Starting question
Could the completed fan and shaft arrangement ventilate every floor and preserve usable escape conditions during the selected fires?
3Basement levels
1,135Vehicle capacity
59Jet fans
22.57MMesh elements
4 MWVehicle fire
180 sControl activation
01 The brief
The question was not whether air moved. It was whether it reached the right floor.
The study examined a three-level underground car park under routine ventilation and fire conditions. The model had to distinguish local circulation created by jet fans from the net supply and extract flow that renews each floor.
Model scope
Levels B2, B3, and B4; ramps, cores, openings, six main shaft boundaries, 59 induction fans, and three fire scenarios.
Performance checks
Six air changes per hour in normal operation, ten in fire mode, occupied-height velocity, smoke movement, and escape-path exposure.
My role
Prepare the CFD model, define the numerical method and boundary conditions, interpret the results, and turn them into design recommendations.
02 Decision path
Six decisions turned plans into an engineering diagnosis.
The useful result was not a colourful contour. It was the chain of checks that connected architecture, fan duties, numerical physics, and evacuation risk.
-
Turn the building into a fluid domain
What geometry controls the air?
The architectural and mechanical plans were inverted into connected air volumes. Ramps, cores, columns, shafts, vents, fan locations, and escape routes were retained where they could alter the flow.
Output · Three connected floor domains representing B2, B3, and B4.
-
Define success before looking at contours
What has to be true?
Average air renewal and local conditions answer different questions. The model therefore combined floor-level air-change targets with velocity and species checks along occupied and escape-path elevations.
Output · 6 ACH in normal operation, 10 ACH in fire mode, plus occupied-height flow and tenability review.
-
Put resolution where gradients matter
Where can a coarse mesh hide the answer?
Local refinement was concentrated around inlets, jet-fan outlets, fires, openings, and geometry changes. Less sensitive regions used larger elements to keep the three-floor model computationally practical.
Output · 4.19 million nodes and 22.57 million elements, with reported element sizes from 50 to 500 mm.
-
Separate routine flow from fire physics
Which physics does each operating mode require?
Normal ventilation was solved as a steady-state airflow problem. The fire cases were transient and added gravity, temperature-dependent density, energy, and species transport to represent buoyant smoke and combustion products.
Output · A standard k–ε ventilation model and transient fire models with heat and species transport.
-
Compare circulation with mass balance
Are the fans solving the same problem as the shafts?
Velocity plots showed that the jet fans formed useful local paths. The floor flow balance revealed a different system-level result: B2 and B3 received excess air while B4 remained below target.
Output · The governing failure moved upstream from individual jet fans to floor-by-floor supply and extraction.
-
Test the system where people depend on it
Does the control sequence preserve escape time?
Three 4 MW vehicle-fire cases were positioned to challenge long escape paths and the principal fire-fan influence. Conditions were compared before and after the reported 180-second control activation.
Output · A time-based review of smoke control, exposure, evacuation assumptions, and recovery after fan activation.
03 Air balance
Two floors had excess airflow. B4 missed both targets.
The same imbalance appeared in both operating modes. This made the diagnosis stronger: the issue was not a single fan setting but the way the shared supply and extract system distributed capacity across the building.
Normal ventilation · target 6 ACH
| Level | Calculated | Result |
|---|---|---|
| B2 | 16.75 ACH | Pass |
| B3 | 11.76 ACH | Pass |
| B4 | 3.98 ACH | Fail |
Fire operation · target 10 ACH
| Level | Calculated | Result |
|---|---|---|
| B2 | 38.75 ACH | Pass |
| B3 | 27.39 ACH | Pass |
| B4 | 9.12 ACH | Fail |
Governing insight
Local velocity can look healthy while the floor still receives too little replacement air.
The jet fans redistribute momentum inside a level. They cannot correct a supply-and-extract system that sends too much capacity to B2 and B3 and too little to B4.
The source record uses different floor volumes in the normal and fire calculations and also contains conflicting total-area values. The professional report preserves the stated results and identifies those discrepancies for confirmation.
04 The diagnosis
The fan count was not the governing design variable.
B2 was closest to the principal mechanical influence and received the largest flow. B4 depended on shared shafts and indirect paths, yet had the longest modeled escape routes and the lowest renewal rate.
05 Fire scenarios
A 4 MW fire turned air balance into evacuation time.
Fire locations were selected near long escape paths and areas expected to challenge extraction. Each source represented a sprinklered-car-park vehicle fire over a 2 m by 5 m footprint.
Case 01
One fire on B4
A single 4 MW source tests a critical lower-level escape path.
Case 02
Two fires on B4
Two simultaneous 4 MW sources test the most severe lower-level case.
Case 03
One fire on B2
A single 4 MW source tests the level closest to the principal fire-fan influence.
4 MW
Each modeled vehicle fire used a 0.153 kg/s fuel mass-flow rate. The source record states concentration limits of 50 ppm CO and 5,000 ppm CO₂ for its escape-path review.
Time-history finding
By 180 seconds, the reported limits were exceeded in every case.
The activated controls did not recover the modeled environment. The source narrative says two cases were acceptable at 90 seconds and the remaining case was near the boundary, but its floor labels conflict with the listed scenarios; that ambiguity is retained rather than guessed away.
06 Design response
Redesign distribution before adding more machinery.
Because B2 and B3 already had substantial excess capacity, the first response is to control where the air goes, how smoke is collected, and when the system changes mode.
07 What comes next
The model found the design question. A validation phase would close it.
A professional design-stage update would test whether the diagnosis survives numerical, operational, and fire-safety uncertainty.
Professional project report
The complete engineering record.
The original 2021 analysis has been rewritten as a focused 18-page report with the model setup, calculations, results, design diagnosis, recommendations, and source-data reconciliation.