Personal project · 2026.7-2026.10

Data center thermal simulation
and operations analysis

I built a six-rack thermal and airflow model to compare airflow, rack loads and cooling-route changes for maintenance assessment.

COMSOL 6.2Java APIPython

Room model: 6racks
Cooling scenarios: 4scenarios
Monitoring data: 32points

Project context

Assessing rack temperatures
with one cooling route shut down

I modeled six racks supplied by two cooling routes, then compared reduced airflow, adjusted rack loads and shutdown of one route.

Room dimensions: 6 × 5 × 3 m
Total rack load: 30 kW
Supply temperature: 20°C
Normal airflow: 3.6 m³/s

Six-rack floor plan, central intake aisle, and supply and return port heights — Cool air enters at low level and returns at high level; racks draw air from the central aisle.

My contribution

01
Built the thermal and airflow model
Built a six-rack room model with two cooling routes in COMSOL, defined rack loads, materials and airflow boundaries, and modeled airflow, heat transfer and buoyancy.
02
Designed and calculated four scenarios
Used normal airflow as the reference and varied total airflow, rack load distribution and cooling routes. Completed 0–23 s transient calculations for four scenarios and compared intake temperatures and local hotspots.
03
Automated model calculations
Used Java API for modeling, solving and data export, and Python to manage calculation runs, check saved times, units and data completeness, and generate comparison plots.
04
Documented monitoring data and maintenance workflows
Mapped temperature and airflow data to 8 equipment assets and 32 monitoring points, compiling 3,072 simulation records. Used rack temperature differences to document events, problems, risks and maintenance-change workflows.

Results

Higher rack temperatures with one route shut down

At 23 s, R4’s average intake temperature is 11.71°C above normal airflow; R5’s peak intake temperature is 44.94°C.

Higher local peaks after load redistribution

After redistributing rack loads, average intake temperatures fall across six racks, while the highest sampled room temperature rises from 34.15°C to 38.26°C.

One route shut down

Route 1 supply and return are closed; route 2 maintains 1.8 m³/s.

Average rack intake temperature · °C

R1

37.31°C

Average intake temperature · 23 s

vs. normal airflow: +11.33°C
Peak intake temperature: 43.90°C

At 23 s, R4’s average intake temperature is 11.71°C above normal airflow; R5’s peak intake temperature is 44.94°C.

Average rack intake temperatures · 23 s

Normal airflowOne route shut down

R1 · intake temperature

Normal airflowOne route shut down

Coarse-grid simulations over 0–23 s, with each scenario initialized separately.

Scenario temperature summary at 23 s
Scenario	Supply · m³/s	Highest average intake · °C	Highest intake peak · °C
Normal airflow	3.6	25.98	31.67
Reduced airflow	2.7	26.79	33.31
Adjusted rack loads	3.6	25.19	31.31
One route shut down	1.8	37.31	44.94

Temperature samples and horizontal velocity at 23 s and z = 1.1 m for four scenarios, with a common temperature scale — Temperature and horizontal velocity · 23 s · z = 1.1 m · shared color scale.

Average intake temperatures of six racks in four cooling scenarios at 23 s, with numeric values and temperature shading — Average intake temperature · °C · simulation at 23 s.

Average intake temperature curves for six racks across four cooling scenarios, 0–23 s — 0–23 s transient calculations, each scenario starting from its own initial temperature field.

Data center thermal simulation
and operations analysis

Project context

My contribution

Built the thermal and airflow model

Designed and calculated four scenarios

Automated model calculations

Documented monitoring data and maintenance workflows

Results

Higher rack temperatures with one route shut down

Higher local peaks after load redistribution

One route shut down

R1

Temperature and airflow distribution

Rack temperature overview

Six-rack temperature curves

Account and Access