3DAirQualityPrediction

1. Objectives and goals

Tha main goal is to Develop 3DAirQualityPrediction, an accurate, fast, modular, easy-to-use software framework for urban AQ prediction.

2. Main features of the framework

  • Modular framework with exchangeable solvers for modules

  • Standardized input-output data formats for each module

  • Availability in HPC oriented cloud e-infrastructure

  • Detailed modelling for AQ prediction at urban scale

  • Real 3D geometry

  • Vehicular traffic

  • CFD for dispersion → RANS and LES/DNS

  • Data integration (e.g. meteorology, vehicular traffic, authorities)

  • Validation

3dairq1
Figure 1. Fig.1. Overview of the present framework and its modules
Fig.2. 3D CAD from shape files with Blender scripts

3dairq2

2.1. Preprocessing of the geometrical data: meshing

3dairq3
Figure 2. Fig.3. Meshing with ANSA, ANSYS and/or in-house parallel octree mesher

2.2. Preprocessing of the geometrical data: parametric CAD and meshing by scripts

  • Parametrization of OSPM geometry to 3D CAD geometry

  • OSPM street configuration converted to CAD geometry by script using some additional parameters (for the 3D model size)

  • CFD compatible mesh generated from CAD geometry

2.3. Preprocessing of the geometrical data: fitting the traffic geometry to the CFD mesh

Emission source location fit from traffic model to CFD mesh, calculated with in-house Java program Parametric lanes defined by some measures (distances) OpenStreetMap and the national road authority’s format

3dairq4
Figure 3. Fig.4. Openstreetmap data

2.4. The traffic and emission modules

  • Traffic is modelled by SUMO based on calibrated historical data or by interpolating measurements from operational road traffic data

  • Emission model Copert 4 implemented in Java based on real regional fleet data

2.5. The meteorology modules

AROME model running by OMSZ (the Hungarian Meteorological Services) for the whole Carpathian basin Resulting wind fields at different heights for the demonstration area, selected automatically

2.6. The dispersion simulation

Used models (until now): ANSYS Fluent, OpenFOAM and Parmod CFD model components:

  • 3D RANS k-ε turbulence model with calibrated coefficients

  • Humidity

  • Parks and groves as porous zones

  • Meteorological wind data as inlet boundary condition

  • Initialized with wind and temperature profiles

  • Polyhedral and hex core meshes, with/without boundary layer resolutions

  • Full transient and frozen flow field models

3dairq5
Figure 4. Fig.5. The dispersion module: NOx