale models.
Technology is valuable when it disappears into the lives of the people it serves.
Common Questions
What software was used for CFD simulation?
OpenFOAM, an open-source C++ toolbox for computational fluid dynamics. The simulations solved Reynolds-Averaged Navier-Stokes (RANS) equations with the standard k–ε turbulence model.
What is the LLM app prototype?
Built with Python + LangChain + Qwen-plus, it accepts natural-language questions like ‘How will air quality near Building 4 change?’ and generates physics-informed, context-aware answers using CFD rules, sensor data, and map layouts.
How does atmospheric stability affect pollution?
Stable stratification suppresses turbulent mixing, trapping pollutants near the ground and transporting high-concentration plumes far downstream. Unstable conditions promote dilution but widen the affected area.
What are the policy implications?
Cities should design ventilation corridors and building setbacks based on dominant stability regimes, not just emission targets. This could reduce residential exposure during stable winter nights by 30–50%.
发表回复