Hi everyone,
I’m a 16-year-old developer and aspiring architecture student. Ladybug Tools has been my absolute gold standard for understanding environmental building design. Your ecosystem inspired me to dive deep into building physics.
While learning about CFD workflows (like Butterfly/OpenFOAM), I realized that while they offer unmatched accuracy, they can be too computationally expensive for the very first days of a conceptual sketch, where architects just need a quick directional intuition.
To explore a faster, lighter alternative for that specific phase, I built Polymère (AHI) — a building physics engine that runs entirely in the browser using WebGPU compute shaders.
The Technical Stack:
- Takes an IFC model and runs a GPU-accelerated conservative voxelization pipeline.
- Solves airflow using a custom Lattice Boltzmann Method (LBM D3Q19) written completely in WGSL.
- Couples the fluid dynamics with a Conjugate Heat Transfer (CHT) solver to calculate buoyancy-driven flows.
- Calculates thermal comfort (PMV/PPD) per ISO 7730.
I know perfectly well that a browser-based voxel solver cannot replace the rigorous RANS/LES models of OpenFOAM for final validation. My goal is strictly to provide instant, real-time feedback (~60fps) during early massing.
Links:
Live engine: https://polymère.com/
Whitepaper (math & equations): Polymère - Revealing the Invisible
My question to you (the true BPS experts):
Is there a place for ultra-fast, lower-fidelity WebGPU solvers in the conceptual design workflow? Or is the risk of “garbage in, garbage out” too high without proper meshing and traditional CFD?
I would be incredibly grateful if you could tear apart my methodology or test the UI. I’m here to learn from the best!
You can also text me directly: ilyatrofimov@polymere.press