Introduction: The Critical Role of Ventilation in Subway Stations

Subway stations are confined, high-traffic environments where thousands of passengers pass through every hour. The combination of limited fresh air supply, high occupancy, and the constant movement of trains generates complex airflow patterns and elevates the concentration of airborne particulates—ranging from dust and brake wear particles to exhaled droplets carrying pathogens. In the wake of global health concerns and tightening air quality regulations, optimizing ventilation systems has become a top priority for transit authorities. Computational fluid dynamics (CFD) software such as ANSYS Fluent offers a powerful, cost-effective means to simulate and analyze airflow and particulate transport, enabling engineers to design safer, more energy-efficient subway stations. This article explores the methodology, benefits, and practical considerations of using ANSYS Fluent to model the flow of air and particulates in subway stations, providing a comprehensive guide for engineers and researchers.

The Physics of Airflow and Particulate Transport in Subway Stations

Understanding the underlying fluid dynamics and particle behavior is essential before building a simulation. Subway stations feature complex geometries with platforms, staircases, tunnels, and ventilation shafts. The flow is typically turbulent, driven by forced ventilation (fans and jets) and natural forces such as train-induced piston effects. Air velocities can range from near-zero in stagnant zones to several meters per second near train openings. Particulates, which are often polydisperse with sizes from 0.1 to 100 µm, are transported by the airflow but also subject to gravitational settling, Brownian motion, and inertial impaction. Pathogenic droplets (e.g., from coughs or sneezes) typically range from 1–50 µm, making them susceptible to both airborne suspension and surface deposition.

Turbulence Modeling Approaches

Accurate simulation requires selecting an appropriate turbulence model. The Reynolds-Averaged Navier-Stokes (RANS) approach, particularly the realizable k-ε model, is widely used due to its computational efficiency and reasonable accuracy for indoor airflow. However, for capturing transient phenomena like piston effects or buoyancy-driven flows, the Shear Stress Transport (SST) k-ω model may be preferred. In some cases, Large Eddy Simulation (LES) can provide higher fidelity but at a much higher computational cost. ANSYS Fluent supports all these models, allowing the engineer to balance accuracy and resource constraints.

Particle Dynamics and Deposition

Particulates are tracked using the Lagrangian Discrete Phase Model (DPM) in Fluent. The equation of motion includes drag, gravity, Brownian force (for submicron particles), and Saffman lift forces. Deposition on surfaces can be modeled using empirical wall boundary conditions or by employing a near-wall treatment that resolves the viscous sublayer. For larger particles, turbulent dispersion is accounted for using stochastic tracking (random walk). The accuracy of particle tracking relies on mesh resolution near walls, especially for particles <10 µm, as they are influenced by boundary layer turbulence.

Building a Computational Fluid Dynamics Model with ANSYS Fluent

Creating a reliable CFD model of a subway station involves several systematic steps. Each step must be carefully executed to ensure the simulation results are both accurate and actionable.

Geometry and Mesh Construction

The first step is to generate a 3D computer-aided design (CAD) model of the station, including platforms, tunnels, escalator voids, ventilation ducts, and train bodies. Local geometry, such as columns and advertising panels, can significantly affect local airflow patterns. The geometry is imported into ANSYS Meshing, where an unstructured tetrahedral or polyhedral mesh is created. A boundary layer mesh (prism layers) is essential near walls to resolve the viscous sublayer for accurate turbulence and particle deposition modeling. Mesh independence studies should be performed by refining the mesh until key output variables (e.g., average air velocity at platform, particle concentration in breathing zone) change by less than 2–5%.

Boundary Conditions

Boundary conditions define the physical drivers of the flow. Typical conditions for a subway station include:

  • Inlets: Fresh air supply from ventilation shafts, often specified as velocity inlets with prescribed turbulence intensity (e.g., 5–10%).
  • Outlets: Exhaust fans modeled as pressure outlets or mass flow outlets.
  • Train-induced flow: Can be modeled as moving walls or by prescribing velocity profiles at tunnel openings. Alternatively, a sliding mesh approach can simulate train motion, but for many studies a steady-state approximation with a momentum source is sufficient.
  • Passenger presence: Passengers are often modeled as heat sources (body heat of ~100 W per person) and as obstacles that affect flow. Some studies use porous media to represent crowds.
  • Particle sources: Defined based on particle size distribution, density, and mass flow rate. For pathogen modeling, cough or sneeze events can be modeled as transient puff injections.

