Methods & citations
Eddy3D writes cases for solvers other people wrote — the physics belongs to the engines, and results cite the engine (OpenFOAM, urbanMicroclimateFoam, OpenLB, FluidX3D, Radiance, EnergyPlus; see the credits). What Eddy3D's own papers own is the workflow wrapped around those engines: how the mesh is built, how a year of weather becomes a handful of solves, how a facade pressure reaches an airflow network. A study that relies on one of these methods cites the paper next to the engine.
The preferred general citation for Eddy3D is Kastner & Dogan (2022), Building and Environment 212:108639, doi:10.1016/j.buildenv.2021.108639. The repository ships a machine-readable CITATION.cff (GitHub's Cite this repository button), and the full list is on the publications page.
Where a citation reaches you
Each published method is cited in three places, all fed from one table in the code (MetaFOAM.Lib/Eddy3DCitations.cs), so they cannot drift apart:
- The component's description — the tooltip in Grasshopper ends with
Method: …, and the component pages on this site mirror that text (the exporter renders them from the live components). - This page — the map from paper to components to dictionaries below.
- The generated OpenFOAM dictionaries that implement a method carry a
// Method: …line directly under the Eddy3D banner, so a case folder found in the wild names the paper next to the plugin version that wrote it. Line comments only — the parser strips them on re-read.
// ------------------------------------------------------------------------- //
// Generated by Eddy3D 1.10.0.827 — https://www.eddy3d.com
// ------------------------------------------------------------------------- //
// Method: Kastner & Dogan (2020), Journal of Building Performance Simulation 13(1):59-68, doi:10.1080/19401493.2019.1692906
/*--------------------------------*- C++ -*----------------------------------*\
========= |
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
...
Published methods → components → dictionaries
| Paper | Published method | Components | Generated dictionaries |
|---|---|---|---|
| Kastner & Dogan (2020), J. Build. Perform. Simul. 13(1):59–68, doi:10.1080/19401493.2019.1692906 — precursor: Kastner & Dogan (2018), eSim 2018, IBPSA | Cylindrical meshing. One cylindrical blockMesh serves every wind direction; the cylinder's side patches switch between inlet and outlet per direction, so a study of n directions meshes once instead of n times. Validated against wind-tunnel data; recommended from eight directions up. | Cylinder Domain, Outdoor Case | mesh/system/blockMeshDict; every case_NNN/0/* field file that carries the switched side patches |
| Kastner & Dogan (2022), Build. Environ. 212:108639, doi:10.1016/j.buildenv.2021.108639 — precursor: Kastner & Dogan (2019), Building Simulation 2019, doi:10.26868/25222708.2019.210458 | Decoupled annual outdoor thermal comfort. Hourly wind directions are clustered (k-means) into a simulation budget; each RANS solve becomes a spatial lookup table; velocity amplification factors turn the lookup into an hourly wind-speed matrix; trees are Darcy–Forchheimer porous zones; MRT comes from a Radiance two-phase DDS shortwave chain plus view-factor longwave exchange; hourly UTCI closes the loop. Validated against on-site weather stations. | Wind Rose Cluster, Velocity Amplification Factors (VAF), Tree, MRT, MRT Solve, UTCI (Simulation) | case_NNN/constant/fvModels (tree porosity) |
| Dogan, Kastner & Mermelstein (2021), Build. Environ. 196:107762, doi:10.1016/j.buildenv.2021.107762 | Surfer. Discretises every model surface, computes probe-to-surface view factors with a ray-cast occlusion test, simulates exterior surface temperatures in EnergyPlus, and assembles MRT from surface temperatures, sky temperature and a solar-gain ΔMRT (effective radiant field). RMSE 1.8 °C (surfaces) and 2.0 °C (MRT) against measurements at ~80× the speed of a full simulation. | MRT View Factors, SurfaceTemp (EnergyPlus), MRT Solve, MRT | — (Radiance and EnergyPlus inputs, not OpenFOAM) |
| Dogan & Kastner (2021), Build. Simul. 14(4):1189–1200, doi:10.1007/s12273-020-0727-x — precursor: Dogan & Kastner (2018), IBPC 2018, doi:10.14305/ibpc.2018.ms-5.05 | Facade pressure coefficients for airflow networks. An automated OpenFOAM wind study writes the surface pressure coefficient Cp = (p − p∞)/(½ρU∞²) for arbitrary building shapes and context, probed on the facades and exported as the wind-pressure-coefficient arrays an EnergyPlus AirflowNetwork expects. | Run Settings → Pressure Coefficient, Airflow Network Cp | case_NNN/system/Cp; the pressureCoefficients function object in case_NNN/system/controlDict |
| De Simone, Kastner & Dogan (2021), Building Simulation 2021, doi:10.26868/25222708.2021.30632 | Airborne-pathogen passive scalar. An isothermal OpenFOAM ventilation case carries an airborne-pathogen tracer as a passive scalar injected from an emitter zone, so plan layouts and inlet/outlet positions can be compared for exposure. Benchmarked against a CFD test case and a tracer-gas restaurant study. | Indoor Case, Viral Emitter | system/controlDict (scalarTransport_viral), constant/fvModels (the source) |
| Kastner & Dogan (2023), Build. Environ. 242:110384, doi:10.1016/j.buildenv.2023.110384 | GAN wind surrogate. An image-to-image GAN trained on RANS results predicts a pedestrian-level wind field from a rasterised building layout in a fraction of a second — the hosted Yel 1.0 model. | Wind Predictor (Cloud) | — |
| Kastner, Chen, Manivasakan, Michalatos, Karadag & Dogan (2026), SSRN 6401886, doi:10.2139/ssrn.6401886 | Physics-informed GAN with arbitrary input and output channels — the 8-channel ONNX family (Yel 2.0, Esen, Poyraz) that runs locally over arbitrary points and several wind directions. | ML Model, Wind Predictor, Wind Comfort Predictor (ML) | — |
| Kastner & Dogan (2020), SimAUD 2020, 405–412, PDF | Thermal bridges with OpenFOAM. Conjugate heat transfer through layered assemblies solved with chtMultiRegionFoam, set up from Grasshopper geometry. | CHT Case | system/controlDict of the CHT case |
| Rahimi, Alvarez, Stone, Berardi & Kastner (2025), Build. Environ. 285:113637, doi:10.1016/j.buildenv.2025.113637 — precursor: Rahimi, Berardi, Stone & Kastner (2024), IBPC 2024 | Validation, not a method. urbanMicroclimateFoam (heat, air, moisture, radiation, vegetation), run through Eddy3D Outdoor+, against weather stations on the Georgia Tech campus: temperature RMSE down to 1.03 °C, humidity RMSE 4.78 %, vegetation-driven cooling up to 2.1 °C. The physics stays Kubilay, Derome & Carmeliet's — cite the solver for results. | Outdoor+ Case | — (the dictionaries are the solver's) |
Related work that is not an Eddy3D feature
- Rahimi, Kastner & Berardi (2025), Building and Environment 276:112858, doi:10.1016/j.buildenv.2025.112858 — convective heat transfer and humidity in unsteady microclimate CFD with a customised
buoyantHumidityPimpleFoam. A solver study; Outdoor+ models humidity through urbanMicroclimateFoam's own HAM physics instead. Also: Rahimi, Kastner & Berardi (2024), How much computational complexity is necessary to model relevant aspects in microclimate urban physics?, IABP 2024. - Karadag, Smyl & Kastner (2026), Sustainable Cities and Society 148:107614, doi:10.1016/j.scs.2026.107614 — the review the 11 | ML panel sits in: where surrogate wind models stand, and what they can and cannot replace.
Applications of the annual comfort workflow
Studies that used Eddy3D rather than extended it — useful as worked examples of the decoupled workflow at neighbourhood and city scale:
- Young, Kastner, Dogan, Chokhachian, Mokhtar & Reinhart (2021), Building and Environment 208:108577, doi:10.1016/j.buildenv.2021.108577 — outdoor thermal comfort along cycling routes at varying levels of physical accuracy, against bike-ridership counts in Cambridge, MA.
- Natanian, Kastner, Dogan & Auer (2020), Energy and Buildings 224:110283, doi:10.1016/j.enbuild.2020.110283 — annual outdoor thermal comfort autonomy and energy balance across Mediterranean typologies, using the hourly wind factors.
- Kastner & Dogan (2020), SimAUD 2020, 85–91, ResearchGate — predicting space usage on a university campus by multi-objective assessment of outdoor thermal comfort.