Solver Settings and Numerical Schemes

ANSYS Fluent offers both pressure-based and density-based solvers. For incompressible low-speed flow typical of subway stations, the pressure-based solver (coupled or segregated) is standard. Second-order upwind schemes for momentum and turbulence equations improve accuracy. Convergence is typically monitored with residuals below 1e-4 for continuity and 1e-5 for momentum. Additional monitors (e.g., average velocity at key locations, total particulate mass in the domain) help assess solution stability.

Turbulence Model Selection

As mentioned, the realizable k-ε model is a robust baseline. For more accurate near-wall behavior, the SST k-ω model is recommended. In ANSYS Fluent, the Transition SST model can also be used if laminar-to-turbulent transition areas are important, such as in regions of very low velocity. For transient simulations of train movement, the Scale-Adaptive Simulation (SAS) or Detached Eddy Simulation (DES) can resolve larger eddies while keeping computational cost lower than LES.

Simulation of Particulate Dispersion

Modeling particulates in ANSYS Fluent involves the Discrete Phase Model (DPM), which treats particles as a dilute second phase. The coupling between the continuous phase (air) and particles can be one-way (particles affected by flow but do not affect flow) or two-way (particle mass and momentum affect the flow). For typical passenger densities, one-way coupling is often sufficient because the volume fraction of particles is below 1e-6. However, if droplets from many passengers cause significant evaporative cooling or mass loading, two-way coupling may be necessary.

Particle Injection and Size Distribution

Particles are injected from sources such as train brakes, station floor dust, and passengers. A realistic size distribution (e.g., lognormal with a geometric mean of 2.5 µm and standard deviation of 2.0 for fine particulates) should be used. ANSYS Fluent allows injection of multiple particle types with different densities and diameters via the Rosin-Rammler or lognormal distribution functions. For pathogen modeling, a typical injection could be a cough event producing ~3,000 droplets per cough with sizes ranging from 0.5 to 50 µm.

Particle Tracking and Deposition

Particle trajectories are computed by integrating the force balance in a Lagrangian reference frame. Fluent's DPM offers several integration options: the automatic tracking scheme selects the best method based on particle Courant number. For turbulent dispersion, the discrete random walk model (DRW) provides stochastic velocity fluctuations. Deposition is modeled with the trap boundary condition where particles are removed upon wall contact, while the reflect condition simulates rebound. For more realistic deposition, the wall-film model can be used for liquid droplets. Studies have shown that deposition velocity depends heavily on particle size, with <1 µm particles depositing slower than larger ones.

Resuspension Considerations

Deposited particles can be resuspended by airflow disturbances (e.g., train passage, walking passengers). This phenomenon is complex and depends on particle adhesion, surface roughness, and shear stress. Advanced models like the Rock'n'Roll or Monte Carlo resuspension model can be implemented via user-defined functions (UDFs) in Fluent, but for first-order assessments, a critical shear stress threshold is often used.

Analyzing Simulation Results for Safety and Ventilation Optimization

Once the CFD simulation converges, the vast amount of data must be transformed into actionable insights. ANSYS Fluent provides robust post-processing tools, but dedicated software like ANSYS CFD-Post or third-party tools can generate detailed reports.

Key Performance Indicators (KPIs)

Typical KPIs for subway ventilation optimization include:

  • Air age: The average time air has spent inside the station. Lower air age indicates better ventilation efficiency.
  • Particle removal efficiency: Percentage of injected particles that exit via exhausts versus depositing on surfaces or remaining suspended.
  • Infection risk index: For pathogen-laden particles, the concentration in breathing zones (1.5–2.0 m above platform) weighted by exposure time.
  • Velocity uniformity: Standard deviation of velocity magnitude on the platform; high uniformity reduces stagnant zones.
  • Energy consumption: Total fan power required to achieve target KPIs.

Identifying Stagnant Zones and Short-Circuiting

Visualization of streamlines and particle tracks often reveals regions of recirculation or short-circuiting where supply air exits directly via exhausts without mixing in occupied areas. For example, supply diffusers placed near exhaust grilles can create a short circuit, wasting energy. ANSYS Fluent's contour plots of mean age of air and particle concentration clearly highlight these problem areas. Engineers can then relocate or reorient diffusers, increase supply air velocity, or add jet fans to eliminate stagnant zones.

Optimization via Parametric Studies

ANSYS Fluent's parametric analysis tools allow users to sweep multiple variables (e.g., supply flow rate, exhaust location, number of fans) and compare KPIs. Design of Experiments (DOE) methodologies can reduce the number of runs. For each configuration, the simulation is set up using journal files, and results are automatically collected. The optimal configuration minimizes particle concentration in the breathing zone while meeting energy budget constraints. For instance, a study might find that increasing the supply flow rate by 20% reduces the infection risk index by 60% while only increasing fan power by 15%.

Case Studies and Real-World Applications

Numerous transit agencies have leveraged ANSYS Fluent to improve subway ventilation. One notable example is the New York City Transit Authority, which used CFD to evaluate the effectiveness of existing ventilation fans in controlling particulate matter (PM2.5) on platforms. Simulations showed that fan placement near tunnel portals was more effective in diluting train-related particulates than fans located solely on platforms. Another study modeled droplet dispersion from passengers across an entire station and found that increasing the air change rate from 6 to 12 ACH could reduce the peak concentration of viral droplets by 85% within 10 minutes. In London, the subway system used CFD to optimize the positioning of portable air purifiers during the COVID-19 pandemic, achieving up to 50% reduction in particle count in high-traffic areas.

These real-world applications demonstrate the power of CFD in making evidence-based decisions. A well-validated model can replace costly physical experiments and guide retrofits with confidence. For further reading, see ANSYS Fluent product page for technical capabilities and the WHO Guidelines on ventilation in indoor spaces for benchmark air change rates.

Considerations for Model Validation and Verification

A CFD model is only as good as its validation. Engineers should compare simulation results with field measurements (velocity profiles, particle number counts) from actual stations. Use test cases such as the mock-up station experiments or literature data (e.g., from the ASHRAE standards). Uncertainty quantification should be performed by perturbing input parameters (e.g., particle density, boundary layer mesh height) and observing the variation in KPIs. ANSYS Fluent's mesh adaption tools can help ensure grid independence. Confidence in the model is achieved when simulation-to-measurement agreement is within ±10% for key metrics.

Limitations and Challenges in CFD Modeling of Subway Stations

Despite its power, CFD modeling of subway stations faces several limitations. The computational cost of simulating the entire station at high resolution, especially for transient events like a train passing every two minutes, can be prohibitive. Many studies simplify by using steady-state or coarse meshes. Additionally, the model may not capture all real-world complexities such as variations in passenger movement, door opening patterns, or external weather conditions (wind, temperature). The polydisperse nature of particulates and their chemical composition (hygroscopic growth, evaporation) require additional modeling, often via UDFs. Finally, the choice of turbulence model and particle tracking parameters can introduce significant uncertainty if not carefully calibrated. Despite these challenges, CFD remains the most practical tool for optimizing ventilation when combined with field validation.

Future Directions: Coupling CFD with Machine Learning and Real-Time Monitoring

Emerging trends involve integrating CFD with machine learning to create digital twins of subway stations. A real-time sensor network feeding data into a reduced-order model (ROM) built from ANSYS Fluent simulations can predict particulate distributions and automatically adjust ventilation setpoints. This approach promises both energy savings and adaptive safety measures. Additionally, advances in high-performance computing (HPC) now allow full transient LES of entire stations, enabling more accurate predictions of peak contaminant exposure. As ANSYS Fluent continues to evolve with GPU solvers and cloud-based HPC, the barriers to high-fidelity subway ventilation modeling will further diminish.

Conclusion

Modeling the flow of air and particulates in subway stations using ANSYS Fluent is an indispensable methodology for improving passenger safety and ventilation efficiency. By systematically constructing geometry, setting boundary conditions, selecting appropriate turbulence and particle models, and analyzing performance indicators, engineers can optimize ventilation designs to reduce contamination and energy use. While challenges remain in computational cost and model validation, the benefits—enhanced public health, regulatory compliance, and lower operational costs—make CFD a cornerstone of modern subway infrastructure planning. As urban transit networks expand, the application of advanced simulation tools will only grow in importance, helping to create safer, healthier underground environments for millions of commuters worldwide.

For engineers new to the domain, starting with small-scale models and gradually increasing complexity is recommended. Resources such as the ANSYS Academic Program offer tutorials and case studies specific to indoor airflow. Investing in proper validation against field data ensures that the simulation results are trustworthy, leading to design decisions that genuinely improve air quality and safety in subway stations.